Total Recall: One man’s loss is another man’s gain? An analysis of the effectiveness of marketing mix instruments during a competitor product-harm crisis.

Total Recall: One man’s loss is another man’s gain?

An analysis of the effectiveness of marketing mix

instruments during a competitor product-harm crisis.

by

Parameswaram, D. S3033287

2

Total Recall: One man’s loss is another man’s gain?

An analysis of the effectiveness of marketing mix

instruments during a competitor product-harm crisis.

by

Dion Parameswaram University of Groningen Faculty of Economics and Business MSc Marketing Intelligence & Management

Master thesis January 14, 2018 Parkweg 147a 9727 HB Groningen +31 (0) 6 18 82 96 56 d.parameswaram@student.rug.nl Student number: 3033287

3 Abstract

A product-harm crisis is among the biggest challenges companies face nowadays. The number of product-harm crises has grown exponentially over the last few decades, and are expected to grow in the future. The stricter product-safety legislation, more demanding customers and increasing complexity of products will lead to more frequent product-harm crises in the foreseeable future.

Existing academic literature focuses mainly on the effect of product-harm crises for the focal brand, but not much literature is devoted to the effect of a product-harm crisis on a competitor’s marketing effectiveness in the same category. The author aims to expand upon the existing literature and to provide marketing managers with actionable insights by answering the following research question: ‘How is the effectiveness of a firm’s marketing mix instruments influenced by a competitor’s product harm crisis?’ The follow-up question asked in this study is ‘Which marketing mix instruments are most effective during a product-harm crisis?’

The author proposes an error correction model to determine the effectiveness of marketing mix variables during a competitor product-harm crisis in both the short and long run. The proposed model is used on four years of weekly sales data for the Dutch fast moving consumer goods market between 1994 and 1998.

Pooling is not allowed for the dataset in this study, but by means of the added Z method, across-brand parameter estimates are possible. Findings suggest that a competitor product-harm crisis does not influence the effectiveness of a firm’s marketing mix instruments. The author proposes that this might be due to the non-perishable nature of the product categories. In addition, the occurrence of a product-harm crisis itself does not impact category sales. Based on the results of this study, the author suggests marketing managers to act like business is as usual.

4 Preface

This is it. The fruits of my labor after six and a half years of studying in Groningen. In front of you lies my Master thesis; Total Recall: One man’s loss is another man’s gain? An analysis of the effectiveness of marketing mix instruments during a competitor product-harm crisis. It is the result of pursuing a Master degree in the field of Marketing Intelligence and Management after obtaining my bachelor’s title at the Hanze Hogeschool Groningen.

This Master thesis was written under the supervision of dr. ir. Maarten Gijsenberg. His close supervision, concise feedback, and attention to detail have made my thesis live up to the Master title. Looking back, I can safely say that it was a good choice to pick his topic ‘Timing is money’ as the subject for the last part of my education. The topic covered time series modeling, which has not been discussed extensively in our curriculum. For this reason, the methodology was overwhelming at first, but due to close cooperation with my fellow group members and the excellent guidance from Maarten, I was able to succeed.

I would like to thank Melvin Bredewold and Marc Boels for spending good times with each other during our time at the University of Groningen and for spending most of our disposable income on soup in the cafeteria (even though they re-introduced the croutons mere days before we hand in our Master thesis). These two young fellows helped make the Master more tolerable and definitely a lot more fun. In addition, I’d like to thank my fellow group member Pim Menkveld for the close

cooperation during our Master thesis. We have spent many hours on our thesis and I firmly we can be proud of the results.

I hope you have as much fun reading my thesis as I put in the effort for the final version. Dion Parameswaram

5 Table of Contents

1. Introduction ... 6

2. Theoretic framework ... 9

2.1 Product-harm crises and the focal brand ... 9

2.2 Competitors and product-harm crises ... 10

2.3 Conceptual model ... 12

3. Data and Methodology ... 13

3.1 Data description ... 13

3.1.1 Data cleaning and variable operationalization ... 13

3.1.2 General descriptive insights ... 16

3.2. Error Correction Model ... 16

3.2.1 Model ... 16 3.2.3 Generating insights ... 19 4. Results ... 20 4.1 Model diagnostics ... 20 4.2 Model results ... 21 4.2.1 Short-term effects ... 21 4.2.2 Long-term effects ... 22

5. Conclusion and recommendations ... 24

5.1 Summary of findings ... 24

5.2 Managerial implications ... 25

5.3 Limitations ... 25

5.4 Future research directions ... 26

6 1. Introduction

On the 29th of March, 1994 Unilever proudly presented its new solution for removing particularly persistent stains: Omo Power (Persil Power in other parts of the world). Unilever claimed that Omo Power was able to remove stains, even at low temperatures, by using a new chemical accelerator. This claim lead to a fast increase in market share of 8%, which increased revenue by millions of dollars (Volkskrant, 2002). Shortly after the introduction of their new laundry detergent Unilever’s biggest competitor, Procter & Gamble claimed they had found that Omo Power could create holes in the clothing of consumers. This claim resulted in one of the largest recalls in the Netherlands and tainted Unilever’s brand name for years in Dutch households. Newspapers of that time were already referring to this product-harm crisis as one of the biggest marketing blunders in recent history (NRC, 1994).

A product-harm crisis is among the biggest challenges companies face nowadays. Product-harm crises can be defined as well publicized events wherein products are found to be defective or even dangerous (Dawar and Pillutla 2000; Siomkos and Kurzbard, 1994). Cleeren, Dekimpe, and Van Heerde (2017) expand upon this definition by taking into account that media attention to the crisis differs among product categories. In addition, some crises do not affect all potential customers, such as erroneous labeling of ingredients within products which may lead to allergic reactions for a part of a company’s customers base. Based upon the reasoning by Cleeren, Dekimpe, and Van Heerde (2017) this research adopts their definition of a product-harm crisis: a product-harm crisis is a discrete event in which products are found to be defective and therefore dangerous to at least part of the product’s customer base. A follow-up action resulting from a product-harm crisis is a product recall, which is defined as an action by a manufacturer or distributor to remove a product from the market.

