9
Does nutrition information on food products lead to healthier food choices? Jesús Barreiro-Hurlé a, * , Azucena Gracia b , Tiziana de-Magistris b a Agricultural Economics Area, Andalusian Agricultural Research Institute (IFAPA) – Junta de Andalucía, Centro Camino de Purchil, PO BOX 2027, 18080 Granada, Spain b Agro-Food Economics and Natural Resources Unit, Centre for Agrofood Reseach and Technology (CITA), Avda. Montañana 930, 50059 Zaragoza, Spain article info Article history: Received 14 January 2009 Received in revised form 26 November 2009 Accepted 21 December 2009 Keywords: Nutrition label use Nutrition knowledge Claims Facts panel Spain abstract This paper examines the link between nutrition label use and consumers’ healthier food choices. Label use is considered for the two main types of labels currently found on food products; nutrition facts panels and nutrition/health claims. This link is tested using a three-equation multivariate probit model. Data were obtained from an ad hoc survey conducted in two medium-sized Spanish cities. The results indicate that the use of nutrition information by consumers, whether this involves the fact panel or the claim labels, does influence consumer choice of healthier food products to the same extent, although different types of consumers use the various types of labels considered. Ó 2010 Elsevier Ltd. All rights reserved. Introduction Poor diet and physical inactivity are among the leading causes of death in the US (Mokdad et al., 2004) and EU studies also show that five of the six leading risk factors for ill health are linked to poor nutrition (World Health Organization (WHO), 2002). It is also recognised that changing lifestyle and dietary habits may confer better benefits than medical care (Wansink, 2006) yet modifying individual dietary habits is a challenge involving trade-offs be- tween nutrition and taste, price, convenience and cost (Blaylock et al., 1999). Access to knowledge, education and information is key to informed consumer choice (Ippolito, 1999), and a variety of options are available, of which the most widely known and used is that of product labelling. This can be an important instrument for changing and influencing dietary habits and behaviour (Grunert and Wills, 2007). Product labels currently allowed for in the Euro- pean Union’s legislative framework are the nutrition facts panel, nutrition claims and health claims (OJ, 1990, 2006). The nutrition facts panel and nutrition/health claims are now broadly found on food products within the EU, with over 50% of products carrying a nutrition facts panel, yet rates may vary significantly per country (EAS, 2004). A review of 250 packaged products in Spain reported that over 70% carried a nutrition facts panel, 43% nutrition claims and 23% health claims (CESNID, 2007). Despite numerous studies focused on consumer use of nutrition and health labels and the impact thereof on consumer choice, sig- nificant gaps in our knowledge remain. Most studies have focused on a specific region or location, i.e. Northern Europe and the Unites States, and have been limited in their scope and size to consumer understanding and use of labels and which factors influence that use (Cowburn and Stockley, 2005; Williams, 2005; Drichoutis et al., 2006 and Grunert and Wills, 2007). Yet, in order to assess the impact of product information on behaviour change and come to a conclusion that would apply for the EU as a whole, we need to gain a deeper understanding of regional differences, in addition to establishing a clear link between the use of the information and food choice. Whether different label types have a different impact is also crucial as more and more labels become available. In this manner we can gain an understanding how healthier food choices can be better promoted. This paper therefore aims to fill some of these gaps and provides answers to the following questions: firstly, what type of nutrition labels are used by consumers, and in the second place whether la- bel use promotes healthier food choices. To date only a few studies have analysed both issues simultaneously making it difficult to draw meaningful conclusions. Moreover the Spanish consumer environment is one for which published research to date has been scarce. In this study, two three-equation systems have been simul- taneously estimated. Firstly, consumers’ nutrition knowledge and the factors explaining this knowledge are analysed. Secondly, the factors that explain consumers’ use of nutrition labels (nutrition facts panels and nutrition/health claims) are investigated, includ- ing nutrition knowledge as an explanatory factor. Finally, the will- ingness of consumers to trade-off better tasting, more convenient and cheaper food products for healthier ones is examined. This is done using a multivariate limited-dependent variable model which 0306-9192/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodpol.2009.12.006 * Corresponding author. Tel.: +34 958 895 260; fax: +34 958 895 203. E-mail addresses: [email protected] (J. Barreiro-Hurlé), agracia@ aragon.es (A. Gracia), [email protected] (T. de-Magistris). Food Policy 35 (2010) 221–229 Contents lists available at ScienceDirect Food Policy journal homepage: www.elsevier.com/locate/foodpol

Does nutrition information on food products lead to healthier food choices?

Embed Size (px)

Citation preview

Page 1: Does nutrition information on food products lead to healthier food choices?

Food Policy 35 (2010) 221–229

Contents lists available at ScienceDirect

Food Policy

journal homepage: www.elsevier .com/locate / foodpol

Does nutrition information on food products lead to healthier food choices?

Jesús Barreiro-Hurlé a,*, Azucena Gracia b, Tiziana de-Magistris b

a Agricultural Economics Area, Andalusian Agricultural Research Institute (IFAPA) – Junta de Andalucía, Centro Camino de Purchil, PO BOX 2027, 18080 Granada, Spainb Agro-Food Economics and Natural Resources Unit, Centre for Agrofood Reseach and Technology (CITA), Avda. Montañana 930, 50059 Zaragoza, Spain

a r t i c l e i n f o a b s t r a c t

Article history:Received 14 January 2009Received in revised form 26 November 2009Accepted 21 December 2009

Keywords:Nutrition label useNutrition knowledgeClaimsFacts panelSpain

0306-9192/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.foodpol.2009.12.006

* Corresponding author. Tel.: +34 958 895 260; faxE-mail addresses: [email protected]

aragon.es (A. Gracia), [email protected] (T. de-Ma

This paper examines the link between nutrition label use and consumers’ healthier food choices. Labeluse is considered for the two main types of labels currently found on food products; nutrition facts panelsand nutrition/health claims. This link is tested using a three-equation multivariate probit model. Datawere obtained from an ad hoc survey conducted in two medium-sized Spanish cities. The results indicatethat the use of nutrition information by consumers, whether this involves the fact panel or the claimlabels, does influence consumer choice of healthier food products to the same extent, although differenttypes of consumers use the various types of labels considered.

� 2010 Elsevier Ltd. All rights reserved.

Introduction

Poor diet and physical inactivity are among the leading causesof death in the US (Mokdad et al., 2004) and EU studies also showthat five of the six leading risk factors for ill health are linked topoor nutrition (World Health Organization (WHO), 2002). It is alsorecognised that changing lifestyle and dietary habits may conferbetter benefits than medical care (Wansink, 2006) yet modifyingindividual dietary habits is a challenge involving trade-offs be-tween nutrition and taste, price, convenience and cost (Blaylocket al., 1999). Access to knowledge, education and information iskey to informed consumer choice (Ippolito, 1999), and a varietyof options are available, of which the most widely known and usedis that of product labelling. This can be an important instrumentfor changing and influencing dietary habits and behaviour (Grunertand Wills, 2007). Product labels currently allowed for in the Euro-pean Union’s legislative framework are the nutrition facts panel,nutrition claims and health claims (OJ, 1990, 2006). The nutritionfacts panel and nutrition/health claims are now broadly found onfood products within the EU, with over 50% of products carryinga nutrition facts panel, yet rates may vary significantly per country(EAS, 2004). A review of 250 packaged products in Spain reportedthat over 70% carried a nutrition facts panel, 43% nutrition claimsand 23% health claims (CESNID, 2007).

