Product crises and harm

Product crisis effects for different harm types on brand attitudes

Product crises and harm

Product crisis effects for different harm types on brand attitudes

Martijn Hazelhorst

Executive summary

Little has been written about product crises and damage. There are hardly any categories or frameworks to use regarding product crises. This paper hopes to contribute to the knowledge about product crises by investigating the effect of product crises on brand attitudes, which are used to measure the negative impact of a product crisis. Two moderating variables are taken into account; harm types and brand commitment. Harm types might have differing impacts on the relationship between product crises and brand attitudes. Brand commitment is said to have a mitigating effect on the impact of negative information, such as that from a product crisis. The relationships are tested by holding a survey among consumers regarding the Apple iPhone. The results state that product crises indeed affect brand attitudes. Secondly, the different harm types are found to have a significant negative impact on brand attitudes. The harm type with the most negative impact is direct non-physical harm (e.g. trauma, fear). Direct physical harm and indirect harm do not seem to differ for the average consumer. Higher brand commitment has a mitigating impact on the harm types, leading to higher purchase intentions. However, brand commitment seems less effective for indirect harm types. Thus, in indirect cases there could be a lot to be gained by organizations, by more deliberately linking or influencing cause-and-effect relationships. This is because brand commitment can improve sales in no-crisis situations and it can mitigate the impact of negative consequences from occurring crises. Indirect harm does not seem to be affected greatly by commitment, but it is well possible that (positive) indirect consequences can benefit from high commitment in the form of higher sales and the reduced impact of negative events. Therefore it would be wise to more deliberately link indirect consequences to brand commitment so that an organization can reap the benefits in these situations as well. This should be done by improving communications and marketing.

Keywords

Preface

Introduction

Product crises are not something new to the consumer and to a lot of companies. It is heard with some regularity that products are not complying to the standards they are supposed to which subsequently could pose a risk to the consumer. Problems with regard to contamination (poisonous ingredients or some other ingredients that do not belong in the product), safety and other damage might occur. The range of possibilities for problems is almost endless. For example, take the case of British Petroleum (BP). On April 20, 2010 there was an explosion at the oil drilling rig the “Deep Horizon” positioned in the Gulf of Mexico1. This explosion resulted in several employee casualties, and also caused a large leak. This caused a lot of oil to spill into the sea. This oil eventually ended up at the beaches of the American continent, generating large pollution. Besides the pollution there was also the risk that people might get ill. In 2011, several cases of people getting ill were also reported2. This is almost a year after the actual disaster. Since BP produces oil, this can be considered a product crisis. Because BP had a problem at its production process (collecting oil) a lot of people now run the risk to get ill because of the chemical contamination that resulted because of the oil spill. This subsequently led to damage for BP’s brand image and its stock price, which decreased by 52% in 50 days3. So, even though the actual disaster caused contamination and sick people, this also fired back on the brand by making the brand less attractive. This is a very extreme example, but also smaller problems or mistakes seem to damage a brand and its performance.

Currently, product crises are talked about as a one-dimensional concept. But is this really the case? A product crisis is a well-known event where the product has some kind of defect or harm associated with it (Siomkos & Kurzbard, 1994). This definition already states the vagueness of the concept because it contains “some kind of defect or harm 1 http://en.wikipedia.org/wiki/Deepwater_Horizon_oil_spill [Accessed at 15.50 on 23-02-2011] 2 http://bpoilleak.org/category/health-effects-odor/ [Accessed at 16.00 on 23-02-2011] 3

associated with the product”. The definition does say that there is some kind of damage or harm, but no prerequisite range is stated. This could be caused by the fact that different industries differ with regard to the damage their products can cause. No general range of damage for product crises can be given. This might be because some industries are more vulnerable to product crises than others. This could be due to industry characteristics (e.g. resources used, such as chemicals) or due to consumer characteristics (e.g. some crises frighten customers more in comparison to others). Current research especially focuses on possible consequences and effects of product crises for certain products or brands. But a lot of these assumptions and the strength of these relations regarding differences have hardly (if not at all) been investigated. This might also lead to the fact that there is contradictory research in this field, such as the spillover effect on competitors versus competitor advantage. This issue will be discussed into greater depth in paragraph 2.3.2.2.

Because types of risk and industrial structures will differ between product categories this should be researched to shed some new light on the existing knowledge of product crises. This is relevant because the consequences of a product crisis can damage a brand’s performance (Ma et al., 2010). Having more knowledge about this subject could increase a brand’s ability to prevent or resolve a product crisis. Nothing has been written about product categories and the crisis severity that might be inherent to them. By measuring crisis severity with different levels this could lead to the discovery that not all product categories will have the same product crisis severity levels. Because of these differences the nature of the product crisis and its effect on brand attitudes might also differ. Current research does describe the possibility of different levels of crisis severity, but it does not show that these could be related to product category factors and the industries behind those product categories. Furthermore, the harm types provided thus far are very basic, only varying from low to high. But this does not clearly define the different harm types that could follow from different product categories. This would require dimensions or factors that entail all product categories.

The problem statement appropriate for this issue is:

What is the effect of product crises on brand attitudes and how do harm types and brand commitment moderate this effect?

Academic relevance

This research will add something to the existing knowledge of product crises. As it is stated in the definition of product crisis, it is a well-known event. It therefore seems to affect a lot of consumers that possibly buy the brand. Do all these consumers respond the same when a product crisis occurs? Or are there other factors playing a role regarding the impact? This article investigates two such variables; brand commitment and product crisis harm types. The outcomes of this research could assist in telling which product categories are more susceptible to damage from a product crisis. Another issue hopefully explained is whether consumers respond differently for different levels of crisis severity which eventually leads to a different effect on brand attitudes. Since this is researched by looking at different harm types this would show that these harm type characteristics play a role and that it could be different per product category. The lesson to be learned here is that some industries have to be more careful in managing their crises since the impact of their crises is larger.

Managerial relevance

because the product characteristics (e.g. chemical, fragile etc.) are likely to put the emphasis on a certain harm category. Providing the knowledge that there are different harm types should assist managers in dealing with product crises more effectively. Managers can link the characteristics of their products (e.g. flammable, electricity-driven etc.) to the harm types shown in this research. If that specific harm type has a large impact according to the research this teaches the manager that he has to be more careful with the product and that he or she has to invest more in preventing the occurring of these harm events.