The number of product-harm crises has grown exponentially over the last few decades, and are expected to grow in the future. The stricter product-safety legislation, more demanding customers and increasing complexity of products will lead to more frequent product-harm crises in the foreseeable future (Dawar & Pillutla, 2000; NRC, 1995). Occasionally, a product-harm crisis affects the entire product category such as the fipronil crisis for the eggs category in 2016.

7 from brands such as Toyota and Tesla (USA Today, 2012). In the case of Omo Power, it can lead to an irreversible loss of market share.

Existing academic literature focuses mainly on the effect of product-harm crises for the focal brand, i.e., the brand experiencing the crisis. Not much literature is devoted to the effect of a product-harm crisis and the subsequent product recall on a competitor’s marketing effectiveness in the same category. Since product-harm crises are more likely to occur in the future, the potential effect on the effectiveness of competitor’s marketing instruments is of increasing importance. After all,

companies do not act in a vacuum. A product recall happening to a competitor may have detrimental effects on the product category and lead to negative category demand or vice versa. In the case of product harm crises, is one man’s loss really another man’s gain?

This leads to the following research question and sub-question:

- How is the effectiveness of a firm’s marketing mix instruments influenced by a competitor’s product harm crisis?

- Which marketing mix instruments are most effective during a competitor’s product-harm crisis?

From a managerial point of view, it is crucial to understand the potential effects of a competitor’s product-harm crisis on the company’s brand. According to Siomkos, Triantafillidou, and Tsiamis (2010) competitors should have a crisis plan available which can be executed directly after a crisis occurs. But is increasing marketing efforts effective during a competitor’s product-harm crisis? Which marketing mix instruments are most effective in such a situation? Whenever a product-harm crisis occurs, the affected company’s marketing instruments become less effective and its competitiveness also decreases (Lock, 2008). These consequences can potentially turn the affected brand into easy prey. By increasing marketing efforts the competitors might be able to increase their own market share at the cost of the affected brand. Previous studies in this field have only looked at competitor effects in a controlled lab experiment. This research aims to provide managers insight into whether marketing effectiveness increases when a product-harm crisis occurs for a competing brand using real-life product-harm crises.

8 2011). As such, these findings lack generalizability (Cleeren, van Heerde & Dekimpe (2012). By

looking at multiple product-crises in different product categories, this current research aims to expand upon the current literature and academic findings and to provide marketing managers with actionable insights.

Secondly, other studies that are not based on the Kraft peanut butter incident focus mainly on lab experiments (Dawar & Pillutla, 2000; Siomkos, Triantafillidou, Vassilikopoulou & Tsiamis, 2010) Lab experiments often suffer from low external validity by creating a situation which is not true to life. As a consequence, results from lab experiments are often not generalizable (Bracht & Glass, 1968). According to Kempthorne (1961), lab experiments have to make a distinction between the

experimentally accessible population and the actual target population. Generalizing results from lab experiments is being made more difficult due to this distinction since characteristics of both

populations differ (Kempthorne, 1961). This research is one of the few in the area of product-harm crises that uses real-life data and is thus able to add empirical knowledge to the existing academic literature.

Lastly, Cleeren, Dekimpe, and van Heerde (2017) state that product harm crises may spill over to other, non-affected brands, but not much is known about the impact of the crisis on competitors and their marketing effectiveness. To better understand current research and their results, it is needed to replicate findings in order to understand generalizations better. To the best of my knowledge, this research is the first to address all of these issues and thus adds considerable empirical knowledge to the existing literature.

9 2. Theoretic framework

This chapter will elaborate upon concepts given in the introduction and also look at the current academic landscape concerning product-harm crises. First, the relevant literature of product-harm crises and the focal brand will be discussed. Secondly, the existing knowledge regarding competitors and product-harm crises will be discussed in detail. The chapter closes by providing a conceptual model based on the theoretic framework.

2.1 Product-harm crises and the focal brand

The existing academic literature is very much focused on the focal brand, i.e., the brand that is experiencing the crisis. Less is known about the impact of a product-harm crisis on competitors in the same product category. This research first looks at relevant existing theory concerning crises in general and competitor actions. During a crisis, firms are likely to only look out for their own interests, especially when it threatens their survival (Thomas, 1979). A crisis offers competitors an opportunity to look out for their own interests, at the expense of the focal firm. However, taking advantage of a firm during a crisis might not be viewed as ethical and can have an adverse effect on the firm (Cleeren, Van Heerde, and Dekimpe, 2012). In addition, corporate social responsibility is becoming increasingly more important over the last few years and is high on the agenda for both firms and consumers (Lindgreen & Swaen, 2010). Consumers nowadays hold corporate social responsibility in high regard. Due to the increased level of globalization, firms are more likely to be placed in a bad spotlight when undertaking actions at the expense of other firms (Carroll, 2015). Thus, customers are more likely to punish companies that act irresponsibly by not engaging in repeat purchases (Bhattacharya & Sen, 2004).

When zooming into the relevant marketing literature, we find that existing academic literature focuses mainly on the focal brand that is dealing with a product-harm crisis. Van Heerde et al. (2007) found significant results for a quadruple jeopardy for the focal brand following a product-harm crisis: loss of baseline sales, loss of effectiveness of own marketing instruments, increased cross-sensitivity to competitor’s marketing instruments, and decreased cross-impact of own marketing instruments. The loss of baseline sales and loss of effectiveness of own marketing instruments directly affect the focal brand’s competitors. These consequences offer competitors a chance to capitalize on the product-harm crisis by increasing their own marketing efforts.