Despite numerous studies focused on consumer use of nutritionand health labels and the impact thereof on consumer choice, sig-

ll rights reserved.

: +34 958 895 203.(J. Barreiro-Hurlé), agracia@

gistris).

nificant gaps in our knowledge remain. Most studies have focusedon a specific region or location, i.e. Northern Europe and the UnitesStates, and have been limited in their scope and size to consumerunderstanding and use of labels and which factors influence thatuse (Cowburn and Stockley, 2005; Williams, 2005; Drichoutiset al., 2006 and Grunert and Wills, 2007). Yet, in order to assessthe impact of product information on behaviour change and cometo a conclusion that would apply for the EU as a whole, we need togain a deeper understanding of regional differences, in addition toestablishing a clear link between the use of the information andfood choice. Whether different label types have a different impactis also crucial as more and more labels become available. In thismanner we can gain an understanding how healthier food choicescan be better promoted.

This paper therefore aims to fill some of these gaps and providesanswers to the following questions: firstly, what type of nutritionlabels are used by consumers, and in the second place whether la-bel use promotes healthier food choices. To date only a few studieshave analysed both issues simultaneously making it difficult todraw meaningful conclusions. Moreover the Spanish consumerenvironment is one for which published research to date has beenscarce. In this study, two three-equation systems have been simul-taneously estimated. Firstly, consumers’ nutrition knowledge andthe factors explaining this knowledge are analysed. Secondly, thefactors that explain consumers’ use of nutrition labels (nutritionfacts panels and nutrition/health claims) are investigated, includ-ing nutrition knowledge as an explanatory factor. Finally, the will-ingness of consumers to trade-off better tasting, more convenientand cheaper food products for healthier ones is examined. This isdone using a multivariate limited-dependent variable model which

Page 2: Does nutrition information on food products lead to healthier food choices?

Fig. 1. Conceptual model for healthier food choices.

222 J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229

takes into account the issue of endogeneity between the depen-dent variables of the model, an issue often overlooked.

Literature review and the definition of the model

The existing literature was reviewed in order to inform themodel used in the study presented in this paper and as such pro-vides important background on the relationship between use of la-bels and healthiness of food choice. Based on the findingssummarised below a conceptual model for healthier food choicesis developed which incorporates the factors influencing label usepaying particular attention to the role of nutritional knowledge.

Healthiness of consumer choice is influenced by label use. Dric-houtis et al. (2005) theoretically demonstrated the positive link be-tween nutrition label use and purchase behaviour through theinfluence nutritional label use has on consumer values and percep-tion. This model is seldom tested in practice as statistics do not col-lect data on knowledge, attitudes and consumption. However,three studies that use one of the few data sets that does (USContinuing Survey of Food Intakes for Individuals – CSFII- andthe accompanying Diet and Health Knowledge Survey – DHKS-)conclude that nutrition fact panels use leads consumers to health-ier diets (Guthrie et al. 1995; Variyam, 2008); and that, regardlessof the type of nutrition information provided, label use improvesthe overall quality of consumer diets (Kim et al., 2001a).

Secondly, the review leads us to conclude that label use1 isinfluenced by a core group of factors identified as: (i) nutritionknowledge, (ii) individual characteristics, (iii) economic conditionsand time constraints, (iv) health concerns and habits, (v) productinvolvement and (vi) other factors such as need for informationand lifestyle. First, consumers with higher nutrition knowledge aremore likely to use nutrition labels when shopping for food.

Whilst the literature supports the assumption that increasedhousehold income and size (used as proxies for economic condi-tions and time constraints) would impact negatively on label use(i.e. Guthrie et al., 1995; Drichoutis et al., 2006; Gracia et al.,2007), interestingly some studies have found that income and sizeactually have a positive effect on use (i.e. Wang et al., 1995; Nayga,

1 Unless stated, references to nutrition label use include both facts panels andclaims.

1996; Kim et al., 2000) indicating a lack of consensus in this area.Yet health status and awareness are recognised as positivelyimpacting use (i.e. Wang et al., 1995; Kim et al., 2001a; Kimet al., 2001b;.Drichoutis et al., 2006). The importance given by con-sumers to price, nutrition, taste and convenience when shopping isused to capture consumer involvement with food products. Impor-tance attached to price is negatively correlated with label use(Nayga et al., 1998; Drichoutis et al., 2005; Gracia et al., 2007),whereas the opposite applies to importance given to nutrition(i.e. Nayga et al., 1998; Kim et al., 2000; Gracia et al., 2007). Yetthere is no consensus in the literature surrounding taste (Guthrieet al., 1995; Nayga, 1996; Nayga et al., 1998; Nayga, 2000; Dric-houtis et al., 2005).

Finally, factors influencing label use also influence nutritionalknowledge (i.e. individual socio-demographic characteristics, eco-nomic conditions and health awareness, habits and status). In par-ticular, education, sex (female), level of income and health statuspositively influence nutritional knowledge (Guthrie et al., 1995;Szykman et al., 1997; Nayga, 2000; Kim et al., 2001b; Drichoutiset al., 2005; Drichoutis et al., 2006 and Gracia et al., 2007).

Using these findings, the proposed conceptual model of sequen-tial equations is graphically presented in Fig. 1. The model reflectsthe relationship between the endogenous variables (nutritionknowledge, nutrition label use and healthier food choices) andthe major groups of independent variables that may affect eachof the equations, as well as allowing for potential bi-directionalrelationships between the endogenous variables (see below).

Data and definition of variables

Questionnaire design and sample characteristics

Data was gathered using a questionnaire where consumerswere asked questions related to health habits and status, nutritionknowledge, food label use and willingness to trade-off healthy foodproducts for better tasting, easier to prepare and cheaper optionswhen purchasing food. The questionnaire also contained questionson socio-demographic characteristics (sex, family size and compo-sition, age, education level, income) and lifestyles. Before under-taking the main survey, the questionnaire was validated using a

Page 3: Does nutrition information on food products lead to healthier food choices?

Table 1Sample characteristics (% unless stated) and definitions of exogenous variables.

