Changing the world 2 words at a time
Researching the effects of alternative food expiration labels
and expiration label education on consumers’ risk perception.
Changing the world 2 words at a time
Researching the effects of alternative food expiration labels
and expiration label education on consumers’ risk perception.
Author: Anke van den Brink
Het Uding 27
7496 DC Hengevelde Tel: +31652889500
ankevandenbrink@hotmail.com Student number: 2546574
University: University of Groningen
Faculty of Economics and Business – Department of Marketing Qualification: Master thesis
MSc. Marketing Intelligence
Research theme: Alternative expiration dates and education
Supervisors: First supervisor: Dr. J. Van Doorn Second supervisor: Dr. F. Eggers
Preface
This report was written as a master thesis for the study MSc. Marketing at the University of Groningen. This topic was selected after numerous struggles and family debates on whether or not to still eat some leftovers found in the fridge. As it turned out, this was not only an issue we were facing, but millions of people in the world were asking themselves the same question. Throughout this research I enjoyed learning more about the topic, and about my own capabilities.
Writing this thesis was a very challenging process, but in the end I can say I am happy with the final paper that is presented here for you. All of this would not have been possible without the help and guidance from my supervisors, dr. J. Van Doorn, and dr. F. Eggers. I also would like to thank my thesis group for providing me with support and feedback, and all the respondents of my surveys. Finally, a special thanks to my family and friends for their consistent support and believe in me throughout my studies.
Kind regards,
Summary
This research explorers two possible solutions to decrease consumer’s level of perceived risk of perishable food products two days after their expiration date. Products carrying a ‘best before’ expiration label can still be safely consumed within a reasonable amount of time, but consumers perceive them as being risky and are therefore less likely to consume them. The first proposed solution is to change the ‘best before’ expiration label wording to ‘quality guaranteed till’ or ‘preferably use before’. The second proposed solution is to provide consumers with education about expiration labels.
A conjoint analysis was used in this study. Data was gathered by means of two surveys presenting three alternatives. The first two alternatives consisted of one of the three aforementioned labels, and either low-fat yoghurt or chocolate & vanilla custard representing virtue and vice product types respectively. Vice foods are pleasurable foods but with negative long-term effects, virtue products are more beneficial on the long-term but are not always as pleasurable. Vice foods are expected to have a higher level of perceived risk. Taste is very important for these products, and older products do not taste as good as fresher ones. The third alternative was a none-option indicating both presented alternatives are perceived as being unsafe. Consumers were asked to indicate the option they perceived to be the safest.
The results show that both label and product type do not significantly affect consumers’ level of perceived risk. Consequently, no inferences can be made as to which label or product type is perceived safer by consumers. The none-option, however, is significant which is in conflict with other consumer statements saying they perceive the options to be safe. A possible explanation is that consumers want to avoid having to select one option and therefore selected the none-option. The effect of education on perceived risk is as well non-significant. Consumers in the with-education survey only took approximately 5 seconds longer to complete the survey, which may indicate they did not carefully process the information and thus mitigating the effects of education.
Limitations to this research are that the results may not be generalizable to other countries, as label meaning linguistically differs when translated. Furthermore, only label and product were used which is a very simplistic setting. Further research is advised to ensure respondents better process information, and to use more variables to create a more realistic setting.
Table of Contents
1. Introduction ... 3
2. Literature review ... 7
2.1 Perceived risk ... 7
2.2 The expiration label ... 8
2.3 Label education ... 10
2.4 Virtue vs. vice foods ... 11
3. Conceptual model ... 13
4. Hypotheses ... 14
4.1 Expiration labels ... 14
4.2 Label education ... 16
4.3 Virtue vs. vice products ... 17
5. Research design ... 18
5.1 Method ... 18
5.2 Attributes and levels ... 19
5.3 Control variables ... 19 5.4 Research design ... 20 5.5 Model specification ... 21 5.5 Method of analysis ... 22 5.6 Data collection ... 23 6. Results ... 24 6.1 Descriptive statistics ... 24 6.2 Risk perception ... 24
6.3 Education on perceived risk ... 25
6.4 Conjoint analysis ... 26
6.5 Interaction effects ... 28
6.6 Model validation ... 29
6.6.1 Likelihood ratio test ... 29
6.6.3 Mean absolute error ... 31
6.7 Hypothesis overview ... 32
7. Conclusion and discussion ... 33
8. Theoretical and managerial implications ... 35
1. Introduction
Two words is all it takes to make European shops throw away 90 million tons of still consumable food each year (European Commission, 2014). These are the expiration labels ‘best before’ and ‘use by’. Confusion about expiration labels is mentioned by almost half of the Dutch consumers as the main reason for food waste (Onderzoeksbureau Stroom, Tetra Pak, 2015). The average Dutch consumer disposes of 7kg of eatable food, not including peels, bones, etc., per year (Soethoudt, Sluis, Waarts, & Tromp, 2013). This may not sound like much, but when multiplied by 16.8 million people, this results in over 117 million kg of unnecessary food waste in the Netherlands alone.
It is interesting to see how so few words can have so much impact on the way consumers handle food products. To better understand the impact of expiration labels on consumer behaviour, the expiration label meaning and the reasons why it influences consumers’ behaviour will be discussed.
The ‘best before’ label indicates the date until the product keeps is specific properties, such as taste and colour. After passing the expiration date, the product can still be safely consumed within a reasonable time frame (European Commission, 2011). The ‘use by’ label on the other hand is a safety indicator used for highly perishable food products, and the product should not be consumed after passing this date (European Parliament, Council of the European Union, 2011). A majority of the Dutch consumers, however, tends to be unaware of the difference between the two expiration labels. 53% of Dutch consumers claim to know the difference between the ‘best before’ (‘ten minste houdbaar tot’ or THT in Dutch) and the ‘use by’ (‘te gebruiken tot’ or TGT in Dutch) label, however, of the people who claim they could only 37% of the people actually can correctly identify the differences between the labels (Temminghoff & Damen, 2013).
this. These measure are, among others, avoid purchasing products close to their expiration date, or consuming products after the expiration date as they believe these products are unsafe for consumption (Yeung & Morris, 2001). Furthermore, the misconception exists that the quality of a perishable food item decreases over time, thus heightening its perceived risk toward the products’ ‘best before’ expiration date. However, the quality of these products remains relatively constant until the expiration date (Tsiros & Heilman, 2005). An increase in perceived risk thus results in a decrease of consumers’ likelihood of buying or consuming a product (Sen & Block, 2009).
