CONSUMER UNDERSTANDING AND USE OF NUMERIC INFORMATION IN PRODUCT CLAIMS by NAMIKA SAGARA A DISSERTATION Presented to the Department of Marketing and the Graduate School of the University of Oregon in partial fulfillment of the requirements for the degree of Doctor of Philosophy December 2009 11 University of Oregon Graduate School Confirmation of Approval and Acceptance of Dissertation prepared by: Namika Sagara Title: "CONSUMER UNDERSTANDING AND USE OF NUMERIC INFORMATION IN PRODUCT CLAIMS" This dissertation has been accepted and approved in partial fulfillment of the requirements for the Doctor of Philosophy degree in the Department of Marketing by: Peter Wright, Chairperson, Marketing Lynn Kahle, Member, Marketing Ellen Peters, Member, Not from U of 0 Robert Madrigal, Member, Marketing Paul Slovic, Outside Member, Psychology and Richard Linton, Vice President for Research and Graduate Studies/Dean of the Graduate School for the University of Oregon. December 12,2009 Original approval signatures are on file with the Graduate School and the University of Oregon Libraries. © 2009 Namika Sagara 111 in the Depatiment of Marketing Namika Sagara An Abstract of the Dissertation of for the degree of to be taken IV Doctor of Philosophy December 2009 Title: CONSUMER UNDERSTANDING AND USE OF NUMERIC INFORMATION IN PRODUCT CLAIMS Approved: Dr. Peter Wright Numeric information is often presented to consumers in order to communicate impOliant and precise infonnation that is not well communicated through non-numeric information', The assumption of marketers, then, seems to be that numeric infonnation is useful for consumers in evaluating products. Do' consumers understand and use such numerical information in product claims? Recent research suggests that many people are "innumerate" and about half of Americans lack the minimal mathematical skills needed to use numbers embedded in printed materials. This suggests that many Americans lack the minimal mathematical skills needed to use numbers embedded in product claims and other marketing communications, In a series of five experiments. I investigated if and vhow consumers understand and use numeric infonnation presented in product claims in their evaluation of consumer goods. The results demonstrated that participants, and especially less numerate individuals, were susceptible to an Illusion-of-Numeric-Truth effect: they judged false claim as hue when numeric meaning was inaccurately translated (e.g., "30% of consumers" inaccurately translated to "most consumers"). Mediation analysis suggested that highly numerate participants were better at developing affective reactions toward numeric infonnation in product claims and using these affective reactions as infonnation when they were faced with truth judgments. Highly numerate individuals were also more sensitive to different levels of numeric infonnation in their product evaluations. This sensitivity also seemed to depend on their drawing affective meaning from numbers and number comparisons and using this infonnation in product evaluations. Although less numerate individuals reported that numeric infom1ation is impOliant, they were less sensitive to numeric infonnation unless they were encouraged to process numeric infonnation more systematically. The results from this dissertation indicate that not all numeric infOlmation will be used and be useful to all consumers. Therefore, simply presenting numetic infonnation may not be sufficient for numeric infonnation to be useful for all consumers. VI CURRICULUM VITAE NAME OF AUTHOR: Namika Sagara PLACE OF BIRTH: Tokunoshima, Kagoshima, Japan· GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED: University of Oregon DEGREES AWARDED: Doctor of Philosophy in Marketing, 2009, University of Oregon Master of Science in Psychology, 2009, University of Oregon Bachelor of Science in Psychology, 2002, University of Oregon AREAS OF SPECIAL INTEREST: Affect Decision Making Numeric Information Processing and Numeracy Human Life Valuation PROFESSIONAL EXPERIENCE: Graduate Teaching and Research Fellow, Department of Marketing, University of Oregon, Eugene, 2005-2009 Research Associate, Decision Research, Eugene, Oregon, 2004-2009 Graduate Teaching and Research Fellow, Department of Psychology, University of Oregon, Eugene, 2005 Vll GRANTS, AWARDS AND HONORS Merle King Smith Marketing Scholars Award, 2005-2009 Lundquist College of Business Scholarship. Lundquist College of Business Summer Research Fellowship 2007-2009 2009 AMA Sheth Doctoral Consortium Fellow, Georgia State University 2008 Robert Mittelstaedt Doctoral Symposium Fellow, University of Nebraska 2007 Robert Mittelstaedt Doctoral Symposium Fellow, University of Nebraska Marketing Science Institute Doctoral Student Travel Award at Transfom1ative Consumer Research, 2007 Marketing Science Institute Travel Award at Transformative Consumer Research, 2007 Calvin Reed Smith Research Grant. Lundquist College of Business, 2005 PUBLICATIONS AND CONFERENCE PROCEEDINGS: Dickert, S., Sagara, N., & Slovic, P. (forthcoming). Affective motivations to help others: A two-stage model of donation decisions. In D. M. Oppenheimer & C. Y Olivola (Eds.), Experimental approaches to the Study ofCharitable Giving. Dickert, S, Sagara, N., & Slovic, P. (Under second review). Affective motivations to help others: A two-stage model of donation decisions. Olivola, C. Y, & Sagara, N. (forthcoming). The psychophysical and cognitive foundation of human life valuation. Proceedings ofthe National Academy of Sciences. Sagara, N., & Peters, E. (2007). Affect, affective precision, and primacy effects in stock choices. Advances in Consumer Research, 34,461. V111 AKNOWLEDGEMENT First of all, I would like to thank my committee members-Ellen Peters, Peter Wright, Lynn Kahle, Bob Madrigal, and Paul Slovic-for providing invaluable help and feedback. I am also grateful for their understanding and support regarding the timing of my dissertation-related work. I especially thank Ellen Peters and Paul Slovic, who have been my greatest mentors and role models through my undergraduate, masters and doctoral trainings. Without their guidance, support, and patience, few of my academic accomplishments, including two graduate degrees, would have succeeded. I also would like to thank Stephanie Carpenter, Courtney Boerstler, and Michael Kyweriga for their editorial feedback on my dissertation, and also Bob Madrigal and Nathan Dieckmann for their advice on some of the analysis. In addition, I would like to thank the Department of Marketing and the Department of Psychology for providing me with opportunities to recruit participants. I would like to thank Misao Makino for strongly encouraging me to pursue graduate degrees in the United States. lowe my family a special debt of gratitude for their continuous support and for believing in me. Finally yet importantly, I would like to thank my friends who are always willing to show their support, and who have made my life in Eugene a full and happy one. Although the list could go on forever, notably among them are Michael Kyweriga, Stephanie Carpenter, Johnny Chen, Sonia Venkatraman, Seraphine Shen-Miller, Brian Merrell, and John Ahlen. IX TABLE OF CONTENTS Chapter I. INTRODUCTION . Page Research Objectives............................................................................................... 3 II. LITERATURE REVIEW........................................................................................ 5 NUlneracy 5 Affect 11 The Illusion-of-Truth Effect 18 Hypotheses 21 III. STUDIES 24 Overview..... 24 Study 1: Numeric Memory for a Product Claim....... 26 Method 26 Results.............................................................................................................. 29 Discussion 33 Study 2: Use of Numeric Information in Affective Product Evaluation 35 Method............................................................................................................. 36 Results :............................................... 36 Discussion 39 Chapter Page x Study 3: Use of Numeric Information in Affective Product Evaluation and in the Illusion-of-Numeric-Truth Effect.......................................................... 40 Method 40 Results.............................................................................................................. 41 Discussion 52 Study 4: Systematic Processing of Numeric Information in Affective Product Evaluation................................................................................................. 54 Method 54 Results.............................................................................................................. 56 Discussion 64 Study 5: Number Comparison and Numeracy....................................................... 65 Method 66 Results.............................................................................................................. 67 Discussion ,.................................................................................. 71 IV. GENERAL DISCUSSION 76 APPEJ\JDICES 82 A. STUDY 1 PRETEST.. 82 B. STUDY 1 SURVEY INSTRUMENT............................................................... 83 C. STUDY 2 SURVEY INSTRUMENT............................................................... 86 D. STUDY 3 SURVEY INSTRUMENT. ·89 E. STUDY 4 SURVEY INSTRUMENT 92 Chapter Xl Page F. STUDY 5 SURVEY INSTRUMENT 96 G. NUMERACY MEASURE 101 H. DEMOGRAPHIC QUESTIONS 103 REFERENCES 105 XlI LIST OF FIGURES Figure Page 1. Percentages of new and false claims judged true by the less and the highly numerate groups ,. 32 2. Affect towards product with unfavorable and favorable numeric information reported by the less numerate and highly numerate groups............... 38 3. Affect towards product with unfavorable and favorable numeric information reported by the less numerate and the highly numerate groups 43 4. Predicted affect towards the product in the unfavorable numeric condition depicted by three different levels of numeracy.. 