Investigating the effect of a consumer's level of involvement on the strength of the zero price effect

(1)

Investigating the Effect of a Consumer’s Level of

Involvement on the Strength of the Zero Price Effect

Master Thesis, final draft

Msc. Business Administration – Marketing

Kim van Zon (10443134) 23rd of June, 2017

(2)

1

List of Figures

Figure 1: Elaboration Likelihood Model ... 21

Figure 2: Conceptual model ... 24

Figure 3: Low priced product advertisement in the zero pricing condition ... 36

Figure 4: High priced product advertisement in the zero pricing condition ... 37

Figure 5: Low priced product choice across the pricing conditions... 51

Figure 6: High priced product choice across the pricing conditions ... 53

Figure 7: Low priced product choice based on pricing conditions and involvement ... 55

Figure 8: High priced product choice based on pricing conditions and involvement ... 56

List of Tables

Table 1: Pre-test: correlations between variables ... 70

Table 2: Pre-test: sample demographics ... 70

Table 3: Correlations between model variables ... 43

Table 4: Normal distribution of involvement towards the advertisement ... 45

Table 5: Normal distribution of involvement towards the advertisement (2) ... 46

Table 6: Success of the involvement manipulation (independent sample t-test) ... 47

Table 7: Effect of pricing conditions on low priced product choice ... 50

Table 8: Effect of the control variables on low priced product choice ... 84

Table 9: Effect of pricing conditions on high priced product choice ... 53

Table 10: Effect of the control variables on high priced product choice ... 84

Table 11: Moderation effects of pricing strategy x involvement on demand distribution, for low priced products ... 54

Table 12: Moderation effects of pricing strategy x involvement on demand distribution, for low priced products ... 55

(6)

5

Statement of originality

This document is written by Kim van Zon who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Date: 23rd_{of June, 2016}

(7)

6

Abstract

Products bought through price promotions account for a quarter of a company’s sales, and are an important part of doing business. However, pricing strategies are expensive and need to be optimized. This study examines the effectiveness of zero pricing strategies in creating irrational demand. Specifically, whether a product bundle with a free component can lead consumers to forego a product option they should prefer. This concept is termed the zero price effect. To help marketers to effectively apply zero pricing strategies, more consistent insights into the strength of the zero price effect for utilitarian product bundles are necessary. Specifically, it is hypothesized that a zero pricing strategy creates changes in the demand distributions for both low and high priced utilitarian product bundles. Moreover, it is hypothesized that this effect is moderated by a consumer’s level of involvement towards the persuasive message of ‘free’. High involved consumers cognitively process a message, which weakens the zero price effect. Low involved consumers use affect heuristics when processing the message, which strengthens the effect. The hypotheses are tested based on data collected amongst 207 Dutch consumers. The data is analysed through multinomial and binary logistic regression. Weak support for the existence of the zero price effect for the low priced utilitarian product bundle is found. None of the other hypotheses are supported.

Key words: zero price effect; involvement; free; affect; heuristics

(8)

7

Many retailers, manufacturers, and service providers, incorporate price promotions in their marketing mix. For example, walk into any Dutch drugstore and the signs ‘1+1 free’, ‘buy 2 get 1 free’ or ‘50% off’, are immediately evident. In fact, 26 percent of the volume of sales in the Netherlands (2015) results from products that are sold on promotions (Eales, 2016). These discounted products have an average deal depth of 27 percent, which is an increase of 1 percent point in comparison to 2014 (Eales, 2016). These numbers indicate that products bought through promotions continue to be an important aspect of a firm’s sales. However, such promotional activities are also costly. Eales (2016) therefore argues that one of the challenges that firms face, is to optimize promotional pricing strategies.

One of these promotional strategies is one in which products, or product bundle components, are offered for free. The success of this type of price promotion partly depends on the extent to which the campaign leads to increases in profitable sales and demand. Shampanier, Mazar and Ariely (2007) argue that the signal ‘free’ leads consumers to forego an option they find preferable, to obtain the free product or product bundle component. In other words, the signal ‘free’ has the potential to create irrational demand (Shampanier et al., 2007; Baumbach, 2016; Nicolau and Seller, 2012). Irrational demand resulting from a zero pricing strategy is termed the zero price effect, and occurs when two elements are present in a demand distribution. First, when there is a significant increase in the proportion of consumers choosing the free product. Second, when there is a significant decrease in the proportion of consumers that choose an alternative product with a similar discount, but that remains positively priced (Shampanier et al., 2007). The zero price effect is thus present when the distribution of demand changes when a product is offered for free. However, the literature is inconsistent regarding the existence of this effect. Whereas some scholars argue that the zero price effect is present for low priced products (Shampanier et al., 2007; Nicolau and Seller, 2012) and high priced products (Shampanier et al., 2007), others find no effect for high priced

(9)

8

products, and only a weak effect for bundled high priced products (Baumbach, 2016). This indicates that there may be some additional variables that influence the strength of the zero price effect. This idea is supported by Hossain and Saini (2015), who argue that the zero price effect is particularly strong for hedonic or hedonically framed products, and is weak for utilitarian or utilitarian framed products. Their explanation is that utilitarian products inherently contain lower levels of affect. This is relevant, because affect is the driving mechanism behind the zero price effect (Shampanier et al, 2007). However, their theory does not fully explain the differences in the strength of the zero price effect for the researched high priced products, all of which are more utilitarian than hedonic in nature. For example, the zero price effect was found to be present for a television, but not for a car sound system (Shampanier et al., 2007; Baumbach, 2016). What then drives differences in the strength of the zero price effect?