10 & Pons, 2009). Van Heerde et al. (2007) state that due to this loss in brand equity consumers have an increased sensitivity to competitors’ marketing efforts. The increased sensitivity to competitor’s marketing efforts might lead to consumers switching from the focal brand to a competitor brand. Cleeren, Dekimpe & Van Heerde (2017) state that while there have been studies about the effect of product-harm crises on the focal brand, less is known about the impact of the crisis on competitors. In addition, Borah and Tellis (2016) state that: ‘Brands may not only face product-harm crises through their own wrongdoing, but they may also be affected negatively by crises occurring with their competitors. When looking at online chatter behavior, their study found that when one brand in a product category suffered from a product-harm crisis, negative chatter for competitor brands increased. Roehm & Tybout (2006) show that brand scandals affect the entire product category when other brands are perceived as guilty by association. The authors also argue that category spillover effects are more likely for negative events. This argument can be expanded to the context of

product-harm crises. Freedman et al. (2012) support these findings and also find that product recalls can have industry-wide spillovers in the form of a loss in baseline sales for the entire industry. These claims are further supported by Cleeren, Dekimpe & Helsen (2013), who find similar results. These studies reinforce the theory that product-harm crises not only affect the focal brand but competitor brands as well. Therefore, this study expects that the occurrence of a product-harm crisis negatively impacts the product category.

2.2 Competitors and product-harm crises

Competitors seem to have the tendency to view a product-harm crisis for the focal brand as an opportunity to gain more market share. Examples of this include the recall of faulty Firestone Tires in the U.S. Michelin, one of their biggest competitors, used this recall to become one of the leading brands in tires in the U.S. (New York Times, 2000). Despite this anecdotal evidence, there is little academic research on the effect of product-harm crises on competitors (Cleeren, Dekimpe & Van Heerde, 2017). Exceptions are Cleeren, van Heerde & Dekimpe (2013), Siomkos, Triantafillidou & Tsiamis (2010) and Bala, Bhardwaj & Chintagunta (2015).

11 Cleeren, van Heerde & Dekimpe (2013) look at price and advertising and not at other marketing mix variables such as the use of display. Findings from Ailawadi et al. (2009) suggest that the general consensus regarding display advertising is that is positive contributes to the sales elasticity.

Therefore, this research expects that not only price and advertising have an increased positive effect on sales during a product-harm crisis, but that the same will hold for display.

Secondly, Siomkos, Triantafillidou, and Tsiamis (2010) state that not only consumers of the affected company, but also consumers of competitor firms are affected by the product-harm crisis in the form of a loss in trust for the product category. Their study also identifies potential opportunities and threats for firms that experience a competitor’s product-harm crisis. After a product-harm crisis occurs, competitors might see an opportunity to increase their own market share by increasing spending on their own marketing instruments. Based on these findings the question can be raised if the increase in market share can indeed be attributed to the increased effectiveness of the

marketing mix instruments or if the increased expenditures compensate for the reduced

effectiveness after a product-harm crisis. Another consequence of a competitor’s product-harm crisis is that it might lead to negative brand spillover. This refers to a product-harm crisis negatively impacting other products operating under the same umbrella brand name (Sullivan, 1990). As a result, consumers might be inclined to switch to a competitor's’ brand due to the damage in brand equity. Siomkos, Triantafillidou, and Tsiamis (2010) state that a considerable percentage of

customers switch brands following a product-harm crisis. Thus, this research expects that when a product-harm crisis occurs, consumers will be more susceptible to marketing efforts by a competitor which will results in a positive effect on category sales. It should be noted that this spillover effect does not count in cases where all brands are perceived as guilty by association. This occurs when all products in a category are deemed too similar to each other. In this case, the crisis (or scandal) results in a decrease in sales for the entire product category and competitors are not able to benefit from a competitor’s product-harm crisis (Roehm & Tybout, 2006).

12 2.3 Conceptual model

The existing literature suggests that there is little academic literature about the effect of a product-harm crisis on the effectiveness of a competitor’s marketing mix instruments. This research, therefore, aims to shed light on this potential effect. In their conceptual framework, Cleeren, Van Heerde, and Dekimpe (2013) only include price and advertising as marketing variables. They find that increasing advertising can be considered a double-edged sword as their results show that an increase in advertising leads to adverse effects since consumers view the increase in advertising as ‘chasing ambulances’. This research expands upon their findings by including the display marketing mix variable, which has not been featured prominently in product-harm crisis literature as of this moment in time. In line with findings from Cleeren, Van Heerde, and Dekimpe (2013), this research expects to find a negative effect of price and a positive effect of advertising on category sales during a competitor’s product-harm crisis. Since there is little to be found on the marketing mix variable display in a product-harm crisis, this research turns to a study from Siomkos, Triantafillidou, and Tsiamis (2010). They state that, due to a loss in brand equity, consumers might be inclined to switch brands. Brand switching can occur due to altering beliefs about brands (Deighton, Henderson & Neslin, 1994). When looking for an alternative to the recalled product, display advertising might influence purchase decisions. Thus, this study expects a positive effect of the display marketing mix variable on sales during a competitor’s harm crisis. This research focuses on the product-harm crisis having an interaction effect with the effectiveness of marketing mix instruments and sales. The effectiveness of the marketing mix instruments is determined by the effect they have when a competitor’s product-harm crisis occurs. Figure 1 provides the conceptual model for the relationship between marketing mix instruments price, advertising, and display on sales for a product category during a competitor’s product-harm crisis.

Figure 1

13 3. Data and Methodology

This chapter will briefly discuss the dataset used for this study, the proposed time series model this research will use, elaborate on its individual elements, and touch upon the methodology behind the error correction model.

3.1 Data description

This study uses data provided by marketing research company IRI with data available for the entire Dutch fast moving consumer goods market. Four years of weekly data are available on 560 different product categories, starting week 29 of 1994 through week 28, 1998. The dataset includes a wide selection of product types including frozen pizza, animal shampoo, and pre-packaged hamburgers. A key advantage of this wide selection of product types is that it allows us to make empirical

generalizations. The product types will be further referred to as product categories in this study. The dataset has been complemented with advertising data from the BBC research agency in the

Netherlands, adding media such as newspapers, magazines, television, etc. It is important to note that advertising is aggregated over all media in this research.

3.1.1 Data cleaning and variable operationalization

This study will use a subset of the total available data because not in every product category a product-harm crisis has occurred. Dutch newspapers were searched using the Lexis Nexis newspaper search engine between the period of week 29, 1994 through week 28, 1998 to look for any product recalls on the Dutch market.

14 Table 1

List of product recalls used in this study.