Variable Name (type) Value

Individual characteristicsGenderMale FEMALE (dummy) 28.0Female 72.0Age (average from total sample) AGE (continuous) 45.5Education of respondentUniversity studies UNIVER (dummy) 36.6

Economic conditions and time pressureAverage net household incomea

Households with income below €1500/month LOW_INCOME (dummy) 17.6Household size (average from total sample) HSIZE (continuous) 3.1Household with children less than 6 years old (1 = Yes) CHILDREN (dummy) 19.0Household with adults more than 65 years old (1 = Yes) ELDERLY (dummy) 15.6

Health habits and statusConsumers’ self-assessment of the importance of the effect of food intake on their health status (ten-point increasing scale) FOOD_HEALTH (continuous) 8.2Some household member is following a special diet recommended by a physician (1 = Yes) DIET (dummy) 26.9Consumer suffers or has suffered health problems related to food intake (1 = Yes) HEALTH_PROB (dummy) 31.3Consumer undertakes annual health checks ups (1 = Yes) CHECKUPS (dummy) 46.8Some household member or relative suffers or has suffered cardiovascular diseases (1 = Yes) DISEASE (dummy) 29.5

Importance given by consumers to different attributes when shoppingPrice (ten-point increasing scale) PRICE (continuous) 7.9Convenience (ten-point increasing scale) CONVENIENCE (continuous) 6.8Nutrition attributes NUTRITION b

Image attributes IMAGE b

Consumer’s lifestylesHealthy lifestyle LS_HEALTHY c

Hedonist lifestyle LS_HEDONIC c

Consumer’s information searchConsumer reads list of ingredients in food labels. . ..:Always (4) INGREDIENTS (ordered) 10.63Often (3) 22.38Once in a while (2) 39.88Never (1) 27.13

a 14% of respondents do not provide information on the income level and have been assigned average income for further analysis.b Results of a factor analysis (see Table A1 in Appendix A).c Results of a factor analysis (see Table A2 in Appendix A).

J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229 223

pilot survey of 20 consumers in each town to test for understand-ing and interview length.

The interviews were targeted to primary food buyers in thehousehold and conducted face to face by market research compa-nies2 in two medium-sized Spanish cities, Cordoba and Zaragoza,during March and April 2007. Sample size in both cities was set at400, resulting in a sampling error of ±5%, and a confidence level of95.5% when estimating proportions (p = q = 0.5; k = 2). A stratifiedrandom sample of consumers was made on the basis of districtand age. A number of representative grocery stores and supermar-kets were selected in each district, and shoppers were randomly se-lected outside these food outlets. In order to take into account thechanges in shopper characteristics that occur between differenttimes and days of the week, interviews were spread from Mondayto Saturday covering the full range of opening hours in each foodoutlet.

Summary statistics for the characteristics of the full sample arepresented in Table 1. The majority of respondents were female(72%). The respondent’s average age was about 45 years, living ina household with an average of three members. Around 52% ofrespondents stated that they had a net household monthly incomebetween €1500 and €3500 and over one third of the sample hasuniversity studies. Finally, the proportion of households with chil-dren younger than 6 years old was 19%, and with adults older than65 was 15.6%.

2 Identical questionnaires were used and survey protocols followed. Moreover, theresearch team de-briefed and trained the interviewers to assure maximum homo-geneity between the survey procedures employed in each city.

Definition of variables

Model constructsOne of the reasons for the limited amount of research focusing

on the analysis of factors affecting healthiness of consumer choiceis that of data availability and/or acquisition. It is very rare, evenunique, to be able to obtain in a single data set variables that reflectconsumer choice together with the possible factors explainingthem (Guthrie et al., 1995), while obtaining such information fromconsumer surveys is a difficult task. In order to overcome thisproblem we constructed a health trade-off index (HTI) using theinformation obtained from the questionnaire regarding trade-offsbetween health and other food product attributes. The trade-offquestions were worded as ‘‘When buying food products, do you usu-ally choose a healthier food product instead of another one whichis. . .?” This question was repeated three times for better tasting,easier to prepare and cheaper options.3 This index can be inter-preted as follows: a HTI value of one means that consumers givenegative answers to the all three trade-off questions, a value oftwo implies that they reject the health trade-off versus one of theother attributes (taste, convenience or price), a value of three thatthey reject the health trade-off health versus two of the other attri-butes; and the maximum value of four that they would always reject

3 From focus groups discussions and monitoring of the pilot survey we found thatwhen answering this question interviewees considered changes between productscategories to shift to healthier choices in a general diet selection setting. Thus, whatour index is measuring is whether consumers declare that they are willing to trade-off health for other attributes in their diets.

Page 4: Does nutrition information on food products lead to healthier food choices?

Table 2Definition of dependent variables and sample figures (%).

Variable Name (type) Value

Consumer’s nutrition knowledgeHigh (4) KNOWLEDGE (ordered) 21.63Medium (3) 30.25Low (2) 33.88Null (1) 14.25

Consumer uses the nutrition facts panel label when shopping. . ..:Always (4) PANEL_LABELUSE (ordered) 12.9Often (3) 29.8Once in a while (2) 32.9Never (1) 24.4

Consumers uses the nutrition or health claim label when shopping. . ..:Always (4) CLAIM_LABELUSE (ordered) 9.1Often (3) 21.3Once in a while (2) 41.7Never (1) 27.9

Health trade-off index (taste, convenience and price)4 HTI (ordered) 29.43 17.72 19.41 33.5

224 J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229

the health trade-off (i.e. they would always choose the healthy prod-uct).4 The descriptive statistics of this variable can be found inTable 2.

Nutrition label use is measured by two variables which reflectthe frequency of use of nutrition facts panel (PANEL_LABELUSE)and claims (CLAIM_LABELUSE). Finally, nutrition knowledge wasmeasured using an ‘‘objective knowledge” perspective to avoidbias from self-assessed nutrition knowledge measures. Buildingon existing literature on knowledge scales (Drichoutis et al.,2005; Barreiro-Hurlé et al., 2008) three questions were put to allinterviewees. Two of them were pair-wise comparisons regardingindividual nutrient content in different food products, that re-quested consumers to indicate which product had more choles-terol (butter or margarine) and which product had more fat(fried chicken or boiled chicken). The third was an open-endedquestion designed to elicit the number of daily servings of fruitand vegetables recommended by the health authorities.5 Answersto the three questions were used to create a knowledge index con-structed as follows. Individuals who gave correct answers to all threequestions were given a KNOWLEDGE value of four: those who gavecorrect answers to both pair-wise comparisons for specific nutrientcontent in foods and provided a close estimate of the recommendednumber of fruit and vegetable servings (i.e. 4 or 6) got a KNOWL-EDGE value of three; individuals who correctly answered bothpair-wise comparisons for specific nutrient content in foods6 wereassigned a KNOWLEDGE value of two and those who did not provideany correct answers were given a KNOWLEDGE value of one.

4 Notwithstanding the fact that the index is based on intentions and self-reporteddata it can be considered to measure the extent to which consumers view thehealthiness of foods to be a dominant feature in making food choices. From atheoretical perspective intentions have been accepted as the best predictor ofbehaviour (Ajzen, 1991) and factor affecting intentions coincide with those affectingreal behaviour (Fishbein and Manfredo, 1975). Moreover, ANOVA analysis, notreported here, HTI scores and self-reported dietary habits recorded in the survey (i.e.avoiding fat, reducing salt content) reveal a positive relationship between HTI andactual healthy eating habits.

