The Impact of Inequality on Consumer Numerosity Bias

The Impact of Inequality on Consumer

Numerosity Bias

Edgina Christabel

Master Thesis MSc. Marketing Management

The Impact of Inequality on Consumer Numerosity Bias

By: Edgina Christabel

University of Groningen

Faculty of Economics and Business

Department of Marketing

Supervisors

First Supervisor: AlBalooshi, S. Second Supervisor: Fennis, B. M.

Edgina Christabel H.W Mesdagstraat 61

9718 HD Groningen Ph: +31 616 5711 03

Abstract

The thesis reviews the impact of the different level of inequality to consumer’s numerosity bias. The theoretical framework discusses this relationship to be mediated by consumer’s sense of control. It is elaborated that by experiencing high level of inequality, consumer will perceive as if their self-control is threatened. This then drives them to make decision, which could restore their sense of control. The result shows that the main relationship and the mediating relationship are both not significant. This was concluded due to the low correlation between all variables, except for the hedonic and utilitarian product categories. This paper further discusses the possible explanations of the results and the limitations of the current study, which lead to the direction and recommendations for further research. Hence, this research contributes by giving a more specific discussion of the impact of inequality to consumer behavior.

Preface

In order to round up my Master program in Marketing Management, with this I present to you my Master Thesis. This thesis may only take five months to be completed but has been the result of my five years bachelors and one year masters education. On this part of the paper, I would like to use the opportunity to thank my thesis supervisor, Sumaya AlBalooshi, Ph.D. for her help from the start, during and until the end of the process. Especially for her constructive feedback on my work and her flexibility in her timeline, which helps to reduce the amount of stress during the process of writing this thesis. Also, I would like to thank my second supervisor, prof. dr. Bob Fennis, who has given me the final evaluation on my thesis.

Next, I would like to thank my family, my friends, and my boyfriend who have given me greatest moral supports during the course of writing this paper and who also have always been proud of me. Even though my family lives on the other side of the world, they have always been there for me. The past year and especially this past five months have been quite challenging especially once I have started my internship. Nonetheless, I am satisfied with the result, and all the challenges are worthwhile.

Table of Contents

Introduction 7

Theoretical Background 8

Inequality 8

Numerosity Bias 8

Inequality and Numerosity Bias 9

Mediating Role of Sense of Control 10

Methodology 13

Experimental Design 13

Data Collection 13

Variable Manipulation: Inequality 14

Measuring the Sense of Control 14

Measuring the Numerosity Bias: Hedonic and Utilitarian Products 15

Results 16

Variable Manipulation Check: Inequality 16

Reliability Analysis: Dependent variable 16

Discussion 19

Inequality Manipulation Limitation 19

Lack of Knowledge of the Inequality Concept 19

Numerosity Bias Scale Limitation: Likert Scale 20

Numerosity Bias Scale Limitation: The Product Display 21

Hedonic and Utilitarian Perception 22

Initial Packaging Preference 22

Introduction

"Better cut the pizza into six pieces, I do not feel hungry enough for eight pieces". Various well-known persons have claimed this infamous pizza-ordering anecdote. Despite the initiator of the joke, the evidence of this kind of misconception has been found in our daily lives, and the process of consumers’ purchase decision is no exception. As demonstrated by this classic anecdote, people may presume that the same whole quantity of pizza as more when it is cut into eight pieces — than when it is cut in six pieces. Even further, this mind fallacy could be experienced by any individual in various situations, even to those who are aware of its possibility to occur.

In modern life, people objectively quantify their environment more and use that as a consulting point in decision-making (Bagchi and Davis, 2016). People believe that numbers will provide a stronger argument for their decisions. However, most of the time this argument is not supported by an adequate numeracy skill and motivation to elaborate the reasoning, which contributes to the tendency of using the peripheral route as discussed in the Elaboration Likelihood Model, the Dual Processing Theory of Petty and Cacioppo (1986). Supplementary to this, the situation in which the consumers are in plays an essential role in this phenomenon.

One of the situations could be when consumers are exposed to economic inequality (hereafter referred to as inequality) in their environment. Inequality has caused people to be in the state of uncertainty of their economic situation due to their income risk (Blundell and Preston, 1998). Recently, the issue of inequality has been increasing significantly in most parts of the world. Many researchers have investigated the impact of inequality on different aspects of lives, especially to consumer behavior. However, the existence of literature regarding the effects of inequality on the tendency of numerosity bias is scarce. Hence, this thesis aims to contribute to that literature gap and trigger more research in this area.

Theoretical Background

Inequality

Deininger and Squire (1996) from the World Bank have found that the different definitions of inequality have caused the incomparable results between countries. World Bank has also mentioned that inequality has a broader concept than poverty as it revolves around the entire population (Haughton and Khandker, 2009). In his research, Ravallion (1997) found that inequality could be the result of disparities in income distribution, in spite of the existence of growth. Moreover, Kuznets (1955) had theorized that inequality comes in a U-shaped pattern, which shows that in spite of economic development, inequality will still exist due to unequal income distribution. In other words, even though countries are continuously developing and improving their economic status, inequality still exists among the sizeable society. This finding is intriguing and has led this paper to look further to the change of the consumer behavior within the inequality environment. Inequality can be defined using multiple perspectives, but for the current paper, the definition of inequality we use is the "extent to which wealth is concentrated in the hands of a small proportion of the population" (Côté, House and Willer, 2015, p. 1).