This study argues that the strength of the zero price effect is influenced by a consumers’ level of involvement towards the persuasive advertising message of a free product or bundle component. According to the Elaboration Likelihood Model (ELM), low involved consumers are more likely to use heuristic cues in their decision making, whereas high involved consumers pay less attention to such cues (Petty and Cacioppo, 1983). The cue ‘free’ is associated with positive affect, and may serve as affective heuristic (Shampanier et al., 2007). Thus, the zero price effect is likely to be stronger for low than high involved consumers. Th current study will investigate this effect for bundled products. Bundled products are of more practical relevance to marketers, since the concept of ‘free’ is usually applied in combination with a positively priced product. Think of ‘buy one and get one free’ or ‘order now and get free shipping’. The concept of ‘free’ is less often applied in a single product context, in which the consumer does not have to pay any monetary amount to obtain a product (Baumbach, 2016). Furthermore, previous research is consistent for hedonic products,

(10)

9

but inconsistent for high priced utilitarian products. To shed additional light on the main relationship between a pricing strategy and changes in the demand distributions, high priced utilitarian products are investigated. To the best of my knowledge, little research has been conducted on utilitarian low priced product bundles. These will also be included in this research to improve robustness of the results, and to explore differences across product types within utilitarian product categories. This leads to the following research question:

“To what extent does a consumer’s level of involvement towards a persuasive message of a product bundle, moderate the relationship between a zero pricing strategy and changes in demand distributions (zero price effect) for both high and low priced, utilitarian, bundled product offerings?”

A quantitative approach is taken to investigate the research question. Data is collected through experimental vignette surveys, resulting in a sample of 207 participants. This data is analysed through multinomial and binary logistic regression. The results of these analyses contribute to the existing literature by combining two theories and models: the zero price effect and the elaboration likelihood model. Additionally, it extends the concept of the zero price effect to a low priced utilitarian product category. The research is also of practical relevance to managers and marketers, since it helps them to obtain a better understanding of when the promotional strategy of ‘free’ works best. This information can be used to optimize costly promotional activities involving free components in a bundle.

This paper will continue with an overview of the existing literature on the zero price effect in section 2. Section 3 discusses the methodology used to provide an answer to the research question. This is followed by an overview of the results in section 4 provides. Finally, the study concludes with the discussion in section 5.

(11)

10

2. Theoretical Framework

The purpose of this section is to review the existing literature to provide a more complete overview of the research problem. The first section introduces different perspectives on the value of free goods and services, including the zero price effect. This is followed by an explanation of the mechanisms underlying the zero price effect. Next, the elaboration likelihood model, and the role of involvement are considered. The section concludes with an overview of the control variables that are included in the model.

2.1. The value of free

Scholars have different opinions about the value of a free product or product bundle component. Specifically, on how consumers evaluate the quality of a free product, and how this can affect demand. Both topics are subsequently discussed in the following sections.

2.1.1 The relationship between price and quality

There are multiple perspectives on how a product’s price influences quality evaluations. Not all scholars support a consistent association between price and quality. Friedman (1967) argues that the relationship between the two variables is low, and others find that the relationship is dependent on individual characteristics (Shapiro, 1973), or product types (Gardner, 1971). On the other hand, Kardes, Cronley, Kellaris and Posavac (2004) argue that price and quality are consistently positively correlated. In other words, the lower the price, the lower the perceived quality. This is also known as the price-quality heuristic (Gneezy, Gneezy and Lauga, 2014). The price-quality heuristic is particularly prevalent when information load is high, consumers have a high need for cognitive closure, and price is the only extrinsic cue (Kardes et al., 2004; Cronley, Posavac, Meyer, Kardes, and Kellaris, 2005; Dodds, Kent,

(12)

11

Monroe and Grewal, 1991). A lower perception of quality can, but does not necessarily, lead to a decrease in demand for a product (Zeithaml, 1988). If a product is of high quality, but exceeds the consumer’s budget, the perceived value of a quality product may be lower than the perceived value of a lower quality product with a lower price (Zeithaml, 1988; Dodds et al., 1991). Applying this theory to a zero priced product would indicate that the product can be perceived to be of poor quality, but still create demand (depending on the price of the bundle). Chandran and Morwitz (2006) take this concept a level further, by arguing that a zero priced product can serve as harness against negative quality information, and that consumers perceive the quality of a product only as lower when promotions involve monetary discounts. ‘Free’ is viewed as something nonmonetary, and is therefore mentally processed independent of the price of a product or service (Chandran and Morwitz, 2006). Specifically, the ‘free’ cue itself has so much salience, that consumers pay less attention to negative quality cues. When the promotion involves a monetary discount, the price is included in the processing, and negative quality cues will affect value evaluations (Chandran and Morwitz, 2006). In line with this, Baumbach (2016) finds no evidence of negative quality inferences when product or bundle components are offered at a zero price. This combination of a free product or component, without the negative quality inferences that are associated with low priced products, can lead to larger demand increases for a zero priced product, than a discounted product.

2.1.2 The effect of a pricing strategy on demand

According to standard economic theory, demand for a product increases if its price decreases, as long as the price elasticity of demand is not below one (Pindyck and Rubinfeld, 2013). If the price of two products decrease with the same amount, demand may increase for both products, but consumers’ preferences for a specific product should not change (Shampanier et

(13)

12

al., 2007). In other words, if consumers prefer product X before the price reduction, they should not prefer product Y after the price reduction, if both products receive the same discount. To illustrate this rational behaviour, consider the following model, adopted from Shampanier et al., (2007). In the model, 𝑃𝑖 represents the product price for product i. 𝑉𝑖

represents a consumer’s perceived product value for product i.

Consumers will choose product X over Y when:

𝑉_𝑋 > 𝑃_𝑋 and 𝑉_𝑋− 𝑃_𝑋 > 𝑉_𝑌− 𝑃_𝑌 (1)

Consumers will choose product Y over X when:

𝑉𝑌 > 𝑃𝑌 and 𝑉𝑌− 𝑃𝑌 > 𝑉𝑋− 𝑃𝑋 (2)

Consumers choose nothing when:

𝑉_𝑌 < 𝑃_𝑌 and 𝑉_𝑋 < 𝑃_𝑋 (3)

Now, assume that both products X and Y get the same price reduction that equals amount 𝜀.