5 At European level, health authorities encourage the daily intake of five servings offruit and vegetables.

6 Parmenter and Wardle (1999) state that individual items with more than 80% ofcorrect answers are not useful for nutritional knowledge scales. The fat-related pair-wise comparison, obtained a percentage of correct answers well above this threshold(98.5), while grouping allows correct answers to provide more information aboutindividuals nutritional knowledge (percentage of correct answers for the combinedfat-cholesterol-related pair-wise comparisons = 66.75).

Exogenous variablesThree variables capture consumer health habits and status.

Importance given to the effect diet choice has on their health statuswas rated using a ten-point increasing scale (FOOD_HEALTH). Tomeasure their health status, consumers were asked to state if theyhave a health problem related to dietary habits (HEALTH_PROB).Additionally, consumers were asked whether any member of thehousehold was following a special diet (DIET) and second, whetherany household member or close relative suffered from cardiovas-cular diseases (DISEASE). Finally, they were asked whether theycarried out voluntary annual health check ups (CHECKUPS), withless than half of the respondents stating that they did.

As far as involvement with food products consumers rated theimportance given to different food product attributes when shop-ping, such as price, convenience, etc. using a ten-point increasingscale. As the importance attached to each attribute was highly cor-related with that attached to others (except for price and conve-nience), instead of using them directly as explanatory variables, afactor analysis was carried out to identify the main underlying fac-tors (see Table A1 in Appendix A). Two factors were identified, onerelated to nutrition attributes (NUTRITION) and another related toimage attributes (IMAGE), and were included as explanatory vari-ables in the model.

Information reflecting lifestyles was obtained through individ-ual’s agreement with different statements based on a ten-pointincreasing scale, and due to the same correlation problem theywere included in the models after a factor analysis had been per-formed. This analysis identified two distinct factors (Table A2 inAppendix A) one related to healthy lifestyles (LS_HEALTHY), posi-tively related to avoiding fat and salt, following a balanced dietand not eating between meals, and another to a more hedonisticlifestyle (LS_HEDONISTIC), related to eating what one likes and fre-quent meals outside the home. Last, Consumers were also asked toreport their frequency of use of additional nutrition information la-bels such as list of ingredients (INGREDIENTS).

Econometric model specificationTwo models have been estimated, each of them comprising

three equations: (i) consumers’ choice of healthier food products,(ii) nutrition label use and (iii) nutrition knowledge. Both modelsdiffer only with regards to the endogenous variable used in the sec-ond equation, having considered the use of nutrition facts panel forone model and the use of nutrition/health claims for the other.

The first equation, specifies consumers’ choice of healthier foodproducts as an ordered probit as follows:

HTI�i ¼ kLU�i þ bXi þ ui ð1Þ

where LU�i designates the nutrition label use variables defined be-low (nutrition facts or claim use depending on the model), Xi is avector of all exogenous variables, and ui is the error term normallydistributed N(0, r2

u).The nutrition label use equation for both nutrition information,

nutrition facts panel and nutrition/health claims, is defined asfollows:

LU�i ¼ dK�i þ aZi þ ei ð2Þ

where K�i is the consumer’s nutrition knowledge defined below; Zi

contains all exogenous variables and ei is the error term normallydistributed N(0, r2

e ).Finally, nutrition knowledge is defined as:

K�i ¼ -Yi þ ni ð3Þ

where, Yi represents all the exogenous variables and fi is the nor-mally distributed error term N(0, r2

f ).All dependent variables are unobserved and what is observed is

the ranking of the index with regards to threshold parameters

Page 5: Does nutrition information on food products lead to healthier food choices?

J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229 225

which are estimated. To estimate the three equations we have as-sumed that the error terms (ui, ei and fi) may be correlated. Then,instead of independently estimating them, they are considered tobe a multivariate limited-dependent variable model, in which thethree error terms (ui, ei and fi) follow a multivariate normal distri-bution with mean zero and variance and covariance matrix X.Exogenous variables included in each equation (Xi, Zi and Yi) are de-fined in the literature review section and displayed using arrows inFig. 1.

Using cross-sectional data to estimate this model raises the is-sue of causality identification (i.e. whether or not some of the vari-ables considered to be exogenous are indeed endogenous). If suchbi-directional effects are presented estimating the models leads tobiased and inconsistent estimates. In our case, it is difficult to findout whether label use affected the choice of healthier food prod-

Table 3Endogeneity tests results for KNOWLEDGE, PANEL_LABELUSE, and CLAIM_LABELUSEvariables.

Variables Test

Hausmann(1978)

Davidson & McKinnon(1993)

Nutrition facts panel usemodela

8.00** 14.00***

KNOWLEDGE-PANEL_LABELUSEb

�1.98** �5.72***

PANEL_LABELUSE-HTIb �2.8*** �4.34***

Claims label use modela 10.00*** 18.00***

KNOWLEDGE-CLAIM_LABELUSEb

0.53 �0.09

CLAIM_LABELUSE-HTIb 7.74*** �7.97***

a Follows a v22 distribution.

b Follows a Student’s t distribution.** Denotes statistical significance at the 5 per cent significance level.

*** Denotes statistical significance at the 1 per cent significance level.

Table 4Estimates of the multivariate probit model: nutrition facts panel use.

Coefficients Knowledge N

Estimates t-ratio Es

Intercept 1.1045 16.99*** �CHILDREN 0.2743 2.91***

ELDERLY �0.1945 �1.75*

LOW_INCOME �0.2301 �2.22**

FOOD_HEALTHDIETHEALTH_PROBCHECKUPSDISEASEPRICE �CONVENIENCENUTRITIONIMAGELS_HEALTHY 0.1113 2.92***

LS_HEDONISTIC �INGREDIENTSCLAIM_LABELUSEKNOWLEDGEPANEL_LABELUSEN 800Log Likelihood �2.899

l2 1.0471 18.94***

l3 1.8973 28.24***

CorrelationsKnowledgeNutrition facts panel use

* Denotes statistical significance at the 10 per cent significance level.** Denotes statistical significance at the 5 per cent significance level.

*** Denotes statistical significance at the 1 per cent significance level.

ucts or whether people who chose healthier food products actuallyuse nutrition labels more (Grunert and Wills, 2007; Drichoutiset al., 2006). The same concern can be raised when we try to under-stand the effect of knowledge on label use; higher knowledge willlead to label use, yet the greater use of labels may increase con-sumer knowledge.