Numerosity Bias

More than half of purchase decisions are made at the point of sale (Ampuero and Vila, 2006). This finding shows that all marketing strategies that companies try to push to the market —preceded with the decision of purchase— are very easily overestimated. According to Shrivastava, Jain, Nayakankuppam, Gaeth, and Levin (2017), numerosity bias could be one of the factors that influence consumers in their decision-making. This condition shows that the numerical value, which exhibits the amount of resource, might systematically bias consumers' decision making by perceiving it as "less than adequate" or "more than adequate" (Shrivastava et al., 2017). As defined by Pelham, Sumarta and Myaskovsky (1994; 2), “numerosity itself is the number of units into which a stimulus is divided”. Hence, for instance, some customers will perceive €4 as less than adequate in comparison to 400 cents, which can result in making customers over or under infer the quantity. 500grams of sugar in one cup might seem less than when it is put in 5 different cups.

their thinking process to analyze that more pieces of something most of the time turns out to be more of that something (Pelham et al., 1994). Further, when the benefit is separated into two or more different units will become more satisfying than when it is not. This is due to the sum total judgmental impact of each unit that is perceived as higher than the judgmental impact as a whole (Thaler, 1985).

The Dual Process approaches on consumer information processing such as, the Elaboration Likelihood Model (Petty and Cacioppo, 1986) and the Heuristic-Systematic Model (Chaiken, 1987), have been extensively researched and have underlined the point that consumers tend to process the information that they are exposed on two different continuum cognitive processing poles. One of them is systematic which use more cognitive elaboration, and the other one being automatic processing that uses less cognitive elaboration and tends to rely on the heuristics. Pelham et al. (1994) had found that numerosity bias is categorized as automatic processing that is difficult to be resisted by consumers. It includes the perceptual process prior to the cognitive process that is impossible to ignore. Hence, it becomes a heuristic that consumers rely upon as an information source to go through their peripheral route automatically (i.e., Numerosity heuristic).

Furthermore, it is understood that consumers’ cognitive resources capacity is bounded (Szmigin and Piacentini, 2014; Fennis and Stroebe, 2015). There is a particular limit of information processing that consumers can have in their cognitive processing. That is when their cognitive resources have reached the capacity limit due to making difficult, rapid and multitasked judgments (Pelham et al., 1994), consumers tend to rely on the heuristics processing; one of them could be numerosity heuristic.

Inequality and Numerosity Bias

According to Tversky and Kahneman (1974), when creating decisions in uncertain situations, people are found to be relying on themselves to heuristics due to cognitive biases. One of these judgmental heuristics employed is the adjustment from an anchor, which mostly used in related numerical predictions. This heuristic occurs when different starting points result in different estimation that is usually biased against the initial value (Bagchi and Davis, 2011). Correspondingly, this heuristic could lead consumers to make an incorrect prediction and could impact their perception towards the benefits (i.e., numerosity bias).

Moreover, numerosity can be used as one of the price communication strategies. Especially at the point of purchase, packaging plays one of the most critical aspects of marketing tools (Underwood and Klein, 2002). A significant number of retailers have been using the packaging that is bigger and consists of multiple numbers of items inside. Wansink (1996) had found that not only it influences consumers’ purchase decision; packaging has also been long known to influence consumer behavior. Furthermore, businesses that are selling subscriptions (e.g., Fitness Center, Newspaper) have been applying their price communicating strategies by relying on consumers’ numerosity bias. These businesses make the prices difficult to understand, which then guide the consumers to use heuristics to estimate their decision-making process (Bagchi and Davis, 2011).

Mediating Role of Sense of Control

Research has found that having the sense of control is one of the essential elements in life as people are continually seeking the ability to manage what happens in their lives (Higgins, 2011) and to produce the desired outcome (Leotti, Iyengar and Ochsner, 2010). The desires to demonstrate competence and superiority over events have been established as significant motivations in one’s life (Adler, 1930) and are parts of the fundamental human needs (Kelley, 1987). As pointed out by Leotti et al. (2010), humans have the ingrained motive to exert control over one’s surrounding environment personally to produce the expected results.

1975), which enforces them always to seek control over the outcome of their lives. Furthermore, it was found that individuals and/or households, who are categorized as the lower income group of population, perceive less mastery over the events in their lives (Lachman and Weaver, 1998b). This motive to have the control to get the best of every decision can be attributed not only to social interactions but also the consumption behavior.

Another way to look at it would be that inequality has found to be affecting consumer behavior in an unusual way. For instance, contrary to popular belief, people who earn a lot less do not necessarily consume less (Christen, Morgan and Ruskin, 2005). People are trying to get the best for their lives not only to fulfill their needs and wants but also to show their social status to their environment.

The Social Comparison theory or the need to compare self with others could be one of the explanations for this phenomenon (Buunk and Gibbons, 2007). One of the most important drivers of this theory is the lack of perceived control (i.e., when there is something that can be done to improve and or avoid failure). One of the significant theoretical developments on the topic of Social Comparison theory is fear affiliation theory, which shows that people have the need to affiliate themselves with their social environment when they are faced to an uncertain situation (Gerard and Rabbie, 1961). Hence, people might fear that they are not getting the most benefit as they can and also fear of not being a part of the society, due to inequality, which can be one of the drivers that enforce them to strive for control over their lives.