Risk perception levels may even differ between product categories. Products can be categorized as either being a vice product, pleasurable foods but with negative long-term effects, or a virtue product, more beneficial on the long-term but not always as pleasurable (Van Doorn & Verhoef, 2011). Having a good taste is especially important for vice products compared to virtue products (Chernev & Gal, 2010). Taste is also an influencer of a products’ risk perception, and a bad taste increases this perceived risk (Vermeir & Verbeke, 2006). As expired products do not taste as good as fresher ones (Wansink & Wright, 2006), it is expected that consumers respondent differently to vice products compared to virtue products with regard to risk perception.
There could be a link between perceived risk and the decision whether or not to buy, consume, or dispose perishable food products. This research will examine two possible ways to influence the level of perceived risk of perishable food products. A lot of food past its expiration date can still be safely consumed, but is thrown away regardless as people perceive it as being too risky (Tsiros & Heilman, 2005; Sen & Block, 2009).
even one in every three persons (GFK-Rabobank, 2015). These proposed labels may take away some of confusion around expiration labels, and as a result decrease the level of perceived risk consumers have.
The second option explored in this paper is to provide consumers with more information and education about expiration labels. This label information will consist of information about the actual meaning of the ‘best before’ expiration label, and informing consumers how long certain products can still be consumed after their expiration date. Providing consumers with more information is a commonly mentioned suggestion to reduce the level of perceived risk (Cox & Rich, 1964; Dowling & Staelin, 1994; Murray, 1991). Additional information allows consumers to better understand and make use of expiration labels (Marietta, Welshimer, & Anderson, 1999; Hawthorne, Moreland, Griffin, & Abrams, 2006; Misra, 2007).
To see how the level of perceived risk for food items past their expiration date can be altered, the following research questions have been formulated:
1. What is the effect of changing the current ‘best before’ expiration label to ‘quality guaranteed till’ on consumers’ risk perception of perishable food products two days past their expiration date?
2. What is the effect of changing the current ‘best before’ expiration label to ‘preferably use before’ on consumers’ risk perception of perishable food products two days past their expiration date?
3. What is the difference between the effects of virtue and vice products on the risk perception of perishable food products two days past their expiration date?
4. What is the effect of providing expiration label education on the risk perception of perishable food products two days past their expiration date?
2006; Misra, 2007), and it is therefore worth it to test the effects of label education on perceived risk.
As a practical contribution, the outcomes of this research could serve as an input for decision makers. With the results the decision makers will have more information at their disposal to make decisions regarding the labels on products, and education regarding labels, in order to reduce food waste. Therefore, the outcomes of this research could potentially lead to a reduction of the amount of food wasted, which in turn may benefit consumers, industries, and nature.
2. Literature review
A lot of confusion exists among consumers about expiration labels and their meaning (GFK, 2012; Temminghoff & Vlerken, 2013). Consumers perceive products close to or past their expiration date as having an increased level of perceived risk, and are therefore less likely to consume the product (Tsiros & Heilman, 2005; Sen & Block, 2009). This section will elaborate on how consumers handle perceived risk, provide information on expiration labels and the way label information could affect consumers, and will be finalized by reviewing the effect of virtue and vice products on perceived risk.
2.1 Perceived risk
Bauer (1960), was the first to propose consumer behaviour could be seen as risk taking (Taylor, 1974). He defined perceived risk as “the expected negative utility associated with the purchase of a particular brand or product” (Bauer, 1960). Consumer behaviour is greatly influenced by perceived risk, as it is something consumers wish to minimize (Smith & Paladino, 2010). It is important for consumers to avoid making mistakes, which may cause them to rather pick a safer option than the best option (Mitchell, 1999; Tsiros & Heilman, 2005; Chen & Chang, 2012). Risk related to food products can take a variety of forms, varying from underperformance risk such as a change in flavour or texture, to more severe consequences from safety risk such as getting ill (Hoffmann, 2009; Yeung & Morris, 2001).
Since perceived risk strongly impacts consumers’ decision-making processing (Mitchell, 1999; Tsiros & Heilman, 2005; Chen & Chang, 2012), the next section will examine what the antecedents and consequences of perceived risk are.
When consumers perceive risk in the purchase of a product, the willingness-to-pay (WTP) for that product decreases (Grunert, 2005; Heiman & Lowengart, 2008; Sen & Block, 2009). This decreased WTP, as a result of consumers’ wish to avoid negative outcomes (Mitchell, 1999; Tsiros & Heilman, 2005; Chen & Chang, 2012), can be linked to the six types of perceived risk being: physical risk, performance risk, financial risk, time loss, social risk, and psychological risk (Roselius, 1971; Jacoby & Kaplan, 1972; Kaplan, Szybillo, & Jacoby, 1974). In the context of food being bought or consumed close to, or right past its expiration date, the main types of risks faced are physical risk and performance risk (Tsiros & Heilman, 2005). Physical risk is the “negative health impact on consumers, associated with a decline in food safety, associated with microbiological, chemical or technological factors”, while performance risk is “The taste and/or nutritional value of a food product is adversely affected by the food hazard” (Yeung & Morris, 2001).
To minimize the perceived risk in the buying process, consumers can take several steps. These steps include: 1) postponing their purchase, 2) purchase a well-known brand, 3) seek advice from a trusted source, or 4) in the case of perishable products search for visual and other freshness cues such as expiration dates (Tsiros & Heilman, 2005). The action steps of Tsiros & Heilman (2005), with regard to information and advise seeking, are consistent with the studies of Cox & Rich (1964), Murray (1991), Dowling & Staelin (1994), and Mitra, Reiss, & Capella (1999), who also indicate perceived risk and information search are correlated. Consumers can seek information about the supplier, ask friends or family about the service or product, observe advertisements, or seek out information provided in the media (Mitra, Reiss, & Capella, 1999).
2.2 The expiration label
Vlerken, 2013). Label interest can be covered by the fact that 64% of Dutch consumers read food labels, and of those who do 95% looks at the expiration date, while only 72% looks at the price of the product (GFK, 2012).
The research by GFK (2012) indicated that knowing the correct meaning of the ‘best before’ label does not impact the likelihood of throwing away meat when it is past its expiration date. This research specifically asked about meat products, it did not test for other food categories. Other researches found that even though the majority of Dutch consumers still consume products past their ‘best before’ date, approximately 25% disposes of a product on the ‘best before’ expiration date itself (GFK-Zuivelbarometer, 2015; GFK-Rabobank, 2015).
