Hostname: page-component-7c8c6479df-995ml Total loading time: 0 Render date: 2024-03-26T16:38:00.551Z Has data issue: false hasContentIssue false

Testing consumer perception of nutrient content claims using conjoint analysis

Published online by Cambridge University Press:  15 January 2010

Adam Drewnowski*
Affiliation:
Nutritional Sciences Program and Center for Public Health Nutrition, School of Public Health, University of Washington, Box 353410, Seattle, WA 98195-3410, USA
Howard Moskowitz
Affiliation:
Moskowitz Jacobs Inc., White Plains, NY, USA
Michele Reisner
Affiliation:
Moskowitz Jacobs Inc., White Plains, NY, USA
Bert Krieger
Affiliation:
Moskowitz Jacobs Inc., White Plains, NY, USA
*
*Corresponding author: Email adamdrew@u.washington.edu
Rights & Permissions [Opens in a new window]

Abstract

Objective

The US Food and Drug Administration (FDA) proposes to establish standardized and mandatory criteria upon which front-of-pack (FOP) nutrition labelling must be based. The present study aimed to estimate the relative contribution of declared amounts of different nutrients to the perception of the overall ‘healthfulness’ of foods by the consumer.

Design

Protein, fibre, vitamin A, vitamin C, calcium and iron were nutrients to encourage. Total fat, saturated fat, cholesterol, total and added sugar, and sodium were the nutrients to limit. Two content claims per nutrient used the FDA-approved language. An online consumer panel (n 320) exposed to multiple messages (n 48) rated the healthfulness of each hypothetical food product. Utility functions were constructed using conjoint analysis, based on multiple logistic regression and maximum likelihood estimation.

Results

Consumer perception of healthfulness was most strongly driven by the declared presence of protein, fibre, calcium and vitamin C and by the declared total absence of saturated fat and sodium. For this adult panel, total and added sugar had lower utilities and contributed less to the perception of healthfulness. There were major differences between women and men.

Conclusions

Conjoint analysis can lead to a better understanding of how consumers process information about the full nutrition profile of a product, and is a powerful tool for the testing of nutrient content claims. Such studies can help the FDA develop science-based criteria for nutrient profiling that underlies FOP and shelf labelling.

Type
Research paper
Copyright
Copyright © The Authors 2010

Front-of-pack (FOP) labelling is intended to convey the nutritional attributes of a food product at a glance. It rests on sets of nutritional criteria developed by researchers(Reference Drewnowski and Fulgoni1), food manufacturers(2), grocery stores and health organizations. FOP summary symbols, based on nutrient profiles(Reference Drewnowski3), are intended to communicate the overall nutritional quality of the food product to the consumer, whereas nutrient-specific symbols provide information on selected nutrients, generally calories (energy), fats, sugar and sodium(4, 5).

Concerned that the proliferation of FOP labelling will make people less likely to read the full Nutrition Facts Panel, the US Food and Drug Administration (FDA) has moved to ensure that the new FOP labelling schemes not be false or misleading(4). A proposed regulation will define the nutritional criteria that will have to be met by manufacturers making claims about the nutritional quality of foods, whether expressed as symbols or as text(4). In order to assemble the evidence base, the FDA plans to use an online consumer panel, established by a contractor, to study how different nutritional attributes of a given product, presented with or without FOP symbols, influence perceived product healthfulness(4). To help understand consumer reactions, the FDA plans to collect data on the participants’ background, health literacy and health status. In the meantime, the FDA will proceed with enforcement action against products that bear FOP labels that are implicit nutrient content claims and are inconsistent with the FDA’s existing requirements.

At this time, the Nutrition Facts Panel is still the chief means of conveying information about the nutrient composition of foods to the American consumer(Reference Borra6). Mandated by the 1990 Nutrition Labeling and Education Act (NLEA), it was intended to promote healthier food choices. Manufacturers were permitted to post nutrition claims on the front of the package, a form of advertising regulated by the FDA(5). Stringent guidelines exist as to what nutrition claims are permitted and how they must be communicated(5).