Exogeneity was tested for using Hausman’s Specification Test(Hausman, 1978) and the Durbin-Wu-Hausman (DWH) aug-mented regression test (Davidson and MacKinnon, 1993). If thenull hypothesis of exogeneity is rejected, the multivariate lim-ited-dependent variable models defined above should be esti-mated using some method that takes into account theendogeneity. Estimations have been made using Hajivassiliouand McFadden (1998) procedure implemented in Proc QLIM inthe SAS� 9.1 statistical software package.

Results

Table 3 shows the Likelihood Ratio test for the joint insignifi-cance of the predicted/residuals values in the augmented regres-sion for the two endogenous variables and the t-ratio test for theindividual insignificance of the predicted/residual values for eachvariable. For the nutrition facts panel use model, the null hypoth-esis of exogeneity is rejected by both tests. However, in the claimlabel use model, the individual insignificance of nutrition knowl-edge in the label use equation cannot be rejected. This leads to con-clude that knowledge and label use variables cannot be treated asexogenous in the right-hand side of Eqs. (1) and (2) and the estima-tion method suggested by Gourieroux (2000) has been applied. Aninitial model with all exogenous variables from Table 1 was esti-mated; however Tables 4 and 5 report the results for models whereonly exogenous variables statistically different from zero at a sig-nificance level of 0.10 or less are included. Correlations betweenequations are significant and thus, the simultaneous estimationof the three equations for both models is statistically justified. Also,

utrition facts panel use Health trade-off index

timates t-ratio Estimates t-ratio

0.5979 �1.89* �0.5461 �2.02***

0.1162 3.60***

0.2126 2.29** �0.1913 �2.10**

�0.2456 �2.77***

0.0605 �2.85***

0.0485 2.68***

0.2747 5.65***

0.2319 5.31***

0.0824 �2.00**

0.6903 12.98***

0.3300 6.46***

0.1956 2.19**

0.1491 2.90***

1.2261 18.90*** 0.5366 13.64***

2.6100 26.97*** 1.0570 20.21***

0.0746 1.74* 0.1325 3.18***

�0.0410 �0.91

Page 6: Does nutrition information on food products lead to healthier food choices?

Table 5Estimates of the multivariate probit model: claims label use.

Coefficients Knowledge Claims use Health trade-off index

Estimates t-ratio Estimates t-ratio Estimates t-ratio

Intercept 1.1086 17.10*** �0.4425 �1.24 �0.4780 �1.77*

CHILDREN 0.2651 2.81***

ELDERLY �0.2142 �1.94*

LOW_INCOME �0.2257 �2.18** �0.2055 �1.76*

FOOD_HEALTH 0.0651 1.99** 0.1148 3.48***

DIET �0.1745 �1.94*

HEALTH_PROB �0.2572 �2.95***

CHECKUPS 0.2447 2.94***

DISEASE 0.1944 2.27**

PRICE �0.0337 �1.63CONVENIENCENUTRITION 0.2468 5.21***

IMAGE �0.0810 �2.00**

LS_HEALTHY 0.1125 2.94*** 0.2499 5.86***

LS_HEDONISTIC 0.0907 2.25**

INGREDIENTS 0.1517 2.80***

PANEL_LABELUSE 0.3864 7.04***

KNOWLEDGE 0.0243 0.24CLAIM_LABELUSE 0.1223 1.75*

N 800Log Likelihood �2,948

l2 1.0459 18.93*** 1.3357 21.59*** 0.5332 13.64***

l3 1.8966 28.23*** 2.3415 27.10*** 1.0511 20.20***

CorrelationsKnowledge 0.0218 0.51 0.1363 3.28***

Nutrition facts panel use 0.1980 4.61***

* Denotes statistical significance at the 10 per cent significance level.** Denotes statistical significance at the 5 per cent significance level.

*** Denotes statistical significance at the 1 per cent significance level.

226 J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229

all threshold parameters in the equations are positive and signifi-cant at the 5% level, indicating that these variables indeed followan ordered sequence.

Nutrition knowledge

Significant variables and signs in the knowledge equation areconsistent for both types of labels, and include individual charac-teristics (LS_HEALTHY), economic conditions (LOW_INCOME) andconcerns with nutrition (CHILDREN, ELDERLY). As expected, CHIL-DREN is positively related to KNOWLEDGE, suggesting that house-holds with children are more concerned by nutrition issues andmore motivated to search for nutrition information, thus have ahigher level of nutrition knowledge. Also observing good healthhabits (LS_HEALTHY) positively affects nutrition knowledge, whilelower incomes do so negatively. A striking finding is that of EL-DERLY, being responsible for older people should also require moreattention to be paid to nutrition issues; however the negative signobtained leads us to believe this variable is acting as a proxy forrespondents’ age, which was not found to be significant. If these re-sults are compared with other is the Mediterranean region (Dric-houtis et al., 2005; Gracia et al. 2007) we can try to identify acommon characteristics related to behaviour on this issue. How-ever, the only common variable that increases knowledge in allthree studies is the combination of education and income, withhealthy lifestyles being significant in the two Spanish studies.

Nutrition labels use

Nutrition knowledge has a statistically positive significant ef-fect on nutrition facts panel use but a non significant effect onthe claim use. These results indicate that consumers with highernutrition knowledge are more likely to use the nutrition facts panelwhen shopping, while they show no difference with regards to thefrequency of claim use. The impact of nutrition knowledge on facts

panel use is consistent with previous studies (Guthrie et al., 1995;Szykman et al., 1997; Kim et al., 2001b; Drichoutis et al., 2005;Drichoutis et al., 2006 and Gracia et al., 2007), while the findingof lack of impact of nutrition knowledge on claim use is novel tothis study. The results of the endogeneity tests show that the useof nutrition facts panel increase nutrition knowledge while claimuse does not. This last finding contradicts the limited positive ef-fect of claim use on nutritional knowledge reported by Szykmanet al. (1997). It seems that consumers with higher levels of nutri-tion knowledge are more likely to use the nutrition facts panel be-cause they are more capable of understanding it.

Additionally, the variables influencing the use of each type ofinformation are different. None of the individual and economiccharacteristics of consumers were statistically significant inexplaining the level of nutrition panel use while income affectsthe use of claims. The negative parameter for LOW_INCOME indi-cates that as consumers’ income increases, so does the frequencyof using claims when shopping for food products. This result canbe explained due to higher opportunity costs, with high incomeconsumers relying more often on the more accessible informationpresented by claims. On the other hand, consumer health habitsand status influence the use of both the nutrition facts panel andclaims.

With regards to health status, objective measurement (DIET)promote the use nutrition facts panel while subjective ones (DIS-EASE and FOOD_HEALTH) promote the use of claims. Consumers’involvement in food products was also important in explainingthe use of both nutrition information labels. The negative valueof the PRICE coefficients indicates that those consumers who paymore attention to price when shopping are less likely to use nutri-tion facts panels. This result, consistent with Nayga et al. (1998),Drichoutis et al. (2005) and Gracia et al. (2007), might indicate thatprice sensitive consumers pay less attention to the nutrition con-tent of the food when shopping than non-price sensitive consum-ers. On the other hand, consumers who highly value the nutrition

Page 7: Does nutrition information on food products lead to healthier food choices?