Name of the recalled brand Category Source

Felix Vlees Smaken Mix Dry cat food NRC, 1996

Tom Poes Variantjes Dry cat food NRC, 1996

Heinz Tomato Frito Tomato paste conserves Branch Uitgevers, 2008

Amstel Malt Light beer Branch Uitgevers, 2008

Olaz New Skin Discovery Facial cream NRC, 1995

Guhl Shampoo Shampoo Algemeen Nederlands Persbureau, 1997

Euroshopper eye makeup remover Eye/lip makeup remover Algemeen Nederlands Persbureau, 1997

Edelzwicker wine White wine Brabants Dagblad, 1997

This research uses a time series model to assess the effectiveness of a firm’s marketing instruments during a competitor’s product-harm crisis. Usage of a time series model requires adding additional variables to the data frame. Specifically, a trend variable and a control variable to account for

15 week of the second season of 1994 and the last week is the second week of the second season in 1998.

To be able to interpret the coefficients of the model estimates as a percentage in- or decrease, the variables in the model are transformed to elasticities. The transformation is done by taking the natural logarithm of the continuously measured variables Sales, Advertising, Display, and Price.

Indicator variables were created to account for the effect of the product-harm crisis. The indicator has a value of 1 in the week that the product-harm crisis was first announced in the newspapers and keeps that value until x number weeks to capture the full effect of the announcement of the product-harm crisis. The amount of weeks the indicator variable keeps the value of 1 will be further discussed in chapter 4.1. Finally, dummy variables were created for each brand within a category to account for possible brand effects. Table 2 shows an overview of the variables within the dataset.

Table 2

Explanation of the variables included in the dataset.

Variable name Explanation variable

Week Indicator of the number of weeks within the data frame starting at 1 and ending at 208

Display Natural logarithm of the sum of category sales in stores where a product is displayed,

divided by the sum of category sales in all stores

Sales Natural logarithm of the sales per volume, expressed in, for example, liters

Advertising Natural logarithm of the guldens (ƒ) spent on advertising for the entire Dutch market

aggregated over all media

Price Natural logarithm of the average price per volume, which is calculated by dividing value

sales by volume sales

Indicator PHC Indicator variable for the occurrence of a product-harm crisis

16 3.1.2 General descriptive insights

Table 3 provides general descriptive insights into the marketing mix instruments for each category that are available within the dataset, as well as insights into category sales. Due to the extensive nature of the dataset, the marketing mix instruments and sales for each category differ greatly. The average use of display is the greatest for shampoo and white wine products. The average price per volume is highest for shampoo and facial cream, the price might be higher for these products as the care products category tend to include more luxury products than other product categories in the fast-moving consumer goods industry. The category with the highest advertising mean is light beer contrasted by the tomato paste conserves, which has the lowest. Concluding, the sales per category is highest for eye/lip makeup remover. Overall, the key insight Table 3 provides is that the use of marketing mix instruments and sales differ greatly for each category.

Table 3

General descriptive insights of the marketing mix variables for each category

Category Average Sales Mean Advertising Average Price Mean Display

Dry cat food 31719,7 7326 4.990 0.9386

Tomato paste conserves 6180,75 50.23 4.658 0.115

Shampoo 9057,7 47308 18.358 2.1192

Facial cream 2170,17 27072 14.246 0.2953

Eye/lip make-up remover 1110,834 1746 6.874 0.0655

Light beer 61975,1 17820.6 2.753 0.7871

White wine 38298,1 395.8 5.294 2.3285

3.2. Error Correction Model

This study aims to provide empirical generalizations that contribute to the existing product-harm literature. Specifically by looking at the effectiveness of marketing mix instruments during a

competitor product-harm crisis. In addition to direct short-term effects, marketing mix variables also tend to have an impact on sales in the long run (Pauwels, 2004; Slotegraaf & Pauwels, 2008). This study proposes an error correction model to capture both the short and long-term dynamics of the marketing mix variables advertising, price, and display.

3.2.1 Model

17 One major assumption of time series modeling is stationarity. If stationarity criteria are violated, it is not statistically wise to use a time series model. The first stationarity criterion is that the mean of the data should not be time-dependent, but constant instead. Secondly, the variance of the data should also not be dependent on time, otherwise known as homoscedasticity. The third and final criterion is that the covariance should not be a function of time. The stationarity of the data will be discussed in section 3.2.2.

To be more precise, this research will use an error correction model. Error correction models are often used when confronted with non-stationary time series data. However, as shown by Keele & De Boef (2004), error correction models can also be used on data that is stationary. They argue that usage of an error correction model is both a theoretically desirable and empirically feasible approach to stationary data. Another argument to use an error correction model on stationary data is that short and long-term dynamics no longer have to be lumped together (Keele & De Boef, 2004). Thus, using an error correction model allows for making preliminary inferences about the short and long-term effects of different marketing mix instruments during a competitor’s product-harm crisis. Based on the methodology in the previous sections and the conceptual model in the previous chapter, the following econometric model can be specified:

∆ ln 𝑆𝑎𝑙𝑒𝑠_𝑗𝑏𝑡 = 𝛽₀+ 𝛽₁𝑠𝑡_{∆ ln 𝑃𝑟𝑖𝑐𝑒} 𝑗𝑏𝑡+ 𝛽2𝑠𝑡∆ ln 𝐴𝑑𝑣𝑗𝑏𝑡+ 𝛽3𝑠𝑡∆ ln 𝐷𝑖𝑠𝑝𝑗𝑏𝑡+ 𝛽4𝑠𝑡𝐼𝑗𝑏𝑡 + 𝛽₅𝑠𝑡_𝐼 𝑗𝑏𝑡∆ ln 𝑃𝑟𝑖𝑐𝑒𝑗𝑏𝑡+ 𝛽6𝑠𝑡𝐼𝑗𝑏𝑡∆ ln 𝐴𝑑𝑣𝑗𝑏𝑡+ 𝛽7𝑠𝑡𝐼𝑗𝑏𝑡∆ ln 𝐷𝑖𝑠𝑝𝑗𝑏𝑡 + II_𝑏[ln 𝑆𝑎𝑙𝑒𝑠_{𝑗𝑏𝑡−1} + (𝛽8𝑙𝑡ln 𝑃𝑟𝑖𝑐𝑒𝑗𝑏𝑡−1+ 𝛽9𝑙𝑡ln 𝐴𝑑𝑣𝑗𝑏𝑡−1+ 𝛽10𝑙𝑡 ln 𝐷𝑖𝑠𝑝𝑗𝑏𝑡−1+ 𝛽11𝑙𝑡𝐼𝑗𝑏𝑡 + 𝛽₁₂𝑙𝑡_𝐼 𝑗𝑏𝑡ln 𝑃𝑟𝑖𝑐𝑒𝑡−1+ 𝛽13𝑙𝑡𝐼𝑗𝑏𝑡ln 𝐴𝑑𝑣𝑡−1+ 𝛽14𝑙𝑡𝐼𝑗𝑏𝑡ln 𝐷𝑖𝑠𝑝𝑗𝑏𝑡−1 )] + 𝛽₁₅𝑙𝑡_{𝑇𝑟𝑒𝑛𝑑 + 𝛽} 16𝑙𝑡𝑆𝑒𝑎𝑠𝑜𝑛𝑎𝑙𝑖𝑡𝑦 + 𝜀𝑗𝑏𝑡 Where:

Δ = First difference operator

Βb0 = The intercept of brand b

Salesbt = Number of volume sales brand b in week t

Pricebt = Regular price of brand b week t

Advbt = Weighted distribution of total advertising per brand b week t

18 Ijt = Indicator variable for the product-harm crisis equaling 1 in condition j

Trend = Indicator for the number of weeks (=1 through 208)

Seasonality = Indicator for the yearly quarters (=1, 2, 3 or 4)

Πb = Adjustment effect for brand b

εbt = Error term / Residuals

3.2.2 Stationarity tests

As previously mentioned, a major assumption of time series modeling is stationarity. If stationarity criteria are violated, it is statistically not wise to build a time series model. This study refers to both individual unit root tests and panel unit root tests to assess the stationarity of a series of data. The Phillips Perron individual unit root test (1988) allows for a very wide class of weakly dependent and possibly heterogeneously distributed data (Phillips & Perron, 1988). However, recent studies have shown that test based on individual series lack power compared with panel-based unit-root tests (Levin, Lin, and Chu, 2002; Im, Pesaran, Shin, 1997). In addition to lacking power, the Phillips Perron test also requires the individual computation of 70 x 10 series which would take up a considerable amount of time. Thus, this study favors panel-based unit-root tests over individual unit root tests.

19 3.2.3 Generating insights

This research aims to generate insights that contribute to the existing empirical knowledge about marketing mix effectiveness during a competitor’s product-harm crisis. In order to make

generalizable conclusions, it is first needed to check if the proposed model is allowed for all

categories (a pooled model) or that each brand within a category needs to be estimated individually (an un-pooled model). A Chow test (Leeflang, 2015 p. 147) allows us to check if pooling is indeed allowed, or a unit by unit analysis should be conducted for each brand within each category. The Chow test looks at the sum of squared residuals of both the pooled model and the unit by unit models and calculates an F-statistic. The calculated value of the F-statistic was much higher than its critical value, and therefore we reject the null hypothesis in favor of the alternative hypothesis; pooling is not allowed for this dataset. Thus, the model has to be estimated individually for each brand within each category.

The result of the Chow test conflicts with the main goal of this study, which is to provide empirical generalizations. In this case, the added Z method (Rosenthal, 1991) offers a solution. The added Z method allows for combined insights based on individual brand models (or unit by unit models) and can be applied to all variables in the model.

20 4. Results

This chapter will dive into the robustness of the error correction model specified in the previous chapter. After that, the results of the estimated error correction model will be discussed in detail by means of the added Z method (Rosenthal, 1991).

4.1 Model diagnostics

According to Little (1970), one of the criteria of a good model is robustness. Little (1970, p470) defines robustness as a quality characteristic which makes it difficult for a user to obtain bad answers. As mentioned in section 3.1.2 the effect of the product-harm crisis is captured over a number of weeks. To determine what number of weeks captures the effect of the product-harm crisis the best, we turn to the Akaike Information Criterion (Akaike, 1987) and the Bayesian

Information Criterion (Schwarz, 1978). These information criteria can tell us which model performs better by changing the value of the number of weeks to either 4 (one month), 8 (two months), or 13 (one season).

The Akaike Information Criterion is one of the most commonly used information criteria and included in the standard output of many statistical software packages. Thus, it is well-suited to judge the number of weeks the variable should include. For AIC, the model with the lowest score is the preferred model (Leeflang, Wieringa, Bijmolt & Pauwels, 2015). In addition, this study also looks at the Bayesian Information Criterion. As stated by Leeflang et al. (2015), the parameter penalty BIC is larger, so that model selection based on this criterion leads to a preference for more parsimonious models. Also, BIC is more often preferred when dealing with smaller datasets. As with the AIC, the model with the lowest BIC score is the preferred model. As shown in Table 4, the variable that has 8 weeks for capturing the effect of the crisis has both the lowest AIC and BIC scores and is, therefore, used in the final model.

Table 4

AIC and BIC scores for the model with a different value for the product-harm crisis variable.

Model 1 (4weeks) Model 2 (8 weeks) Model 3 (13 weeks)

AIC 290873 290867.4 290871.6

21 4.2 Model results

Table 5 shows the overall across-brand parameter estimates, indicated by the weighted beta, its associated Z-score (Rosenthal, 1991), and p-value.