The present experimental study used an online consumer panel and a within-subject design to collect information on how different nutrient content claims, couched in FDA-approved language, affected consumer perceptions of the overall healthfulness of a hypothetical food product. Conjoint analysis, a mathematical technique used to assess the key influences on consumer choice(Reference Krieger, Cappuccio and Moskowitz7), was then used to determine whether consumers were more influenced by the presence of nutrients to encourage, such as protein, vitamins and minerals, or by the declared absence of nutrients to limit. The FDA’s stated intent is to develop uniform nutrition criteria on which FOP nutrition labelling must be based(4). Studies on how consumers perceive the overall healthfulness of a food, using modern online techniques, can help the FDA develop science-based criteria for nutritional profiling, leading to a standardized approach to food labelling(5, 8).

Experimental methods

Participants

The study was conducted using an online marketing i-Novation panel. Three thousand invitations were sent to potential participants by email. Panellists were recruited by a double opt-in method, where the prospective panellist had to affirm interest in participating in two different portions of a panel questionnaire. Approximately 70 % of the respondents completed the 15–18 min survey, consistent with previous studies on health and well above the 40 % completion rate for studies on the use of financial services. A summary of panellist demographics is provided in Table 1.

Table 1 A summary of participant characteristics: volunteers in an online consumer panel

FDA nutritional criteria

For a food to be considered ‘healthy’ by the FDA, it has to contain at least 10 % of the Reference Daily Intake (RDI) or Daily Value (DV) of one or more of these six nutrients: protein, fibre, vitamin A, vitamin C, calcium or iron per reference amount and prior to fortification. The present study therefore used protein, fibre, vitamin A, vitamin C, calcium and iron as the six nutrients to encourage.

A ‘healthy’ food also has to be low in total fat (<3 g), saturated fat (<1 g), cholesterol (<60 mg) and sodium (<480 mg), also per reference amount. The present study therefore used total fat, saturated fat, cholesterol and sodium as nutrients to limit. Given concerns about the high sugar content of several food products, total and added sugars were also included for a total of six nutrients to limit. Twelve nutrient elements were the initial input for the study. With the exception of added sugars, all these nutrients are currently listed on the Nutrition Facts Panel.

FDA-approved nutrient content claims

A nutrition claim is defined as any representation, stated, suggested or implicit, that a food has particular nutritional properties, including but not limited to its content of macronutrients, vitamins and minerals(5). Although such claims are based on information provided on the Nutrition Facts Panel, mere declaration of presence or quantity of a nutrient does not qualify as a nutrition claim(5).

Nutrition claims can be divided into nutrient content claims, describing the level of a nutrient contained in a food (e.g. ‘low in fat’), and comparative claims that compare the nutrient level to a pre-defined standard (e.g. ‘reduced in fat’). The descriptive language of nutrient claims must be approved by the FDA. Foods that provide 10–19 % of the DV per serving amount can be called ‘good sources’, whereas foods that provide >20 % of the DV can be called ‘excellent sources’ of a nutrient. Foods that contain more than 20 % of the DV for protein, fibre, vitamins or minerals are also entitled to terms such as ‘high’ or ‘rich in’. The FDA language guidelines specify that the terms ‘good source of’ or ‘excellent source of’ can only be used to describe protein, vitamins, minerals, dietary fibre or potassium.

Conversely, such terms as ‘free,’ ‘low’ or ‘reduced/less’ apply only to calories, total or saturated fat, sodium, sugars or cholesterol. These too are based on serving sizes; or on 50 g if the serving size is small. Depending on amounts per serving, foods can be described as being ‘low in’ or ‘free of’ calories, total fat, saturated fat, cholesterol, sugars and sodium. Each of the twelve nutrient elements was presented at two nutrient content levels, using FDA-approved claim language (see Table 2), for a total of twenty-four nutrient elements in all.

Table 2 Nutrient content claims tested (FDA definitions(5))

FDA, US Food and Drug Administration; DV, Daily Value.

*Per 50 g if reference amount is small.

†With <15 % of calories from saturated fat.

Consumer elements

Twelve additional control elements, six related (in a non-specific way) to nutrients and six related to lifestyles and eating, were selected. The intent was to test whether general concepts about healthful eating, including some featured in the 2005 Dietary Guidelines for Americans and the 2005 MyPyramid(9), would contribute to the perception of healthfulness.