Table 6Marginal effects of the nutrition label use ordered probit equation.

Variable Prob.LU = 1

Prob.LU = 2

Prob.LU = 3

Prob.LU = 4

KNOWLEDGEPanel label use

model�0.0448* �0.0080 0.0250 0.0278

Claim use model – – – –

PANEL_LABELUSE/CLAIM_LABELUSEPanel label use

model�0.0755* �0.0134 0.0421 0.0469

Claim use model �0.1067** �0.0134 0.0578* 0.0519

INGREDIENTSPanel label use

model�0.1580* �0.0281 0.0880 0.0981

Claim use model �0.0415** �0.0012 0.0135* 0.0122

LS_HEDONISTICPanel label use

model0.0189* 0.0034 �0.0105 �.0117

Claim use model �0.0250** �0.0007 0.0135* 0.0122

J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229 227

characteristics of the product (NUTRITION) are more likely to usenutrition fact panel and claims. Consumers who highly appreciateease of preparation (CONVENIENCE) are more likely to use thenutrition facts panel. This finding would reflect the fact that con-sumers concerned with convenience pay additional attention tonutrition information in order to be able to combine ease of prep-aration and healthy eating.

Consumers who pay more attention to the attractiveness andbrand of the product (IMAGE) are less likely to use the nutritioninformation and claims, as they rely more on the overall imageprovided by the brand (Drichoutis et al., 2005). Consumers’ life-styles also influence the type of nutrition label used, in that hedo-nists’ consumers are less likely to use the nutrition facts panel butmore likely to use the claims. Finally, if consumers use one labelthey will also use others. The estimated parameter for the variableINGREDIENTS is positive in both label use equations, indicatingthat consumers who read the list of ingredients more frequentlyare more likely to use either label. Moreover, the use of the nutri-tion facts panel positively influences the use of the claim and viceversa. This supports the majority of findings regarding the jointinfluence of claims and nutrition facts panel, which identify com-plementary use of both sources of information (Keller et al.,1997; Mazis and Raymond, 1997; Mitra et al., 1999; Szykmanet al., 1997; Garretson and Burton, 2001; and Kozup et al., 2003).

While most consumers use both nutrition facts panels andclaims, still a significant number of consumers use only one(33%) and of these, a majority reads only the nutrition facts paneland pay no attention to claims (68%).7 Both groups have been char-acterised using bi-variate analysis and from the results we can con-clude that consumers using only claims face stricter time constraints(higher income [INCOME p < 0.01] and a higher percentage of uni-versity education [p < 0.05]), consider that diet has a more importantimpact on health (FOOD_HEALTH p < 0.10), have healthier lifestyles(LS_HEALTHY p < 0.10) and are more willing to trade-off food attri-butes for healthy eating (HTI p < 0.01). On the other hand thosewho only use nutrition facts panels also more often read the list ofingredients (INGREDIENTS p < 0.01) and undergo more frequenthealth checkups (CHECKUPS p < 0.01). However, no significant dif-ferences were detected for the level of nutrition knowledge(p = 0.56). Thus, claims seem to be used not only by lower educa-tional strata as previously reported (Ippolito and Mathios, 1991),but also by individuals with higher health concerns and who are un-der more time constraints.

Together with findings from other studies in the Mediterraneanarea, we can identify a common pattern of nutrition label use in theMediterranean area. First, consumers with higher nutrition knowl-edge use nutrition facts panels more often, even when the issue ofendogeneity is taken into account. Moreover, consumers who usenutrition labels seem to be already using other types of labels,whether or not these are nutrition-related. On the other hand,the importance given to other non-nutrition-related food attri-butes and time constraints limit their use.

NUTRITIONPanel label use

model�0.0629* �0.00112 0.0351 0.0391

Claim use model �0.0682** �0.0019 0.0369* 0.0332

PRICEPanel label use

model0.0139* 0.0025 �0.0077 �0.0086

Claim use model 0.0093** 0.0030 �0.0578*** �0.0519***

For continuous variables, marginal effects are the partial derivatives of theprobabilities.For dummy variables; marginal effects are calculated taking the difference betweenthe predicted probabilities when the variables change from zero to one, everythingelse is held constant.

Health trade-off index

In both models, the coefficients of label use were statisticallysignificant, indicating that the higher the consumer use of eitherlabel the higher self-reported healthy food choices. Other factorsaffecting the health trade-off index are: consumer health habitand status (FOOD_HEALTH, DIET and HEALTH_PRO) and lifestyle(LS_HEALTHY). Perceived diet effectiveness has a positive impacton following healthier food choices, however, consumers who are

7 One hundred and eighty-three consumers read only nutrition facts panels, while85 read only claims.

following a special diet (DIET) and those who declare they suffera health problem related to food intake (HEALTH_PROB) show low-er intentions to do so. However, this finding must be taken withcaution as there could be an adverse selection problem betweenHTI and health status (i.e. people who are less willing to trade-off traditional food attributes for health may be following lesshealthy diets and therefore have diet-related health problems)and prescription diets do not necessarily require the consumptionof less tasty or convenient food.

In order to assess the effects of the independent variables, mar-ginal effects on event probability can be calculated. These marginaleffects provide a clearer interpretation of the results, which in turnallows comparisons to be made between the impacts of theexplanatory in both models. Tables 6 and 7 show the marginal ef-fects for those continuous and dummy variables which are signif-icant at 10% or better both for label use models.

The results presented in Table 6 indicate that the impact ofnutrition knowledge is only significant for the level representingno use of facts panel and not significant for claim use. Even correct-ing for endogeneity, nutrition knowledge still has an impact onnutrition facts panel use. Complementarity use of the differentnutrition labels is further reinforced by these results. In particularclaims and facts panel use reinforce the use of the other type of la-bel. However, the marginal effect of claim use on facts panel use ismuch stronger. The evidence thus appears to support that consum-ers who read claims read them in conjunction with the nutritionfacts panel, something further confirmed by the fact that fewerconsumers read only claims than only facts panels. An increasein the importance assigned by consumers to the price attribute de-creases the probability of reporting higher use of both facts panels

* Denotes statistical significance at the 10 per cent significance level.** Denotes statistical significance at the 5 per cent significance level.

*** Denotes statistical significance at the 1 per cent significance level.

Page 8: Does nutrition information on food products lead to healthier food choices?

Table 7Marginal effects of the health trade-off index ordered probit equation.