Consumers with a higher need to have sense of control will have a higher preference for products that potentially enhance their sense of control (Faraji-Rad et al., 2017). The need to restore their control drives consumers to constantly using their cognitive capacity to look for ways that can compensate the sense of no control. As discussed earlier, this phenomenon leads to people relying on their heuristics processing when rounding up the decision in their lives. This shows that there is a psychological effect of not having control -in this case due to experiencing inequality- results in cognitive depletion that leads them to rely on peripheral routes more than elaborate thinking. As shown on the conceptual model in figure 1, having low sense of control will therefore mediate the impact of inequality on consumer’s numerosity bias. Thus, the hypothesis that will be tested in this research is:

(Figure 1: Conceptual Model)

Inequality Numerosity Bias

Methodology

Experimental Design

The experiment is a 2x1 factorial design, which combines both between-subject groups and within subject group designs. The subjects were randomly divided into two groups that are exposed two different hypothetical conditions; one group is exposed to a high inequality condition and the other group is exposed to low inequality condition. The study also have conducted a within-subject test, which means that the respondents in both groups were exposed and had to choose between high or low numerosity, to be able to see the whether there is a tendency for numerosity bias used or not. In addition, the study has tested the hedonic and utilitarian nature of the product to both groups (within-subject test), to investigate whether the numerosity bias occurs in both nature, one of them, or none of them.

Data Collection

A survey was developed on Qualtrics, which was divided into 2 parts; Inequality manipulation (IV) and test for sense of control (mediator), and the numerosity bias test (DV). At the end of the survey, some control variables are included such as age, gender, mood, ethnicity, political orientation and income level. Also, due to the load of the question, an

attention and comprehension check was presented to see how much distractions and attention

were involved during the process of filling in the survey. This then shows whether the respondents pay attention to the study instructions or not (Oppenheimer, Meyvis, and Davidenko, 2009) and to make sure that the effect of the manipulation took place. In addition to that, the mood level of the respondents is also measured to use it as one of the control variables. At the end of the study, the respondents were asked whether they know the aim of the study or if they have done a similar study before. The respondents who confirm these facts have to be excluded from the study to reduce the possibility of bias responses.

Variable Manipulation: Inequality

In manipulating the inequality as the independent variable, the Nationally Representative Survey Study by Côté et. al. (2015) was used. To make it more convincing to the respondents, some questions about their demographic (i.e., gender, ethnicity, age, income level and the state they live in) are presented. The participants were asked to wait for a couple of seconds for their data to be processed. This is however was not real and a part of the manipulation process, which aims to increase the perceived credibility of the later data. Based on the information given, the respondents will then be randomly shown one of the graphs from Nationally Representative Survey Study (the high inequality or low inequality graph) and will be presented with some manipulation check questions, as can be seen on Appendix 1 In this case, the graph was shown as if the data is taken from the area where each of the respondent live, which has been asked earlier.

As mentioned before, one group of the subjects will be presented with a high inequality graph (i.e., the economic wealth difference between population segments is high), and the other group will see the low inequality graph (i.e., the economic wealth difference between population segments is low). Afterwards, the respondents were asked to answer a couple of comprehension questions to make sure they grasp the insights of the graph and increase their exposure frequency to (in)equality, which would potentially increase their feelings of equality or inequality. The respondents who do not demonstrate their comprehension of the survey are excluded from the data analysis. Therefore, three questions were presented (i.e., perceived economic suffer, dissatisfaction and inequality) to enable manipulation check to be done in the following analysis.

Measuring the Sense of Control

To measure the mediating role of sense of control, the respondents were asked to fill in a set of personality question. The scale used is the MIDI Sense of Control scale, which is a self-reporting scale developed by Lachman and Weaver (1998b). This scale consists of in total 12 questions with equal weight with two different categories, which are perceived

constraints and personal mastery.

questions. The scale use 7-point Likert scale asking how each statement describes the respondents with 1= Strongly disagree and 7=Strongly agree. The lower the score in these questions, the lower their sense of control will be. Hence, they will seek for control (i.e., due to low sense of control), as discussed in the theoretical part.

Measuring the Numerosity Bias: Hedonic and Utilitarian Products

The third part of the study aims to measure the dependent variable. The researcher had developed a scale to see respondents’ numerosity bias by providing a number of products with different numerosity level has been presented. This was done by presenting two kinds of the same product side by side and with different packaging sizes, while the other product attributes (i.e., total quantity, packaging design, relative price and the type of product) remain the same. Furthermore, the respondents are asked to indicate their preferences by selecting the radio button closer to the product that they prefer. The measurements of the preference include the liking of the product, product preference and the willingness to buy. The same products are asked to each measurement. There are two products used on each of the product category (hedonic and utilitarian) of the scale and the survey display can be seen on Appendix

3. The significance of these measurements will later be evaluated in the results and analysis

section.

Results

Variable Manipulation Check: Inequality

To be able to find out whether different level of inequality could impact respondents’ bias on numerosity, it has to be made sure that the study has actually succeed in making them believe that they are in either high or low inequality environment. As mentioned in the methodology part earlier, the respondents were randomly presented some wealth distribution data to manipulate their view on the inequality level in their area. To check the effectiveness of the Inequality manipulation in this research, a one-way ANOVA test has been done. There are three different measures (i.e., economic dissatisfaction, perceived inequality and economic suffer) to check this and they are factored by a categorical independent variable, which is the randomized inequality manipulation (0 = low inequality, 1 = high inequality).

The result of this one-way ANOVA is illustrated on the ANOVA table of Appendix 4. It can be seen from the table that out of three measures, a significant effect (p < .001) was found on the perceived inequality measure to the inequality manipulation (F (1,86) = 61.121,

p = .000). The F-test is also significant, which means that the means of each independent

variable group (i.e., low or high) are significantly different from each other. In addition, as it is also shown in the table, the variances of between and within groups are different. Therefore, it can be concluded that the inequality manipulation has successfully create the different effect to respondents to each condition and influence their perception of their economic inequality in their area.