When looking at the expiration labelling system in the US, Canada, Australia, and New Zealand, as translations can cause labels to be perceived differently (Soethoudt, Sluis, Waarts, & Tromp, 2013), it becomes clear that they use the same ‘best before’ and ‘use by’ labels as they do in Europe (Food Standards Australia New Zealand, 2012; Canadian Food Inspection Agency, 2014; United States Department of Agriculture, 2015). Besides these two labels, the US also uses a ‘sell by’ label. The ‘sell by’ label is an indicator of how long the store should display the product, although a majority (61%) of the American consumers believes it represents the last day the product can be consumed (Tsiros & Heilman, 2005).
Most researches conducted about the impact, use, and understanding of expiration labels took place either in the US or in the UK (Tsiros & Heilman, 2005; Harcar & Karakaya, 2005; Sen & Block, 2009; Milne, 2013). Researches on labels in English can be perceived differently by Dutch consumers, as linguistically speaking the term used for ‘best before’ differs from the Dutch version ‘ten minste houdbaar tot’. Even though both express the same thing, the lower limit of durability, the English version sounds milder than the Dutch one (Soethoudt, Sluis, Waarts, & Tromp, 2013).
(Onderzoeksbureau Stroom, Tetra Pak, 2015). A majority of respondents believed that this label is a better fit, as it clarifies that many products can still be consumed after the ‘best before’ expiration date.
2.3 Label education
Consumers perceive a situation to be risky when they are unsure about the outcome of a decision. When there is risk associated with making a purchasing decision, a consumer can either do something to reduce the risk, or do something to reduce the amount at stake. In the context of grocery shopping reducing the amount at stake mostly means they would forgo the purchase (Cox & Rich, 1964). Perceived risk is said to explain phenomena such as brand loyalty, comparing alternatives, and relying on the advice of others (Stone & Grønhaug, 1993; Laroche, Nepomuceno, & Richard, 2010). These phenomena are all ways to gather more information, which is one commonly mentioned solution to reduce perceived risk (Cox & Rich, 1964; Murray, 1991; Dowling & Staelin, 1994; Mitra, Reiss, & Capella, 1999).
The solution of providing additional information to reduce perceived risk has been explored previously in the field of grocery shopping. These studies provided consumers with information about nutrition labels, and show that providing additional label information increases their understanding of these labels (Marietta, Welshimer, & Anderson, 1999; Hawthorne, Moreland, Griffin, & Abrams, 2006; Misra, 2007).
Besides increasing label understanding, providing additional label information also increases the use of labels. The research by Marietta, Welshimer & Anderson (1999) found that consumers’ use of nutrition labels increased after receiving label education. Furthermore, a study performed in 2012 showed that providing consumers with information on making low-salt choices and increasing their understanding on how to use the nutrition label, successfully reduced the average salt intake in the UK over a five-year period (Wyness, Butriss, & Stanner, 2012). Another study about calorie labelling helped consumers to better use the information and reduce their calorie intake initially, but a review of these effects five years later showed that there were no lasting significant effects (Cantor, Torres, Abrams, & Elbel, 2015).
as this results in an increase of WTP and quality perception (Fox, Hayes, & Shogren, 2002; Heiman & Lowengart, 2008; Dillaway, Messer, Bernard, & Kaiser, 2011). The effects of negative information are stronger than those of positive information, thus in the case where both types of information are provided the negative effects will have the greatest impact (Dillaway, Messer, Bernard, & Kaiser, 2011).
Searching for information as a way to reduce risk is mainly present in high involvement purchasing situations (Kujala & Johnson, 1993). In the case of a routine, low-involvement purchasing situation, such as grocery shopping, an extensive information search usually does not take place, therefore information may need to be provided to them (Beharrell & Denison, 1995; Verbeke & Vackier, 2004).
2.4 Virtue vs. vice foods
Food can be categorized as either being a virtue or a vice product (Van Doorn & Verhoef, 2011). Vice products, also known as ‘wants’, are products that provide immediate pleasure, like chocolate, but contribute to long-term negative outcomes, like weight gain. Virtue products, also known as ‘shoulds’, such as fruits and vegetables, are less gratifying and appealing now, but have less negative long-term consequences than vice products (Van Doorn & Verhoef, 2011). Virtue and vice products are usually conceptualized in relation to each other as relative virtues and relative vices. For example, “reduced-fat milk can be represented as a virtue when compared with whole milk and as a vice when compared with fat-free milk” (Chernev & Gal, 2010).
immediate gratification of consuming the product (Wertenbroch, 1998). Another type of compromise is ‘licensing’. After purchasing virtue products, consumers are more likely to purchase vice options. Purchasing a virtue option boosts a consumers’ self-concept, reducing the negative self-attributions associated with vice products. When consumers buy vice products first, the opposite effect takes place. Buying vice products first reduces consumers’ self-concept, and increases the negative self-attribution associated with additional vice product purchases (Khan & Dhar, 2006; Hui, Bradlow, & Fader, 2009). A third option is by pairing the vice product with a guilt reducing complement, such as contributing it to a good cause (Van Doorn & Verhoef, 2011).
3. Conceptual model
Based on literature discussed above, a conceptual model displaying the variables that are hypothesized to influence consumers’ risk perception was created as shown in figure 1.
H1a & H1b (+)
H4 (-)
H3 (+)
H2 (+)
Figure 1. Conceptual model
The decision whether to purchase or consume a product is a type of risk taking. “Since the outcome of a choice can only be known in the future, the consumer is forced to deal with uncertainty, or risk” (Taylor, 1974). The expiration label has an influence on consumers’ risk perception (Sen & Block, 2009), and using alternative versions of the ‘best before’ label are expected to alter this risk perception. The ‘preferably use before’ label will be tested as it does not sound as strong as the current label (Soethoudt, Sluis, Waarts, & Tromp, 2013), and the ‘quality guaranteed till’ label will be used as Dutch consumers indicated this label better fits the label meaning (Onderzoeksbureau Stroom, Tetra Pak, 2015). The type of product also plays a role, especially for vice products as taste is the most important attribute for these products (Vermeir & Verbeke, 2006), and products past their expiration date do not taste as good as fresher products (Wansink & Wright, 2006). Receiving information about labels is likely to make labels an even more important attribute in determining the level of perceived risk compared to other types of information (Suter & Burton, 1996). Finally, searching and
Alternative expiration labels
- Best before
- Preferably use before - Quality guaranteed till
being provided with information tends to lower the perceived level of risk in general (Cox & Rich, 1964; Dowling & Staelin, 1994; Murray, 1991).
4. Hypotheses
This section provides an overview of the hypotheses researched, based on the conceptual model discussed above.
4.1 Expiration labels
As a replacement for the current ‘best before’ expiration label, the alternatives ‘quality guaranteed till’ and ‘preferably use before’, as introduced earlier, are proposed to be examined.