Consumers will choose product X over Y when:

𝑉_𝑥 > (𝑃_𝑥− 𝜀) and 𝑉_𝑥− (𝑃_𝑥− 𝜀) > 𝑉_𝑦− (𝑃_𝑦− 𝜀) (4)

(14)

13

𝑉_𝑦 > (𝑃_𝑦− 𝜀) and 𝑉_𝑦− (𝑃_𝑦− 𝜀) > 𝑉_𝑥− (𝑃_𝑥− 𝜀) (5)

Consumers choose nothing when:

𝑉𝑦 < (𝑃𝑦− 𝜀) and 𝑉𝑥< (𝑃𝑥− 𝜀) (6)

This model illustrates that consumers could only switch from nothing to one of the products, but not from one product to the other, when faced with price reductions that are equal for both products. When a price reduction causes one of the products to become free, consumers overvalue that product, because it has no or downside risk (Shampanier et al., 2007; Baumbach, 2016). This is explained in more detail in section 2.2. Let this additional perceived value of free be 𝛼 (Shampanier et al., (2007), and let the product that becomes zero be product X:

A consumer who initially preferred Y will prefer X when:

(𝑉_𝑥+ 𝛼) − 𝑃_𝑥 > 𝑉_𝑦− 𝑃_𝑦 (7)

Equation (7) shows that overvaluation of a free product can lead to switching behaviour from product Y to product X. This causes a change in the demand structure, in comparison to a pricing situation in which both X and Y are offered at their normal prices, or at discounted (non-free) prices. This phenomenon is termed the zero price effect. The zero price effect formally defined as “the combination of the increase in the proportion of consumers choosing

(15)

14

X and the decrease of those choosing Y when prices fall from [PX,PY] to [0, PY − PX]” (Shampanier et al., 2007, p. 745).

2.1.3 Evidence for the zero price effect

2.1.3.1 Evidence for the zero price effect: hedonic versus utilitarian products

Hossain and Saini (2015) find that the zero price effect is stronger for hedonic or hedonically framed products, than for utilitarian or utilitarian framed products. Hedonic products “relate to the multisensory, fantasy and emotive aspects of one’s experience with products” (Hirschman and Holbrook, 1982, p. 92). This includes imagined or real scents, tastes, sounds, visuals, and touches, as well as emotional feelings, such as joy or fear (Hirschman and Holbrook, 1982). Hedonic products are aimed at fulfilling the need for fun or pleasure and are affect-rich (Hossain and Saini, 2015). Moreover, consumers are more likely to base purchase decisions for hedonic products on emotional heuristics, rather than rational cost-benefit analyses (Hossain and Saini, 2015). As will be discussed in a subsequent section, the use of emotional heuristics in decision making may positively influence the strength of the zero price effect. However, these observations do not mean that the zero price effect cannot be triggered for utilitarian products. Utilitarian products are cognitively driven, rather than emotion based, goal oriented, and instrumental (Strahilcvitz and Myers, 1998 in Dhar and Wertenbroch, 2000). Utilitarian products are aimed at fulfilling functional needs, and thus have a very different nature than hedonic products (Hossain and Saini, 2015). Hossain and Saini (2015) find that the zero price effect is not present for utilitarian products, due to the lack of affect contained within them. This is inconsistent with results by Shampanier et al. (2007), who confirm the existence of the zero price effect for a utilitarian product. Baumbach (2016) finds mixed results, arguing that there is a weak effect for bundled utilitarian products, but no effect for non-bundled utilitarian products.

(16)

15

Overall, the amount literature on the zero price effect is limited. However, the existence of the zero price effect is generally confirmed for hedonic products. For utilitarian products the results are inconsistent, meaning that more research is necessary to be able to draw conclusions about the zero price effect for these products.

2.1.3.2 Evidence for the zero price effect: low priced utilitarian products

Scholars provide support for the existence of the zero price effect in a low priced single product setting (Shampanier et al., 2007; Hossain and Saini, 2015; Votinov, Aso, Fukuyama, and Mima, 2016), as well as in a low priced bundled setting (Nicolau and Sellers, 2012). However, all of these studies investigated products with hedonic characteristics, including chocolates, cupcakes, hotel breakfasts, and other foods (Shampanier et al., 2007; Hossain and Saini, 2015; Votinovet al., 2016; Nicolau and Seller, 2012). Given the differences between hedonic and utilitarian products, the previous results may not be generalizable to utilitarian products. To the best of my knowledge, only Hossain and Saini (2015) looked at the effects of a zero pricing strategy on the demand for utilitarian low priced products. This study found that the zero price effect was not present (Hossain and Saini, 2015). However, the utilitarian product under investigation was sugar. Sugar is a commodity product, and may not be representative for all types of utilitarian products. Based on the notions that most of the researched low priced products (although hedonic) are subject to the zero price effect, and that previous studies did find an effect for utilitarian products that are bundled, the following is hypothesized:

H1a: When compared to a normal pricing strategy, a zero pricing strategy changes the distribution of demand preferences for low priced, utilitarian product bundles.

(17)

16

To improve the robustness of the study, a discounted pricing strategy is included as well. This is in accordance with previous research on the zero price effect (Shampanier et al., 2007; Baumbach, 2016). According to the model as described in section 2.1.2, the zero price effect is caused by a product becoming free, and not by a regular discount. There should thus be a difference between the discounted and zero pricing condition as well. This leads to the following hypothesis:

H1b: When compared to a discounted pricing strategy, a zero pricing strategy changes the distribution of demand preferences for low priced, utilitarian product bundles.

Based on the model by Shampanier et al. (2007), no differences are expected between the normal and discounted pricing strategies (null hypothesis).

2.1.3.3 Evidence for the zero price effect: high priced utilitarian products

The high priced products that are investigated in the literature are mainly utilitarian in nature, including televisions, car radios, and trolleys (Shampanier et al., 2007; Baumbach, 2016). The results are therefore similar as described previously in section 2.1.3.1: the literature is contradictory. Shampanier et al. (2007) support the existence of the zero price effect, whereas Baumbach (2016) delivers mixed evidence. Looking at the evidence for bundled products alone, a weak effect is found for two high priced utilitarian products (Baumbach, 2016). Based on the latter information, the following hypotheses are developed:

H2a: When compared to a normal pricing strategy, a zero pricing strategy changes the distribution of demand preferences for high priced, utilitarian product bundles.

(18)

17

H2b: When compared to a discounted pricing strategy, a zero pricing strategy changes the distribution of demand preferences for high priced, utilitarian product bundles.

Again, Based on the model by Shampanier et al. (2007), no differences are expected between the normal and discounted pricing strategies (null hypothesis).

2.1.1.4 Differences between low and high priced utilitarian product bundles

Most of the investigated low priced products are hedonic, whereas most of the investigated high priced products are utilitarian. Therefore, no clear comparisons between low and high priced products can be made based on the available zero price effect literature. Related research on price-quality heuristics indicates that price-quality inferences are stronger for high- than low priced utilitarian products (Zielke and Komor, 2015). The negative effect of quality inferences based on a zero price, may therefore cause the zero price effect to be less prevalent for the high priced category. However, section 2.1.1 already provided evidence that this price-quality heuristic may not hold for free products.