4.2.1 Short-term effects

The estimated coefficients of the error correction model give us an intercept that is highly significant. The results indicate that there is no significant short-term impact of the occurrence of a competitor’s product-harm crisis on the sales of a category. Table 1 in chapter 3 shows us that all categories included in this study can be considered non-perishables, meaning their expiry date is typically longer than a year. The reason that a product-harm crisis occurring within these categories might not have a significant effect on the sales elasticity might be attributed to their non-perishable nature as

common knowledge suggests these categories consists of products consumers need on a daily or weekly basis and sales within a category are therefore not directly influenced by a competitor’s product-harm crisis as they will purchase any brand regardless. In addition, another possible explanation for not finding any statistically significant result for the occurrence of a competitor’s product-harm crisis might be that the differences between brands and categories included in this study (Table 3 in section 3.1.2) can make it difficult to give across-brand parameter estimates. The weighted beta of the short-term elasticities of display (=0.0097), advertising (=0.0024), and price (=-2.3075) are significant at the 0.01 level. These findings are in line with existing marketing literature (Chevalier, 1975; Tellis, 1988; Baghestani, 1991). The negative price elasticity is consistent with the findings from Bijmolt, Heerde, and Pieters (2005) in their meta-analytic price elasticity study, who find an average price elasticity of -2.62. In addition, Baghestani (1991) finds that, using an error correction model, advertising has a significant positive short-term effect on sales. Specifically, Baghestani finds this effect when advertising is a percentage of the sales revenue. While this study does not specifically measure advertising effects in this way, the results of this study partially support his findings.

Concluding the short-term effects, the results indicate that no significant interaction effects exist between the occurrence of a competitor’s product-harm crisis and advertising, price, and display. In addition, the results of this study do not support findings from Van Heerde, Helsen, and Dekimpe (2007), who find that a brand experiencing a product-harm crisis are more vulnerable to competitor actions. Direct academic evidence to support the findings of this study is rather sparse. For a possible explanation, this study turns to the field of applied psychology. Specifically, the

22 experiencing the crisis, are inferred and applied to brand B, i.e., a competing brand, when both brands have attributes that are closely shared. The results suggest a weak linkage between the attributes of the competitor brand that is experiencing the product-harm crisis and the rest of the category as no interaction effects are statistically significant. In addition, the results of this study suggest that the short-term effects of a product-harm crisis might be isolated to the focal brand and thus the effectiveness of marketing mix instruments from competitors is not influenced.

4.2.2 Long-term effects

When looking at the long-term effects, we also find no significant effect of the occurrence of a product-harm crisis on the sales within a category. The non-significance of the long-term effect of the occurrence of a product-harm crisis can also be explained by the non-perishable nature of the

product categories included in this study as mentioned for the short-term effect in paragraph 4.2.1. The weighted beta coefficients of display (=0.0050) and price (=-0.2472) elasticities are significantly positive and negative respectively. Thus, the display and price elasticities keep their short-term effect in the long run. These findings are, again, supported by existing marketing literature (e.g., Pauwels, 2004; Slotegraaf & Pauwels, 2008). Also, the advertising elasticity (=0.0027) is significantly positive. In his theoretical framework, Schmalensee (1972) suggests that a long-term equilibrium relationship exists between advertising and sales as a general economic rule. The results of this study support his theoretical framework as the long-term advertising elasticity has a significant positive effect on sales. Also, findings from Baghestani (1991) support Schmalensee’s theoretical framework.

As for the long-term interactions between a competitor product-harm crisis and advertising, price, and display, this study finds no significant effects. The reasoning for not finding short-term

interaction effects as stated in section 4.2.2 seems to hold for the long-term effects as well. The results of this study suggest that the long-term effects of a product-harm crisis might be isolated to the focal brand as no significant interactions between the marketing mix instruments and a

competitor product-harm crisis exist.

23 Table 5

Overall across-brand parameter estimates.

Weighted 𝛽 Z-score p-value

Intercept 3.7348 13.3136 0.000*

ST - Display 0.0097 8.5091 0.000*

ST - Price -2.3075 -43.4878 0.000*

ST - Advertising 0.0024 3.7030 0.001*

ST - Product-harm crisis -0.2105 -0.0988 0.4607

ST - Product-harm crisis x Advertising 0.0131 1.4174 0.0782

ST - Product-harm crisis x Price -0.2565 -0.6757 0.2496

ST - Product-harm crisis x Display -0.0071 -0.7498 0.2267

ST - Product-harm crisis x Advertising x Price x Display 0.0010 0.6325 0.2635

LT - Display 0.0050 4.3174 0.000*

LT - Price -0.2472 -6.0805 0.000*

LT - Advertising 0.0027 4.1329 0.000*

LT- Product-Harm Crisis 0.1571 0.4483 0.3270

LT - Product-harm crisis x Advertising 0.0089 0.8882 0.1872

LT - Product-harm crisis x Price -0.3708 -0.0499 0.4801

LT - Product-harm crisis x Display -0.0014 -0.1289 0.4487

LT - Product-harm crisis x Advertising x Price x Display -0.0033 -1.241 0.1073

LT - Sales -0.2330 -41.1239 0.000*

24 5. Conclusion and recommendations

This chapter answers the research questions as specified in the introduction, followed by managerial implications based on the results specified in the previous chapter. The chapter ends with limitations this study dealt with and possible directions for future research.

5.1 Summary of findings

A product-harm crisis is among the biggest challenges a company faces nowadays. These crises are expected to occur more frequently in the foreseeable future due to stricter product-safety

legislation, more demanding customers and increasing complexity of products. As a consequence, the occurrence of a product-harm crisis can have a serious impact on the firm’s bottom line, either in terms of loss of sales or a loss of market share.

However, most of the existing academic literature concerning product-harm crisis looks at the effectiveness of marketing mix instruments of the focal firm experiencing the crisis. Not many insights and academic results have been available to marketing managers of firms when a product-harm crisis occurs at a competitor. Does the marketing mix effectiveness increase or decreases following such a crisis? In other words, is one man’s loss another man’s gain?

This study aimed to provide empirical generalizations that add to the existing literature concerning product-harm crisis by answering the following research question: ‘How is the effectiveness of a firm’s marketing mix instruments influenced by a competitor’s product harm crisis?’ The follow-up question asked in this study is ‘Which marketing mix instruments are most effective during a product-harm crisis?’. These questions were answered using a dataset consisting of 12,471 observations, representing 70 brands from 7 product categories in the Dutch fast moving consumer goods industry, covering 4 years of weekly sales.

Based on the dataset, this study did not find a statistically significant impact of the occurrence of a competitor's product-harm crisis on sales. However, the product categories studied in this research are of a non-perishable nature, which might explain as to why no significant interactions between product-harm crisis and marketing mix instruments were found. Products of this nature are usually needed on a daily or weekly base and thus consumers might pick the next best alternative despite the occurrence of a product-harm crisis in the category.