The first alternative that is examined in this study is the ‘quality guaranteed till’ label. This label has been selected as the most suitable alternative by a study with 1026 respondents, representative of the Dutch population between the ages of 18-65 (Onderzoeksbureau Stroom, Tetra Pak, 2015). Consumers believe this label better explains that many products can still be consumed after the expiration date compared to the current label. The research stated that three in five consumers would throw away less food with a clearer label, and more than half of the respondents (54%) stated that they would be less afraid of becoming ill after consuming expired products with the ‘quality guaranteed till’ label. Furthermore, 55% indicated they would use their senses of sight, taste, and smell more often before disposing a product with the proposed label. The ‘quality guaranteed till’ label seems to reduce the level of perceived risk as evidenced by the fact that people better understand the label, perceive their chances of becoming sick to be smaller, and are less likely to simply dispose a product without inspecting the product first.
framed messages. Stronger messages are therefore likely to be perceived as more dangerous and risky compared to milder labels. Products carrying a ‘preferably use before’ label are therefore expected to have a lower level of perceived risk compared to products with a ‘best before’ label.
The proposed expiration labels ‘quality guaranteed till’ and ‘preferably use before’ are expected to have a lower risk perception compared to the ‘best before’ label, although some consumers may still favour the current label. One reason why consumers may prefer the ‘best before’ label is because they are familiar with it. This reasoning can be classified as the familiarity or affect heuristic, and can influence consumers’ behaviour when assessing risk (Fischer & Frewer, 2009). Heuristics are cues or simple decision rules to make judgements about information without thoughtful analysis of the information (Cacioppo, Petty, Kao, & Rodriguez, 1986). Familiarity or affect heuristics state that when people are familiar with something, or if they had a pleasant experience with it previously, they may select it now as well without giving the decision much thought (Fischer & Frewer, 2009). This may result in consumers selecting the current ‘best before’ label, as it is a well-recognized label with which they have experience, contrary to the new proposed labels. In the case of assessing risk, however, consumers tend to rely more on deliberate information processing instead of heuristic information processing (Fischer & Frewer, 2009). As the proposed alternative labels of ‘quality guaranteed till’ and ‘preferably use before’ are expected to have a lower level of perceived risk compared to the current ‘best before’ label as discussed above, consumers are expected to favour the new labels. After carefully assessing risks and selecting the safest option consumers may, however, return back to using heuristics as grocery shopping still remains a low-involvement activity (Beharrell & Denison, 1995; Verbeke & Vackier, 2004).
H1a. The ‘quality guaranteed till’ label will have a lower level of perceived risk compared to the current ‘best before’ label.
4.2 Label education
A lot of misconception exists among consumers about the correct meaning of the ‘best before’ expiration date (Temminghoff & Damen, 2013). Consumers’ risk perception of products increases when they are closer to their expiration date, and rises fast as soon as that date is reached, even when the actual risks do not follow the same pattern (Tsiros & Heilman, 2005; Wansink & Wright, 2006). To counter this misconception, consumers could be educated about what the labels actually mean, as this enhances their ability to correctly process the information provided on the labels. Proving consumers with additional information is one of the most commonly mentioned ways to reduce perceived risk (Cox & Rich, 1964; Murray, 1991; Dowling & Staelin, 1994; Mitra, Reiss, & Capella, 1999). Especially in the case of positive information, such as that products can be consumed longer than consumers think and they are not as dangerous as one might think, have the ability to reduce the level of perceived risk (Fox, Hayes, & Shogren, 2002; Heiman & Lowengart, 2008; Dillaway, Messer, Bernard, & Kaiser, 2011). Studies on the topic of nutrition label education have already shown positive effects in allowing consumers to better understand, and more often use the labels (Marietta, Welshimer, & Anderson, 1999; Hawthorne, Moreland, Griffin, & Abrams, 2006; Misra, 2007).
The positive effects of providing consumers with more information can be linked to the Elaboration Likelihood Model (Petty & Cacioppo, 1983) (Schumann, Petty, & Clemons, 1990), which states that the way consumers process information depends on their motivation and ability. Peripheral route processing is characterized by low motivation and ability, while central route processing has high motivation and ability (Lazard & Atkinson, 2015). Expiration labels are considered peripheral cues (Saha, Vemula, Mendu, & Gavaravarapu, 2013), which means that consumers are unlikely to actually read and think about the label meaning as consumers believe they already know what it means (Zuckerman & Chaiken, 1998). This allows the misconceptions and misunderstandings to continue existing. When consumers receive new information, they are more likely to critically think about the expiration label (Trumbo & McComas, 2003), this increases their knowledge, which in turns reduces their level of perceived risk (Klerck & Sweeney, 2007).
information about compared to other information available. It is thus likely that consumers who received expiration label education will use this information more, and attach a greater importance to this information compared to consumers who did not receive this type of education.
H2. Consumers receiving label education will have a lower level of perceived risk compared to consumers who did not receive this information.
H3. Consumers receiving education will express a greater importance for label in determining the level of perceived risk than consumers who did not receive this information.
4.3 Virtue vs. vice products
When consumers have to make a food choice, the dominant motives are the expected quality and taste of the product (Vermeir & Verbeke, 2006). These motives are especially important for vice products as consumers only consider the immediate effects, such as a good taste, and ignore the long-term negative effects such as a bad health (Wertenbroch, 1998). When a product passes its expiration date, the product may lose its flavour, colour, or texture and thus result in a decrease of the product quality (European Commission, 2011). But besides an actual decrease in product quality, some of these effects may also be caused by something else. People generally taste what they expect to taste, and labels influence our taste perception (Wansink & Park, 2002; Wansink & Wright, 2006). Research by Wansink & Wright (2006) states that the negative perception of expired dates has a negative influence on consumers’ taste evaluation and acceptance. This negative perception will become stronger the closer the product is to its expiration date, and at the point of actually expiring it will show a big drop in consumers’ taste evaluation and acceptance. Consumers actually experience a product close to, or past its expiration date to taste worse than a fresher product, even though they actually tasted the same product (Wansink & Wright, 2006).
H4. Vice products in general will have a higher level of perceived risk compared to virtue products for all types of expiration labels.
5. Research design
In this chapter the research methodology will be discussed. First, the decision to use the choice-based conjoint (CBC) will be explained. Thereafter, the different attributes and levels used will be provided, followed by discussing the control variables. Finally the research design, model specification, and method of analysis will be examined, followed by the data collection method.
5.1 Method
A conjoint analysis is a technique for measuring trade-offs among multi-attributed products and services (Green & Srinivasan, 1978; Green & Srinivasan, 1990). The main goal of conjoint analyses is to obtain estimates of the relative influence different attributes have on consumers’ evaluation of products (Netzer, et al., 2008). This technique has been widely used, and deemed to be especially appropriate for evaluating hypothetical products or attributes (Baker & Burnham, 2001).