Creation of forty-eight composite test messages

The combination of nutrient elements (n 24) and consumer elements (n 12) yielded a total of thirty-six elements in all. These are summarized in Table 3. The thirty-six elements were sampled in such a way that a test message never contained two elements of the same type that might be contradictory. For example, a test message could not simultaneously include such statements as: ‘This product is a good source of protein’ (A1) and ‘This product is an excellent source of protein’ (A2)(Reference Moskowitz, Gofman and Itty10). Composite test messages (n 48) were then created using two, three or four elements each(Reference Box, Hunter and Hunter11).

Table 3 Elements contributing to the test message

All of the elements appeared across the different test concepts in a statistically balanced, independent fashion. All thirty-six elements were randomly permuted, independently of each other(Reference Gofman12). Each element appeared exactly three times in each set of forty-eight test messages(Reference Moskowitz, Gofman and Itty10), allowing us to observe how different elements influenced the response to the test message by each respondent. Different permutations of message elements were tested across multiple respondents, with each respondent providing forty-eight different ratings. As the permutations varied across respondents, any sequence effects or biases that might be caused by a specific juxtaposition of elements were cancelled out.

Online panel procedures

The orientation page set up the panellists’ expectations about the study, while providing relatively little detail. Panellists were told that they would be given nutritional information for a hypothetical food product. The instructions read:

Providing nutrition information on food and drink product labels is an important way of conveying the message about diets and health to the consumer. It is important that such information accurately reflect the nutrient composition of the product in a simple manner that is easily understood. However, nutrition labels do not always get the right message across. You are invited to review a selection of messages and rate each product on a ‘healthfulness’ scale.

Panellists were asked to rate the product using a nine-point category scale, anchored at each end with labels ‘1 = least healthy’ and ‘9 = most healthy’. Each new message was to be rated independently of any previous message that had just been presented. The specific combinations and the order of presentations were systematically permuted across respondents. Panellists then completed a questionnaire detailing who they were (demographics) and what aspects of foods were important to them (attitudes). A total of 320 panellists completed the 15–18 min survey, a completion rate in line with previous studies on consumer panels(Reference Bosnjak and Tuten13). Following completion of the survey, they were entered into a prize draw.

Panellists were instructed to inspect the entire test message and to react to the message in its entirety, much as what happens in the supermarket at the point of sale. Panellists were not permitted to parse the message into its components, and had to assign a rapid overall rating to each particular combination of concept elements. Although such panels may begin the task by reading and responding to each element in a message, they soon adopt a holistic attitude, inspecting the concept and quickly integrating the information to provide their rating(Reference Moskowitz, Gofman and Itty10). One advantage of within-subject design and conjoint analysis is that the rules underlying the response need not be articulated by the respondent and indeed often cannot be.

Conjoint analysis

Healthfulness ratings obtained along a 9-point scale were first converted to binary form, a prerequisite for logistic regression. ‘Healthy’ was defined as ratings of 7–9, whereas ‘not healthy’ was defined as ratings of 1–6. The equation is written as

The statistics of primary interest in logistic regression are the beta coefficients. The interpretation of the beta coefficients for dichotomous independent variables Ai with values of 1 or 0 (present, absent) is that the beta coefficients represent the log odds that a food is perceived as healthy or nutritious when the message is present (Ai = 1) as opposed to when it is absent (Ai = 0). In a multivariate model, this beta coefficient is the independent effect of variable Ai on Rating (0, 1) after adjusting for all other covariates in the model.

The coefficients bi are a measure of the conditional probability or percentage of respondents who considered the hypothetical food to be ‘healthy’ if the element Ai was present. The constant, b 0, was the baseline probability or percentage of respondents who may think that the food is ‘healthy’ even if there are no elements present. The logistic regression model allowed us to rank the relative contribution of the thirty-six elements to the perceived healthfulness of foods.

Results

Online consumer panel

Consistent with the FDA’s future plans, the present study collected data on consumer demographics, nutrition attitudes and practices, and self-reported health status. Table 1 shows that the 320 participants were predominantly female (78 %), middle-aged (57 % in 45–64 year age range), educated (75 % had at least ‘some college’), and most had no small children or adolescents living at home (76–80 %). More than half of the participants reported being in very good or in excellent health (60 %). More than half (53 %) stated that their diets were healthy or very healthy. However, 62 % were overweight and 42 % were on a calorie-restricted diet at the time of the study.