This paper will start with a description of the concept of product crises and the other relevant variables for this research, which are brand attitudes, harm types and brand commitment. The second part will describe the research model and the relationships between the concepts which follow from the research gap found in the current literature.

Literature review

This chapter will elaborate about the current literature written about product crises and its related concepts. As mentioned before, the concepts that will be included in this research are product crises, brand attitudes, brand commitment and harm types.

2.1 Product crisis

2.1.1 Definition of product crisis

Nowadays, but also in the past numerous brands came into problems because something went wrong with their products or production processes. Think of the cars made by Toyota, which contained certain defects with the accelerator pedals and therefore were recalled4. During the process of writing this paper BMW is recalling thousands of cars

4

because there might be a defect in the braking systems5. In 2005, Nestlé sold milk powder in China that contained more iodine than Chinese standards allowed. A different case is British Petroleum (BP), which still has major problems because one of their oil pipes sprung a leak leading to massive outflows of oil into the sea. These kinds of issues we can call product crisis.

A product crisis is a well-known event related to some kind of product defect or harm associated with some brands (Siomkos & Kurzbard, 1994). However, this definition is relatively vague. Defects or harm are terms which can include a wide variation of situations. It can range from increased safety risks to actual casualties (Ma et al., 2010). This variety also involves issues such as ailments or physical damage.

Often, caused damage can be relatively small or be caused over a long period of time, which can be solved quickly without too much public involvement (Kinghorn, 1985). However, sometimes the scale of the damage is large enough so that it may create a serious threat towards a company and its stakeholders (i.e. consumers, government etc.). Product crises have been becoming more important because they are occurring more often (Shrivastava et al., 1988). A remark could be made here that this might be due to the growth and multinational character of many companies these days. This is because these companies can serve a lot of customers in comparison to companies that are smaller-scaled or nationally-oriented. Multinational companies are not restricted to the consumers of only one country. This means that if a defect or harm is caused by a multinational company, the consequences are likely to reach a large group of people. This might be a reason why crises are more often “large” and why they more regularly become public events. Even if an organization is not multinationally oriented it might cause damage on a larger, even global, scale like what happened at Chernobyl (Shrivastava et al., 1988).

Also, industrial technology is becoming more complex and potentially harmful which implies that the scope or frequency of product crises will increase (Perrow, 1984).

5

Shrivastava et al. (1988), give characteristics of a product crisis. Eight characteristics are mentioned, but these also involve conflict resolution issues, which are not taken into account in this research. Therefore only six characteristics are relevant here:

- (1) Triggering event

- (2) Large-scale damage to human life and environment - (3) Large economic costs

- (4) Large social costs - (5) Causes of crises

- (6) Multiple stakeholder involvement and conflict - (7) Responses to crises

- (8) Crisis resolution and crises extension

It can thus be said that there is a cause leading to the scandal and that the effect of this scandal is large. The effects are a large (external) reach of the problem and the high costs that accompany it. Furthermore, there is some kind of damage and there are also costs. But damage is a broad concept and can be elaborated upon. There may be differences regarding the types of damage and the outcomes of a product crisis.

2.1.2 Consequences of product crises

Even after the actual problem with the product might have occurred the consequences can still be noticed, which indicates that there can be a long-term effect. It could also take a certain period of time before the problems occur. For example, because it requires long-term or regular exposure to the consequences of a product crisis. Therefore the indirect consequences of a product crisis are important too.

The second consequence is that a company can become more susceptible to a competitor’s market activities which might lead to an advantage for competitors (Ma et al., 2010). However, there is also the possibility that one brand’s scandal can damage another brand (because of strong linkages) or even the entire category in the case of a prototypical brand (i.e. a brand that people see as typical for the category) (Roehm & Tybout, 2006). In this so-called case of spillover the damage for the industry is likely to be even larger and relationships between players in the market might be endangered since one company can damage another. This effect might differ per category, because as Roehm and Tybout (2006) state in their research, is that it depends on the proximity and equality of competitors. Some industries and product categories are more differentiated, possibly reducing the likelihood of spillover. For example, if the aluminum of Gazelle bicycles is found to be brittle, leading to frames breaking down, this could well spill over to Giant bicycles, since these are also made out of aluminum. The crisis is related to something both companies have in common, which are their production resources and their end products. The output is very similar and this could lead people to believe that Giant bicycles are also more likely to break down because of poor quality input materials. On the other hand, if Pepsi Cola would have contaminated ingredients in its soda this is not very likely to affect the Coca Cola-owned Fanta. Even though both companies produce soft drinks, the end product of each company is rather different, which could lead consumers to believe that the crisis did not spill over, even though they might have some of the ingredients in common.

organization. Causes can be manufacturer negligence or product misuse (possibly because of lacking usage information) (Vassilikopoulou, 2008). These causes can be numerous. Since the crisis is a negative event concerning the performance of the organization (i.e. less sales, reduced consumer trust) this negatively affects the stock price. However, this research is not focusing on the specific cause of the product crisis, but at the factors influencing the relationship between the product crisis and brand performance in the form of sales.

2.2 Brand attitudes

2.2.1 Brand attitudes

about which consumers can become aware of. This provides information to these consumers about the brand or products. This makes it likely that a product crisis negatively influences brand attitudes and therefore also the performance of the brand. This is because when brand attitudes deteriorate the likelihood of a reduction in purchase intentions is also possible.

attitudinal commitment is indirectly influencing these sales and is thus included in the outcomes, but not explicitly.

The hypothesis following from this is as follows:

H1: A product crisis has a negative impact on brand attitudes

2.3 Crisis severity

2.3.1 Definition of crisis severity

Product crisis may vary greatly. Some might do much more harm than others, which can influence consumer attitudes and favorability towards the brand (Mowen and Ellis, 1981). This means that the more severe the product crisis is, the larger the effect on the brand. This severity in defect or harm might be related to the product category (e.g. physical consumption might cause more perceived harm). So the product category might have a direct influence on the product crisis with regard to harm, which makes it interesting to investigate. Since an exact classification of harm types is difficult to find, a newly designed classification will be introduced in this study. This classification will be elaborated on later.