Multiple versions of the conjoint exist, although the choice-based and rating-based conjoint are the main types used (Hervé & Pelliot, 2015). In a rating-based conjoint consumers have to rate their preference for the different products presented to them, while in a choice-based conjoint consumers only select their most preferred option (Karniouchina, Moore, Van Der Rhee, & Verma, 2009). Rating products is not the type of behaviour a person would usually adopt in a typical purchase situation, whereas making a choice whether or not to do or buy something for example is a more natural manifestation of a persons’ preference (Toubia, Hauser, & Simester, 2004; Eggers & Sattler, 2011). Furthermore, the CBC also allows for interaction effects, and the possibility to add a none-option in which consumers indicate they would not select any of the presented alternatives like in a real purchasing situation (Orme, 2014). The CBC is among the most used techniques for measuring consumer preferences (Eggers & Sattler, 2011; Wlömert & Eggers, 2014).
respondents find the presented alternatives to be unsafe, this provides valuable information as to how effective the label, product type, and education are on influencing consumers’ risk perception. Furthermore, the limited number of alternatives, consisting of three labels and two products, results in only six possible and very similar combinations. It may therefore be easier for consumers to compare two similar options at a time, as is the case in a CBC.
5.2 Attributes and levels
For the conjoint analysis different attributes and levels have been used, as can be seen in table 1. To represent the virtue and vice product categories, the products low fat yoghurt and vanilla & chocolate custard have been selected respectively. Both products have a similar shelf life, and belong to the dairy category. Furthermore, Dutch consumers commonly consume these types of products so they are likely to be familiar with them.
The number of levels has been kept low to keep the standard error per parameter as low as possible. To avoid the number-of-levels effect (Eggers & Sattler, 2011), a similar amount of levels were used for both attributes.
Attribute Level 1 Level 2 Level 3
Label Best before
Ten minste houdbaar tot
Preferably use before
Bij voorkeur gebruiken voor
Quality guaranteed till
Kwaliteit gegarandeerd tot
Product Low fat yoghurt
Halfvolle yoghurt
Vanilla & chocolate custard
Dubbel vla Vanille & Chocola
None option I do not consider these options to be safe Table 1. Attributes and levels
5.3 Control variables
consumers perceive dairy products to be. The questions asked are very direct, and do not ask about aspects that are not connected to the products used for the experiment.
Other control variables asked for were respondents’ age, gender, and level of education to gain a better understanding about the demographic characteristic of the respondents. Another control variable was the questions whether the respondents ever had food poisoning before (never, once, more than once). This question was added as respondents who have suffered from the effects that could potentially be caused by eating food that has gone bad, may have a higher risk perception towards products past their expiration date. These questions were asked to be able to control for their effects. The surveys used can be found in appendix A.
Question Answer scale
I consider consuming dairy products as 1 = not at all risky, 10 = highly risky When consuming dairy products I am
exposed to
1 = no risk at all, 10 = very high risk
Eating dairy products is risky 1 = strongly disagree, 10 = strongly agree Score < 2.5 Perceive dairy as safe Score 5.01 -7.5
Score 2.51 – 5.0 Score > 7.50 Perceive dairy as unsafe
Table 2. Risk perception
5.4 Research design
Per choice set respondents are presented with two alternatives and a none-option, and they have to select the option they perceive as being the safest. Afterwards they have to indicate on a 1-5 scale how safe they perceive the chosen option to be.
The moderating effect of label information will be implemented by using two different surveys. One group receives a page with information before answering the questions. On this page information is provided stressing that the ‘best before’ expiration label is only a quality indicator, and that products past this date do not necessarily have to be thrown away immediately. The product can still be consumed within a reasonable amount of time if stored properly. The quality, however, can deteriorate after the date as the product could possibly lose its flavour, its crispiness, or a change in colour could occur. Examples will be provided as well stating that milk, if stored properly in the fridge, could still be consumed approximately three days after the ‘best before’ date, mayonnaise up to two months, and a product like rice could be consumed for years after the expiration date. The survey displaying this information can be found in appendix A.
For both surveys, the alternative labels will be introduced. Stating that since a lot of misunderstanding exists among consumers about what the current expiration label means, two new labels have been designed which could be used as a replacement for the current label. Both new labels will be mentioned here with their Dutch names. The reason for doing so is to make clear that alternative labels are indeed expiration labels, and consumers should treat them this way.
5.5 Model specification
Based on the attributes and levels discussed earlier, the utility function for the model is displayed in the function below. This function indicates that the systematic utility of respondent n for alternative i is the sum of the part worth utilities. The model uses effect-coding to obtain a beta for each attribute level.
V
i= ß
LabelBB
i+ß
labelQGT
i+ ß
labelPUB
i+ ß
productYOG
i+ ß
productVC
i+ ß
noneNone
i Where:Vi = Utility
QGT = Quality guaranteed till label PUB = Preferably use before label YOG = Low fat yoghurt
VC = Vanilla & chocolate custard None = None-option
5.5 Method of analysis
The risk perception scale consists out of three separate questions, and to determine whether or not they can be combined into a single variable several tests will be conducted. A factor analysis will determine which variables can be combined. To determine if a factor analysis is appropriate, the KMO test of sampling adequacy will be performed. This test predicts if the items are likely to factor well, based on correlation and partial correlation, and needs a value of above 0,5 (Parsian & Dunning, 2009). Also the Bartlett’s tests of sphericity needs to be significant, which then rejects the null hypothesis stating that the variables are uncorrelated (Leech, Barret, & Morgan, 2005). If these requirements are met, a factor analysis can be performed in SPSS, which then will determine how many variables will be created. The number of times the Eigenvalue is above 1 indicates the number of variables that should be created, and in this case this should be one to create a single variable. Also the total variance explained by the to be created variable should be above 60% (Leech, Barret, & Morgan, 2005). Next, the communalities, which represents the proportion of variance of the item that is accounted for by the new variable, should all have a loading of 0,4 or higher (Broome, Pryor, Habermann, Pulley, & Kincaid, 2005). And finally, the Cronbach’s alpha, a measure of internal consistency, should be at least above 0,7 (Peterson, 1994). When these conditions are met, the risk perception scale questions can be combined into a single variable.