Survey responses were first tested for goodness-of-fit to the individual-level models. Most of the 320 respondents showed an R 2 > 0·80 for the individual models, suggesting that the data from the respondents were consistent. Similar goodness-of-fit values have been shown in many other studies(Reference Krieger, Cappuccio and Moskowitz7, Reference Moskowitz, Gofman and Itty10).

Results of conjoint analysis

Table 4 shows the best twenty-two elements sorted by their importance. The range of utilities skewed more towards the ‘healthy’ side and no elements were perceived to be ‘unhealthy’. This is what one might expect since all messages were phrased in a manner that connoted health. Based on data from previous studies using conjoint analysis, the important elements were defined as having a utility >8. The unimportant elements were defined as having a utility <−5. In previous studies, differences of five points on the utility scale tended to be significant at the 90 % confidence level.

Table 4 Performance of nutrient elements among respondent segments

Important elements (≥10) in bold.

Nutrient content claims based on the declared presence of protein, fibre, calcium and vitamin C contributed the most to the overall judgement of the food as healthful. Equally influential was the declared absence of nutrients to limit: saturated fat, cholesterol and sodium. Total fat, total and added sugars, and vitamin A had lower utilities and contributed less to the overall perception of healthfulness. The declared presence of iron did not contribute significantly to the perception of healthfulness of the product.

For nutrients to encourage, the declared presence of the nutrient was more important than the amount that was claimed. Comparable utilities were obtained for statements about ‘good’ as opposed to ‘excellent’ sources of protein. That was not the case for nutrients to limit. Utilities were higher for statements that described the products as being completely ‘free’ of saturated fat, cholesterol or sodium. Claims that the food product was ‘low in’ a nutrient to limit did not alter the perceived healthfulness rating.

Nutrient content claims had much higher utilities than the consumer elements, which were largely based on more generic appeals to lifestyle and to good nutrition. Consumer perceptions of the healthfulness of foods appeared to be based largely on the declared nutrient content of foods rather than by statements than might be found on advertising copy.

Demographic segmentation

There were major differences between women and men. Men were far more likely to call a product healthy even in the absence of nutrition content claims (b 0 = 35); whereas women were less likely to do so (b 0 = 18). Whereas women showed more consistent judgements, men did not always do so, and their range of utility values was consequently greater (21 v. 14). Men were less likely to be swayed by nutrient content claims than were women.

Contrary to expectations, the impact of nutrient content claims on the perception of healthiness was not affected by age. The big exception was calcium. Statements ‘good source of calcium’ and ‘rich in calcium’ had higher utilities among older respondents (age 45+ years).

Discussion

How nutrition information on product labels is conveyed to the consumer is subject to stringent regulations by the FDA(5). FDA regulations specify which nutrients need to be listed on the Nutrition Facts Panel, in what format and in what order. Total calories, calories from fat, and the amounts of total fat, saturated fat, cholesterol, sodium, total carbohydrate, dietary fibre, sugars, protein, vitamin A, vitamin C, calcium and iron must be included. Mono- and polyunsaturated fats, potassium and other essential vitamins and minerals must be listed only in the event of a nutrient content claim. The Nutrition Facts Panel lists the amounts in grams or milligrams of nutrients per serving, as well as percentage of the DV, based on a diet supplying 8368 kJ/d (2000 kcal/d)(Reference Drewnowski and Fulgoni1, 5).

New point-of-purchase food labelling schemes developed by manufacturers and grocery stores are based on diverse nutritional criteria. Some are based on nutrients to encourage and nutrients to limit, whereas others are based on calories and nutrients to limit only: fat, sugar and sodium. One important question is whether consumer perception of product healthfulness, based on the elements of the Nutrition Facts Panel, is driven by the presence of beneficial nutrients or by the declared absence of nutrients to limit. A better understanding of how consumers perceive the relative healthfulness of foods would help shape national policies for nutrition education and dietary guidance.