Furthermore, the other two measures (i.e., economic dissatisfaction and economic suffer) are not statistically significant (all p > .05). However, the mean square of these measures show that the value of the between groups variance is still higher than the within groups variance. This signals a small impact of the manipulation to these measures, but might not be statistically significant. Hence, even though the manipulation did not work for all of the measures, it has succeeded to make the survey respondents perceive the economic inequality in their living area according to each treatment.

Reliability Analysis: Dependent variable

low, 7 = high) that shows the higher the point, the higher the numerosity bias of the participant. In the scale also, a within variable test was included by incorporating both hedonic (chocolate and wine) and utilitarian (toothpaste and shampoo) products.

A reliability test was conducted to check the reliability of this dependent variable scale, which then will show the internal consistency of the scale. As it can be seen on

Appendix 5, both the hedonic (Cronbach’s Alpha = .891) and the utilitarian parts

(Cronbach’s Alpha = .958) of the scale are both higher than 0.6, which show their high internal consistency in the dependent variable scale. This shows that the preference, liking and purchase intention of these products on each category are consistent, which can be summed up and further labeled as Numerosity Hedonic and Numerosity Utilitarian.

Testing the Hypothesis

The causal step approach by Barron and Kenny (1986) explains that in order to see the mediating effect of a variable, the researcher should look at the paths of the relationship between each of the variable and evaluate whether they have carried out according the statistical criteria or not. Therefore, a one-way ANOVA test has been conducted in order to see the main relationship of the variables in the hypothesis. In this case, the main effect of inequality to numerosity bias is measured. As it can be seen on Appendix 6, the effects of inequality variable to respondents’ numerosity bias on both within variable categories are not significant.

Specifically, the output shows that both p-value of the categories are higher than the significant level (all p > .05). The numerosity bias of hedonic products (Numerosity Hedonic:

p = .59) and the numerosity bias of utilitarian products (Numerosity Utilitarian: p = .568),

which also makes the F-statistics to become not significant on either case.

Hayes (2009) has argued that even though according to the causal path approach, the analysis should not be continued when the main relationship is not significant, the indirect effect might still present. However, in his paper, Hayes started the discussion by illustrating that the mediator variable should yet be significantly correlated with the other two variables; independent and dependent variable, for the mediation effect to occur. Therefore, a correlation check among all variables was performed. As shown on Appendix 7, all variables do not have significant correlation, except for the numerosity between hedonic and utilitarian products, which signals the non-existence of mediation.

Control and Desire for Control vs. both Numerosity relationships are not significant. It is important to note that the Sense of Control variable (mediator) is recoded to accurately demonstrate that the less of sense of control that the respondents have, the higher the level of desire for control will be. Base on this results, it can then be concluded that the hypothesis developed earlier is not proven and H1 is rejected. In other words, desire for control does not mediate the relationship between inequality and consumer numerosity bias.

Furthermore, even though there are two product categories in the dependent variable; hedonic and utilitarian products, it has been found that there are no significant differences between the results between them. As exhibited on Figure 2 and Appendix 6, the means of numerosity hedonic in low inequality has approximately the same means in the high inequality. The same case happens to the numerosity utilitarian. This means that the respondents respond more or less uniformly regardless of the inequality level, which in line with the earlier conclusion in rejecting the hypothesis (H1).

Discussion

This paper has focused on investigating the relationship between inequality and consumer behavior. Particularly, the study looked into the impact of different level of inequality to customers’ behavior on numerosity bias. As briefly discussed in the result part earlier, even though the manipulation of inequality variable turned out to be successfully impact the perceived inequality, the result of the main relationship and the indirect relationship are insignificant. In this section, some possible explanations for the insignificant result will be presented, such as; the study design limitation, the initial consumption behavior and the various perceived hedonic and utilitarian value.

Inequality Manipulation Limitation

To start off, based on the manipulation check performed earlier in this paper, the respondents’ reaction was only significantly impacted the perceived inequality, but did not give statistically significant result to the two other manipulation check factors; economic dissatisfaction and perceived economic suffering. This means that the variable manipulation in the survey did not demonstrate different impact on respondents’ satisfaction on their economic condition, nor they identify any different level of suffering of their economic situation. These findings might have reduced the overall effectiveness of the inequality manipulation. Perhaps, if the manipulation has checked off all of the inequality factors, the respondents would feel more exposed to inequality and possibly increase the impact of inequality in the study.

Lack of knowledge of the Inequality Concept

Moreover, in the study, the level of education was not controlled. This could impact the results because respondents might not acknowledge completely of the society and individual consequences of inequality in general. In the list of questions also, the topic inequality was not elaborated and specified. Hence, the respondents might have different understanding and it is hard to establish the grounding fact that the inequality

Numerosity Bias Scale Limitation: Likert Scale

Furthermore, the scale used for the dependent variable has made it easy for the respondents to compare the different product samples that are tested. The products are put side by side to make the difference in (i.e., size and number of pieces) and the same characteristics that they share (i.e., packaging design, overall quantity, quality) explicitly (Charness, Gneezy and Kuhn, 2012). The scale however, has not been extensively researched prior to the execution of the study. This could have made the scale to be less powerful in measuring numerosity bias.