2.3.2 The role of product categories

important too because non-physical damage might generate considerable hesitancy towards purchasing a brand because a consumer might remember the negative experience (e.g. fear) from the previous usage of the product. Because fear of the physical damage might be so intense, the non-physical damage by itself might lead to purchase hesitancy. Because of differences in product category characteristics the initial outcomes of a product crisis might differ. Consumer perception of a crisis might be completely different for another product category. This, in turn, can influence the amount of conflict resolving and repairing actions required. Whereas most other research only looks at the repairing actions companies take after a crisis to find out which one is most effective, it is also important to look at the roots of a crisis and whether this might make a difference regarding the choice of a certain type of conflict resolve. The range of catastrophes might be different per product category. It is unlikely for a computer to cause severe physical harm. When the producer would say the defect is removed, the hesitancy to buy of consumer’s would be expected to disappear rapidly. However, if a food product has been containing ingredients that are able to make people ill that have an allergy for a certain ingredient, then it is likely that the consumer will be more hesitant for a while. They want absolute certainty that they will not become ill. This can be because they have to risk their physical health when consuming a product. The computer, however, can only break down. This is likely to lead to a refund by the producer if it falls within the guarantee. This would cause annoyance, but no physical risk.

2.3.2.1 Differences regarding purchase frequency between product categories

resolved this could lead to the fact that consumers are using the defect-less product again and hopefully are satisfied again with using the product. Also, in the case of product crisis that occur after regular and frequent consumption the purchase (and consumption) frequency is also important since this allows for certain types of harm to happen, especially harm on the long run. Another risk is that because of the high frequency purchases the consumer runs into the defect again, possibly increasing grudges against the company. This factor shows how many impacts only one factor can have on a consumer’s perception of the company and its products and that it may sometimes take a long time before a consumer begins to notice the crisis events which will eventually lead to negative emotions with regard to the producer of the product. It could be small events that, when put together, form a large issue which will have an effect on the consumer. However, there could also be events where the defects or negative consequences become visible after a longer period of time. This could be explained by the fact that the crisis event is not able to cause immediate damage or it can be caused by the fact that it requires a lot of subsequent contacts with the product for the problems to show up. In the real world this is also likely to be very complex because there are no set “incubation” times for product crisis. Nonetheless, this does make clear that the consequences of product crises can show up after a substantial period of time and could therefore be indirect.

2.3.2.2 Spillover and prototypicality

by consumers associating different brands. Reasons could be prior cooperation or strong competition between the brands which then links them together in consumers’ minds. However, little research has actually been done about the fact whether product categories all experience the same issues with regard to product crises. Is the intensity of all relationships the same? Common sense would suggest not, because of inter-industrial and inter-product differences. For example, industries where intense competition is the norm should experience a different crisis than an industry where a lot of differentiation is going on and where each organization is serving a different niche, leading to reduced competition.

2.3.3 Approach product crises by looking at harm types

As can be seen in the previous paragraphs products can differ a lot from one another. Therefore it would be difficult to approach product crises by looking at individual products or by attempting to generate general characteristics. This is because these general characteristics are not very likely to cover all products or because there is an endless list of characteristics that might cause harm. This paper approaches product crises by looking at the harm types, and not at the products. The paper tries to generally define harm that can be caused with dimensions which should be independent of product (category) characteristics which allows for a more generalizable framework.

2.3.3.1 Categories of crisis severity

(Crafton et al., 1981; Souiden & Pons, 2009). Crafton et al. (1981) also state that a recall that occurs because of safety reasons (and thus not with injured consumers per se) has a less negative impact on brand performance. For example, think of the Toyota car recalls between 2009 and 2011 because of possible technical malfunctioning. This has been estimated to have cost Toyota two billion US dollars in sales6. Furthermore, time might play a role, which includes both the fact that regular consumption could be required and also the fact that crisis outcomes show up after a longer period of time or are unsure at the moment the crisis occurs. Oil spills and also the Fukushima nuclear disaster have resulted in decreased incomes for these companies78 (Marketing Week, 2010; Gehani, 1990). A lot of the consequences from these disasters might occur in the future and are therefore indirect. Because consumers can fear these long-term results this could also lead to a reduced brand performance on the long term for these companies. The product crisis occurs over a long period of time and more damage and costs could occur later. But because the consequences can be hard to notice the impact of these effects is likely to be reduced and only the initial disaster has the largest impact on brand performance. It will be difficult in the future to state that consumers that got sick from eating fish because of the Fukushima disaster or because industrial pollution contaminated the water. It is thus difficult to trace the consequences back to a specific brand unless they are very obvious.

If we combine these factors (physical/non-physical harm and direct/indirect harm) this would lead to four combinations. However, indirect harm is difficult to divide into physical or non-physical harm because the cause and effect are farther removed from one another than is in the case for direct harm, making it more difficult to say whether physical or non-physical damage could be caused. Also, indirect non-physical harm such as fear could be caused after a longer period of time because a consumer did not receive or look for information about the product crisis even though it was available. This paper

6

http://news.bbc.co.uk/2/hi/business/8493414.stm [Accessed at 17.00 on 23-05-2011]

7

Mason, Rowena (29 Jun. 2010). "Exxon or Shell should buy BP for £88bn, says analyst". The Telegraph (London). http://www.telegraph.co.uk/finance/newsbysector/energy/oilandgas/7862274/Exxon-or-Shell-should-buy-BP-for-88bn-says-analyst.html. Retrieved 6 Jul. 2010. [Accessed at 16.10 on 23-02-2011]

8

assumes that consumers receive the information about product crises when it becomes publicly available. Another problem with indirect harm is that non-physical harm could strongly be based on the fact that consumers think that people might have physically suffered because of the product crisis, but that this simply was never noticed or became public. The crisis might therefore, and possibly genuinely, be related to physical harm. The example in the introduction illustrates this where people got sick after the BP oil spill. There is evidence that physical harm has been done, but possibly not to everyone who somehow was exposed to the product crisis. The final harm that has been done is therefore vague and it may take many more years before the actual damage becomes visible and even parts of it may never be related to the product crisis at all. Because these two issues are so closely intertwined indirect non-physical and indirect physical damage will be combined in one category named indirect harm. These categories may be applicable in different combinations with regard to product categories, because product category characteristics may enable or disable the possibility of certain types of crises from occurring.