51 5. Predicted affect towards the product in the favorable numeric condition depicted by three different levels of numeracy 51 6. Mean affect towards the target product in the unfavorable numeric condition depicted by numeracy 59 7. Participants' preference towards the target product in the unfavorable and the favorable numeric conditions depicted separately for the less numerate and the highly numerate groups............................................................................. 61 8. Participants' preference towards the target product in the unfavorable and the favorable numeric conditions depicted separately for the fluent and the disfluent conditions 61 9. Affect towards the product by the less numerate and the highly numerate groups in each condition 70 10. Affect towards product reported by the less numerate and the highly numerate groups in each condition 71 XliI LIST OF TABLES Table Page 1. Examples of Product Claim 29 2. Percentages of Participants Who Judged True for Each Type of Claims 31 3. Significance Test from ANOVA and Mean Affect Towards the Products by Numeric Conditions and Numeracy....................................................................... 38 4. Significance Levels................................................................................................ 42 5. Results from Linear Regression Analysis.............................................................. 46 6. Results from Logistic Regression Analysis 46 7. Coefficients and Significance Levels for Predicting Participants' Affect Towards Products. 50 8. Significance Levels of MANOVA...................................................................... ... 56 9. Means and Standard Deviations............................................................................. 57 10. Significance Levels of MANOVA.................................................................... ..... 63 11. Results from a MANOVA 69 12. Correlations Between Affective Product Evaluations and the Use of Star Ratings in Affective Product Evaluations for Each Condition Crossed by Numeracy 74 CHAPTER I INTRODUCTION Numeric information is often presented to consumers in order to communicate important and precise information that is not well communicated through non-numeric information. For example, marketers use numeric information intending to convey favorable information about their products. PepsiCo, in a recent television campaign, claimed that their Diet Pepsi tastes more like real cola than Diet Coke because in a test comparing the two colas, 56% believed Diet Pepsi tasted most like real cola. In addition, interactive Web sites allow customers to evaluate their products using numeric information. Amazon.com, for example, features star ratings ranging from 0 to 5 stars. The average star rating for each product appears as an icon, and sample size and distribution (both in frequency and percentage format) are also available. On dell.com you can find claims similar to "89% (85 out of 96) of customers would recommend this product to a friend" (dell.com, 2009a) and "Avg Customer Rating 4.3 of 5" (dell.com, 2009b). These consumer reviews are available to any customer who visits their Web site. The assumption of marketers, then, seems to be that consumers value numeric information, and that they can understand and use such numeric information when evaluating a product. There are several reasons why, however, consumers may be less sensitive to numeric information in their product evaluations. One is consumers' basic ability to conduct and understand simple math (e.g., 15% off of $30). In addition, 2although consumers may recognize each piece of numeric intormation, they may fail to draw meaning out of numeric information in given contexts and may unsuccessfully use the numeric information in their judgments. In fact, data trom the National Adult Literacy Survey indicates that about half of Americans lack the minimal mathematical skills necessary to use numbers embedded in printed materials (Kirsch et aI., 2002). This suggests that about half of all Americans may lack the skills to understand simple numerical infoD11ation used in product infoTI11ation and other marketing communications. For these reasons, some consumers may read, "35% of consumers preferred Diet Pepsi" as "only few consumers preferred," whereas other consumers may read it as "consumers preferred Diet Pepsi" because they fail to incorporate the numeric information into the product claim. In addition, some consumers may realize that more infoTI11ation may be needed to truly understand the meaning of numeric information in this claim. For example, "35%" has a different meaning if 35% of consumers preferred Diet Pepsi between two diet cola drinks, or among ten different diet cola drinks. Another reason consumers may be insensitive to numeric information is a lack of motivation to process numeric information in depth. Consumers may feel overwhelmed to work with complicated numbers, or they may feel numeric infonnation is not useful tor their decisions. They may also prefer to make judgments using a heuristic due to, for example, time pressure or limited cognitive ability. In addition, they may prefer and weigh nonnumeric infonnation part of marketing communication more than numeric information in their judgments. For example, they may focus on how they feel about brand images or the wording of product claims rather than on factual numeric 3information. Some consumers may trust the source or the numeric information more than others (Gurmankin, Baron, & Armstrong, 2004). Insensitivity to numeric information may influence not only online judgments of a product and product claims but also later judgments about a product and product claims. Research has demonstrated that when consumers engage in low-involvement information processing they tend to rely on familiarity of claims when they later judge truthfulness of claims. This truth effect was also observed when claims were explicitly identified as true when consumers were evaluating them for the first time (cf. Illusion-of-Truth effect). If people engage in low-involvement numeric information processing, then consumers may be susceptible to the Illusion-of-Truth effect when judging the truthfulness of numeric product claims. That is, consumers may use familiarity of non-numeric information when judging the truthfulness of claims involving numeric information. For example, claims like "most consumers preferred Diet Pepsi" may be judged accurate when in fact only 35% of consumers preferred Diet Pepsi, because the nonnumeric part of information seems familiar. This familiarity effect may be particularly strong for people who are unable to develop precise feelings about numeric information (e.g., "I am not sure how good or bad I feel about the numeric information "35%" given the context of this claim"). Research Objectives The main objective of this dissertation is to investigate if and how consumers understand and use the numeric information presented in product claims and consumer polls. A second interest is to explore how we can help consumers, especially consumers 4with limited numerical ability (cf. less numerate consumers), use crucial numeric information more in their judgments and decisions. There are three major research goals in this dissertation. First, this research will investigate evidence for a novel version of the Illusion-of-Truth effect (Skumik, Yoon, Park, & Schwarz, 2005) using product claims that contain crucial numeric information that mayor may not be consistent with the rest ofthe information written in the text. In this study, the Illusion-of-Numeric-Truth effect is observed when participants judge claims to be true even if the numerical meanings of the claims are inaccurately translated (e.g., "30% of consumers" inaccurately translated to "most consumers"). The relationship between the Illusion-of-Numeric-Truth effect and numeracy (the ability to process basic probability and numerical concepts: Peters et a!., 2006) will be investigated as well. Second, the influence of numeric information in product evaluations among participants that are lower or higher in numeracy will be investigated. Last, I test ways to help less numerate individuals improve their use of numeric information in their product evaluations. 5CHAPTER II LITERATURE REVIEW Numeracy Conceptualization and Development of Measurement Numeracy may be broadly defined as a basic ability to understand and work with numbers. Numbers may be expressed in various forms, such as in probability, proportion, time, money, and measurement. In order to work with numbers, we may need to understand the absolute and relative magnitude of numbers and the contextual information around the numbers, and be able to compare numbers and engage in simple calculation. In marketing contexts, we see numbers used in the descriptions of discounts, price, consumer polls, rebates, and product and service attributes. Therefore, in order to be competent with numbers in the marketplace, we may need to, for example, understand the magnitude of price and discount, be able to compare prices and product atttibutes, and calculate change, tax, and tips. One interesting characteristic of numeric information is its dependency on context: the meanings of numeric information change dramatically from one context to another. For example, the following infonnation has very different meanings even though they have exactly the same numbers: $32, 32 Fahrenheit, 32%, and 32 out of 250. 