Using online consumer panels and conjoint analysis, the present study showed that the consumer perception of product healthfulness was largely driven by the declared presence of protein, fibre, calcium and vitamin C and by the total absence of saturated fat and sodium. Total fat, iron and vitamin A, all which are part of the FDA definition of healthy foods, had less impact on the consumer utilities. The utilities for total and added sugar were much lower than for either fat or fibre. For nutrients to encourage, the declared presence was more important than the claimed amount (‘good’ v. ‘excellent’ source). By contrast, for nutrients to limit, participants responded more strongly to the total absence of a problematic nutrient (e.g. ‘no fat’).

The data suggest that an effective nutrient profiling system, broadly consistent with the current definition of healthy foods, can be constructed using a limited number of key nutrients. One important question is whether protein should be included in a nutrient profile model, given that it is not currently a shortfall nutrient in the American diet. The present data clearly show that the protein content of foods is one of the key determinants of the perceived nutritional quality of foods by the consumer.

Numerous systems, including the Nutrient-Rich Foods (NRF) index, have been based on the nutrients studied here. The NRF index is based on protein, fibre, vitamins A, C, and E, calcium and iron as well as potassium and magnesium(Reference Fulgoni, Keast and Drewnowski14). The French SAIN sub-score, based on nutrients to encourage, uses protein, fibre, vitamin C, calcium and iron(Reference Darmon, Vieux and Maillot15). Both the NRFI and the French LIM score use saturated fat, added sugar and sodium as the nutrients to limit. Showing how each nutrient contributes to the perception of the overall healthfulness of foods in a quantitative manner, has implications for the design of standardized criteria as planned by the FDA.

These findings have major implications for the emerging technique of nutrient profiling, described as a science of ranking individual foods based on their nutrient composition. Nutrient profiling provides a basis to quantify the overall nutrient value of foods and convey it to the consumer at a glance(Reference Drewnowski and Fulgoni1).

The FDA is also planning to explore the use of FOP labels by different consumer groups. Different households may use nutrition information in different ways(Reference Drichoutis, Lazaridis and Nayga16, Reference Ippolito and Mathios17). The present analyses by gender showed that women were much more influenced by nutrient content claims than were men. These data are consistent with past studies showing that both women and better educated consumers were more likely to read and use nutrition labels(Reference Cowburn and Stockley18Reference Kim, Nayga and Capps20). Some studies found that the elderly used nutrition labels less but others did not(Reference Burton and Andrews21, Reference Nayga22). In the current sample of mostly educated women, education and age played a relatively minor role.

Several previous studies have explored ways in which consumers interpreted favourable nutrition claims for foods that also contained nutrients to limit(Reference Andrews, Netemeyer and Burton23Reference Kozup, Creyer and Burton25). Participants evaluated a limited number of mock product labels. Typically, such studies featured between-subject factorial designs, followed by ANOVA(Reference Andrews, Netemeyer and Burton23, Reference Garretson and Burton24). Other studies explored interactions between the provision of specific nutrient information, consumer demographics, and knowledge(Reference Kim, Nayga and Capps20, Reference Mhurchu and Gorton26).

The present study represents an application of epidemiological methods to marketing research. Analyses made use of dichotomized variables and an equivalent of logistic regression, with participants making as many as forty-eight rapid judgements, with all stimuli presented online. Conjoint analysis could become the chief tool in the design of FOP signposting, other graphics, or function or health claims, since it can rapidly test the impact of diverse elements on consumer response. Conjoint analysis explores the dimensions underlying consumer perception and their integration into the judgement of nutritional value. Such studies can create prototypes of possible food labels, assess responses and draw conclusions about the underlying dimensions of decision making. All of this can be performed rapidly online.

Some limitations of the methodology must be noted. First, the study was based on a hypothetical food product as opposed to a specific food or beverage. The present results must therefore be treated as a general response to nutrient content claims, and not a response that is tied to a specific product. For example, a much stronger response to added sugars might be elicited with cereals. Future studies will need to focus on specific food categories in turn. Second, the test messages were presented in text format and were not integrated into graphics, labels or logos. Placement of nutrition content claims relative to other advertising copy may well influence consumer response(Reference Burton and Andrews21). That too is a matter for future research.