25 the crisis to other brands in the category as supported by the accessibility-diagnosticity framework (Feldman and Lynch, 1988).

Thus, based on this dataset, the conclusion to the research question ‘How is the effectiveness of a firm’s marketing mix instruments influenced by a competitor’s product harm crisis?’ is that the marketing mix instruments of a firm are not influenced by a competitor’s product-harm crisis. Subsequently, the sub-question ‘Which marketing mix instruments are most effective during a product-harm crisis?’ cannot be answered based on the results of this dataset.

5.2 Managerial implications

Marketing managers experience increasing pressure from the management team to account for the effectiveness of the implemented marketing mix instruments. The results of this study can be translated into actionable insights for marketing managers of firms that are experiencing a

competitor product-harm crisis. Based on the results of this study managers are advised to act as if business is as usual during a competitor product-harm crisis. The occurrence of a competitor product-harm crisis does not have a direct impact on sales. Subsequently, the effectiveness of marketing mix instruments is not influenced by a competitor product-harm crisis.

5.3 Limitations

As with any research, this study had to deal with a few limitations. The product categories included in this study are all non-perishable products. This study did find product-harm crises for perishable products within the timeframe of this dataset, but they were unfortunately not included in the dataset. Therefore, the results are not generalizable to perishable product categories. Using a combination of both perishable and non-perishable products in future studies might lead to different results.

Second, marketing mix instruments have evolved drastically over the past two decades. The rise of digital communication has drastically changed the marketing landscape. However, the dataset used in this study dates from the period 1994-1998 where digital communication wasn’t as commonly used as an advertising platform as it is nowadays. Therefore, using a recent dataset where these new digital media are included might lead to different results.

26 Finally, this research was written as a Master thesis and was therefore under considerable time pressure. As discussed in section 3.2.2, the Philips Perron individual unit root test was not conducted due to the computing time of 70 x 10 series of data. Conducting these tests might influence the conclusion based on the panel unit root test concerning stationarity.

5.4 Future research directions

As previously mentioned, actionable insights for marketing managers of firms experiencing a

competitor product-harm crisis are rather sparse at the time of writing. Therefore, any contributions to this side of product-harm crisis literature are greatly appreciated. Giving managers more insights into which marketing mix variables are most effective during a competitor product-harm crisis might make the marketing department more accountable within more firms.

Concerning future research related to this study, further studies into the effectiveness of marketing mix variables during a competitor product-harm crisis are needed for perishable products either separately or in combination with non-perishable products. Testing the proposed conceptual model in the context of perishable products might generate different results.

27 References

Ailawadi, K.L., Beuchamp, J.B., Donthu, N., Gauri, D.K., and Shankar, V. (2009). Communications and Promotions Decisions in Retailing: A Review and Directions for Future Research. Journal of Retailing, 85 (1), 42–55.

Akaike, H. (1987). Factor analysis and AIC. Psychometrika, 52(3), 317-332. Algemeen Dagblad. (1997). AH verwisselt ‘removers’. Retrieved from:

https://www.nexis.com/results/enhdocview.do?docLinkInd=true&ersKey=23_T26766205148&forma t=GNBFI&startDocNo=0&resultsUrlKey=0_T26766205174&backKey=20_T26766205175&csi=149014 &docNo=32

Algemeen Nederlands Persbureau. (1997). Guhl haalt verkeerde shampoo terug. Retrieved from: https://www.nexis.com/results/enhdocview.do?docLinkInd=true&ersKey=23_T26765493598&forma t=GNBFI&startDocNo=0&resultsUrlKey=0_T26765499226&backKey=20_T26765499227&csi=149022 &docNo=7

Baghestani, H. (1991). Cointegration analysis of the advertising-sales relationship. The Journal of Industrial Economics, 671-681.

Baum, C. F., Schaffer, M. E., & Stillman, S. (2003). Instrumental variables and GMM: Estimation and testing. Stata Journal, 3(1), 1-31.

Bell, D. R., Chiang, J., & Padmanabhan, V. (1999). The decomposition of promotional response: An empirical generalization. Marketing Science, 18(4), 504-526.

Bhattacharya, C., & Sen, S. (2004). Doing better at doing good: When, why, and how consumers respond to corporate social initiatives. California Management Review, 47, 9-24.

Bijmolt, T. H., Heerde, H. J. V., & Pieters, R. G. (2005). New empirical generalizations on the determinants of price elasticity. Journal of marketing research, 42(2), 141-156.

Borah, A., & Tellis, G. J. (2016). Halo (Spillover) Effects in Social Media: Do Product Recalls of One Brand Hurt or Help Rival Brands?. Journal Of Marketing Research (JMR), 53(2), 143-160.

Brabants Dagblad. (1997). Elzas-wijn uit de winkel. Retrieved from:

28 Bracht, G. H., & Glass, G. V. (1968). The external validity of experiments. American Educational Research Journal, 5, 437-474.

Branche Uitgevers. (2008).LMT 4b informatie recall. Retrieved from: http://www.tglweb.nl/www-tgl/documenten/4-3%20De%20Dis%20Voedselproducent/LMT%204b.12%20informatie%20Recall.pdf Buckland, K. 2017, June 26th. Takata Bankruptcy Clouds Auto Industry’s Biggest Recall. Retrieved from: https://www.bloomberg.com/news/articles/2017-06-25/takata-bankruptcy-poses-a-risk-to-auto-industry-s-biggest-recall

Carroll, A. (2015). Corporate social responsibility: The centerpiece of competing and complementary frameworks. Organizational Dynamics, 44(2)

Chevalier, M. (1975). Increase in sales due to in-store display. Journal of Marketing Research, 426-431.

Cleeren, K., Dekimpe, M. G., & Helsen, K. (2013) Rising from the ashes: How brands and categories can overcome product - harm crises. Journal of Marketing, 77(2), 58 - 7

Cleeren, K., Dekimpe, M. G., & Helsen, K. (2008). Weathering product-harm crises. Journal of the Academy of Marketing Science, 36(2), 262-270.