6Numeracy has been operationalized slightly differently by different researchers. Paulos (1988) defined innumeracy as "inability to deal comfortably with the fundamental notions of number and chance" (p. 3). Schwartz, Woloshin, and Rimer (2001) defined numeracy as facility with basic probability and numerical concepts, and measured numeracy with three simple math-like questions. One of the questions asked: "Imagine that we flip a fair coin 1,000 times. What is your best guess about how many times the coin would come up heads in 1,000 flips?" Only about half of the women (54%) recruited from communities in the U.S. answered this question correctly. Their study results demonstrated that the accurate use of numeric information in assessing perceived risk related to breast cancer was more strongly associated with numeracy than how the information was presented. Lipkus, Samsa, and Rimer (2001) defined numeracy as "how facile people are with basic probability and mathematical concepts," and added eight questions to the three items from the Schwart7. et al. (2001) study to measure numeracy. The additional items were designed to assess individuals' ability to compare risks, and move between decimal representations, proportions, and fractions (e.g., If the chance of getting a disease is 20 out of 100, this would be the same as having a __ % chance of getting the disease). In their paper, Peters, Dieckmann, Dixon et aI., (2007) added four more items that are more difficult to the II-item scale developed by Schwartz et ai. and Lipkus et ai. The items were added to test the understanding of base rate as well as the ability to make more complex likelihood calculations. The additional four questions helped the measure to be more normally distributed. Previous Findings: Influence of Numeracy on Judgments and Decisions Dieckmann, Slovic, and Peters (2009) tested if people with different levels of numeracy focus on different information sources-likelihood assessments in numeric or narrative evidence without numeric likelihood estimates-in judging risk of terrorist attack forecast. They demonstrated that individuals with lower numeric skills used their perceptions of the narrative evidence more, whereas respondents with higher numeric skills focused more on the numeric likelihood assessment. They concluded that factors that influence the judgments of less and highly numerate individuals may be different. In Peters's et al. (2006) study, participants were presented with a statement that included a probability. For example, a student, "Emily," was described as receiving 74% correct on her exam in one condition and 26% incorrect on her exam in another condition. They found that judgments made by those who were lower in numeracy were more sensitive to how the numeric information was framed: "26% incorrect" was perceived more negative than "74% correct." They argued that this is because highly numerate individuals are better at transforming a number in one format (e.g., 74% correct) into another fomlat (e.g., 26% incorrect). There were several studies conducted to test the associations of numeracy and the understanding and the use of numeric information in risk and health domains (e.g., Hibbard, Peters, Dixon, & Tusler, 2007; Lipkus et al., 2001; Nelson, Fagerlin, Lipkus, & Peters, 2008; Peters, 2008; Peters et al., 2009; Peters, Hibbard, Slovic, & Dieckman, 2007; Peters & Levin, 2008; Woloshin, Schwartz, Black & Welch, 1999). Sheridan and Pignone (2002) investigated medical students' numeracy level, and its association to the 7 8ability to interpret risk infom1ation. The results demonstrated that, although 94% reported that they thought they were good with numbers, only 77% of the participants answered all three relatively simple numeracy questions correctly. In addition, they demonstrated that numeracy and interpretation of risk infoTI11ation were related: students with perfect numeracy scores did better in both risk comparison tasks and quantitative interpretation tasks than those who did not receive perfect numeracy scores. Feldman-Stewart et al. (2000) tested whether fonnats of displaying quantitative infonnation, such as probabilities of treatment risks and benefits, influence patients' accuracy and speed regarding the use ofquantitative infonnation. Their results suggest that the fonnats that are best for making a choice are different from the fonnats that are best for estimating the size of an amount. Gunnankin, Baron, and Annstrong (2004) investigated whether patients trust and are more comfortable with doctor's verbal and numeric risk estimates. In the experiment, participants were presented with scenarios that discussed the likelihood of a certain cancer, and were asked to rate (a) how likely they think they have the cancer, (b) how comfortable they were with the infonnation they were given about the risk of the cancer, and (c) how much they trusted the infonnation given by the doctor. Each scenario contained either only verbal (cf. non-numeric) infonnation, or verbal infonnation along with numeric infonnation (either percentages or fractions). They found that participants were more likely to trust the scenarios with both numeric and verbal estimates more than the scenarios with only verbal estimates. However, this effect was qualified by numeracy: trust and numeracy were positively correlated. This suggests that people with lower number proficiency were more likely to trust verbal infonnation than 9numeric infonnation, whereas people with higher number proficiency were more likely to trust numeric infonnation. One factor that has been identified to influence individuals' judgment processes is "evaluability." Hsee and associates demonstrated that individuals tend to put more weight on the attributes that are easily evaluated than attributes that are not easily evaluated (Hsee, 1996; Hsee, Blount, Loewenstein, & Bazerman, 1999). Participants in Hsee's (1996) study evaluated two used dictionaries-one contained 10,000 words and looked like new, and another contained 20,000 words and had a tom cover. Half of the participants were presented with one of the dictionaries and asked how much they would pay for it (cf. separate-evaluation condition), and the rest of the participants were presented with both dictionaries and asked how much they would pay for each dictionary (cf. joint-evaluation condition). In the separate-evaluation, participants gave a higher price for the dictionary with 10,000 words than for the dictionary with 20,000 words, whereas participants gave a higher price for the dictionary with a 20,000-word entry than with a 1O,OOO-word entry in the joint-evaluation condition. Hsee (1996) argued that number of entries was hard to evaluate in the separate-evaluation condition because the evaluator does not know how good a 10,000-word entry is. Physical conditions of the dictionaries are, on the other hand, relatively easier to evaluate (e.g., a new cover is good and a tom cover is bad). Therefore, participants weighed the aesthetics of the dictionary cover in their judgments more than the number of word entries when the two dictionaries were evaluated separately. In the joint-evaluation condition, the dictionaries with a 10,000-word entry and a 20,000-word entry were presented together. Participants could 10 therefore compare the number of word entries. Because participants were able to evaluate how good a 20,000-word entry was compared to a 10,000-word entry and, arguably, word entry was more important factor than the cosmetics of the dictionary, they were willing to pay more for the dictionary with a 20,000-word entry. Peters et al. (2009) manipulated "evaluability" by providing affective labeling to numeric ratings. They presented pmiicipants with hard-to-evaluate healthcare ratings (e.g., score of 93 out of 100 for a survival rate), and provided affective categories (e.g., good, poor) to help the healthcare ratings become more "evaluable." They demonstrated that participants were more likely to focus on a more important attribute than a less important attribute when they were presented with affective categories. Interestingly, most ofthe hard-to-evaluate attributes were expressed in the numeric information in these studies. Numeric information may be chosen as hard-to- evaluate attributes, partially because numeric information is often completely dependent on its context (e.g., 20% correct on exam vs. 20% wrong on exam), and many individuals are "innumerate" (Paulos, 1988). This suggests that simply presenting numeric information may not be sufficient for consumers to effectively use numeric information in their judgments and choices. In summary, one of the consistent themes in the numeric cognition literature is that people may differ substantially in numeric ability (Lipkus et aL, 2001; Woloshin et a1., 1999), and that many may be "innumerate" (Paulos, 1988). Data from the National Adult Literacy Survey also indicates that about half of Americans lack the minimal mathematical skills necessary to use numbers embedded in printed materials (Kirsch, II Jungeblut, Jenkins, & Kolstad, 2002). For example, only 23% of participants in the work force could detennine correct change using infonnation in a menu. In addition, many individuals are insensitive to numeric infonnation, and different levels of numeric ability may lead to different judgments and risk perceptions (e.g., Peters et a1., 2009). The underlying mechanisms that lead to differential judgments and risk perception are not entirely clear. To communicate with consumers effectively, it is important to have an understanding of the underlying mechanisms of consumers' judgments and decisions. However, consumers' numeric infonnation processing, with the exception of pricing cognition, has not yet received much attention in the marketing literature. Some research demonstrate that affect towards numeric infonnation may play an important role in the use of numeric infonnation in individuals' judgments and decision makings (Peters et aI., 2006; Peters et aI., 2009). Affect Definitions of Affect Broadly speaking, affect includes discrete emotions, feelings, and mood. Slovic, Finucane, Peters, and MacGregor (2002) defined affect as the "special quality of goodness or badness." Affect is to be "experienced as a feeling state," and people experience affect rapidly and automatically, with or without consciousness. Affect is further categorized into two different types (Peters, 2008). Incidental affect is defined as positive or negative feeling (e.g., mood state) without any specific target objects. Although incidental affect is not directly elicited from a specific stimulus, it has been 12 shown to be misattributed to a stimulus (Peters et al., 2009). On the other hand, integral affect is defined as "positive and negative feelings about a stimulus that are generally based on prior experiences and thoughts and are experienced while considering the stimulus" (Peters et al., 2006). It is demonstrated that integral affect is an essential part of individuals' judgment and decision making (e.g., Damasio, 1994; Epstein, 1994; Peters, Slovic, & Gregory, 2003; Slovic et al., 2002). Integral affect and its relationship to judgment, decision making, and numeracy is the focus of this dissertation. Discrete emotions, such as anger and fear, are short-lived and more intense, and have salient cause (Forgas, 2000). Each discrete emotion provides a tendency to perceive events and objects in ways that are consistent with the cognitive-appraisal dimensions of the emotion (Lerner & Keltner, 2000). For example, Lerner and Keltner demonstrated that fearful people are more pessimistic in judging future events than angry people. Unlike emotions, moods are usually viewed as relatively low-intensity and do not have salient cause (Ekman, 1999; Forgas). Different moods with the same valence are demonstrated to have differential effects on information processing and choice tendency. More specifically, for example, individuals in happy moods, compared to those in sad moods, were demonstrated to rely more on heuristics and other easily accessible infonnation, such as stereotypes and expectations (e.g., Bodenhausen, Kramer, & Susser, 1994). Some researchers argue that emotions and moods have different functions. Davidson (1994) argued that mood biases cognition while emotion biases behavior. Other researchers suggested that, whereas emotions direct behavior and result in action tendencies (Lerner & Keltner, 2000), moods bias cognition by influencing information 13 processing--moods can hider or accentuate the accessibility of certain cognitive information. Although affect may have broader meanings in some literature, in this dissertation I use the definition developed by Slovic et al. (2002). Affect and Attitude Eagly and associates (Eagly & Chaiken, 1993a, 1993b) defined attitude as "a psychological tendency that is expressed by evaluating a particular entity with some degree of favor or disfavor" (p. 1). Several researchers (Crites, Fabrigar, & Petty, 1994; Eagly & Chaiken, 1993a, 1993b; Lutz, 1981) have discussed the tripartite view of attitude; specifically, that attitude has three underlying components-cognition, affect, and behavior (cf. conation). Cognition refers to all beliefs that an individual holds with respect to the object, and affect refers to positive or negative emotional reactions towards the object. The behavior part pertains to intended and actual behaviors with regard to the object. Based on the tripartite view of attitude, every attitude has greater or lesser degrees of each component. One major criticism of this tripartite view is that it lacks empirical evidence, and with a few exceptions (e.g., Peters & Slovic, 2007), researchers often measure only the overall attitude, particularly the affective component, rather than each component. Conceptualization and Functions of Integral Affect The role of affect in everyday decisions has received an increasing amount of attention in recent literature in the last 15 years. Epstein (1994), for example, proposed 14 that we have analytical and experiential systems that are interrelated but separable, and that we generally use both systems to make decisions. It has also been suggested that affect is central to human cognitive processing and acts as information (Damasio, 1994; Peters, Lipkus, & Diefenbach, 2006; Slovic et al. 2002; Zajonc, 1980). Zajonc claimed that all perception contains some affect, and this affect can influence the ensuing cognitive processing to a significant degree. Further, affective reactions are argued to be hardly escapable and are often the most remembered facet of an experience. Damasio argued that life experiences lead options and attributes to be "marked" by positive or negative feelings linked to somatic or bodily states. When the positive somatic marker is activated, we are drawn towards options. Whereas, when the negative somatic marker is activated, it acts as a warning signal to drive us away. Relying on somatic markers can lead to better and more efficient decisions. Slovic et al. proposed the affect heuristic, stating that we often use affect as a shortcut to guide judgment and decision making processes in our information-rich and complex world. Similarly, Peters et al. conceptualized affect as information: affect may serve as cues for many judgments, such as probability and risk. This is consistent with Damasio's somatic-marker hypothesis, and is substantially similar to models of "risk as feelings" (Loewenstein, Weber, Hsee, & Welch, 200 I) and "mood as information" (Schwarz & Clore, 1996). According to these models, affect can be experienced immediately as individuals encounter the events and objects, or it can be experienced after some cognitive processing. Reliance on affect is also thought to be quicker, less effortful, and more efficient for making decisions in a complex and uncertain world. 15 Peters et al. (2006) discussed four functions of affect. The first function is the previously discussed affect as information. The second function proposed is affect as a motivator of information processing and behavior. Stronger affect, for example, was demonstrated to lead to more effort in choosing which lottery to play (Peters et al., 2003). Another function proposed is affect as common currency. Because affect is much simpler than cognition, affect allows us to compare apples to oranges (Cabanac, 1992). Peters et al. also proposed that affect plays a role as a spotlight in a 2-step process (cf. affect as spotlight). First, the extent or type of affective feelings focuses consumers on certain new information, and then the new information is processed to guide their further judgments or choices. Once people experience affective states, the affective states influence subsequent information searching and infOlmation processing. Peters et al. provided an example that a cancer patient who is worrying and hence in a negative affective state may spend more time looking at risk infonnationrather than benefit information of a certain treatment. Affect and Memory Some research suggests that affect plays an important role in the formation and retrieval of memory (Graf & Mandler, 1984; Graf, Mandler, & Haden, 1982; Posner & Snyder, 1975). Zajonc (1980) discussed that, although the cognitive basis of affective reactions may be forgotten, the affective reaction itself can be dissociated from its cognitive basis and still be retJieved. Kida, Smith, and Maletta (1998), for example, found that affective memories of responses to numerical data, compared to memories of actual and approximate numerical data, were the most enduring and accessible. This 16 suggests that when we try to recall events, people, or other objects, the affective quality is among the first elements to emerge, and its emergence can occur with very little effort. For example, you might not be able to remember the details of the product description in ads you saw last week, but you probably automatically remember whether or not you liked the product and the ads. Often, we use this affect to further guide our judgments and decisions. For example, although you do not remember the details of ads and the product, you may be more likely to choose the advertised product over its competitors if you remember liking the product and the ads. Affect and Numeric Information As discussed earlier, recent research on affect and social decision making suggests that affect is an essential part of preference formation, judgment formation, decision making, and more. However, until recently, numerical information was considered to be processed purely cognitively and, thus, free from affect. A few exceptions exist. Participants in Slovic, Finucane, Peters, and MacGregor's (2004) study were asked to rate how attractive a bet is. Half of the participantsreceived a simple bet- 7 out of36 chances to receive $9, or otherwise win nothing. The remaining participants received a bet in which they may lose a small amount of money-7 out of 36 chances to receive $9, and 29 out of 36 chances to lose $.05. Slovic et al. demonstrated that participants rated the bet with a chance to lose $.05 more attractive than the bet with no loss. This is an interesting finding because it violates economic theory that people should prefer a bet with the highest expected return. Peters et al. (2006) extended this study by demonstrating that this effect is driven by highly numerate individuals: only the highly 17 numerate group rated the bet with a chance to lose significantly more attractive than a bet without a loss. In addition, Peters et al. demonstrated that highly numerate individuals, compared to those who were less numerate, had a clearer feeling about the goodness or badness of their feelings toward numeric information. Peters et al. argued that highly numerate individuals are more likely to draw affective meaning from probability (e.g., 7/36 chances) and numeric comparison (e.g., $9 and $.05). In evaluating the bet, highly numerate individuals may find $9 very attractive in the presence of a $.05 