An example of the contributing factor to this the use of the Likert scale (in this case 7-points Likert scale) could have possibly affected their decision making in filling in the survey. In the earlier paper by Chimi and Russel (2009) have found a couple disadvantages of employing Likert scale in a survey. Based on their paper, there are some limitations that are pertinent to this research, such as; the answers recorded are limited to a small number of discrete categories and the ambiguity of the middle response.

The Likert scale that was used in this research is a 7-points Likert scale. This means there are merely seven categories of preference, liking and purchase intentions that the respondents could choose from. They are compelled to pick their choice from a roughly granulated number of categories (Chimi and Russel, 2009). Specifically for this study, the respondents were asked to pick their choice between low numerosity and high numerosity products. According to this paper the distance between them are not made clearly and very fuzzy. Therefore, the depths of the numerosity between the products are greatly dependent to the respondents’ subjective judgment, which will make it hard to measure and analyze the results objectively.

neutral. There are in fact more possible motivations for the middle point answers. For instance, the respondent does not understand the scale or is unfamiliar with the products.

Another potential motivation for this is that even though the respondent understands the instructions of the scale, he/she might have chosen neutral because for some reasons he/she is not interested at the products presented. They then choose neutral for these motivations due to the lack of category for their reasons. These reasons do not fall into the ‘neutral’ category and should not have to be considered as ‘neutral’. This shows that the Likert scale does not necessarily take into account these cases when respondents actually do not have any specific response towards the scale and cause them to pick one of the options.

Numerosity Bias Scale Limitation: The Product Display

In the dependent variable scale, some of the sizes of the products displayed were not all common in the market. For instance, the high numerosity wine is presented as five individual bottles and each has 30ml quantity (Appendix 3). This size of bottle is not widely popular in the market, which might influence participants’ response. Their unfamiliarity with the products might trigger them to select the other product. This indicates that the results might be affected simply by unfamiliarity instead of by the independent or the mediating variables focused in the study.

Additionally, the brand name of the products were not removed and included in the survey. Initially, it was done to reduce the possibility of confusion among the respondents when identifying the products. The brands that are used in this scale are well-known brands in their categories (e.g., ‘Snickers’ brand for chocolate). Nonetheless, consumers original attitude towards the brands might have interfere the tested products in the numerosity scale.

Hedonic and Utilitarian Perception

Another aspect that might be caused by the scale is respondents’ different perception on hedonic and utilitarian products. The increasing trend of globalization has made some products more accessible and affordable. This trend might have changed consumers’ perception on the inequality and their buying habit. Consumers spending infrastructure has been experiencing some transition the past decade. For instance, in the beginning of their launching, smartphones were considered as luxury product that will only be purchased after all of the necessity products are fulfilled. However, as commonly known, having a smartphone has been one of the necessity needs that have to be fulfilled even before the basic needs (i.e., food, shelter and clothes) are realized.

This phenomenon may have happened in this study too. An article by Dhar and Wertenbroch (2000) discussed about how every product has its own hedonic (i.e., fun, excitement and pleasure) and utilitarian (i.e., instrumental and functional) parts of it. Some people may perceive one part more than another. Consequently, it is possible for the product examples in the scale have different meaning between respondents. Personal value, economic conditions and needs priorities might have played some roles in this effect.

Initial Packaging Preference

As explained earlier, the findings in the current paper were gathered through Amazon’s MTurk. The whole respondents are either citizens or people who are residing in the United States. Unlike in the Asia Pacific area, the use of small sachet packaging can hardly be found in the United States market. Instead, the large economic packaging is very popular among the market, especially among the household products.

Accordingly, the result of this research might have been impacted by respondents’ familiarity to the larger packaging, which they immediately associate with relatively cheaper price due to their pre-existing product association. This might still happen even when the fact that the relative price between high and low numerosity products are the same, has already been specifically mentioned in the survey instruction. Supporting this, World Bank national data (2016) shows that the consumption expenditure in the United States has been substantially increasing in the past decade. One of the reasons is due to the vast availability of large size products packaging in the market that stimulates the usage volume (Wansink, 1996).

Compensating Behavior

Immediately after being exposed to their economic situation, whether it is low or high inequality, the respondents were presented with the Sense of Control scale. This fact might be another probable explanation of the low correlation relationship between the independent variable (inequality) with the mediating variable (sense of control). The small gap and no delay of exposure to this scale could potentially describe the findings.

An example would be when a participant learned in the survey that the economic inequality in their area is high, according to the hypothesis; they will have to have lower sense of control. As elaborated in the theoretical background of this paper, these participants would feel fragile and uncertain that leads them to compensate their shortcomings, which then lead them to choosing the products that enable self-improvement (Allard and White, 2015), through heuristics. However, this was not the case for everybody in the survey.

Conclusion

This research has investigated the impact of inequality on consumers’ behavior, especially on their numerosity bias. The relationship was hypothesized to have consumers’

sense of control as the mediator, which was predicted to be triggered by the different level of

inequality that they are exposed to. To investigate the hypothesis developed, a survey has been spread among 102 participants from the United States derived from M-Turk. Afterward, the valid data (n=87) were processed on SPSS program, and some tests have been done to find out whether the hypothesis can be proven or not.

After the data was run, this study has found that the relationship between the independent variable (inequality) and the dependent variable (numerosity bias) is not significant. Also, the predicted mediator (desire for control) does not mediate the relationship both fully or partially. Therefore, the hypothesis (H1) must be rejected, and some possible explanations for this have been discussed. However, this study has interestingly found that the inequality manipulation (Côté et al., 2015) did not ultimately work for the participants. The manipulation check shows that the manipulation affects participants’ perception of their environment inequality, but does not change their feelings of dissatisfaction nor suffering.