Additionally, some higher priced products (a car or house) may trigger higher levels of involvement for all consumers making a decision on the specific product (Desphane and Hoyer, 1983), causing them to use less affective heuristics in their decision making (Petty and Cacioppo, 1983). Since these heuristics are related to the zero price effect, as is discussed in the following sections, this can lead the zero price effect to be weaker for high priced products. Whereas involvement levels may be high, inherently, for some specific high priced products, involvement is generally context-dependent, and varies across individuals (Petty and Cacioppo, 1986). Therefore, it cannot be concluded that the zero price effect is not present for all utilitarian products. Note that this concept is different from the extent to which

(19)

18

involvement at an individual level influences the strength of the zero price effect. This is discussed in a subsequent section.

To summarize, there is no clear evidence on whether the zero price effect is different for low and high priced products. This is also the reason why the variables are included separately in the model, rather than as moderator variable. An exploration of the differences will be discussed in the Discussion in section 5.

2.2 Explaining the zero price effect

Although the literature is inconsistent regarding the existence of the zero price effect for different type of products, scholars do agree on the mechanism that explains the zero price effect: affect. The following section discusses the affect heuristic, and the relevance of the heuristic in the context of free product offerings.

When making decisions, consumers engage in risk-benefit analyses (Pindyck and Rubinfeld,2013). Slovic and Peters (2006) focus on the risk part of these analyses, and argue that consumers discern two types of risk: risk as analysis and risk as feelings. Risk as analysis assumes that people determine uncertainty levels in purchase decisions through logical and deliberative thinking (Slovic and Peters, 2006). This type of risk therefore fits well within theories of rational choice. Risk as feelings, on the other hand, assumes that individuals use instinctive responses to identify uncertainties in purchase decisions (Slovic and Peters, 2006). One of the main characteristics of the risk as feelings mechanism is affect (Slovic and Peters, 2006). This means that consumers link specific emotions to objects or events, and sometimes consider these emotions in purchase decisions (Finucane, Alhakami, Slovic and Johnson, 2000). They do so, because an emotion-based impression of a product, object, or service is readily available, whereas a careful trade-off between risks and benefits is not (Finucane et al., 2000). It is thus faster, easier, and sometimes more efficient to rely on affect instead of

(20)

19

deliberative thinking (Slovic and Peters, 2006). The process of using affect as a shortcut to make decisions, is also referred to as the affect heuristic (Finucane et al., 2000). This heuristic is particularly strong when people’s opportunities for analytical reflections are limited, and when efficient decision-making is necessary (Finucane et al., 2000). For example, when a consumer feels pressured due to resource or time constraints.

The risk as feelings mechanism, and the affect heuristic, are likely to be used more often than usual, when a zero pricing strategy is applied. Specifically, a free product or service has no costs and therefore carries no downside risks. This invokes a positive affective response, which consumers use as decision making cue (Shampanier et al., 2007). Scholars find that consumers view a free low priced hedonic product as significantly more attractive than the same product when it costs one cent (Shampanier et al., 2007; Baumbach, 2016). However, these results are found in a single product context. From a theoretical perspective, it can be argued that for a free product contained in a bundle, the downside risk remains present: a consumer still has to pay for the other components in the bundle. Nonetheless, Nicolau and Sellers (2012) still find an increase in affect for a low priced hedonic product bundle containing a free component. Baumbach (2016) finds the same result for two high priced utilitarian product bundles. Additional support for the affect mechanism is provided by neuroscientific research. When consumers choose a free over a positively priced product, the medial prefrontal cortex is activated. This brain area is positively correlated with consumers’ happiness level (Votinov et al., 2016). Moreover, consumers choosing the free products in a zero pricing condition show higher levels of activation in the vPCC (posterior congulate cortex) than participants choosing the higher value products in a non-zero pricing condition (Votinov et al., 2016). The vPCC brain region is associated with the evaluation of emotional content. This indicates that a free product actually contains emotional content, thereby

(21)

20

providing further evidence for the importance of affect in the zero price effect (Vogt, Vogt, and Laureys, 2006).

The increased level of affect associated with a free product can, however, be diminished. When prices are expressed as ratios, there is no increase in affect, nor is the zero price effect present (Baumbach, 2016). An example of a ratio price is 11:1. This means that the more expensive product is 11 times more expensive than the cheaper product. The increase in affect, and the zero price effect, also disappear when consumers are forced to cognitively elaborate on the different price options (Shampanier et al., 2007). The importance of elaboration, in combination with the affect heuristics, in the relationship between pricing strategies and changes in demand distributions, are discussed in more detail in the following section.

2.3 Elaboration likelihood model

The discussion in the previous section indicates that differences in the strength of the zero price effect may be related to the extent to which consumers consider affect heuristics in their purchase decisions. To determine when and why consumers base decisions on affective cues, the Elaboration Likelihood Model (ELM) is discussed. An overview of this model is provided in Figure 1 below. The parts of the model that this study focuses on have a light grey fill. The ELM shows two different processes underlying attitudinal changes, as a result of persuasive messages (Petty and Cacioppo, 1986). An example of a persuasive message is the sign ‘free’.

(22)

21

Figure 1: Elaboration Likelihood Model (adapted from Petty and Cacioppo, 1986)

The following sections discuss the model’s two routes that can lead to persuasion: the central route (dark coloured in Figure 1) and peripheral route (light coloured in Figure 1). This is followed by the role of the level of involvement within the ELM. This section concludes with an overview of how the ELM may play a role in determining the strength of the zero price effect.

(23)

22

2.3.1 Central route to persuasion

When a consumer goes through the central route to persuasion he or she elaborates and actively thinks about a persuasive message. Attitudinal changes as a result of a persuasive message are, amongst others, determined by argument quality (Petty and Cacioppo, 1983).