Variable Prob.HTI = 1

Prob.HTI = 2

Prob.HTI = 3

Prob.HTI = 4

PANEL_LABELUSE/CLAIM_LABELUSEPanel label use

model�0.0500*** �0.0049 0.0074 0.0476***

Claim use model �0.0413*** �0.0041 0.0061 0.0393***

LS_HEALTHYPanel label use

model�0.0778*** �0.0077 0.0115 0.0740***

Claim use model �0.0843*** �0.0083 0.0124 0.0803***

HEALTH_PROBPanel label use

model0.0823*** 0.0081 �0.0122 �0.0783***

Claim use model 0.0867*** 0.0086 �0.0128 �0.0825***

DIETPanel label use

model0.0645*** 0.0064 �0.0095 �0.0613***

Claim use model 0.0587*** 0.0058 �0.0086 �0.0559***

FOOD_HEALTHPanel label use

model�0.0391*** �0.0039 0.0058 0.0372***

Claim use model �0.0387*** �0.0038 0.0057 0.0368***

For continuous variables, marginal effects are the partial derivatives of theprobabilities.For dummy variables; marginal effects are calculated taking the difference betweenthe predicted probabilities when the variables change from zero to one, everythingelse is held constant.*** Denotes statistical significance at the 10 per cent significance level.

228 J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229

and claims, while a reverse impact is found for the importance as-signed to nutrition.

The most important conclusion that can be derived from theanalysis of marginal effects on the health trade-off index (Table 7)is that the impact of all the exogenous variables follows the samepattern for both label use models. First we should point out thatthe effect of label use on healthy food choices is stronger for thenutrition facts panel use. However, the presence of both types oflabels relates with a higher intention to follow healthier foodchoices, and authorising both labels broadens the scope of consum-ers that can benefit from them making healthier food choices.

6. Concluding remarks

In accordance with the objectives of this paper, our results pro-vide additional empirical evidence in Europe for the existence of apositive link between nutrition label use and healthy food choices.This is so even when existing bi-directional relationships betweennutrition knowledge, label use and healthier food choices are takeninto account. This effect exists for the two types of nutrition labelsconsidered and thus both nutrition label strategies may be capableof encouraging a shift in choice toward healthier food products.The reported findings are highly relevant to policy at a time whenEU legislation surrounding nutrition information on food labels isbeing updated. A new regulation for nutrition and health claimshas recently entered into force (OJ, 2006) while a new proposalhas been launched by the European Commission to modify thenutrition facts panel regulation (EC, 2008).

For the majority of consumers which use both labels, this willpositively influence the intention to follow healthier diets. Yet,for those using only one label the impact of each label in isolationwill also lead to the intention to improve eating habits. These find-ings provide clear support for a claims labelling strategy. If onlyfacts panels were allowed, consumers who use only claims would

not read these labels, loosing the chance to see their health trade-off intentions increased.

Based on these findings the authors would recommend themandatory introduction of a nutrition facts label as one of the toolsto influence consumer choice towards healthier diets. Yet thismeasure would not be sufficient as consumers with a food relatedhealth issue requiring dietary adjustments, although having agreater potential to benefit from healthy food choices are those lesswilling to trade-off price, taste or convenience above health. Addi-tional measures should be targeted to this group of consumers.

The findings also provide interesting new information on themotives underlying use of labels. Although the factors influencinguse may be different for the respective labels, a common set ofvariables can be identified for both. The well informed consumer(that often reads the list of ingredients or assigns more importanceto nutrition) will more likely use the nutrition label whereas themore price sensitive consumer is less likely to do so. Yet the con-sumer more interested in specific nutrient intake will use the factpanel, those concerned with general health issues will use theclaims panel, the same goes for those with a more hedonistic life-style. In terms of knowledge it is clear that the use of the nutritionfacts panel will lead to an increase in consumer knowledge, thisdoes not apply to the claims panel. Even when correcting for thiscircular impact, the greater the nutrition knowledge the greaterthe use of the facts panel, something we do not see with the claimspanel. In addition, comparing factors affecting label use in Mediter-ranean countries with those of the pan-European review of Grunertand Wills (2007) would indicate that there is no geographical dif-ference in this field.

Therefore to improve the impact of nutrition labelling somecomplementary policies should be considered. Clearly educationalprogrammes on nutrition would be a useful tool to encourage in-creased the use of the nutrition facts panel thereby contributingto the improved dietary habits and behaviour of the European pop-ulation. Yet for a subset group of consumers who only uses claims,authorising clear and truthful nutrition or health claims would leadto similar results.

To conclude that label use does indeed improve dietary pat-terns, these finding must be confirmed using dietary pattern data.It is not given that the declared intention to increase trade-offs be-tween food attributes and health will directly translate to dietarychange. We have treated food choice as declared trade-offs be-tween healthier food products versus better tasting, more conve-nient and cheaper ones, when the optimal data would come fromdietary assessment of real choices. However, this approach is theonly one currently available, as in Spain, and probably in otherEU countries, no database is available that combines real food pur-chase and nutrition knowledge, attitudes and practises. Moreover,we lack data on consumer behaviour before labelling was in wide-spread use and there might still be some individual unobservedcharacteristics that may overestimate the effect of label use onhealthier food choice.

An alternative approach to access real choice data on all theseconcepts for an identifiable sample could be the use of scannerdata. This is a natural extension of our research and would furthervalidate our findings. Yet there is a long way to go before this canbe undertaken and this would not necessarily solve the issue ofobservational bias unless we have access to a control group andlong term follow up. In the meantime, approaches such as theone presented here can provide support for the fact that use ofnutrition labels can lead to healthier diet choice.

Acknowledgements

The authors would like to thank, without implying, the editorand two reviewers for their comments which have made the pub-

Page 9: Does nutrition information on food products lead to healthier food choices?

J. Barreiro-Hurlé et al. / Food Policy 35 (2010) 221–229 229

lished version more focused. Editing by Sofie Bosma is also grate-fully acknowledged. This research is part of the DISOPTIPOL projectfunded by INIA-MEC and EU FEDER. JBH undertook this researchwhile contracted by the INIA-CCAA cooperative research systempost-doctoral incorporation scheme, partly funded by EU-ESF.

Appendix A

See Tables A1 and A2.

Table A1Factor loadings of importance given by consumers to different attributes whenpurchasing food products.

Attribute Averagerating

F1:NUTRITION

F2:IMAGE

Presence of nutrition information 6.53 0.8402 0.0138Presence of list of ingredients 6.68 0.8179 0.0192Organic production 5.40 0.7392 0.0289Presence of list of additives and

preservatives6.54 0.6904 0.0875

Product under Protected Designationof Origin

5.25 0.5901 0.1524

Known brand 6.82 0.0516 0.7968Appearance 7.93 0.0772 0.7665% of total variance 40.180 17.100Eigenvalue 2.810 1.190Cronbach’s alpha 0.710Kaiser–Meyer–Olkin measure of

sampling adequacy0.748

Table A2Factor loadings of consumer lifestyles and habits.