The dependent variable scale was found to be internally consistent and reliable. This shows that the participants’ behaviors of every product presented are consistent throughout the scales (i.e., prefer, liking and purchase intention). Also, due to the self-reporting nature of the survey, the impact of inequality on the sense of control might have come earlier than expected. This immediate or lagging effect could be very interesting to be investigated more in the future. Furthermore, the perception on hedonic and utilitarian products and inequality might have been impacted by globalization and technology, which trigger homogeneity in the market, the transformation of their ideal-self and make people less sensitive to their these concepts.

Limitations

In addition to the factors explaining the insignificant results —which are elaborated in the discussion chapter of this thesis— during the execution of the research, some additional constraints that were found. This part of the paper will mainly review the general limitations that have subjected the study. Based on these limitations, some recommendations will later be presented in the next chapter, which might be useful to direct future research to avoid the same flaws in investigating the similar topic.

As mentioned, this study has incorporated age, gender, mood, ethnicity, and political orientation and income level as the control variable. Nevertheless, all of the respondents come from one homogenous market, which is the United States. This is due to author’s limited time, resources and access to gather bigger and more heterogeneous data. Even though they all live in different states, they have been exposed to the same types of advertisements and social issues. Being in the similar environments might have impacted the homogeneity of the effect on the inequality. Hence, the findings of this research might not be the same when the samples are taken from other regions in the world.

Secondly, the population of the experiment is relatively small. The lack of respondents might be one of the causes for the insignificant result. The study has two different treatments, and the number of respondents randomly exposed to the two conditions is more or less equally divided. By the start of this study, the targeted amount of respondents for each treatment is 50 respondents or 100 in total. However, after the data cleaning, only 87 of them are valid and can be used in the tests. Considering the targeted amount of respondents were in fact very low and considered the minimum, the lack of data could make the results have less power and therefore not significant.

Next, the theoretical framework of this research could have been developed further. Consumer numerosity bias is not a widely researched topic in the field of Marketing. During the academic investigation, it was found that the terminology used to explain this concept varies across different journals, such as; numerosity estimation and numerosity as a part of spatial magnitudes. Therefore it was somewhat challenging to analyze a consistent explanation of this concept based on the existing literature thoroughly. Consequently, it could be one of the reasons why the null hypothesis cannot be rejected.

desirability. This example means that they pick the answer not according to their real present feeling of the scale, but according to their favorable self-image (van de Mortel, 2008).

Further Research

This chapter will confer the direction for the further research in this field based on the recent study. Accordingly, the future research recommendation will be set up around the potential explanations for the insignificant data, which were described in the Discussion chapter, and some general restraints mentioned in the Limitations chapter of this paper. Thus, the same limitations can be avoided and the overlooked factors can later be taken into consideration to get more extensive and accurate results in the future.

As discussed, the hypothesis (H1) of this paper is not significant. Adding more samples in the test might increase the strength of the impact. This research has less than 100 survey participants due to some invalid data. To prevent this, in the future it is useful to have about 220 total participants or more, to make sure at least 200 samples can be used and valid. Besides, since the recent study only involved American citizens, it is recommended to gather a population of more heterogeneous samples to avert potential biases and to have a more comprehensive implementation.

Furthermore, before the actual survey is spread among the participants, the researcher is suggested to do a pre-test beforehand. Even though the scales are taken from various extensive prior researches, a quick test to 10-20 people to check the affectivity of the manipulation and the other scales used in the analysis. This would help to ensure the success of the application of the scales. Moreover, doing pretest will help to identify potential problems with the scale and realizing whether there are more control variables (e.g., education level) that need to be added to the study in general. For instance, the numerosity bias scale in this research could include more products in both hedonic and utilitarian categories, or the brand on the product examples could have been removed to refrain from biases. Another example would be the order of respondents’ exposure that might have impacted the overall results. By doing pre-test, the immediate or lagging effect could be detected and the survey will have the most effective systematic order in presenting the scales.

References

Adler, A. (1930). Individual psychology. Psychologies of 1930. Worcester, Massachusetts: Clark University Press.

Allard, T., & White, K. (2015). Cross-domain effects of guilt on desire for self-improvement products. Journal of Consumer Research, 42(3), 401-419.

Ampuero, O., & Vila, N. (2006). Consumer perceptions of product packaging. Journal of consumer marketing, 23(2), 100-112.

Bagchi, R., & Davis, D. F. (2011). 29 for 70 Items or 70 Items for 29? How presentation order affects package perceptions. Journal of Consumer Research, 39(1), 62-73.

Bagchi, R., & Davis, D. F. (2016). The role of numerosity in judgments and decision-making. Current Opinion in Psychology, 10, 89-93.

Baker, S. M., Gentry, J. W., & Rittenburg, T. L. (2005). Building understanding of the domain of consumer vulnerability. Journal of Macromarketing, 25(2), 128-139.

Bauhoff, S. (2011). Systematic self-report bias in health data: impact on estimating cross-sectional and treatment effects. Health Services and Outcomes Research Methodology, 11(1-2), 44-53.

Blundell, R., & Preston, I. (1998). Consumption inequality and income uncertainty. The Quarterly Journal of Economics, 113(2), 603-640.

Bound, J., Brown, C., & Mathiowetz, N. (2001). Measurement error in survey data. In Handbook of econometrics (Vol. 5, pp. 3705-3843). Elsevier.