To test for model validity, a likelihood ratio test will be performed, and a mean absolute error based on the holdout task and hit rate will be used. The likelihood ratio test will examine if the estimated model is better than a null model that predicts randomly. The hit rate will indicate how well the estimated model predicts respondent choices compared to a random assignment of consumers’ choices. The holdout set will provide information about the predictive power of the model, by comparing how often the model correctly predicted the answers compared to consumers’ actual responses and provides the percentage of times the estimated model is correct.
5.6 Data collection
6. Results
In this chapter the outcomes of the conducted surveys will be analysed. First, some descriptive statistics will be provided about the respondents. Second, the risk perception scale will be combined into a single variable. Third, the results of the conjoint analysis will be reviewed by creating different models for the with-education and without-education model, and possible interaction effects will be examined. Afterwards, the model validity will be examined.
6.1 Descriptive statistics
The with-education survey had 67 respondents (31 male, 36 female), the survey without had 68 respondents (33 male, 35 female) creating a total of 135 respondents. The study had an average response time of 4.7 minutes. The fact that mainly university students participated can be found back not only in the average age of 24, but also in the education level. The majority (77%) has had some form of higher education as can be found in table 3. Furthermore, most respondents (73%) indicated they never
had food poisoning. Table 3. Descriptive statistics
6.2 Risk perception
For both surveys it has to be determined if the three risk perception questions asked can be combined into a single variable. The tests performed have been outlined in chapter 5.5 Method of analysis. First, the with-education model will be discussed, followed by the without-information model.
To determine if a factor analysis is appropriate, the KMO measure of sampling adequacy should be above 0,5. The with-education model has a KMO value of 0,692, which
Item Frequency Percentage
Gender Male 64 47% Female 71 53% Education University 71 53% HBO 32 24% MBO 7 5% High school or below 25 18% Food poisoning Never 99 73% Once 25 19%
is sufficient, indicating the items are likely to factor well. The Bartlett’s test of sphericity is highly significant at P=0,000, which means the null hypothesis ‘the variables are uncorrelated’ can be rejected. After these tests a factor analysis was run. This test confirmed that only one variable had an Eigenvalue of above 1, with a value of 2,343. This item explained 78,112% of the total variance, well above the minimum value of 60%. This indicates that only one new variable would have to be created. Furthermore, the communalities, explaining the percentage of variance in a given variable that is explained by all the extracted factors, are all above the required value of 0,4 with scores of 0,717, 0,770, and 0,857. And finally, the Cronbach’s alpha test was conducted resulting in a value of 0,860, indicating a strong internal consistency. Concluding, the three risk perception question can be combined into a single variable. Therefore, further testing was done using risk perception as a single variable. The mean value for the variable was 2,1, and a score of below 2.5 indicates that consumers perceive dairy products as being safe, as can be seen in table 2.
For the without-education model the same tests were run. The KMO test generated a value of 0,690, which is above the minimum required value of 0,5. The Bartlett’s test of sphericity is highly significant at P= 0,000. In this case also only one item had an Eigenvalue higher than 1 with a value of 2,300, which explained 76,568% of the total variance. The communalities are all well above 0,4 with values of 0,740, 0,843, and 0,717. And finally, the Cronbach’s alpha test indicates a strong internal consistency as well with a value of 0,841. For this survey the three questions can be combined into a single variable as well, and will be used as such for the remainder of testing. The mean value for the variable was 1,9, and as it is below 2.5 this means that also consumers in the without-education group perceive dairy products as being safe.
6.3 Education on perceived risk
To see if receiving education would have an effect on consumers’ level of perceived risk, two tests will be conducted.
to indicate that they selected the none-option. This question was added to see if respondents in one of the two surveys would perceive the alternatives to be riskier than the other group. The scores of the holdout question were used, as this was a fixed choice set. The alternatives displayed were the same for all respondents in both surveys. The first option was the ‘quality guaranteed till’ label combined with the vanilla & chocolate custard, the second option the ‘best before’ label combined with the low-fat yoghurt, and the third alternative was the none-option. Respondents who did receive information had an average score of 0,96, and those who did not had a score of 1,10 on the Likert scale. To test for differences between the two groups, with and without-education, an independent t-test was conducted. This showed that there was no significant difference between the two groups T(135)=0,14, P=0,307. Both groups are very close to the 1-point indicating that they do not perceive the alternatives to be very risky.
The second test was conducted to see if there was a difference in how often respondents selected the none-option. Consumers in both conditions selected this alternative 33% of the time. An independent t-test was conducted with a P-value of 1,00 confirming that there is no difference in how often respondents in both surveys selected the none-option.
The H2 “Consumers receiving label education will have a lower level of perceived risk compared to consumers who did not receive this information” should not be supported based on these results.
6.4 Conjoint analysis
Respondents were asked to select the safest option out of two alternatives, and a none-option indicating they perceived neither product to be safe. In tables 4 and 5 the outcomes of the conjoint analysis can be found.
significant difference in label importance between the two surveys. Hypothesis 3, stating that consumers who received education would have a greater importance for label than consumers who did not receive education, is not supported based on these results.
When looking at the values for the different attributes, for both models the attributes label and product are non-significant, while the none-option is significant. This result indicates that both label and product-type do not significantly influence the risk perception of consumers. When there is no significant effect of label on perceived risk, there cannot be a difference in risk perception for the different labels. This means that hypothesis 1a and 1b, stating that the labels ‘quality guaranteed till’ and ‘preferably use before’ respectively have a lower level of perceived risk compared to the current ‘best before’ label, cannot be supported based on these results for both models. As product type also has no significant results, hypothesis 4 is not be supported for both models, as it states that vice products will have a higher level of perceived risk.
With-education model
Without-education model
Attribute Class 1 Wald P-value Range Importance Label 0,2749 2,2782 0,32 0,2749 12% PUB* -0,0999 BB* 0,175 QGT* - 0,0752 Product 0,0813 0,3119 0,58 0,0814 3% Choco* 0,0960 Yoghurt* - 0,0960 None-option - 1,9871 48,2493 0,00 1,9871 85% Table 5. Conjoint analysis without-education model
*PUB: preferably use before, BB: best before, QGT: quality guaranteed till, None: none-option, Choco: chocolate & vanilla custard, Yoghurt: low-fat yoghurt
6.5 Interaction effects
Besides the main effects, possible interaction effects were also tested for. The interactions tested were: risk perception on label, risk perception on product, gender on label, gender on product, food poisoning on label, and food poisoning on product, see appendix B.
For the with education model, only the interaction effect of gender on label is significant with P=0,00054. To further explorer this effect, two classes were created based on gender. Males select the ‘best before’ label 56% of the time, compared to only 23% of the time for ‘quality guaranteed till’, and 21% for ‘preferably use before’. For females this is ‘best before’ 26%, ‘quality guaranteed till’ 39%, and ‘preferably use before’ 35% of the time. When controlling for the variables food poisoning, and consumers’ risk perception of dairy products in general, they did not have an effect on consumers’ level of perceived risk.