loss compared to $9 alone because $9 has much higher value than $.05. The results from Peters et al. (2006) suggests that highly numerate individuals were more likely to deliberate about and compare numeric quantities, and hence develop a more precise affective reaction to the numeric information. The results from this study suggested that highly numerate individuals may better understand and use the numerical information in product claims and other marketing communications. On the other hand, less numerate individuals' choices have been demonstrated to be influenced by incidental affect (i.e., their moods) when choosing an option that was described with numeric information (Peters et aI., 2009). This suggests that when presented with numeric and non-numeric information, less numerate individuals may rely more on non-numeric information in their judgments and choi ces. In the context of marketing communications, less numerate consumers may rely more on non-numeric information in evaluating a product, such as brand images, pictures in advertisements or wording of product claims. As was argued earlier, consumers often are faced with judgments and decisions that involve numeric information. And the numeric information mayor may not be 18 available to them at the time of their judgments and decisions. Prior research demonstrated that individuals tend to believe information that has been presented to them, regardless of its actual truthfulness (Illusion-of-Truth Effect; Skurnik et al., 2005). Although individuals are often faced with much numeric infonnation, the effect of Illusion-of-Truth on numeric information has never been tested. Because numeric and non-numeric information are processed differently (Gurmankin et al., 2004), and many people are innumerate (Kirsch et al., 2002; Lipkus et al., 2001; Paulos, 1988; Woloshin et al., 1999), the effect of Illusion-of-Truth may work differently for numeric and non- numeric information. The Illusion-of-Truth Effect It has been shown that people are not particularly good at judging truthfulness (cf. accuracy) of product claims when they are asked to rely on their memory (Hasher, Goldstein, & Toppino, 1977; Hawkins & Hoch, 1992). Hawkins and Hoch examined how participants' level of involvement during exposure to consumer information influenced what they learned and what they subsequently came to believe. They found that low- involvement information processing and repetition leads to an increase in the truth effect. Truth effect occurs when individuals are more likely to believe information presented to them is true than false. Familimity was found to mediate this truth effect; the more familiar the information, the more believable it is for participants. Skumik et al. (2005) showed that perceived familiarity leads to credibility even when product information has 19 been explicitly identified as false. They argued that participants lose contextual information (e.g., claim is true or false) or connection between two pieces of information (cf. product information and truthfulness) after some time, and they simply remember having seen the product information. Therefore, participants tend to judge familiar claims as true regardless of actual truthfulness. This is the lllusion-of-Truth effect (Skumik et aI., 2005). Although the lllusion-of-Truth effect has been examined in several research studies, no study has focused on the relationship between numeric information and the lllusion-of-Truth effect. In everyday life, we often encounter product claims that use numeric infOlmation, and we sometimes have to rely on the infOlmation from memory to make our judgments and decisions regarding products and product claims. It has been shown that numeric and nonnumeric infonnation are processed differently (Gurmankin et aI., 2004). This suggests that numeric and nonnumeric information may be stored as two sets of information rather than one set of information. The connection between the two sets of information may be lost after some time. Because many individuals are not very good with numbers (Kirsch et al., 2002; Paulos, 1988), this may lead them to rely more on the nonnumeric part of product claim and neglect the numeric part of product claim. Drawing from the literature on affect and memory, consumers may also rely on affect when they are faced with a truth judgment: if they remember positive feelings about a certain product, for example, they are more likely to believe the favorable statements about the product. Given that highly numerate individuals are more likely to draw affective meanings from numeric information, they may be better able to use 20 affective infonnation derived from numeric infonnation (e.g., good or bad feelings about the meaning of numeric infonnation in a given context) in their truth judgments (Peters et al., 2006). They can do so by checking the consistency of affective infonnation derived from the numeric infonnation encountered earlier and the affective infonnation derived from the claims they need to make truthfulness judgments. If they are consistent, then the claims may seem more likely to be true. However, if they are not consistent, then the claims may seem to be false. For example, ifhighly numerate consumers are presented with the claim that "90% of consumers believed Diet Pepsi tastes more like real cola than Diet Coke," then they are likely to draw positive affective meanings about Diet Pepsi in tenns of its cola taste from the numeric infonnation. Therefore, when faced with the statement that "most consumers believed Diet Pepsi tastes more like real cola," highly numerate consumers are likely to believe the claim because they remember their positive affect towards Diet Pepsi in tenns of its cola tastes. On the other hand, if they were presented with the claims that "35% of consumers believed Diet Pepsi tastes more like real cola than Diet Coke," then they are likely to draw negative affective meaning from the numeric infonnation. As a result, because they remember having negative feelings about Diet Pepsi in tenns of its cola taste, they may be less likely to believe the claim "consumers believed Diet Pepsi tastes more like real cola." The focus on this dissertation is the role of numeracy and affect in truth judgments (Studies I and 3) and product evaluations (Studies 2 through 5). More specifically, the dissertation examines how numeric ability influences the ability to draw 21 affective meanings from numeric information, and in turn influences subsequent judgments, such as truth judgments and product evaluations. Hypotheses One of the main goals of this dissertation is to test the evidence of Illusion-of- Numeric-Truth effect using product claims that contain important numeric information. In the literature, an Illusion-of-Truth effect is observed when familiar claims are judged as true although they were originally presented as false (Skurnik et aI., 2005). In this study, an Illusion-of-Numeric-Truth effect would be observed if participants incorrectly judged false claims as true if the numerical meanings of the claims were translated or remembered inaccurately from the original claims (e.g., if"30% of consumers" was inaccurately translated to "most consumers"). I will also investigate the relationship between the Illusion-of-Numeric-Truth effect and numeracy. It is hypothesized that highly numerate individuals are less susceptible to the Illusion-of-Numeric-Truth effect because they can draw more affective meaning from numeric information than can the less numerate individuals (Gurmankin et aI., 2004; Peters et aI., 2006). HI a: Participants are more likely to judge inaccurate numeric claims as true than judge new claims as true if claims seem familiar. HIb: Highly numerate individuals are able to make more accurate truth judgments than less numerate individuals. Highly numerate individuals have been found to be better at using numeric information in their judgments (e.g., Peters et aI., 2006). In some experiments, ------------------ - - ---- 22 participants were asked to judge the favorability of the target products that were described by important numerical values. It is hypothesized that highly (vs. less) numerate individuals would be better able to integrate numeric information and, thus, their product evaluations would be influenced more by numeric information. When making truth judgments, highly numerate individuals are expected to be better able to rely on feelings towards a product that they developed earlier based on numeric information when engaging in truth judgments. When encouraged to process numeric information more systematically by experimental manipulation, less numerate individuals will be more likely to use numeric information in their judgments than when they were not encouraged to do so. H2a: Highly (vs. less) numerate individuals are more likely to use numenc information when evaluating a product. H2b: Highly (vs. less) numerate individuals are more likely to use their feelings towards a product when judging the truthfulness of claims. H2c: Less numerate individuals are more likely to use numeric information in their product evaluations when they are encouraged to process numeric information more systematically than when they were not encouraged to do so. Meanings ofnumeric information are often context dependent, and we often draw meanings by comparing numbers. For example, scoring 85% correct on an exam may feel better or worse depending on the average score for the exam. There is evidence suggesting that highly numerate individuals are more likely to draw affective meanings from number comparisons and use those affective meanings in their judgments (Peters et al., 2006). Therefore, it is hypothesized that highly numerate individuals are more likely 23 to draw affective meanings from comparing numeric information and to use the meanings in their product evaluations, whereas less numerate individuals will be less likely to draw meanings from number comparisons. H3a: Highly (vs. less) numerate individuals are more likely to draw affective meanings by comparing numbers. 