Cawley, J., and J. Rizzo. 2008. The spillover effects of prescription drug withdrawals. Advances in Health Economics and Health Services Research. 19, 119-144.

Dawar, N. & Pillutla, M (2000) Impact of Product-Harm Crises on Brand Equity: The Moderating Role of Consumer Expectations. Journal of Marketing Research: May 2000, Vol. 37, No. 2, pp. 215-226.

Deighton, J., Henderson, C. M., & Neslin, S. A. (1994). The effects of advertising on brand switching and repeat purchasing. Journal of Marketing Research, 28-43.

Feldman, J. M., & Lynch, J. G. (1988). Self-generated validity and other effects of measurement on belief, attitude, intention, and behavior. Journal of Applied Psychology, 73(3), 421.

Gijsenberg, M. J. (2014). Going for gold: Investigating the (non) sense of increased advertising around major sports events. International Journal of Research in Marketing, 31(1), 2-15.

29 Keele L., De Boef S. (2005). Not just for cointegration: error correction models with stationary data. Nuffield College Working Papers in Politics n. 2005-W7.

Leeflang, P. S. H., Wieringa, J. E., Bijmolt, T. H. A., & Pauwels, K. H., 2015. Modeling Markets. Springer, London.

Leeflang, P. S. H., Wieringa, J. E., Bijmolt, T. H. A., & Pauwels, K. H., 2015. Advanced Methods for Modeling Markets. Springer, London.

Levin, A., C.F. Lin and C. Chu, 2002. Unit root test in panel data: Asymptotic and finite sample properties. Journal of Econometrics 108, 1–25.

Lindgreen, A., & Swaen, V. 2010. Corporate social responsibility. International Journal of Management Reviews, 12: 1-7.

Lock, S. (2008). Protecting brand and bottom line in a crisis. Your Choice Magazine, Vol. 7 No. 32.

Maynard, M. 2000, September 17th. Firestone’s Crisis is Other Dealers’ Opportunity. Retrieved from:

http://www.nytimes.com/2000/09/17/business/business-firestone-s-crisis-is-other-dealers-opportunity

Malkina-Pykh, I.G. & Pykh, Yu. A. The Method of Response Function in Psychology and Sociology. Southampton, Boston: WIT Press, 2013.

NRC. (1995). Procter legt verkoop van creme stil na oogklachten. Retrieved from:

https://www.nexis.com/results/enhdocview.do?docLinkInd=true&ersKey=23_T26765499296&forma t=GNBFI&startDocNo=0&resultsUrlKey=0_T26765507406&backKey=20_T26765507407&csi=259062 &docNo=7

NRC. (1996). Kattenvoer mogelijk oorzaak verlamming. Retrieved from:

https://www.nrc.nl/nieuws/1996/04/20/kattenvoer-mogelijk-oorzaak-verlamming-7307233-a567956

De Raat, F. 1995, January 27th. Mondiger consument pikt produkt met fouten niet meer. Retrieved from:

https://www.nrc.nl/nieuws/1995/01/27/mondiger-consument-pikt-produkt-met-fouten-niet-meer-7254521-a591967

30 Slotegraaf, R. J., & Pauwels, K. (2008). The impact of brand equity and innovation on the long-term effectiveness of promotions. Journal of Marketing Research, 45(3), 293-306.

Persson, M. 2002, November 23rd. Omo wil weer wassen. Retrieved from https://www.volkskrant.nl/archief/omo-power-wil-weer-wassen~a631210/

Phillips, P. C., & Perron, P. (1988). Testing for a unit root in time series regression. Biometrika, 75(2), 335-346.

Roehm, M.L. & Tybout, A.M. (2006) When Will a Brand Scandal Spillover, and How Should Competitors Respond?. Journal of Marketing Research: August 2006, Vol. 43, No. 3, pp. 366-373. Rosenthal, R. (1991). Meta-analytic procedures for social research. Newbury Park, CA: Sage Publications.

Schwarz, G. (1978). Estimating the dimension of a model. The annals of statistics, 6(2), 461-464. Schmalensee, R. (1972). The Economics of Advertising. Amsterdam: North-Holland; 1972 Statsoft. Time Series Analysis: The Basics. Retrieved from:

http://www.abs.gov.au/websitedbs/D3310114.nsf/home/Time+Series+Analysis:+The+Basics

Steenkamp J. B. E. M., Nijs V. R., Hanssens D. M., Dekimpe M. G. Competitive reactions to advertising and promotion attacks. Marketing Sci. (2005) 24(1):35–54

Siomkos, G., Triantafillidou, A., Vassilikopoulou, A., & Tsiamis, I. (2010). Opportunities and threats for competitors in product-harm crises. Marketing Intelligence & Planning, 28(6), 770-791.

Souiden, N., Pons, F. (2009) Product recall crisis management: the impact on manufacturer's image, consumer loyalty and purchase intention. Journal of Product & Brand Management, Vol. 18 Issue: 2, pp.106-114.

Stata. Pperron - Phillips-Perron unit-root test. Retrieved from: https://www.stata.com/manuals13/tspperron.pdf

Sullivan, M. (1990). Measuring image spillovers in umbrella-brand products. Journal of Business, Vol. 63 No. 3, pp. 309-29.

31 Thomas, K.. 1979. Organizational conflict. In Steven Kerr (ed.) Organization behavior, 151--1181. New York: J.

Wiley & Sons Inc. Reprinted in Henry L.. Tosi (ed.)Organizational behavior and management. A contingency approach, 203--2232.

Van Heerde, H., Helsen, K., & Dekimpe, M. G. (2007). The impact of a product-harm crisis on marketing effectiveness. Marketing Science, 26(2), 230-245.

Wessels, P. 1994, June 7th. Omo Power wordt fameuze flop. Retrieved from:

https://www.nrc.nl/nieuws/1994/06/07/omo-power-wordt-fameuze-flop-7227419-a144251 Winter, M. 2012, November 16th. Calif. Meat packer to pay $317M over abuse, recall. Retrieved from: https://www.usatoday.com/story/news/nation/2012/11/16/california-slaughterhouse-fraud-settlement-beef-recall/1710693/