6.6 Model validation
To test the model validity, a likelihood ratio test, hit rate, and a mean absolute error based on the holdout set will be used and examined.
6.6.1 Likelihood ratio test
First, to determine if the estimated model is better than the null model, the likelihood ratio test will be performed. The LL(0), or the null model, has been calculated with the following formula:
LL(0)= 𝑛 ∙ 𝑐 𝑙𝑛 (1/𝑚)
The variables used can be found in table 6.
The LL(ß), or the estimated model, is provided by Latent Gold and the results can be found in table 6. For both models the LL(ß) is lower than the LL(0), which implies that the predictive power of the estimated models is better than that of the null model, which predicts randomly. To see if the estimated model parameters are significantly different from zero, a Chi-square test will have to be performed. The following formula will be used:
Chi-square = -2 (LL(0) – LL(ß))
The results of the Chi-square tests were looked up in a Chi-square distribution table. The cut-off value for a Chi-square test with 4 degrees of freedom at P < 0,001 is 13.28. As can be seen in table 6, the outcomes of the tests are both well above this value. It can thus be concluded that the tests are significantly different from zero.
The next item is the R-square and adjusted R-square. These items provide insights on the predictive strength of the model. The following formulas are used:
R2 = 1 −𝐿𝐿 ß
𝐿𝐿 0
R2adj = 1 −
𝐿𝐿 ß −𝑛𝑟.𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟𝑠 𝐿𝐿(0)
The adjusted R2
punishes for the number of parameters used, as more parameters increases the goodness of fit. The adjusted R2
With-education Without-education
Number of respondents (n) 67 68
Number of choice sets (c) 3 3
Number of alternatives (m) 3 3 Number of parameters (df) 4 4 LL(0) -220.8 -224.1 LL(ß) -183.7 -179.3 Chi-square 74.2 89.6 R2 0,168 0,199 R2 adj 0,149 0,186
Table 6. Model validity
6.6.2 Hit rate
The hit rate calculates the number of times the model predicts the right choice, and divides this by the total number of choices. The outcome indicates how good the estimated model is at predicting respondent choices. The prediction tables used for this calculation are provided by Latent Gold, see tables 7 and 8.
The prediction table of the with-education model shows that for alternative 1, the model predicts 49 out of the 105 responses correctly, for alternative 2 it predicts 54 out of the 96 correct, and for alternative 3 this is 0 out of the 30. This means that the overall model predicts 49+54+0=103 of the 201 observed choices correctly. This creates a hit rate of (103/201)*100=51,2%. The model correctly predicts the respondents’ choice 51,2% of the time.
The hit rate of the with-education model with 51,2% is higher than that of the naïve model with 42%. The same calculations have been run for the without-education model. The hit rate for the without-education model is 47,5%, and the naïve model 43,8%. The estimated models are better at predicting consumers’ choices compared to a random assignment.
6.6.3 Mean absolute error
A holdout set was added to the survey to measure the predictive power of the estimated models. The predicted shares per alternative are compared to the observed shares, and the difference between them is the absolute error rate, see tables 9 and 10. The mean absolute error is the average of the error rates combined. For the with-education model this is (3+6+6)/3 =5%, and the without-education model (6+6+0)/3=4%. The mean absolute error rates for both models are quite low, this means that the models have high predictive power.
Alternative 1 Alternative 2 Alternative 3
Predicted shares 42% 49% 9%
Observed shares 39% 55% 6%
Absolute error 3% 6% 3% Table 9. With education model
Alternative 1 Alternative 2 Alternative 3
Predicted shares 50% 44% 6%
Observed shares 56% 38% 6%
Absolute error 6% 6% 0% Table 10. Without education model
6.7 Hypothesis overview
Hypothesis With
education model
Without
education model
H1a. The ‘quality guaranteed till’ label will have a lower level of perceived risk compared to the current ‘best before’ label.
Not supported Not supported
H1b. The ‘preferably use before’ label will have a lower level of perceived risk compared to the current ‘best before’ label.
Not supported Not supported
H2. Consumers receiving label education will have a lower level of perceived risk compared to consumers who did not receive this information.
Not supported Not supported
H3. Consumers receiving education will express a greater importance for label in determining the level of perceived risk than consumers who did not receive this information.
Not supported Not supported
H4. Vice products in general will have a higher level of perceived risk compared to virtue products for all types of expiration labels.
Not supported Not supported
7. Conclusion and discussion
This study was created to investigate the effects of changing the ‘best before’ expiration label wording, educating consumers about expiration labels, and the effect of product type on the level of perceived risk of perishable food products two days past their expiration date. Confusion about expiration labels is listed as the main reason for food waste by Dutch consumers (Onderzoeksbureau Stroom, Tetra Pak, 2015), resulting in millions of kilos of still consumable food being wasted each year (Soethoudt, Sluis, Waarts, & Tromp, 2013).
As alternatives for the current ‘best before’ label, the labels ‘quality guaranteed till’ and ‘preferably use before’ were suggested (Soethoudt, Sluis, Waarts, & Tromp, 2013; Onderzoeksbureau Stroom, Tetra Pak, 2015). The ‘quality guaranteed till’ label was selected by Dutch consumers as a suitable alternative, as it better explains products can still be consumed after that date (Onderzoeksbureau Stroom, Tetra Pak, 2015). The ‘preferably use before’ was selected for its milder tone compared to the current label, as stronger messages are perceived as more hazardous (Zuckerman & Chaiken, 1998). Both labels are expected to have a lower level of perceived risk compared to the ‘best before’ label. However, this research showed that the attribute label has no significant effect on the level of perceived risk. No claims can thus be made on the difference in effect between the labels, therefore hypotheses 1a and 1b for both studies cannot be supported based on these results.
In both surveys respondents were presented with a none-option to indicate that they perceived both presented options to be unsafe. This effect was significant in both studies. However, consumers in both studies also indicated that they do not perceive the presented alternatives to be risky when asked to rate how risky they perceived their selected option to be. A possible explanation for this may be that the none-option was selected out of convenience to avoid having to select either one of the options. The presented alternatives were very similar, consumers may have found it difficult to carefully examine the differences between the alternatives, or perhaps were not motivated enough to put in the effort. Another possible explanation may be that grocery products are low-involvement goods. When risk is involved consumers are more likely to forgo the purchase than to find ways to reduce the level of perceived risk (Cox & Rich, 1964). A similar effect may have happened during the survey, when consumers perceived there to be risk instead of examining the information available to them, they simply selected the none-option.