24 CHAPTER III STUDIES Overview The main goals of this dissertation are to investigate (a) if and how consumers understand and use numeric information in their truth judgments and affective product evaluations; (b) if and how numeracy influences the understanding and the use of numeric information; and (c) how consumers, especially less numerate consumers, can be helped to use consequential numeric information more in their affective product evaluations. Five studies are proposed to accomplish these goals. Scenario-based surveys were used in all five studies. In general, patiicipants read scenarios that include fictitious product claims, and they answered various questions. All participants were college students recruited from the Psychology and Marketing Departments at the University of Oregon. Studies 1 and 3 rely on a Signal-Detection-Theory paradigm to test an Illusion-of- Numeric-Truth effect (Hypotheses la and lb). Participants were shown a series of numeric product claims that contained either an unfavorable or a favorable numerical value in percentage format. Later, they were asked to judge if the gist of the claims was true, false, or new. It was hypothesized that participants would be more likely to believe that inaccurate numeric claims were true if claims seemed familiar, and further, highly 25 numerate individuals would be better able to make accurate truth judgments than less numerate individuals. Studies 2 and 3 tested whether highly numerate individuals would be better able to use numeric information and, thus, would be influenced more by numeric infonnation in each product claim (Hypothesis 2a). Participants were shown product claims containing different levels of important numeric information, and were asked to judge the favorability of the target claims. In addition, Study 3 tested whether highly numerate individuals' affective product evaluations and truth judgments were associated (Hypothesis 2b). Study 3 also investigated the influence of numeracy on the use of numeric information in affective product evaluation. Study 4 aimed to help participants, and especially less numerate participants, use numeric information more in their product judgments by using methods thought to increase depth of processing (Hypotheses 2c). It was hypothesized that less numerate, as well as highly numerate, individuals would use numeric information more when they are encouraged to engage in more systematic processing of numeric infonnation by presenting the numeric information in a hard-to-read font (Study 4). Study 5 tested whether highly numerate individuals would be influenced by additional numeric information that is not necessarily diagnostic to the affective product evaluations they are making (Hypothesis 3a). More specifically, it tested whether highly numerate participants draw meanings by comparing two ratings expressed as numbers of stars-one for target product and another for the accompanying product-and use the meanings they draw by comparison in their affective product evaluations. Less numerate 26 individuals, on the other hand, are hypothesized to be less influenced by the additional numeric infonnation (Hypothesis 3b). Study 1: Numeric Memory for a Product Claim Study 1 used a Signal-Detection-Theory paradigm to test an Illusion-of-Numeric- Tmth Effect. It was hypothesized that participants would be more likely to believe that inaccurate numeric claims were hue if the claims seemed familiar (Hypotheses 1a), and further, that highly numerate individuals would be able to make more accurate tmth judgments than less numerate individuals (Hypotheses 1b). Method Design. Study 1 was a mixed design. Claim type (Tme, False vs. New) and numerical values (Unfavorable vs. Favorable) were within-subjects factors and numeracy was a between-subjects factor. In the infonnation phase, participants were shown a series of product claims with numeric information in a percentage format. Thirty-six claims were divided into three groups of 12 claims each. For each participant, one third of claims were randomly paired with a percentage ranging from 35% to 45% indicating unfavorable numeric infonnation, and another third of claims were randomly paired with a percentage ranging from 75% to 85%, indicating favorable numeric information. The rest of the claims were never used in the infonnation phase. The claims were worded in such a way that it was always more favorable to have a higher numerical value. In the test phase, participants were presented with a series of product claims including 12 new claims and 24 claims previously seen during the information phase. 27 New claims had not been seen by participants. In the 24 claims from the information phase (with unfavorable or favorable numeric information), the numeric information was replaced by the word "most." Therefore, modified claims were always accurate for those claims with favorable numerical values in the evaluation phase, and inaccurate for those claims with unfavorable numerical values. Stimuli development and pretest. Forty fictitious product claims were created using real product names currently available on the market. All of the products in the claims were consumer products and included beverages (e.g., Diet Pepsi, Samuel Adams Beer), automobiles (e.g., Ford F-150), banks (e.g., Bank of America), and airline companies (e.g., United Airlines). A pretest of the claims was conducted with 68 psychology students (see Appendix A). Each ofthe participants rated their attitude towards 40 products on a 7-point scale ranging from 1 to 7 (-3 was very unfavorable and 3 was very favorable). Fourteen products that received a favorability score of above 5 points were either deleted or replaced with less popular products (e.g., Hilton was replaced with Hampton Inn) in order to avoid using products towards which participants had a strong preexisting attitude. In the end, 36 product claims were retained (see Appendix A). Procedure. Participants were 150 psychology students. They received a link to the study programmed with online survey software Qualtrics, and could take the survey at the time and place of their choosing. After the consent fonn, each participant received 44 product claims one at a time and were asked to engage in a low-involvement comprehension task (Hawkins & Hoch, 1992; Lichtenstein & Srull, 1987) by rating how 28 easy or difficult it was to understand each claim on a 7-point scale ranging from 1 (very easy) to 7 (very difficult) (see Appendix B). Among the 44 claims in this evaluation phase, 12 were described with unfavorable numerical values and another 12 were described with favorable numerical values. The remaining 20 claims were filler items and appeared in a random order among the target claims. The format of these filler claims differed from the format ofthe target claims in order to provide the participants with variety (see Table 1). The first two claims were always filler claims so as to reduce a primacy effect (Law, Hawkins, & Craik, 1998). Following this evaluation phase and after some unrelated tasks that took approximately 15 to 20 minutes, participants proceeded to the test phase. During the test phase, each participant received 38 claims without any numeric information. The first two claims were filler claims. Of the remaining 36 claims, 24 were modifications of the earlier claims presented in the evaluation phase, and 12 claims were completely new. Among the modified claims, 12 were accurately based on the earlier claim described with favorable numerical values, and another 12 were inaccurately based on the earlier claim that was described with unfavorable numerical values. Participants were informed that some of the claims were accurately based and some of the claims were inaccurately based on the claims they saw earlier in the evaluation phase of the study. They were also informed that some of the claims were never presented to them during the study. For each claim presented, they identified whether it was true (i.e., accurately based on the earlier claim), false (i.e., inaccurately based on the earlier claim), or new. They also completed a various demographic 29 questions and numeracy scale, in which participants attempt to solve 11 math~related questions (Lipkus et a!., 2001; Peters et a!. 2006) (see Appendixes G and H). Table 1 Examples ojProduct Claim Target claims in evaluation phase In a double-blind taste test, consumers tasted two cola drinks with a bite of cracker or sip of water before each tasting. Among these consumers, 35% believed that Diet Pepsi tasted most like real cola. 85% of consumers prefelTed the original Nestle's Crunch over the new Snickers Cruncher. Filler claims in evaluation phase A study suggested that drinking eight ounces of cranberry juice cocktail at a time may be better than drinking four ounces for women trying to prevent a bladder infection. Modified target claims in test phase* Most people in double-blind taste test believe that Diet Pepsi tastes most like real cola. Most consumers prefer the original Nestle's Crunch over Snickers Cruncher. *Gist of claims is always consistent with claims with favorable numerical value. Results Recognition performance. Participants' response options were "new," "true" and "false," and these responses were coded based on accuracy. The percentage of COlTect judgments was calculated for each claim type (true, false, and new) for each participant. 