These exploratory data suggest that data collection using online panels followed by conjoint analysis may be a powerful tool for the future testing of nutrition and health claims. Conjoint analysis, a statistical technique that determines how people value different features that make up an individual product, is ideally suited for this purpose. Originating in mathematical psychology(Reference Luce and Tukey27), it is also known as stated preference analysis or multi-attribute compositional model. The method is used in marketing to better understand consumer reactions to concepts in product development(Reference Green, Krieger and Wind28, Reference Wittink and Cattin29) and is becoming increasingly popular in consumer research(Reference Wittink, Vriens and Burhenne30). Conjoint analysis has found applications in studies on the positioning of healthy foods(Reference Krieger, Cappuccio and Moskowitz7); consumer preferences for logos and labelling formats(Reference Harrison and McLennon31); consumer responses to the labelling of genetically modified foods(Reference Hu, Veeman and Adamowicz32); and even to predicting Academy Award winners(Reference Pardoe and Simonton33). In the present version it can be used to deepen the understanding of how consumers process information about the nutrient content of foods.

Acknowledgements

Sources of funding: The data were collected and analysed at the expense of the authors and the studies were not sponsored by any outside source. Conflict of interest declaration: There were no conflicts of interest. Author contributions: H.M., M.R. and B.K. conceived the study, conducted data and modelling analyses, interpreted the results, and participated in the writing of the report; A.D. participated in the conception and design of the study, interpreted the results, and contributed substantially to writing the report.