Based on the previous text product crisis severity is divided into three categories:

1 - Product crises which cause direct physical harm (e.g. illness, burns, deaths)

2 - Product crises which cause direct non-physical harm (e.g. cause fear, mental trauma) 3 - Product crises which cause indirect physical or non-physical harm (e.g. oil-spills, pollution)

non-physical harm, does not cause pain, but can instill fear into consumers that make them very hesitant to acquire the product another time. In the worst case, non-physical damage might even traumatize a person, possibly leading to long-term fear towards a product (category). This is also very likely to have a severe impact on a consumer’s brand attitudes with regard to the product. The third category, indirect harm, does not affect consumers directly, but can still make them afraid or affect their health on the long-term. This is affected to affect consumers the least because the damage is not very visible. The standard consumer is not likely to be able to taste whether some ingredients in the purchased products are contaminated by a chemical spill which occurred two years ago. However, when consumers become aware of the indirect harm cases, or may reason by themselves that such an event could affect them on the long run, their brand attitudes may still deteriorate. Furthermore, because of the improved communication technologies currently available people are more likely to be made aware of the consequences of these events.

2.3.4 Crisis severity levels and brand attitudes

There is the possibility that different harm types might seem more severe from the perspective of the consumer. A harm type that would be perceived as more severe by consumers would therefore be expected to have a greater negative impact on brand attitudes (Mowen & Ellis, 1981; Heerde et al., 2007). This paper will investigate severity by making a distinction between harm types and how these are weighed with regard to severity and their subsequent impact on brand attitudes. Because no previous research has been done with regard to harm types a new framework is introduced.

The hypothesis appropriate for this is:

2.4 Brand commitment

2.4.1 Definition of brand commitment

The current literature states that consumers who used the product (category) before the crisis occurred are more likely to turn back to the product after some time (Cleeren et al., 2008). According to Ahluwalia et al. (2000) commitment to a brand/product has a moderating effect on the processing of negative information received by consumers. Pre-crisis product usage could build up some commitment for the product, especially if the consumer is satisfied and loyal. When consumers are committed the likelihood of attitude degradation (i.e. a decrease in brand preference) is reduced. Even though two different consumers may both like the brand a lot, their commitment to the brand ultimately states how much they will try to counter-argue negative brand information. However, it is unsure whether this holds for different product categories and in which strength it occurs. Product crises in some categories might generate more hesitance and fear among their (former) customers. For example, a product crisis can occur because a production machine has not been cleaned correctly during product switches and therefore the latter product might contain traces of nuts, which might be dangerous for people with an allergy regarding nuts, but not for people without this allergy. The second case might be that instead of traces of nuts, there might be traces of disinfectants in the product which can make all people ill. The first case would imply that only a few consumers will not buy the product or wait for some time. The second case probably has larger consequences regarding sales even if the customers were pre-crisis product users. Between product categories, the nature of the product crises can differ and therefore also the outcome of pre-crisis product usage. Possibly the brand commitment hypothesis only holds for less extreme situations. Furthermore, a company can make a mistake once, but if this happens more often or on a regular basis, this effect might diminish.

is the commitment that matters. Commitment will be interpreted in this research as the loyalty and preference a consumer has with regard to a product. This entails how likely a consumer is to buy a brand or product in comparison to other products in general, but also in the case of better alternatives. A shift in alternatives or the relative position of a brand can be altered because of a product crisis. When the brand’s position in general has deteriorated a consumer might still decide to buy it because there is a certain loyalty or preference towards the products, even though there may be other (better) alternatives which the consumer has access to. This can be called commitment. This commitment can also be seen as the strength of the attitudes consumers hold (Priester et al., 2004). Stronger attitudes seem to be more resistant to new information and also seem to guide (automatic) consumers more than weak attitudes. A high commitment might thus lead to a reduced impact of a product crisis, leaving the purchase intentions of a consumer intact. This would mean that committed customers are less likely to reduce their purchases, making the impact on a company’s brand performance less severe. This variable might therefore influence the relationship between a product crisis and brand attitudes.

A finding from the research of Ma et al. (2010) is that some brands lose a part of their market share to its competitors after the product crisis but that they eventually get it back. So it seems like the consumers after a while decide to buy the brand again. Maybe this can be explained by the fact that they used the product without the problem that recently occurred. They have experiences with the brand before the product crisis came up. So there might be a difference regarding the impact of a product crisis considering the consumer was somehow committed before the product crisis occurred (Mowen et al., 1981). Cleeren et al. (2008) also state that pre-crisis loyalty and familiarity form a buffer for a product crisis. This would support the introduction of the hypothesis that brand commitment would weaken the impact of a product crisis on brand performance. The hypothesis that follows from this is:

Conceptual model

3. 1 Conceptual model

3.1.1 Conceptual model

The model belonging to this research can be found in Figure 1. This also shows that the main relation is between product crisis and brand performance. The expectation is that a product crisis negatively affects brand attitudes. Subsequently this relation is expected to be influenced by two variables, which are brand commitment and product crisis harm type. Brand commitment is expected to weaken the relation between product crisis and brand attitudes. When consumers used the product before the impact of the product crisis it is expected to be less than for non-users. Furthermore differences between levels of product crisis harm are expected. This could lead to either a more negative or less negative influence on the relation between product crisis and brand attitudes depending on the category.

Methodology

4.1 Methodology

4.1.1 The study

Because of the nature of this study (a master thesis) it is not possible to hold a longitudinal study regarding a product crisis, because this might take years to investigate thoroughly. Therefore the emphasis is on short-term effects of a product crisis for different product categories. This means that company’s actions to repair the damage done are excluded, so the actual impact of the crisis can be measured more specifically. Because the differences in repairing actions are excluded, a possible bias factor is removed from the research. Furthermore, prior research mentions that the effects after several months are often minimal (Vassilikopoulou et al, 2009). The fact that this study does not focus on the long term hopefully reduces the possible bias caused by this factor, which might include the fact that consumers simply forget the crisis or that companies can communicate or reduce the damage done in another way.