30 Participants correctly identified repeated claims 87% of the time. A paired sample t-test was conducted with percent correct for new and repeated items (i.e., the hit rate for new items and correct rejection for repeated items). It showed that participants were better at correctly identifying repeated claims (M = .87) than identifying new claims (M = .54, t(150) = -11.9,p < .001). This is consistent with previous research (Hawkins & Hoch, 1992). Truth judgment. The basic Illusion-of- Truth effect was observed (see Table 2). The results of a repeated-measures ANOYA demonstrated that repeated claims (both true and false, M = .66) were judged as true significantly more often than new claims (M = .25, F( 1,151) = 262, p < .001). This suggests that participants are more likely to believe the familiar claims than unfamiliar claims. An Illusion-of-Numeric-Truth effect was also demonstrated. Among the repeated claims, participants accurately judged true claims as true (M= .68) more often than false claims as true (M= .64), F(l,149) = 6.4,p < .01). Further, participants were more likely to inaccurately judge false claims as true (M = .64) than new claims as true (M = .25, F(l, 151) = 231, P < .001). These results support previous Endings ofthe IlIusion-of-Truth effect (Skumik et al., 2005): participants were more likely to believe inaccurate claims as true if they were familiar (Hypothesis 1a). These results also indicated that although participants could correctly judge true claims as true in general, they were likely to judge claims as true ifthey had seen them before regardless of the actual accuracy. Numeracy and truth judgment The mean numeracy score was 8.4 (median = 9) out ofll possible (range = 1-11, a = 67). Because the distribution was highly skewed (= 31 Table 2 Percentages a/Participants Who Judged Truefor Each Type ofClaims False New claims Repeated True claims claims Less numerate participants 30 65 65 65 Highly numerate participants 21 68 71 65 All Participants 25 66 68 65 -1.2, Standard error of skewness = .20), a median split was perfonned (Peters et aI., 2006), therefore participants with numeracy scores of 9 or lower were coded as less numerate (M = 6.7, SD ~~ 1.8) and those with numeracy scores of 10 or 11 were coded as highly numerate (M = 9.7, SD = .8). Results from a repeated-measures ANOYA indicated that both less and highly numerate individuals were more likely to judge repeated claims as true (M = .65 and M = .68 respectively) than new claims (M = .30 and M = .21 respectively) as true (F(I ,149) = 267, p < .001) (see Figure 1). For individuals low and high in numeracy, the proportion of true ratings was higher, on average, for false claims (65% for both less and highly numerate individuals) than for new items (30% for less numerate individuals: F(l ,65) = 73, p < .001, 21 % for highly numerate individuals: F(l ,83) = 179, P < .001; see Figure 1). This indicates that both less numerate and highly numerate individuals are susceptible to the Illusion-of-Numeric-Truth effect. In order to assess the Illusion-of-Numeric-Truth effect further, a measure of d' from signal detection theory was calculated (Law et aI., 1998; Singh & Churchill, 1986). The d' value is a summary report of each participant's truth judgment perfonnance. In 32 100 71 65 ...+ .. Less numel'ate participants --- Highly numerate participants 21 o Newclairns True claims False claims Figure J. Percentages of new and false claims judged true by the less numerate and the highly numerate groups. order to calculate d " the hit rate (HR) and the false-alarm rate (FAR) for each participant was calculated first. The HR is the proportion of times that participants accurately judged true claims as true; the FAR is the proportion of times they inaccurately judged false claims as true. In order to compute d', HR and FAR values of 0% and 100% were converted to 1% and 99% (Law et a1., 1998; Macmillan & Creelman, 1991). Then the d' was calculated using a formula /ar - iI', where /ar is the standardized score for the FAR and zhr is the standardized score for the HR. Because d' represents the difference between standardized HR rate and FAR, a larger number indicates better truth judgment performance. Results of an ANOVA demonstrated that d' scores were higher for individuals higher in numeracy (d' = .25) than for those lower in numeracy (d' = .03, t(148) = -1.8,p < .05). These results indicate that highly numerate individuals were better at truth 33 judgments than less numerate individuals (Hypothesis 2). Results of a one sample (-test indicated that, although d' for highly numerate individuals was significantly different from zero «((83) = 3.5, p < .00 I), d' for less numerate individuals was not significantly different from zero «((65) = A,p = ns). This result indicates that the accuracy of truth judgments made by individuals lower in numeracy was not significantly better than chance. Discussion Consistent with previous research, our participants more often judged repeated claims as true (whether they were hue or false) than they judged new claims as true. In addition, they demonstrated an Illusion-of-Numeric-Truth effect: they were more likely to judge false claims as true than new claims as true. This suggests that people rely on the familiarity of the non-numeric part of the claims to judge the truthfulness ofthe claims. Individuals' numeric ability seems to influence their Illusion-of-Numeric-Truth effect: highly numerate individuals were significantly more successful at judging true claims as true compared to the less numerate individuals. Prior research demonstrated that highly numerate individuals were better at drawing meaning from numeric information (Peters et aI., 2006). Therefore, the meaning of numeric information may be more readily available for them at the time of truth judgments. Judgment accuracy ofthe less numerate individuals was not significantly different than chance. It may be that less numerate participants were less sensitive to numeric information and relied more heavily on the familiarity of non-numeric parts of the claim to judge truthfulness. 34 It is possible that our respondents did not accurately understand the meaning of most. Therefore, a follow-up study was conducted using a similar sample population (n = 130). Participants were asked, in order for the claim ("Most consumers prefelTed Levantra over Phemanide") to be accurate, "What is the smallest percentage of consumers who must have preferred Levantra over PhemanideT The results demonstrated that 95% of participants reported that a minimum of 50% of consumers should prefer Levantra over Phemanide in order for the claim to be accurate. When analyzed separately for less numerate and highly numerate groups, 100% of the highly numerate group stated that 50% ofconsumers should prefer Levantra, and 92% of the less numerate group stated the same. Among 8% of less numerate participants who did not report that 50% of consumer should prefer Levantra, 4% (n = 3) stated that 49% should prefer Levantra. Therefore, only 4% ofiess numerate participants stated a number below 49%. This suggests that both less and highly numerate groups have a basic understanding of what "most" means. Study 1 successfully demonstrated that less numerate individuals were more susceptible to an IIlusion-of-Numeric-Truth Effect. However, the mechanism underlying this effect is still unclear. It is unlikely that participants remembered the exact numeric information because each participant saw more than 24 pieces of numeric information in . a very short amount of time. However, they might have remembered their affective reaction to the product. A number of studies have shown that people can develop affect towards numeric infonnation and later use this affect in making their decisions (e.g., Kida et aL, 1998; Peters et aL, 2006). Further, Kida et aL demonstrated that affective memories 35 of numerical data, as opposed to memories of actual and approximate numerical data, were the most enduring and accessible. This suggests that when decision makers cannot access either the actual or the approximate numerical data, they seem to construct memories that are consistent with their affective responses. In return, they use affect to make choices. This suggests that our participants may have developed affect while reading the product claims, then, in turn, used this affect in their truth judgments. Highly (vs. less) numerate individuals were also shown to develop clearer feelings towards numeric information (Peters et al., 2006). Therefore, affective responses to numeric information may be more accessible to highly numerate individuals than to less numerate individuals. Study 2 tests ifhighly numerate participants, compared to less numerate participants, are more likely to use numeric information in developing affect and, hence, whether their affect towards products are more influenced by numeric information. Study 2: Use of Numeric Information in Affective Product Evaluation Study 2 tested ifhighly numerate individuals, compared to those lower in numeracy, would be better able to use numeric information and, thus, their affect towards product would be more influenced by numeric information in each product claim (Hypothesis 2a). 