Buunk, A. P., & Gibbons, F. X. (2007). Social comparison: The end of a theory and the emergence of a field. Organizational Behavior and Human Decision Processes, 102(1), 3–21. Chaiken, S. (1987). The heuristic model of persuasion. In Social influence: the ontario symposium (Vol. 5, pp. 3-39).

Chimi, C. J., & Russell, D. L. (2009). The Likert Scale: A Proposal for Improvement Using Quasi-Continuous Variables. EDSIG, 1.

Christen, M., & Morgan, R. M. (2005). Keeping up with the Joneses: Analyzing the effect of income inequality on consumer borrowing. Quantitative Marketing and Economics, 3(2), 145-173.

Cobb-Walgren, C. J., Ruble, C. A., & Donthu, N. (1995). Brand Equity, Brand Preference, and Purchase Intent. Journal of Advertising, 24(3), 25-40.

Côté, S., House, J., & Willer, R. (2015). High economic inequality leads higher-income individuals to be less generous. Proceedings of the National Academy of Sciences, 112(52), 15838-15843.

Cutright, K. M. (2011). The beauty of boundaries: When and why we seek structure in consumption. Journal of Consumer Research, 38(5), 775-790.

Cutright, K. M. (2012). The Beauty of Boundaries: When and Why We Seek Structure in Consumption. Journal of Consumer Research, 38(5), 775–790.

Cutright, K. M., Samper, A., Bettman, J., Luce, M. F., Berman, J., & Chan, C. (2014). Doing It the Hard Way: How Low Control Drives Preferences for High-Effort Products and Services. JOURNAL OF CONSUMER RESEARCH, Inc., 41.

Deininger, K., & Squire, L. (1996). A new data set measuring income inequality. The World Bank Economic Review, 10(3), 565-591.

Dhar, R., & Wertenbroch, K. (2000). Consumer choice between hedonic and utilitarian goods. Journal of marketing research, 37(1), 60-71.

Faraji‐Rad, A., Melumad, S., & Johar, G. V. (2017). Consumer desire for control as a barrier to new product adoption. Journal of Consumer Psychology, 27(3), 347-354.

Haughton, J., & Khandker, S. R. (2009). Handbook on poverty+ inequality. World Bank Publications.

Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical Mediation Analysis in the New Millennium. Communication Monographs, 76(4), 408-420.

Higgins, E. T. (2011). Beyond pleasure and pain: How motivation works. Oxford University Press.

Holbrook, M. B., & Hirschman, E. C. (1982). The experiential aspects of consumption: Consumer fantasies, feelings, and fun. Journal of consumer research, 9(2), 132-140.

Kelley, H. H. (1987). Attribution in social interaction. In Preparation of this paper grew out of a workshop on attribution theory held at University of California, Los Angeles, Aug 1969.. Lawrence Erlbaum Associates, Inc.

Kuznets, S. (1955). Economic growth and income inequality. The American economic review, 1-28.

Lachman, M. E. & Weaver, S. L. (1998b). The sense of control as a moderator of social class differences in health and well-being. Journal of Personality and Social Psychology, 74, 763-773.

Leotti, L. A., Iyengar, S. S., & Ochsner, K. N. (2010). Born to choose: The origins and value of the need for control. Trends in cognitive sciences, 14(10), 457-463.

Oppenheimer, D. M., Meyvis, T., & Davidenko, N. (2009). Instructional manipulation checks: Detecting satisficing to increase statistical power. Journal of Experimental Social Psychology, 45(4), 867-872.

Pelham, B. W., Sumarta, T. T., & Myaskovsky, L. (1994). The easy path from many to much: The numerosity heuristic. Cognitive Psychology, 26(2), 103-133.

Posadas, D. (2014, May 22). Sachets help low-income communities but are a waste nightmare. Retrieved from https://www.theguardian.com/sustainable-business/sachet-packaging-low-income-communities-waste-nightmare

Ravallion, M. (1997). Can high-inequality developing countries escape absolute poverty?. Economics letters, 56(1), 51-57.

Roth, S., & Kubal, L. (1975). Effects of noncontingent reinforcement on tasks of differing importance: Facilitation and learned helplessness. Journal of Personality and Social Psychology, 32(4), 680.

Shrivastava, S., Jain, G., Nayakankuppam, D., Gaeth, G. J., & Levin, I. P. (2017). Numerosity and allocation behavior: Insights using the dictator game. Judgment and Decision Making, 12(6), 527.

Sizing Up Savings: Consumers Prefer Package Adjustments Over Sales. (n.d.). Retrieved from http://www.nielsen.com/us/en/insights/news/2016/sizing-up-savings-consumers-prefer-package-adjustments-over-sales.html

Strahilevitz, M., & Myers, J. G. (1998). Donations to charity as purchase incentives: How well they work may depend on what you are trying to sell. Journal of consumer research, 24(4), 434-446.

Szmigin, I., & Piacentini, M. (2014). Consumer behaviour. Oxford University Press, USA. Thaler, R. (1985). Mental accounting and consumer choice. Marketing science, 4(3), 199-214.

Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. science, 185(4157), 1124-1131.

Underwood, R. L. (2003). The communicative power of product packaging: creating brand identity via lived and mediated experience. Journal of Marketing Theory and Practice, 11(1), 62-76.

Van de Mortel, T. F. (2008). Faking it: social desirability response bias in self-report research. Australian Journal of Advanced Nursing, The, 25(4), 40.