4.1.2 Conclusive causal study

causal research design will be followed (Malhotra, 2007). This type of research design requires an experiment to be done. This experiment will be held among consumers by using a survey. One variable will be moderated, while the other variables will remain unchanged. This way, the effects of the moderated variable on the dependent variable can be measured. These different situations will be shown to the consumer by providing them with different situations (where only one variable is different) after which they have to fill in a standard set of questions regarding brand attitudes (the dependent variable).

4.1.3 Sample

Since all consumers can be confronted with a product crises and certain harm types all consumers are suitable for this sample. The participants will be found be using the snowball sampling technique. At first, personal contacts of the author, such as friends, family and acquaintances will be asked to participate by sending them a hyperlink to the survey belonging to this paper. Subsequently, they will be asked to forward this hyperlink to others they know. This should lead to a large and diverse sample. Based on their answer regarding the brand commitment variable they will be divided into two groups, which will be called low- and high commitment. This is because the level of brand commitment should lead to different outcomes regarding brand attitudes.

4.1.4 Measurement of variables

This paper will investigate four variables; product crisis, product crisis harm type, brand commitment and brand attitudes. The variables and their roles will be mentioned hereafter and will be discussed in further detail in the next few paragraphs. This discussion will elaborate about how the variables will be tested and measured.

Independent variable: “Product crisis” Dependent variable: “Brand attitudes”

4.1.4.1 Measurement of brand attitudes

Ma et al. (2010) measure brand performance using purchase-related items. The model they used is called the DNB-Dirichelet-model, which measures several purchase-related factors over a long time. Because this research focuses on the short-term these long-term purchase-related items have to be translated into more short-term items. Brand performance will therefore be defined as the consumer’s willingness the buy the product (after the crisis has occurred) and their expectations to do so in the future. This definition is chosen because that is what is most important for a company is sales. This is also what is influenced by negative publicity regarding a crisis (Vassilikopoulou et al., 2008) (Griffin et al., 1991). According to Heerde et al., (2007) sales can be used to investigate the impact of a product crisis on brand performance. As stated in paragraph 2.4 brand attitudes can be indicative of brand performance. They can show a possible change in brand performance and sales. Since a survey will be held among consumers and not companies, brand attitudes will be used to measure the outcomes.

likelihood. In his research these three elements formed one brand attitude, which were purchase intentions. As discussed before, the effect of a product crisis can be found back in sales, which could be measured by using brand attitudes. Purchase intentions are a brand attitude and can thus be used to measure the effects of product crises. Furthermore, purchase intentions are also closely related to sales. It is likely that other possibly brand attitudes (e.g. brand trust and brand loyalty) are included in this attitude. Therefore, purchase intentions are a good measure to find the outcomes of a product crisis. In Tsao’s (2010) research all of these elements have shown to fit within the factor of purchase intentions because they all had complied with the criteria regarding factor loadings. This means that these elements together are well able to measure purchase intentions together.

After every harm type situation the questions used to measure brand attitudes are:

What is the probability that you will buy the product?

Very low, low, slightly lower, normal, slightly higher, high, very high

What is the willingness for you to buy the product?

Very low, low, slightly lower, normal, slightly higher, high, very high

What is the likelihood that you will buy the product?

Very low, low, slightly lower, normal, slightly higher, high, very high

4.1.4.2 Measurement of brand commitment

likely that this will generate a diverse sample regarding commitment. The product will be discussed in paragraph 4.1.4.4.

According to Raju et al. (2009) brand commitment can be measured by looking at three factors. The first is whether a consumer would go to another store to buy their brand if it is unavailable. The second is whether they see themselves as being loyal to the brand. The final factor is whether they will prefer a brand that is on sale than the brand in question. These questions are measured on a 9-point Likert scale in the research of Raju et al. (2009). However, since this paper utilizes a 7-point Likert scale for the other questions this scale will be used for the sake of consistency. Since brand commitment is about a situation before the crisis occurred this question will only be asked in the no-harm situation. This is because we want to know if pre-crisis brand commitment is relevant and it is not important whether this changes with regard to a product crisis.

Thus, to measure brand commitment the following questions will be asked in case 1 (the normal situation):

Would it make a difference to you to buy another brand if the brand in question was unavailable in the store?

Completely disagree, disagree, slightly disagree, indifferent, slightly agree, agree, completely agree

Do you see yourself as loyal to this brand?

Completely disagree, disagree, slightly disagree, indifferent, slightly agree, agree, completely agree

Would you prefer a brand that is on sale than the brand in question?

4.1.4.3 Manipulation of the harm types

The variable that will be manipulated is that of the product crisis harm types. These will be presented in four different situations. These situations will be: no crisis, direct physical harm, direct non-physical harm and indirect harm. The other variables will be kept the same. The non-crisis situation will be included to compare this situation with the other three situations where harm does take place. This situation can be used to confirm the relationship between product crises and brand attitudes and also to confirm that a product crisis situation rating is more negative than a non-crisis situation.

4.1.4.4 The product

4.2 Manipulation check

4.2.1 Manipulation check

At first, a normal situation is included in the survey to find brand attitude outcomes in a normal situation. In turn, these outcomes can be compared with the different harm type situations to find out if this led to a change in consumer’s brand attitudes. This control group should allow the results to be compared with this control group to find out if there are significant discrepancies.

4.3 Experimental design

4.3.1 Experimental design

The experimental design will be a 2 (high versus low commitment) x 4 (different crisis situations) design. This results in 8 situations as depicted in Table 1.

No crisis Direct physical harm Direct non-physical harm Indirect harm Low commitment

Outcome 1 Outcome 3 Outcome 5 Outcome 7

High

commitment

Outcome 2 Outcome 4 Outcome 6 Outcome 8

4.4 The survey

4.4.1 Survey procedure

All participants to the research will receive the same survey questions and situations and thus a within-group participant design will be used. The distinction regarding brand commitment can be made based on the answers given regarding the crisis situation. Therefore, no separate surveys have to be held. Also, regarding the harm types the participants should give an answer regarding all situations because the goal is to confirm that there are different impacts of the harm types, which should not be solely based on participant characteristics. The goal is to find out whether the different harm types have varying effects on brand attitudes, which can also be shown if participants are exposed to all scenarios, allowing them to more deliberately compare the situations. A causal relationship is being investigated between harm types and brand attitudes, on which consumer characteristics should not have a moderating effect except for the fact that their brand commitment can differ. This can lead to a less severe impact of the harm type, but this should not make a distinction between the different impacts of harm types regarding severity. If brand commitment is shown not to have an effect on brand attitude evaluation this variable can be said to not influence the relationship between product crises and brand attitudes for a given harm type.