Providing consumers with more information is one of the most commonly mentioned solutions to decrease perceived risk (Cox & Rich, 1964; Murray, 1991; Dowling & Staelin, 1994; Mitra, Reiss, & Capella, 1999). It is therefore expected that educating consumers about expiration labels will reduce their level of perceived risk. This research did find that consumers with education had a slightly lower average perceived risk score compared to the group who did not receive information, however, this difference was not significant. Furthermore, no significant difference could be found between the number of times respondents in both groups selected the none-option indicating they perceived the presented alternatives to be unsafe. Therefore, based on these results, hypothesis 2 was not supported. A possible explanation may be, as discussed previously, that consumers do not perceive dairy products in general to be risky. When consumers in both groups indicate they do not perceive diary products to be risky at all, or not very risky, it unlikely that there will be big differences between the groups. The effect of education on the level of perceived risk may be more present in other food categories. Another possible explanation may be that the with-education group only took about 5 seconds longer on average to complete the survey. This may indicate that at least some consumers did not carefully read the additional information provided to them, thus mitigating the effect of education.
Wright, & Brace, 1998). Furthermore, a study by Suter & Burton (1996) showed that when consumers receive label education, they tend to attach a greater importance to those parts they have received information about compared to other information available. It is thus expected that as consumers receive label education, and are better capable of using this information, they will express a greater importance for label in determining the level of perceived risk compared to consumers who did not receive label education. This research showed that consumers in the with-education group indeed expressed a slightly higher importance for label, but this result was non-significant. As a result, hypothesis 3 was not supported based on these results. A reason for this may be that consumers were only provided with a label, product type, and the none-option. If more variables were provided, such as price, or the number of days the product was expired would have varied for example, consumers’ importance per attribute may have varied more, as they would have more information sources to extract information from.
8. Theoretical and managerial implications
Previous studies have provided new information and insights on possible ways to reduce unnecessary food waste caused by food products crossing their expiration date. Examples are: techniques to prolong product’s shelf life (Mastromatteo, et al., 2015; Stratakos, Linton, Patterson, & Koidis, 2015), freshness indicator technology (Fortin & Goodwin, 2009; Rukchon, Nopwinyuwong, Trevanich, Jinkarn, & Suppakul, 2014), or dynamic product pricing, such as expiration date-based pricing (Theotokis, Pramatari, & Tsiros, 2012; Wang & Li, 2012). However, these studies do not touch upon the problem that consumers are confused by expiration labels (GFK, 2012; Temminghoff & Vlerken, 2013), and perceive products close to, or past their expiration date as riskier than they often are (Tsiros & Heilman, 2005), which increases the likelihood of these products being disposed of (Sen & Block, 2009).Griffin, & Abrams, 2006; Misra, 2007), but no studies could be found using education for expiration labels. This study therefore contributes to existing literature by exploring two possible solutions to the problem that a large portion of consumers does not correctly understand the expiration label (Temminghoff & Damen, 2013), although they do make use of expiration labels very often (GFK, 2012). Even though no significant results were found in this research, a more elaborate experiment may generate different results.
Furthermore, respondents in the with-education group did not seem to spend much time on reading the additional information available to them, as evidenced by the fact that this group only took about 5 seconds longer on average to complete the survey. Consumers may have been unmotivated to carefully read and process the information provided to them. If food organizations or other parties are interested in educating consumers about the importance of a proper label understanding, they may benefit from increasing consumers’ motivation to process this information. Possible ways to do so would be by showing consumers what the benefits are from not unnecessarily wasting food, such as the amount of money they could safe or what the effects on nature are from spending the resources on making this food and then being thrown away.
9. Limitations and further research
A number of limitations apply to this research. Some of these issues could be addressed in further research.
Second, it cannot be said for sure that respondents actually selected the option they perceived to be the safest, or simply selected the one they think tastes better. This study did not control for this, and it is therefore advised that further research takes this into account.
Third, consumers in the with-education group only took about 5 seconds longer to complete the survey compared to the without-information group. Consumers may not have carefully read the additional information, which could harm the effects of education this study tried to measure. Further studies could look more extensively into the effect of providing information, and see to which types of information consumers respond best.
Bibliography
Baker, G., & Burnham, T. (2001). Consumer response to genetically modified foods: Market segment analysis and implications for producers and policy makers. Journal of Agricultural and Resource Economics , 26 (2), 387-403.
Bauer, R. (1960, December). Consumer behavior as risk taking. American Marketing Association , 389-398.
Beharrell, B., & Denison, T. (1995). Involvement in a routine food shopping context. British Food Journal , 97 (4), 24-29.
Broome, M., Pryor, E., Habermann, B., Pulley, L., & Kincaid, H. (2005). The scientific misconduct questionnaire - Revised (SMQ-R): Validation and psychometric testing. Accountability in Research , 12 (4), 263-280.
Cacioppo, J., Petty, R., Kao, C., & Rodriguez, R. (1986). Central and peripheral routes to persuasion: An individual difference perspective. Journal of Personality and Social Psychology , 51 (5), 1032-1043.
Canadian Food Inspection Agency. (2014, September 17). Date labelling on pre-packaged foods. Retrieved October 3, 2015 from Canadian Food Inspection Agency:
http://www.inspection.gc.ca/food/information-for-consumers/fact-sheets/date-labelling/eng/1332357469487/1332357545633
Cantor, J., Torres, A., Abrams, C., & Elbel, B. (2015, November). Five Years Later: Awareness Of New York City's Calorie Labels Declined, With No Changes In Calories Purchased. Health Affairs , 1-9.
Chen, Y.-S., & Chang, C.-H. (2012). Enhance green purchase intentions. Management Decision , 50 (3), 502-520.
Chernev, A., & Gal, D. (2010). Categorization effects in value judgements: averaging bias in evaluatin combinations of vices and virtues. Journal of Marketing Research , 47 (4), 738-747.
Cox, D., & Rich, S. (1964). Perceived risk and consumer decision-making: the case of telephone shopping. Journal of Marketing Reserach , 1 (4), 32-39.
Dillaway, R., Messer, K., Bernard, J., & Kaiser, H. (2011). Do Consumer Responses to Media Food Safety Information Last? Applied Economic Perspective & Policy , 33 (3), 363-383.
Dowling, G., & Staelin, R. (1994). A model of perceived risk and intended risk-handling activity. Journal of consumer research , 21, 119-134.
Eggers, F., & Sattler, H. (2011). Preference measurement with conjoint analysis: Overview of state-of-the-art approaches and recent developments. GFK Marketing Intelligence Review , 3 (1), 36-47.