References

1. Drewnowski, A & Fulgoni, F (2008) Nutrient profiling of foods: creating a nutrient-rich food index. Nutr Rev 66, 2339.CrossRefGoogle ScholarPubMed
2.Smart Choices Program™ (2009) Helping Guide Smart Food and Beverage Choices. http://www.smartchoicesprogram.com (accessed December 2009).Google Scholar
3. Drewnowski, A (2005) Concept of a nutritious food: toward a nutrient density score. Am J Clin Nutr 82, 721732.CrossRefGoogle Scholar
4.Food and Drug Administration (2008) Experimental Study of Nutrition Symbols on Food Packages. http://www.thefederalregister.com/d.p/2009-06-01-E9-12669 (accessed December 2009).Google Scholar
5.Food and Drug Administration (2008) Guidance for Industry: A Food Labeling Guide. http://www.cfsan.fda.gov/guidance.html (accessed December 2009).Google Scholar
6. Borra, S (2006) Consumer perspectives on food labels. Am J Clin Nutr 83, 1235S.Google Scholar
7. Krieger, B, Cappuccio, R & Moskowitz, H (2003) Next generation healthy soup: an exploration using conjoint analysis. J Sens Stud 18, 249268.CrossRefGoogle Scholar
8.Food and Drug Administration (2009) [Docket No. 2007N-0277] Food Labeling: Use of Symbols to Communicate Nutrition Information, Consideration of Consumer Studies and Nutritional Criteria. http://www.fda.gov/OHRMS/DOCKETS/98fr/E7-23211.pdf (accessed December 2009).Google Scholar
9.US Department of Agriculture (2005) MyPyramid.gov Steps to a Healthier You. http://www.mypyramid.gov/index.html (accessed September 2009).Google Scholar
10. Moskowitz, H, Gofman, A, Itty, B et al. (2001) Rapid, inexpensive, actionable concept generation and optimization: the use and promise of self-authoring conjoint analysis for the food service industry. Food Service Technol 1, 149167.CrossRefGoogle Scholar
11. Box, GEP, Hunter, WG & Hunter, JS (1978) Statistics for Experimenters: An Introduction to Design, Data Analysis, and Model Building. New York: Wiley.Google Scholar
12. Gofman, A (2006) Emergent scenarios, synergies and suppressions uncovered within conjoint analysis. J Sens Stud 21, 373414.Google Scholar
13. Bosnjak, M & Tuten, TL (2001) Classifying response behaviors in web-based surveys. J Comput Mediat Commun 6, issue 3; available at http://jcmc.indiana.edu/vol6/issue3/boznjak.htmlCrossRefGoogle Scholar
14. Fulgoni, VL 3rd, Keast, DR & Drewnowski, A (2009) Development and validation of the nutrient-rich foods index: a tool to measure nutritional quality of foods. J Nutr 139, 15491554.Google Scholar
15. Darmon, N, Vieux, F, Maillot, M et al. (2009) Nutrient profiles discriminate between foods according to their contribution to nutritionally adequate diets: a validation study using linear programming and the SAIN,LIM system. Am J Clin Nutr 89, 12271236.CrossRefGoogle Scholar
16. Drichoutis, AC, Lazaridis, P & Nayga, RM Jr (2006) Consumers’ use of nutritional labels: a review of research studies and issues. Acad Mark Sci Rev 2006, issue 9; available at http://www.amsreview.org/articles/drichoutis09-2006.pdfGoogle Scholar
17. Ippolito, PM & Mathios, AD (1993) New food labeling regulations and the flow of nutrition information to consumers. J Public Policy Mark 12, 188205.CrossRefGoogle Scholar
18. Cowburn, G & Stockley, L (2005) Consumer understanding and use of nutrition labelling: a systematic review. Public Health Nutr 8, 2128.CrossRefGoogle ScholarPubMed
19. Drichoutis, AC & Lazaridis, P (2005) Nutrition knowledge and consumer use of nutritional food labels. Eur Rev Agric Econ 32, 93118.Google Scholar
20. Kim, S-Y, Nayga, RM Jr & Capps, O Jr (2001) Health knowledge and consumer use of nutrition labels: the issue revisited. Agric Resource Econ Rev 30, 1019.Google Scholar
21. Burton, S & Andrews, JC (1996) Age, product nutrition, and label format effects on consumer perceptions and product evaluations. J Consum Aff 30, 6989.Google Scholar
22. Nayga, RJ (2000) Nutrition knowledge, gender, and food label use. J Consum Aff 34, 97112.Google Scholar
23. Andrews, JC, Netemeyer, RG & Burton, S (1998) Consumer Generalization of Nutrient Content Claims in Advertising. Cambridge, MA: Marketing Science Institute.CrossRefGoogle Scholar
24. Garretson, JA & Burton, S (2003) Effects of nutrition facts panel values, nutrition claims, and health claims on consumer attitudes, perceptions of disease-related risks, and trust. J Public Policy Mark 19, 213227.CrossRefGoogle Scholar
25. Kozup, J, Creyer, E & Burton, S (2003) Making healthful food choices: the influence of health claims and nutrition information on consumers’ evaluations of packaged food products and restaurant menu items. J Mark 67, 1934.Google Scholar
26. Mhurchu, C & Gorton, D (2007) Nutrition labels and claims in New Zealand and Australia: a review of use and understanding. Aust N Z J Public Health 31, 105112.Google Scholar
27. Luce, DR & Tukey, JW (1964) Simultaneous conjoint measurement: a new type of fundamental measurement. J Math Psychol 1, 127.CrossRefGoogle Scholar
28. Green, PE, Krieger, AM & Wind, Y (2001) Thirty years of conjoint analysis: reflections and prospects. Interfaces 31, S56S73.CrossRefGoogle Scholar
29. Wittink, DR & Cattin, P (1989) Commercial use of conjoint analysis: an update. J Mark 53, 9196.CrossRefGoogle Scholar
30. Wittink, DR, Vriens, M & Burhenne, W (1994) Commercial use of conjoint analysis in Europe: results and critical reflections. Int J Res Mark 11, 4152.Google Scholar
31. Harrison, RW & McLennon, E (2004) Analysis of consumer preferences for biotech labelling formats. J Agric Appl Econ 36, 159171.Google Scholar
32. Hu, W, Veeman, MM & Adamowicz, WL (2005) Labelling genetically modified food: heterogeneous consumer preferences and the value of information. Can J Agric Econ 53, 83102.Google Scholar
33. Pardoe, I & Simonton, DK (2008) Applying discrete choice models to predict academy award winners. J R Stat Soc Ser D, The Statistician 171, 375394.Google Scholar
Figure 0

Table 1 A summary of participant characteristics: volunteers in an online consumer panel

Figure 1

Table 2 Nutrient content claims tested (FDA definitions(5))

Figure 2

Table 3 Elements contributing to the test message

Figure 3

Table 4 Performance of nutrient elements among respondent segments