To follow the central route, a consumer must first be motivated to process a message (Petty and Cacioppo, 1986). Motivation to process is dependent on internal factors such as need for cognition. Consumers with a high need for cognition enjoy effortful tasks and thinking, and are more willing to cognitively process a persuasive message (Petty and Cacioppo, 1986). The most important external factor that influences motivation to process, is a consumer’s level of involvement. This is the extent to which a good is personally relevant to a consumer (Petty and Cacioppo, 1986). This will be discussed in more detail in section 2.3.3. In addition to being motivated, consumers must be able to process a message (Petty and Cacioppo, 1986). Ability, as with motivation, can be enhanced and diminished by both internal factors (familiarity) and external factors (distraction). When a person is motivated and able to process a message, the persuasive communication should cause a positive or negative thought to predominate, for the consumer to move to the next stage of the central route. In this follow-up stage, the cognitive structure of the consumer should change, based on the predominating thought. This means that the newly acquired ideas are accepted and stored in memory (Petty and Cacioppo, 1986).

2.3.2 Peripheral route to persuasion

The alternative to the central route to persuasion, is the peripheral route. When consumers go through the peripheral route, they do not carefully process a persuasive message. However, this does not mean that the message cannot lead to an attitude change. Even in the absence of

(24)

23

processing, consumers’ attitudes can be affected by heuristics such as affect (Petty and Cacioppo, 1986; Schumann, 1983).

Consumers move into the peripheral route if they are not motivated to process a message, not able to process the message, when there are no or neutral predominating thoughts after processing, or when the predominating thoughts do not lead to a cognitive structure change (Petty and Cacioppo, 1986).

2.3.3 Level of involvement

The element of the ELM that this research focuses on is a consumer’s level of involvement. The level of involvement is described as the extent to which a product is of intrinsic importance to a consumer (Sherif and Hovland, 1961). In other words, whether the product is of personal relevance or not (Zaichkowsky, 1985). As described previously, this affects the extent to which a consumer is motivated to process a message, and subsequently whether a message is processed centrally or peripherally. Specifically, lower levels of involvement lead to a lower motivation to process a message, and higher levels of involvement lead to a greater motivation to cognitively process a message (Petty and Cacioppo, 1986; Chaikan, 1980). The underlying motivation for systematic processing is that the personal implications of making a wrong decision or drawing an incorrect conclusion, are greater when consumers are highly involved (Petty and Cacioppo, 1986). Similarly, Desphande and Hoyer (1983) argue that a consumer’s level of involvement is dependent on the degree of risk involved for the consumer. They find that high involvement products carry more risks than low involvement products, meaning that personal consequences for a high involved consumer are greater, causing them to elaborate on a message (Desphane and Hoyer, 1983). As the routes in the model show, high involved consumers are pushed further into the central route, whereas low involved consumers are pushed into the peripheral route. Consumers in the peripheral route

(25)

24

will base their attitudes on heuristic cues, such as affect (Chaikan, 1980). Since a free product is associated with affect heuristics, the zero price effect may be stronger for low involved consumers. Additionally, the zero price effect has been shown to disappear when consumers cognitively elaborate on a persuasive message (Shampanier et al., 2007), which occurs in the high involvement condition. Therefore, the zero price effect may be weaker when consumers are high involved. It is thus hypothesized that:

H3: The higher a consumer’s level of involvement towards a persuasive message of a low priced, utilitarian product bundle, the less strong the effect of a zero pricing strategy on the distribution of demand.

H4: The higher a consumer’s level of involvement towards a persuasive message of a high priced, utilitarian product bundle, the less strong the effect of a zero pricing strategy on the distribution of demand.

(26)

25

2.4 Control variables: age, gender, income, education and brand attitudes

Finally, control variables that may impact product choice and demand distributions are briefly discussed. This section is included to provide an indication of how the variables impact the zero price effect, and not to provide a comprehensive overview of all the literature on the variables, because they are not the key focus of this study. Moreover, the variables are not included as hypotheses in the conceptual model, but are considered as covariates in the empirical model. The following sections provide a justification for including age, gender, income, education, and brand attitude as controls.

2.4.1 Age

Age can influence product choice and demand distributions in two ways. First, older consumers use more heuristics in decision making than younger consumers (Kim, Goldstein, Hasher, and Zacks, 2005). Since the zero price effect is explained by affect heuristics, changes in demand distributions may be stronger for older consumers. Whether this effect is truly present cannot be confirmed with certainty, since Sladek, Bond, and Philips (2010) argue that older consumers are more likely to process information rationally, rather than intuitively. This indicates that the zero price effect may be less strong for this age group. Second, Lambert-Pandraud, Laurent, and Lapersonne (2005) argue that older consumers prefer more established brands. This research focuses on a hypothetical choice between an established and less familiar brand, as will be explained in more detail in the methodology. Older consumers are thus more likely to choose the established (non-free) brand in the normal and discounted pricing conditions, while the potential of the increased use of heuristics causes them to choose the less familiar brand in the zero pricing condition. This indicates that there might be an amplified zero price effect for older consumers, unless the brand preferences have a stronger effect on product choice in the zero price condition, than the affect heuristics. If the

(27)

26

latter holds, the effect of heuristics and brand preferences may rule each other out, but this cannot be ensured. Therefore, it is important to control for age.

2.4.2 Gender

Scholars suggest that females are more risk averse than males (Byrnes, Miller and Schafer, 1999; Arch, 1993). Therefore, females may be more strongly affected by zero priced products that have no downside risks (Shampanier et al., 2007). Furthermore, Sladek, Bond, and Philips (2010) argue that males prefer rational reasoning, and females experiential reasoning (intuition), when making decisions (Sladek, Bond, and Philips, 2010). This suggests that women are more likely to use affect heuristics in purchase decisions, which may lead to a stronger zero price effect. Furthermore, males and females show different preferences towards advertisement aspects such as captions, and background colours (Grobelny and Michalski, 2015). Also, the impact of product colour attractiveness and colour attitude on product choice is more important for women. (Funk and Ndubisi, 2006). This indicates that details of the advertisement and chosen products (such as colours) may impact product decisions in all the pricing conditions differently across males and females. It is therefore important to include gender as control variable.

2.4.3 Income

Income is one of the sources of consumers’ budget constraints. Whether the perceived value of a product exceeds its cost, is dependent on such constraints (Pindyck and Rubinfeld, 2013). Since product choice is dependent on product value, income will thus influence consumer’s general product choices in all pricing conditions. Income may also affect product choices across different pricing conditions. For example, Khare, Achtani, and Khattar (2014) find that consumers with a higher income attach less value to discounts, indicating that they are less

(28)

27

likely to be influenced by a zero pricing strategy. However, Mandrik (1996) finds that income affects value consciousness, but that this does not increase the extent to which consumers rely on simple price cues. Results are contradictory, but it cannot be excluded that income does not affect the zero price effect. Therefore, it is included as control variable.