36 Method Design. Study 2 was a mixed design. Numerical values (unfavorable vs. favorable) were repeated as within-subjects factors and numeracy was between subjects. Each product claim contained two fictitious products and three pieces of numeric infonnation (see Appendix C). Numeric infonnation in each claim was always either between 35% and 45% (unfavorable) or between 75% and 85% (favorable) as in Study 1. Each claim contained unfavorable numeric values for halfof the participants and favorable numeric values for the rest of the participants. As in Study 1, favorable numeric infonnation was always preferable to unfavorable numeric infonnation. Procedure. Data were collected in a computer lab from 92 college students. Participants were infonned that they would be presented with claims that compared two products in order to make sure that any values above 50% in the product claims indicated favorable numeric infonnation. Then, each participant was shown six fictitious product claims on a computer screen, one at a time, and asked to rate their affect towards each product on a 7-point scale ranging from 1 (very unfavorable) to 7 (very favorable). At the end of the study, they completed the same demographic questions and numeracy scale used in Study 1 (see Appendixes G and H). Results The mean numeracy score was 9.4 (median = 9.5) out of 11 possible (range = I- ll, Cronbach's a = .67). It was somewhat negatively skewed (skewness = -.69, standard error of skewness = .25), therefore participants with numeracy scores of 9 or lower were 37 coded as less numerate (M = 8.2, SD = .95) and those with numeracy scores of 10 or 11 were coded as highly numerate (M = 10.5, SD = .50). A repeated-measures ANOVA was conducted on the favorability scores with unfavorable (between 35% and 45%) versus favorable (between 75% and 85%) numeric infom1ation claims used as a repeated measure (see Table 3). The median split of numeracy scores was entered as a between-subjects factor. A significant main effect revealed that claims with high values (M = 5.4), were judged significantly more favorably than those with unfavorable numeric values (M= 3.8, F(l,90) = 104.5l,p < .05). A significant interaction with numeracy was also found (F( 1,90) = 4.24, p < .05) (see Figure 2). The means indicate that the difference between highly numerate participants' judgments of favorability towards products with favorable numeric information and claims with unfavorable numeric infonnation was larger than for less numerate participants (MdijJCI"ff1ce = 1.2 for less numerate and Mdi(Jerence = 1.9 for highly numerate; Hypothesis 2a). The additional test revealed that highly numerate participants' feelings towards products with unfavorable numeric information (M = 3.6) was significantly below the midpoint (cf. labeled as "neutral," t(45) = -2.57,p < .05) and their feeling towards products with favorable numeric information was significantly (M = 5.5) above the midpoint (t(45) = l2.3,p < .001). For the less numerate participants, on the other hand, only feelings towards products with favorable numeric infom1ation (M = 5.3) were significantly different from the midpoint (t(45) = 9.97,p < .001) whereas feelings towards products with unfavorable numeric information (M = 4.1) were not significantly different from the midpoint (t(45) = .5). Table 3 Significance Test From ANOVA and Mean Affect Towards the Products by Numeric Conditions and Numeracy 38 Numeric levels (df= 1,90) F= 104.5 (p = .001) Unfavorable numeric 3.84 (1.03) Favorable numeric 5.38 (.85) Interaction with Numeracy F= 4.24 (p = .042) Less numerate Highlj' numerate Less numerate Highly numerate 4.08 (.96) 3.61 (1.06) 5.30 (.90) 5.46 (.80) Parentheses indicate the standard deviations of the mean 7 2 ...... 4.1 ............ ...... ..... 3.6 Unfavorable numeric information 5.5 5.3 Favorable numeric information -+-Low numerate group ___ Highly numerate group Figure 2. Affect towards product with unfavorable and favorable numeric information reported by the less numerate and the highly numerate groups. In a follow-up question at the end of the study, participants were asked how important the numeric and the non-numeric information were to their affective product evaluation. Results of a MANGVA revealed a significant effect of numeracy (F(2,89) = 39 3.5,p < .05). An examination of the means suggested that less numerate participants reported that numeric (M = 4.9) and non-numeric (M = 4.3) parts of the information were equally important (t(45) = 1.7,p = ns); whereas highly numerate participants reported that the numeric part (M = 4.8) of the information was more important than the non- numeric part of the information (M= 3.6, F(2,89) = 3.5,p < .05). Although less and highly numerate participants reported numeric information as equally important, the less numerate participants relied on it less in the lllusion-of-Numeric-Truth effect demonstrated in Study I and in developing feelings towards products as shown in Study 2. Discussion Study 2 demonstrated that highly (vs. less) numerate individuals successfully relied on numeric information more in developing their feelings towards products. This may lead highly numerate individuals to better judge the truthfulness of claims. More specifically, highly numerate individuals may be better at truth judgments because, when making truth judgments, they can use the favorability judgments they made earlier. It may be that they pay more attention to numeric information and translate the numeric information into favorability judgments, and in turn use feelings of favorability in truth judgments. The next study investigates the use of affect towards products among highly numerate individuals when they are asked to make t~th judgments (Hypothesis 2b). It also tests whether less numerate participants become more sensitive to numeric information when they are encouraged to process numeric information more systematically. 40 Study 3: Use of Numeric Information in Affective Product Evaluation and in the Illusion-of-Numeric-Truth Effect In Study 3, I attempt to replicate the basic findings of Studies 1 and 2, and test whether highly numerate individuals are more likely to use their previous favorability judgments when they are asked to judge the truthfulness of a claim (Hypothesis 2b). A fluency of numeric information was manipulated by changing how easy or difficult it is to read numeric information in the text in order to encourage participants to process information more systematically (Schwarz, 2004). Metacognition of disfluency also appears to reduce the impact of heuristics and can activate analytic information processing. Therefore, it was hypothesized that participants, especially less numerate participants, would be more likely to use numeric information in their product judgments if the numeric information was hard to read (cf. disfluent, Hypothesis 2c). Method Design. Study 3 had two between-subject manipulations: two levels of numeric infonnation (unfavorable and favorable) and two levels of fluency or font readability (fluent and disfluent). Numeric information in the claim was either between 25% and 30% (unfavorable) or between 70% and 75% (favorable). Unlike in the previous study, crucial numeric information was presented outside of product claims in a separate table in order to test if participants would still use numeric information in their judgments even when it required extra effort to look up (see Appendix D). For half of the participants, all of the infonnation in the table (cf. summary of claims and numeric information) was in an easy-to-read fluent font (16-point Arial), and for the other half ofparticipants, the 41 infonnation was in a hard-to-read disfluent font (50% gray italicized IS-point TypoUpright BT). The choice of font was based on a pretest, in which participants judged (a) if each of various fonts was readable, and (b) how easy or difficult it was to read each font. The gist of each claim was written in bold. Procedure. Two hundred thirty-one college students participated in this study. At the beginning of the study, participants were explicitly infonned that they would be presented with product claims comparing two products. Then each participant was presented with the fictitious product claim used in Study 2 on a computer screen. Participants were asked to rate how they felt about the target product on a 7-point Likert- type scale ranging from very unfavorable (1) to very favorable (7). After a 1- to 3-minute distracter task, they were asked to engage in truth judgments similar to those of Study 1. In this study, however, no new claims were presented for truth judgments. Participants were then asked various follow-up questions assessing their affect towards each product name and the self-rated importance of the product name and numeric infonnation in their judgments. Finally, they were asked to complete the numeracy scale used in previous studies and demographic questions (see Appendixes G and H). Results The mean and median numeracy score was 9.0 (Cronbach's a = .68). Because the distribution was negatively skewed (-1.3), a median split was used in the subsequent analysis. The fluency manipulation did not have any significant effects on affective product evaluations (F(l ,223) = .14), and it was, therefore, excluded from further analyses. The manipulation offavorability of numeric infonnation was significant 42 (F(1 ,227) = 76.5, p < .001): a product described with favorable numeric information was judged more favorably (M = 5.4) than one described with unfavorable numeric infonnation (M = 4.0, see Table 4). Although the main effect of numeracy was not significant (F(9,227) = 2.4,p = ns), the interaction between numeracy and favorability of numeric infonnation was significant (F(8,227) = 6.2, p < .05). As demonstrated in Study 2, means indicated that highly numerate participants' feelings were more sensitive to numeric infonnation (M = 3.7 in the unfavorable numeric condition and M = 5.5 in the favorable numeric condition, respectively) than less numerate patiicipants (M = 4.3 and M = 5.3, respectively) (see Figure 3). Table 4 Significance Levels (d/'= (1, 227)) Favorability 76.5 (p = .001) Numeracy 2.4(p=.13) Interaction 6.2 (p = .013) Truth effect. The results also replicated the basic truth-effect findings from Study I: highly numerate participants were better at judging the truthfulness of claims than less numerate participants (F(18,396) = .84,p = .01). Unlike in Study 1, however, both less numerate and highly numerate groups were equally able to correctly identify the true claim (92% and 89% respectively). In addition, highly numerate participants (60% correct judgments) were better at correctly identifying false claims than less numerate 43 7 VI .... u :J 1J o ... 0- (1) .c. .... VI 1J ... ro :: o .... .... u (1) :::