4.4.2 Survey design

statement is included which says that the information mentioned in the scenarios is provided during the time the respondents are still deliberating about which item to purchase. This implies that they should take into account the received information to evaluate the possible purchase of the Apple iPhone. Hereafter, the cases will be presented. After this the participant has to answer questions about brand attitudes. The design of the survey will be presented hereafter. The exact survey which will be presented to the participants can be found in Appendix I.

4.4 Plan of analysis

4.4.1 Product crisis and brand attitudes

The relationship can be tested by using a repeated measures ANOVA. This test is suited for comparing a single group on three or more group means. This paper compares a single sample regarding several (harm) scenarios and therefore the repeated measures ANOVA can be used. By utilizing this statistical test, the outcomes regarding brand attitudes for no crisis versus product crisis can be analyzed. The expected outcome is that the no crisis situation differs significant from the situations where a product crisis does occur.

4.4.2 Brand commitment

For every situation a comparison has to be made to find out whether there is a relationship between brand commitment and brand attitudes. If the brand commitment score is high, then the score for brand attitudes should also be relatively high. To test this the sample will be split up in a low (brand commitment rating =<4) and high (brand commitment rating >4) brand commitment group. Independent sample t-tests will be used to tests whether these two groups differ regarding their purchase intentions per harm type.

This is executed by including the two brand commitment groups in the harm types repeated measures ANOVA since this more clearly shows the interactions of the sample.

4.4.3 Harm types

The different harm types can be compared by using a repeated measures ANOVA. This can show whether the different harm types significantly differ from one another regarding the outcomes on brand attitudes.

4.5 Validity

4.5.1 Internal validity

The internal validity should be high because the situations used in the survey are new and is therefore unlikely that consumers have a predisposition regarding the specific situations. Of course, there is the predisposition regarding the harm type, but it is the goal of this research to find out if this predisposition exists and plays a role for evaluating brand attitudes. Furthermore, the situations are clear and the questions used in the survey leave little room for interpretation. The outcomes regarding brand attitudes should score rather high on validity.

4.5.2 External validity

4.6 Reliability

4.6.1Test-Retest Reliability

It might be that the participant’s mood or emotions at the time of filling in the survey can influence the outcomes since these outcomes are strongly based on feelings and emotions. Nonetheless, it is about the distinction between the different outcomes which should not suffer from one’s emotions at that time. The absolute outcomes might differ (e.g. more negative outcomes because of a bad mood) but the relative outcomes do not. For example, a more negative score on direct physical harm, but also on direct non-physical harm should still lead to the same results when comparing the two.

Results

5.1 Results

5.1.1 Results

This chapter will provide the results found during the statistical analysis of the survey results. At first, the factors included in the survey will be tested. Hereafter, the relationships between the different variables and outcomes will be tested.

5.2 Sample characteristics

5.2.1 Gender

There were 50 male and 35 female respondents participating in this research as can be seen in Table 2. This is a ratio of 1 man to 0,7 woman. In The Netherlands, the gender distribution in 2010 existed out of 8.203.476 men and 8.371.513 women which leads to a ratio of 1 man per 1,02 woman9. There thus is a discrepancy between the number of male and female participants. However, the sample is not completely one-sided.

Table 2: Gender distribution

5.2.2 Age

The age distribution as can be found in Table 3. It shows that there is a large proportion of the respondents in the 21-25 and 40+ age categories. If this is compared to the CBS (Central Bureau for Statistics) data it shows that there is a large representation of the

9

40 age group. In The Netherlands the ratio for >20:20-40:40+ is about 1:1:210. In this sample this ratio is 1:4:3. This can be explained by the fact that people younger than 15 often do not use smartphones yet or are not fully aware of issues such as brands and products and whose responses therefore are not fully relevant for this research. At the same time the 40+ category also contains the people who did not grew up with high-tech mobile phones and possible are not using these kinds of phones. The 20-40 age category therefore seems to be the most relevant age category among which to hold this research. Therefore, the disproportionate amount of respondents in this category should not be an issue. This population might not fully represent the Dutch population, but it might very well visualize the customer population of a company such as Apple. Nonetheless it has to be mentioned that all categories are still represented and the statistical tests will be ran separately for the different age groups.

Table 3: Age distribution

10

5.2.3 Household income

Gross salaries in The Netherlands in 2011 were €32.50011. As can be seen in Table 4 the distribution is different in this sample. It includes mostly respondents who are either below or above this average. But this can possible be explained by the age distribution. There are a lot of people below the age of 25, who might not have a job yet and therefore have low incomes. Apart from this there is a large representation of higher incomes. The average income lies in the €10.001-20.000 category. Nonetheless, there is a clear distinction between a low- and a high-income group, which would allow for better results within this population if income differences are tested regarding the variables that will be tested in this research. Furthermore, there were four respondents who did not fill in their household income.

Table 4: Household income

11

5.2.4 Sample summary

It can be stated that the sample is not entirely representative of the Dutch population. However, there are clear distinctions within the sample (e.g. income) that allow testing between these different groups.

5.3 Factor analysis

5.3.1 Factor analysis

This paragraph will describe the factor analysis done regarding the brand commitment and purchase intention factors. The inputs, processing and outcomes will be described.

5.3.1.1 Brand commitment factor

At first, the variable regarding whether the consumer preferred a brand on that was on sale was corrected because the scale was incorrect. Where other variables used high values to describe outcomes that were positive for brand commitment, preference for brand on sale used high values for outcomes that were negative for brand commitment. Therefore, the outcomes were contrasted (i.e. 7 became 1, 6 became 2 etc.) leading to a new variable called CorrectedBrandOnSale.