2.4.4 Education

Creusen (2010) finds that consumers with different educational levels, value different product aspects. For example, higher educated consumers value quality more than lower educated consumers. More expensive products are usually perceived as of being higher quality (Kardes, et al., 2004), thus there is a chance that higher educated consumers are more likely to choose Oral-B and Dell over Dentaclean and iiyama in all pricing conditions. Additionally, more educated consumers seem to have a greater tendency towards rational than impulsedecisions (Nwanko, Hamelin and Khaled, 2014; Goll and Rasheed, 2005). This indicates that higher educated consumers are potentially less affected heuristics such as ‘free’. Income is therefore included as control variable.

3.4.5 Brand attitude

Finally, brand attitude directly influences purchase intentions (Spears and Singh, 2004). This could affect the extent to which consumers prefer a specific product, but also their likelihood of switching towards a different product in a zero pricing condition. Moreover, Walla, Brenner, and Koller (2011), find that brands for which consumers have higher attitudes, elicit more positive emotions. This can influence the interpretations of the results on the moderating variable involvement. Specifically, it cannot be determined whether affect heuristics that might be used for purchase decisions in the low involvement conditions result from the

(29)

28

emotions associated with free or with attitude towards the products. Thus, brand attitude should be controlled for.

3. Methodology

To facilitate the replication of this study, a comprehensive overview of the methodology is provided. The first section describes the research design and -strategy that are used to provide an answer to the research question. This is followed by an overview of the data collection method and the sample. Finally, the measurements, and methods of analyses are discussed.

3.1 Research design and strategy

The first step in defining the research design is determining the research philosophy. The philosophy serves as a basis for the remainder of the methodology, and influences how results are interpreted (Saunders et al., 2012). The adopted philosophy is pragmatism. Pragmatism assumes that there are multiple ways of interpreting information and doing research, and that the most important determinant of whether a more positivist or interpretivist stance is taken, is the research question (Saunders et al., 2012). The research question investigates causal relationships. A philosophy that focuses on collecting data in an objective and value-free manner is most appropriate, since this enables the comparison of results across participants (Saunders et al., 2012). Positivism is thus most fitting.

Furthermore, a deductive approach is used to generate knowledge. Existing theories on the zero price effect and consumers’ levels of involvement are integrated, which forms the basis for the hypotheses that are tested in this study. These hypotheses are evaluated based on primary data (Saunders et al., 2012).

Additionally, a mono-method quantitative research design is adopted. The quantitative design is partially based on the decision to lean towards a positivist philosophy. The latter is

(30)

29

associated with structured data collection, which fits best with quantitative data collection techniques. The choice for a quantitative method further results from the choice for a deductive approach. Deductive approaches usually require data that is standardized, in order to examine relationships between variables. Quantitative research fulfils these requirements the best.

The primary data is collected by means of an experimental design. The experiment has a 3 [zero pricing strategy/discounted pricing strategy/normal pricing strategy] x 2 [high involvement/low involvement] factorial, between-subjects design. A potential disadvantage of the between-subjects design is that differences between groups may not be caused by manipulations, but by inter-group differences. To minimize this risk, demographic control variables are included in the analysis, and participants are randomly assigned to one of the six conditions.

Finally, the study will be cross-sectional. A potential disadvantage of the latter is that external factors, which vary across different points in time, may play a role in consumers’ decision making, such as emotions. However, the choice for a cross-sectional design is in accordance with other studies on the zero price effect (Shampanier et al., 2007 and Baumbach, 2016). Moreover, due to time constraints a longitudinal study is not possible.

3.2 Data collection

The population addressed in this research includes consumers based in the Netherlands that are able to make their own purchase decisions. It is assumed that a consumer is able to make such decisions when they are 16 years old. Even though they are not considered as adults, 16 year olds are allowed by the Dutch law to work a full workweek of 40 hours (Rijksoverheid, 2017). Looking at data from the CBS (2016), 1.38 million consumers between 15 and 25 years old are part of the labour force. 1.25 million of them are actually working, indicating

(31)

30

that many of the 16 year olds are likely to have a job (CBS, 2017). In other words, they have the ability to earn their own money, and use this to make their own purchase decisions. Therefore, they should be included in this research.

A sampling frame is not available for the population. Hence, a combination of non-probability sampling techniques are applied. Convenience sampling is used to generate a sizable sample within a limited time frame. The disadvantage of this method is that it can lead to a non-generalizable sample (Saunders, et al., 2012). Convenience sampling is therefore combined with purposive sampling. Consumer groups that are underrepresented after the convenience sample is collected, will be actively approached to create a sample that is representative of the population.

The data is collected through self-administered vignette surveys. Participants are randomly assigned to one of six hypothetical purchase situations, after which their hypothetical product choices are assessed. This method has an improved level of realism in comparison to pure experiments, but still allows for the manipulation of variables. It is therefore argued that vignettes have a better external and internal validity in comparison to both experimental and non-experimental data collection methods (Aguinis and Bradley, 2014). To further improve the realism of the vignettes, visual advertisements of the different products are added, rather than have the persuasive message (the free price) simply written in text form (Aguinis and Bradley, 2014). The questionnaires are internet-mediated, which increases time-efficiency and reduces errors caused by manually entering pen-and-paper responses into an online database (Saunders et al., 2012). Moreover, it provides access to geographically dispersed areas across the Netherlands (Saunders et al., 2012). This is necessary, since the population includes consumers located in the entire country. Internet-mediated questionnaires have the limitation that some consumers are not familiar with computers or the internet, potentially causing them to exclude themselves. This is limited to

(32)

31

8.2 percent of the Dutch population, who indicated that they have never used internet before (CBS, 2016). Based on this, it is concluded that the advantages of internet-mediated surveys weigh up against its disadvantages.