Wade, R. H. (2004). Is globalization reducing poverty and inequality?. World development, 32(4), 567-589.

Appendix

Appendix 1: National Representative Study Graphs - Low vs. High Inequality

How equally distributed do you perceive the wealth in your home state?

Extremely Unequal Extremely Equal

Correct Answer Please fill in the correct answers:

• What percentage of the private wealth is owned by the wealthiest fifth of the population?

• What percentage of the private wealth is owned by the poorest fifth of the population? • What percentage of the private wealth is owned by the middle fifth of the population? Suffer: To what extent do you feel that your state is suffering economically?

Far too little Far too much

Satisfaction: How satisfied are you with the economic status of your state?

Extremely dissatisfied Extremely satisfied

Comprehension: Do you feel that you understood the information represented in the pie chart about wealth distribution in your state?

Appendix 3: Numerosity Scale

Appendix 4: Manipulation Check Oneway-ANOVA Descriptives N Mean Std. Deviation Std. Error

95% Confidence Interval for Mean Lower Bound Upper Bound Economic Dissatisfaction Low 44 6.07 2.500 .377 5.31 6.83 High 43 6.81 2.822 .430 5.95 7.68 Total 87 6.44 2.675 .287 5.87 7.01 Perceived Inequality Low 44 5.05 2.449 .369 4.30 5.79 High 43 8.72 1.894 .289 8.14 9.30 Total 87 6.86 2.858 .306 6.25 7.47 Suffer Economically Low 44 5.59 2.375 .358 4.87 6.31 High 43 6.44 2.763 .421 5.59 7.29 Total 87 6.01 2.595 .278 5.46 6.56 ANOVA

Appendix 5: Reliability

Reliability - Scale: Numerosity Hedonic

Reliability Statistics

Cronbach's Alpha

Cronbach's Alpha Based on Standardized Items N of Items .891 .891 6 Item Statistics Mean Std. Deviation N

Preference on high numerosity choco 5.06 2.315 87

Liking on high numerosity chocol 4.86 2.436 87

Purchase intention on high numerosity choco 4.97 2.433 87

Preference on high numerosity wine 4.16 2.449 87

Liking on high numerosity wine 4.23 2.400 87

Purchase intention on high numerosity wine 4.18 2.485 87

Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Squared Multiple Correlation Cronbach's Alpha if Item Deleted Preference on high numerosity choco 22.40 100.220 .670 .695 .879

Liking on high numerosity chocol

22.60 96.522 .714 .822 .872

Purchase intention on high numerosity choco

22.49 96.044 .727 .849 .870

Preference on high numerosity wine

23.30 98.235 .667 .770 .879

Liking on high numerosity wine

23.23 94.877 .770 .834 .863

Purchase intention on high numerosity wine

23.28 95.737 .714 .778 .872

Scale Statistics

Mean Variance Std. Deviation N of Items

Reliability - Scale: Numerosity Utilitarian

Reliability Statistics

Cronbach's Alpha N of Items

.958 6

Item Statistics

Mean

Std.

Deviation N Preference on high numerosity toothpaste 3.61 2.512 87 Liking on high numerosity toothpaste 3.52 2.510 87 Purchase intention on high numerosity toothpaste 3.56 2.500 87 Preference on high numerosity shampoo 3.16 2.377 87

Liking on high numerosity shampoo 3.22 2.394 87

Purchase intention on high numerosity shampoo 3.09 2.419 87

Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Preference on high numerosity toothpaste 16.55 123.692 .881 .949

Liking on high numerosity toothpaste

16.64 122.930 .898 .947

Purchase intention on high numerosity toothpaste

16.60 124.801 .862 .951

Preference on high numerosity shampoo

17.00 126.930 .871 .950

Liking on high numerosity shampoo

16.94 125.776 .888 .948

Purchase intention on high numerosity shampoo

17.07 128.786 .812 .957

Scale Statistics

Mean Variance Std. Deviation N of Items

Appendix 6: Testing the Hypothesis, Oneway-ANOVA Descriptives N Mean Std. Deviation Std. Error

95% Confidence Interval for Mean Lower Bound Upper Bound Numerosity Hedonic Low 44 4.44 1.696 .256 3.92 4.95 High 43 4.72 2.188 .334 4.04 5.39 Total 87 4.58 1.948 .209 4.16 4.99 Numerosity Utilitarian Low 44 3.50 2.174 .328 2.84 4.16 High 43 3.22 2.306 .352 2.51 3.93 Total 87 3.36 2.231 .239 2.88 3.84 ANOVA

Appendix 7: Mediation - Correlations Analysis Descriptive Statistics Mean Std. Deviation N Inequality .49 .503 87 Numerosity Hedonic 4.58 1.948 87 Numerosity Utilitarian 3.36 2.231 87 Sense of Control 2.99 1.068 86 Correlations

Inequality Num.Hedonic Num.Utilitarian Sense of Control

Inequality Pearson Correlation 1 .072 -.062 .077

Sig. (2-tailed) .509 .568 .483 N 87 87 87 86 Numerosity Hedonic Pearson Correlation .072 1 .359** -.040 Sig. (2-tailed) .509 .001 .716 N 87 87 87 86 Numerosity Utilitarian Pearson Correlation -.062 .359** _{1 .012} Sig. (2-tailed) .568 .001 .914 N 87 87 87 86 Sense of Control Pearson Correlation .077 -.040 .012 1 Sig. (2-tailed) .483 .716 .914 N 86 86 86 86