(ConsLoyalty). The factor brand commitment only just fails the Kaiser-Meyer-Olkin Measure of Sampling Adequacy test (KMO-test), which should be 0,6 or higher to continue factor analysis (see Table 1 in Appendix II). However, if we look at the communalities, there are no large differences between the lowest two values (see Table 2 in Appendix II). According to the Eigenvalues only the WhatIfNotAvail should be included since this is the only variable with an Eigenvalue higher than 1 (see Table 3 in Appendix II). However, 90% of the variance is explained by the first two factors. The scree plot also states that the first two factors should be used (see Figure 1 in Appendix II). However, the component matrix tells us that all three variables score high on the principal component analysis (see Table 4 in Appendix II). Thus there is evidence that variable 3 (CorrectedBrandOnSale) should be excluded from the factor. A scale reliability test has also been done to verify the internal reliability of the factor. The Cronbach’s Alpha from this test was 0,739, which is higher than the minimum required value of 0,6 (see Table 6 in Appendix II). However, the evidence that the factor is correct is not overwhelming. The KMO-test only failed slightly. And the variance explained by the other two variables is only just 90%. The principal component analysis shows that all variables are suitable for inclusion in the factor. Finally, the Cronbach’s Alpha is good enough. The evidence points both ways. Therefore a correlation analysis is done between CorrectedBrandOnSale and the BrandCommitment factor (see Table 7 in Appendix II). The BrandCommitment factor was generated by summing up the three variables and dividing this by three to provide an average brand commitment outcome. The correlation matrix shows that there is a high correlation (0,737) between the variable CorrectedBrandOnSale and the factor. Because of this outcome and because tests in other research have confirmed the relevance of this factor this paper shall therefore use the factor BrandCommitment made up of three variables.

5.3.2 Purchase intention factor

done for the control group to see if the variables were suited to be included in the factor. The factor passes the KMO-test because the outcome is 0,715, which is higher than 0,6 (see Table 8 in Appendix II). Based on Table 10 from Appendix II the first two variables explain most of the variance in the population already. Furthermore, the scree plot in Figure 2 in Appendix II also shows a sharp curve. However, the component matrix shows that all variables are suited for the factor (see Table 11 in Appendix II). Also, the Cronbach’s Alpha regarding internal reliability is 0,881 which is higher than 0,6 (see Table 13 in Appendix II). Because the factor clearly passes the KMO-test, the principal component analysis and the internal reliability test all three variables will be included in the factor. The factor value is found by summing up the three variables and dividing this by three to come to an average value.

5.4 Harm types and purchase intentions

5.4.1 Testing harm types and purchasing intentions

This paragraph will provide the outcomes regarding the effects of different harm types on purchase intentions. These tests are done to confirm the first hypothesis:

“H1: A product crisis has a negative impact on brand attitudes”

5.4.1.1 Control group versus harm types

A repeated measures ANOVA is done to see if there are significant differences between the scenario without harm and the scenarios with harm. As the within-subject factor name “HarmTypes” is chosen. As the measurement name “PurchaseIntentions” is chosen. The within-subjects variables are the four harm scenarios; no harm, direct physical harm, direct non-physical harm and indirect harm (see Table 1 in Appendix III). A graph plot is also included with HarmTypes on the horizontal axis. The means are displayed for HarmTypes. Main effects are compared by using the Bonferroni method, because this method counteracts the problem of multiple comparisons12. The boxes for descriptive statistics and estimates of effect size are also checked. The test scores significant on Mauchly’s test of sphericity (see Table 3 in Appendix III). Sphericity is the condition where the variances of the differences between all combinations of related groups (levels) are equal13. Violation of sphericity is when the variances of the differences between all combinations of related groups are not equal. Therefore, the test will be continued using the Geisser correction. The outcome is significant after the Greenhouse-Geisser correction with a F(2,539)=48,799 and a p-value of 0,000 (see Table 4 in Appendix III). The next part will continue with describing the differences between the control group and the harm types.

5.4.1.2 Control group versus direct physical harm

The outcomes state that there indeed is a significant difference (p-value of 0,000) between the scenario without harm and the scenario with direct physical harm and that respondents had significantly lower purchase intentions in the case of direct physical harm (see Table 5 in Appendix III).

5.4.1.3 Control group versus direct non-physical harm

Here, the control group is compared to the scenario with direct non-physical harm (i.e. fire). Also, this harm type differs significantly from the control group (p-value of 0,000) and thus direct non-physical damage has a significantly more negative impact on purchase intentions (see Table 5 in Appendix III).

5.4.1.4 Control group versus indirect harm

Finally, the last harm type will be compared with the no harm scenario. This scenario is also significantly different from the control group with a p-value of 0,000 meaning that indirect harm has a significantly more negative effect on purchase intentions (see Table 5 in Appendix III).

5.4.1.5 Test outcomes

Direct physical harm

Direct non-physical harm

Indirect harm

p-value from repeated

measures ANOVA _{0,000 0,000 0,000} µ (control group µ=3,44) 2,21 1,91 2,51 Significantly negative effect on brand

attitudes from no harm control group?

√

√

√

Table 5: Outcomes regarding harm types

5.5 Differences between harm types

5.5.1 Differences between harm types

It is also important to find out whether the harm types differ. This part will test whether the different harm types truly differ in harm from the perception of consumers. The hypothesis belonging to these tests is:

“H2: Different types of damage have a different impact on the relationship between product crises and brand attitudes, where direct physical harm is expected to have the most severe impact, direct non-physical harm a medium impact and indirect harm the weakest impact.”

Just like in the previous paragraph a repeated measures ANOVA can be used to compare the different harm scenarios.

5.5.1.1 Direct physical harm versus direct non-physical harm

p-value of 0,028 (see Table 5 in Appendix III). Since its mean p-value is less this means that direct non-physical harm has a significantly more negative impact on purchase intentions than direct physical harm.

5.5.1.2 Direct physical harm versus indirect harm

Indirect harm does not have a significantly lower impact on purchase intentions than direct physical harm because it has a p-value of 0,091, which is higher than the set α of 0,05 (see Table 5 in Appendix III).

5.5.1.3 Direct non-physical harm versus indirect harm

Direct non-physical harm has a significantly larger and more negative impact on purchase intentions than indirect harm with a p-value of 0,000 (see Table 5 in Appendix III).