In total, 249 responses are collected. 42 responses miss data on the dependent variables and are removed, leaving a dataset of 207 responses. Response rates are not available, because the survey is distributed on social media, and participants are asked to spread the survey as well. It can therefore not be determined how many people saw the survey. Additionally, due to the exclusion of responses, and some errors within the randomization of the survey, participants are not equally spread across the six conditions. The conditions contain a minimum of 26 and a maximum of 40 respondents. All conditions fulfil the preferred minimum of 30 responses, except for one.

The participants’ ages range from 16 to 73, with an average of 26 (𝜎=10.980). Moreover, there are slightly more females (57%) than males (43%) in the sample. Most participants have completed a University Bachelor’s degree (30%), followed by a high school degree(29%), Community College (14%), University of Applied Sciences (13%), University Master’s degree (12%), other educations (2%) and primary school (1%). Finally, most participants have a gross income of below 1000 euros per month (56%). This is followed by participants with an income between 1000 and 2000 euros per month (17%), and those with an income between 2000 and 3000 euros per month (13%). The remainder does not want to say how much they earn (7%), earn between 4000 and 5000 euros (0.5%) or more than 5000 euros (7%).1

3.3 Pre-tests

(33)

32

To identify products that are appropriate for investigating the research question, qualitative unstructured interviews are conducted (n=3). The goal of these interviews is to identify low and high priced product categories that are utilitarian in nature, and allow for product bundling. Based on the interviews two categories are selected: electric toothbrushes (bundled with replacement brush heads), and desktops (bundled with monitors). A pre-test (n=30) is performed to validate these choices. An overview of the respondent demographics, and Pearson correlations between the included variables, are provided in Table 1 and Table 2 in Appendix A. The correlations between the categorical variables only serves as indicator that some relationship might be present, but does not give any indication of what these relationships entail.

3.1.1 Levels of utilitarianism and hedonism

3.1.1.1. Pre-test measurements

To identify whether the categories are perceived as utilitarian, participants rate them on five items, using seven-point semantic scales. An example of an item is: “I find [product category] functional versus not functional” (adopted from Voss, Spangenberg, and Grohmann, 2003). This is translated into Dutch, using parallel translations, as “Ik vind [product categorie] functioneel versus niet functioneel”. An overview of how all scales are translated is provided in Appendix B. The translated scales are found to be reliable for the low priced category (𝛼=0.886) and for the high priced category (𝛼=0.897). Corrected item-total correlations show that all items have a good correlation with the scale (r>0.30), and none of the items substantially affect reliability if they are deleted.

To measure whether categories are perceived as hedonic, participants rate the product categories on five additional items. An example of an item is: “I find [product category] dull versus exciting”, translated into Dutch as “Ik vind [product category] saai versus spannend”

(34)

33

(adopted from Voss, Spangenberg, and Grohmann, 2003). The translated scales are found to be reliable for the low priced category (𝛼=0.727), but not sufficiently reliable for the high priced category (𝛼=0.651). Additionally, corrected item-total correlations show that one item has an insufficient correlation with the scale for the low priced category (r=0.229), and the high priced category (r=-0.147). The item-total correlation for the latter is negative, which indicates a problem. The item is counterindicative, but has been recoded appropriately. Therefore, this should not have caused the issue. It is more likely that the item has been ambiguous or confusing to the participants. Even though the item has proven to be reliable in other studies, it is removed in this study, because there might have been an issue with the translation. The new scales are improved in terms of reliability for both the low priced (𝛼=0.790) and high priced (𝛼=0.840) product category. All items correlate sufficiently with the scale (r>0.30), and do not significantly affect reliability if they are deleted.

3.1.1.2 Pre-test results

The results show that electric toothbrushes (𝑥̅=5.04, 𝜎=1.542) and desktops (𝑥̅=5.64, 𝜎=1.220) score above the threshold of 4 for utilitarianism, thereby indicating that both product categories are utilitarian in nature. A potential issue is that utilitarianism for the high priced product correlates significantly with participants’ income (r=-0.377, p>0.05). This can be problematic, because the income category 0-1000 is overrepresented (40%). Looking at the mean scores on utilitarianism for each group shows that all income groups, except the ‘I don’t want to say’ group, give a score of above 4.9 on utilitarianism of the desktop. Also, the ‘I don’t want to say’ group contains only 2 participants, one of which has an extreme opinion in comparison to the rest of the participants, which causes a low group average (x=2.2). These results should be interpreted with care, because the sample sizes within each category are very

(35)

34

low. However, it likely indicates that, irrespective of income, the high priced product is seen as utilitarian in nature.

Additionally, the results of the pre-test show that electric toothbrushes score below the threshold of 4 on hedonism (𝑥̅=3.24, 𝜎=1.304), and that desktops slightly exceed the threshold (𝑥̅=4.02, 𝜎=1..328). The latter is still accepted for two reasons. First, the threshold for hedonic goods is barely exceeded with 0.017 points. Second, the functional aspects of a desktop carry more weight than the hedonic aspects (1.63 point difference). This indicates that the product is perceived to be more utilitarian than hedonic.

3.1.2 Perceptions of low and high prices

The pre-test is also used to verify whether the chosen products within the low and high priced categories are perceived as low and high priced. Note that these tests are oriented towards products, rather than product categories, since people judge prices relative to other products within a category. The chosen product for the low priced electric toothbrush category is the Oral-B Advanced Power toothbrush, priced at €14.99. This will be bundled with different types of brush replacement heads in the actual survey. The product for the high priced desktop category is the Dell Precision Rack 7910, priced at €3.200,-. This will be bundled with different types of monitors in the actual survey. Participants are asked to rate the product on three items, adapted from Peace, Galletta, and Thong (2003), using seven-point semantic scales for two items, and a seven-point Likert scale for one item. An example item is “I feel that the price for [product X] is very high versus very low”, translated into Dutch as “Ik vind de prijs van [product X] erg hoog versus erg laag”. The translated scales are found to be reliable for the low priced category (𝛼=0.852) and for the high priced category (𝛼=0.827). Corrected item-total correlations show that all items have a good correlation with the scale (r>0.30), and none of the items substantially affect reliability if they are deleted.

(36)

35

The results indicate that the toothbrush is indeed perceived as low priced (𝑥̅=2.92, 𝜎=1.110) and the desktop as high priced (𝑥̅=6.17, 𝜎=1.078). Additional analyses show that age and price perceptions of the high priced product are negatively correlated (r=-4.01, p<0.05), indicating that older consumers perceive the price as less expensive than younger consumers. This can be problematic, since there is an overrepresentation of the age group 21 to 35 in the sample (46.7%). However, the age group with the lowest mean score on price (66+) still gives a rating of 4.67, which is above the threshold.