5.5.1.4 Test outcomes

The hypothesis has not been confirmed (see Table 6). This is because direct physical harm did not have the most severe impact on brand attitudes. Direct non-physical harm was found to do the most severe damage. Also, direct physical harm and indirect harm were not significantly different from one another on the 0,05 level, but they are on the 0,1 level. The stated α for the test was 0,05, so there is no significant difference.

Direct physical harm Direct non-physical harm Indirect harm Expected relative severity among harm types

High Medium Low

µ _{2,21 1,91 2,51}

p-value from repeated measures ANOVA with direct non-physical harm

0,028 - 0,000

p-value from repeated measures ANOVA with indirect harm

0,091 0,000 -

Outcome of relative severity among harm types

- Highest -

Significantly different

from other harm types?

_X

_√

_X

Confirmation of expectations?

X

X

X

Figure 2: Purchase intention means per harm type scenario

5.6 Brand commitment and harm types

5.6.1 Brand commitment and harm types

This part will investigate the fact whether brand commitment has a positive influence on the relationship between harm types and purchase intentions. The following hypothesis is tested:

To analyze this case two subgroups have to be generated. Group one exists out of all respondents who scored normal or lower (i.e. 1 till 4) on brand commitment. Group two exists out of the respondents who scored higher than four on brand commitment and can therefore be considered as highly committed. These two groups were compared based on the different harm types. Independent samples t-tests were performed to come to the results. Also, a repeated measures ANOVA was done where the high and low brand commitment groups were included to see what their outcomes were.

5.6.1.1 No harm scenario

In this scenario high- and low commitment groups were found to be significantly different with regard to their purchase intentions with a p-value of 0,000 (see Table 7 in Appendix III). This makes sense since the “Apple-likers” are initially more likely to buy Apple products. This therefore confirms that committed consumers differ from less committed consumers under normal circumstances regarding their purchase intentions.

5.6.1.2 Direct physical harm scenario

In this scenario the high-commitment group also differs significantly from the lower-commitment group with regard to the scores on the purchase intention variable with a p-value of 0,000 (see Table 9 in Appendix III). It thus seems that committed consumers do not drop to the same level of purchase intentions as their lower commitment counterparts.

5.6.1.3 Direct non-physical harm

intentions for the high-commitment group. It therefore seems that the high-commitment group is also strongly affected by this kind of harm.

5.6.1.4 Indirect harm

There is only a significant difference if α=0.1 is assumed for the case where equal variances are assumed. Thus it would be best to state that there are no significant differences between the high- and low-commitment groups. However, if we look at Table 13 in Appendix III it shows that the purchase intentions for the low-commitment group are higher again. Thus it seems that the low-commitment group is less susceptible to indirect harm, at least not significantly different from the high-commitment group.

5.6.1.5 Brand commitment and harm types

Another repeated measures ANOVA was done. The process is executed the same as in paragraph 5.4.1.1 but this time the two brand commitment groups are included as a between-subjects factor. According to the results there was a significant difference between the high and low commitment groups regarding purchase intentions with a p-value of 0,000 (see Table 18 in Appendix III).

Figure 3: Purchase intention means with brand commitment taken into account

5.6.1.6 Test outcomes

The hypothesis can partially be confirmed. Regarding the independent sample t-tests the results do not support to general applicability of brand commitment regarding the harm types. Only for the no harm and direct physical harm scenarios brand commitment was found to have a moderating effect on the relationship between product crises and brand attitudes (see Table 7). It had no significant influence for the direct non-physical and indirect harm scenarios. However, this was only the case at the independent sample t-tests, which looked at individual harm types without any interaction effects. These can be investigated by using the repeated measures ANOVA. This test provided the result that brand commitment overall caused a significant difference regarding the purchase intentions (see Table 8 and 9). The repeated measures ANOVA showed that there was no significant difference between the direct physical harm and the indirect harm scenario when brand commitment was included (see Table 10). This outcome is similar to the test without brand commitment. However, the p-value increased a lot. The outcome of the test with brand commitment resulted in a p-value of 1,000, whereas the p-value of the test without brand commitment was 0,091. This also shows when Figure 2 and 3 are compared. First there is a more noticeable difference between the two harm types in the graph, whereas they are nearly equal when brand commitment is included.

No harm scenario Direct physical harm scenario Direct non-physical harm scenario Indirect harm scenario Low commitment group µ 3,12 1,96 1,77 2,37 High commitment group µ 4,54 3,07 2,40 2,98

p-value when equal variances are not

assumed 0,000 0,003 0,1 0,13

Significant

difference between high and low commitment groups?

√ √ X X

Outcome

Low commitment group µ 2,308 High commitment group µ 3,250 p-value of pairwise

comparisons

0,000

Significant difference between

high and low commitment?

_√

Table 8: The role of brand commitment based on the repeated measures ANOVA

F-value Greenhouse-Geisser p-value Harm type 44,235 0,000 Harm Type * Commitment 3,092 0,035

Table 9: Repeated measures ANOVA test of within-subjects effects with the Greenhouse-Geisser correction Harm type Compared to: p-value without commitment p-value with commitment No harm Direct physical harm 0,000 0,000 Direct non-physical harm 0,000 0,000 Indirect harm 0,000 0,000 Direct physical harm Direct non-physical harm 0,028 0,004 Indirect harm 0,091 1,000 Direct non-physical harm Indirect harm 0,000 0,006

So even though the groups do not differ significantly on all individual harm types, there does seem to be a general influence of brand commitment. The high brand commitment group seems to be much less susceptible to harm since they have higher purchase intentions in general and they also have significantly higher purchase intentions for the no harm and direct non-physical harm scenarios (see Table 11). Furthermore, the direct physical harm scenario increases a lot for the high commitment group and its mean even surpasses the indirect harm type.

Without commitment µ

Low commitment µ High

commitment µ

No harm 3,44 3,12 4,54

Direct physical harm 2,21 1,96 3,07

Direct non-physical harm

1,91 1,77 2,40

Indirect harm 2,51 2,37 2,98

Table 11: Purchase intentions means for the test without commitment versus the commitment group means

5.6.1.6.1 Test outcomes summary