3.4 Measurements

The following sections will describe how the different variables in the model are measured or manipulated.

3.4.1. Independent variable: pricing strategy

Participants are assigned to one of three pricing conditions: normal pricing (control), discounted pricing, and zero pricing. In each condition they are exposed to two hypothetical purchase situations. In the low priced product category, participants are asked to make a choice between two brands of replacement brush heads for an electric toothbrush they are buying. They can choose between a pack of 12 brush heads from a known Dutch brand, Oral-B, a less familiar brand, Dentaclean, or nothing. The choice between a familiar and less familiar brand was given to ensure that a large enough percentage of the participants chooses the familiar brand in the normal pricing condition. If the majority of the consumers prefer the cheaper brand in the normal pricing condition, than the zero price effect cannot be measured (the consumers prefer the cheaper brand in all conditions). In the normal pricing conditions the products are advertised at their regular prices: €29.99 for Oral-B and €11.99 for Dentaclean. In the discounted pricing conditions Oral-B is priced at €23.99 and Dentaclean at

(37)

36

€5.99. Finally, in the zero pricing conditions Oral-B is priced at €17.99 and Dentaclean brush heads are free. Note that in the discounted and zero pricing conditions participants are told that they are provided with a discount on either one of the two brands, because they are buying the electric toothbrush. Additionally, €29.99 euros for replacement brush heads may not seem to be a low priced product. However, It should be considered that this is less than €3,- per brush, which is not so expensive. To create a realistic purchase situation the replacement brush heads are offered in 12-packs and not per piece. An example of the advertisement in the zero pricing condition is provided in Figure 3 below.

For the high priced products consumers are asked to make a choice between two monitors when buying a desktop. Again, they can choose between a familiar brand, a Dell, and a less familiar brand, iiyama. In the normal pricing conditions the Dell is priced €259,-, and the iiyama at €189,-. In the discounted pricing conditions the Dell is priced at €165,- and the iiyama at €95,-. Finally, in the zero pricing conditions the Dell is priced at €70,- and the iiyama is for free when purchasing the specific desktop. An example of the advertisement in

(38)

37

the zero pricing condition is provided in Figure 4 below. See Appendix C, section 2, for an overview of all the advertisements in the different conditions.

Finally, note that the decreases in price across the different conditions equal half of the original price of the cheapest product. This ensures that the change in prices across the conditions are equal. This allows for better comparisons of the change in demand distributions between the normal versus discounted pricing strategies, and the discounted versus zero pricing strategies.

3.4.2 Dependent variable: change in demand distribution

Changes in the distribution of demand are measured based on methods by Shampanier et al. (2007), and Baumbach (2016). Preferences for the different brands in each of the three pricing conditions are compared. For the distribution to change, and thus demand to be irrational, the demand for the free product in the zero pricing condition should significantly increase, in comparison to the demand for this same product in the normal- and discounted pricing, conditions. Additionally, demand for the non-free product in the zero pricing condition should decrease significantly in comparison to the demand for this same product in the normal

(39)

38

pricing, and discounted pricing conditions. Demand is measured by simply asking consumers: “Based on the above information [i.e. advertisements and description of the purchase situation], which [monitor/replacement brush heads] would you purchase?”

3.4.3 Moderator: level of involvement

To manipulate the level of involvement towards a persuasive message, the level of personal relevance of the product being communicated should be manipulated (Petty, Cacioppo, and Schumann, 1983). Since most studies that investigate the level of involvement are executed through experiments rather than surveys, the number of methods available for this study is limited. Therefore, two approaches from different studies are combined. First, the communication of the gift card that consumers can win is different for the two conditions. For the high involvement condition, consumers are told that they are able to win a gift card that can be spend in one of the product categories discussed in the survey. In the low involvement condition consumers are simply told they are able to win a gift card (adapted from Petty, Cacioppo, and Schumann, 1983). Second, the instructions for answering the questions are varied across the conditions. The high involvement condition asks consumers to evaluate the persuasive message carefully and indicate their preferences as if they are truly going to make a purchase in the category. The low involvement condition only includes a general instruction (adapted from Wright, 1975). To ensure that participants carefully read the instructions, they are marked red and made bold. A full description of these instructions is available in Appendix C, section 1.

To check whether the manipulation is successful, participants rate the extent to which they are involved with the persuasive message on six items, adapted from Andrews and Durvasula (1991). This is done through seven-point Likert scales. An example of an item is: “I have been paying attention to the advertising message”, translated into Dutch as “Ik heb

(40)

39

aandacht besteed aan de advertentie”. The translated scales are found to be reliable for the low priced category (𝛼=0.897) and for the high priced category (𝛼=0.918). Corrected item-total correlations show that all items have a good correlation with the scale (r>0.30), and none of the items substantially affect reliability if they are deleted.

3.4.4 Control variables

This research will control for age, gender, income, education, and brand attitudes. Age is measured through an open question. Gender, income and education are measured through closed questions. Income is a sensitive subject, so an ‘I don’t want to say’ option is added. Additionally, education contains an ‘other’ option that allows consumers to fill out their level of education if it is not present in the standard options. Finally, brand attitude is measured on 7 items (adapted from Spears and Singh, 2004) on 7-point semantic scales. An example of an item is “To me, [brand name] is unpleasant versus pleasant”, translated into Dutch as “Ik vind [brand name] onaangenaam versus aangenaam”. The translated scales are found to be reliable for Dentaclean (𝛼=0.914), Oral-B (𝛼=0.892), iiyama (𝛼=0.938) and Dell (𝛼=0.934) Corrected item-total correlations show that all items have a good correlation with the scale (r>0.30), and none of the items substantially affect reliability if they are deleted.

3.5 Analyses

The following sections describe how the different hypotheses are tested.

3.5.1. Level of involvement manipulation check

An independent t-test is performed to identify whether the involvement manipulation was successful. The independent variable is a categorical dummy variable that indicates whether consumers are manipulated to be low or highly involved. The dependent variable is the extent

Investigating the effect of a consumer's level of involvement on the strength of the zero price effect