Statements Averageagreementscore

F1:LS_Healthy

F2:LS_Hedonistic

I avoid fat in the food I eat 6.71 0.7464 �0.0175I limit my salt intake 6.35 0.6687 �0.1855I follow a balanced diet 6.60 0.6983 �0.1150I avoid eating between meals 5.62 0.6799 0.1934I frequently have meals

outside my home3.57 �0.0466 0.7524

I eat food that tastes good 7.19 �0.0264 0.7377% of total variance 33.05 19.75Eigenvalue 1.98 1.18Cronbach’s alpha 0.430Kaiser–Meyer–Olkin

measure of samplingadequacy

0.687

References

Ajzen, I., 1991. The theory of planned behaviour. Organizational Behaviour andHuman Decision Processes 50, 179–211.

Barreiro-Hurlé, J., Colombo, S., Cantos, E., 2008. Is there a market for functionalwines? Consumer preferences and willingness-to-pay for resveratrol-enrichedquality red-wine. Food Quality and Preference 19 (4), 360–371.

Blaylock, J., Smallwood, D., Kassel, K., Variyam, J., Aldrich, L., 1999. Economics, foodchoices and nutrition. Food Policy 24 (2–3), 269–286.

Centre d’Ensenyament Superior de Nutrició i Dietética (CESNID), 2007. Estudi del’etiquetatge dels aliments. Agencia Catalana del Consumo, Barcelona.

Cowburn, G., Stockley, L., 2005. Consumer understanding and use of nutritionlabeling: a systematic review. Public Health Nutrition 8 (1), 21–28.

Davidson, R., MacKinnon, J.G., 1993. Estimation and Inference in Econometrics.Oxford University Press, Oxford.

Drichoutis, A., Lazaridis, P., Nayga, R., 2005. Nutrition knowledge and consumer useof nutritional food labels. European Review of Agricultural Economics 32 (1),93–118.

Drichoutis, A., Lazaridis, P., Nayga, R., 2006. Consumers use of nutritional labels: areview of research studies and issues. Academy of Marketing Science Review,vol. 9. <http://www.amsreview.org/articles/drichoutis09-2006.pdf>.

European Advisory Services (EAS), 2004. The introduction of mandatory nutritionlabeling in the European Union. Impact assessment undertaken for DG SANCO,European Commission, European Advisory Services, Brussels.

European Commission (EC), 2008. Proposal for a Regulation of the EuropeanParliament and the Council on the provision of food information to consumers.COM(2008)40 Final. Brussels, 85 pp. <http://www.ec.europa.eu/food/food/labellingnutrition/foodlabelling/publications/proposal_regulation_ep_council.pdf> (accessed 29.10.2009).

Fishbein, M., Manfredo, M., 1975. A theory of behaviour change. In: Manfredo (Ed.),Influencing Human Behaviour: Theory and Applications in Recreation, Tourismand Natural Resource Management. Sagamore, Champaign (IL), pp. 29–50.

Garretson, J.A., Burton, S., 2001. Effects of nutrition facts panel values, nutritionclaims and health claims on consumer attitudes, perceptions of disease-relatedrisks and trust. Journal of Public Policy and Marketing 19 (2), 213–227.

Gourieroux, C., 2000. Econometrics of Qualitative Dependent Variables. CambridgeUniversity Press, Cambridge.

Gracia, A., Loureiro, M.L., Nayga, R., 2007. Do consumers perceive benefits from theimplementation of a EU mandatory nutritional labelling program? Food Policy32, 160–174.

Grunert, K.G., Wills, J.M., 2007. A review of European research on consumerresponse to nutrition information on food labels. Journal of Public Health 15 (5),385–399.

Guthrie, J.F., Fox, J.J., Cleveland, L.E., Welsh, S., 1995. Who uses nutritional labelingand what effect does label use have on diet quality? Journal of NutritionEducation 27 (4), 173–192.

Hajivassiliou, V.A., McFadden, D., 1998. The method of simulated scores for theestimation of LDV models. Econometrica 66, 863–896.

Hausman, J.A., 1978. Specification test in econometrics. Econometrica 46 (6), 1251–1271.

Ippolito, P., 1999. How government policies share the food and nutritioninformation environment. Food Policy 24 (2-3), 295–306.

Ippolito, P., Mathios, A., 1991. Health claims in food marketing: evidence onknowledge and behaviour in the cereal market. Journal of Public Policy andMarketing 10 (1), 15–32.

Keller, S.B., Landry, M., Olson, J., Velliquette, A.M., 1997. The Effects of nutritionpackage claims, nutrition facts panels and motivation to process nutritioninformation on consumer product evaluation. Journal of Public Policy andMarketing 16 (2), 256–269.

Kim, S-Y., Nayga, R., Capps, O., 2000. The effects of food label use on nutrient intake:an endogenous switching regression analysis. Journal of Agricultural andResource Economics 25 (1), 215–231.

Kim, S-Y., Nayga, R., Capps, O., 2001a. Food label use, self-selectivity and dietquality. The Journal of Consumer Affairs 35 (2), 346–363.

Kim, S-Y., Nayga, R., Capps, O., 2001b. Health knowledge and consumer use ofnutritional labels: the issue revisited. Agricultural and Resource EconomicsReview 30 (1), 10–19.

Kozup, J.C., Creyer, E.H., Burton, S., 2003. Making healthful food choices: theinfluence of health claims and nutrition information on consumers’ evaluationof packaged food products and restaurant menu items. Journal of Marketing 67,19–34.

Mazis, M., Raymond, M., 1997. Consumer perceptions of health claims inadvertisements and on food labels. Journal of Consumer Affairs 31 (1), 10–26.

Mitra, A., Manoj, H., Ford, G.T., Ringold, D.J., 1999. Can the educationallydisadvantaged interpret the FDA-Mandate Nutrition Facts Panel in thepresence of an implied health claim. Journal of Public Policy and Marketing18 (1), 106–117.

Mokdad, A., Marks, J., Stroup, D., Gerberding, J., 2004. Actual causes of death in theUnited States. Journal of the American Medical Association 291 (10), 1238–1245.

Nayga, R., 1996. Determinants of consumers’ use of nutritional information on foodpackages. Journal of Agricultural and Applied Economics 28 (2), 303–312.

Nayga, R., 2000. Nutrition knowledge, gender and food label use. The Journal ofConsumers Affairs 34 (1), 97–112.

Nayga, R., Lipinski, D., Savur, N., 1998. Consumers’ use of nutritional labels whilefood shopping and at home. The Journal of Consumer Affairs 32 (1), 106–120.

OJ, 1990. Council Directive 90/496/EEC of 24 September 1990 on nutrition labellingfor foodstuffs. Official Journal of the EEC L 276, 06/10/1990 P., 0040–0044.

OJ, 2006. Regulation EC 1924/2006 of the European Parliament and the Council of20 December 2006 on nutritional and health claims made on foods. OfficialJournal of the European Union L 404 30/12/2006.

Parmenter, K., Wardle, J., 1999. Development of a general nutrition knowledgequestionnaire for adults. European Journal of Clinical Nutrition 53, 298–308.

Szykman, L.R., Bloom, P.N., Levy, A.S., 1997. A proposed model of the use of packageclaims and nutrition labels. Journal of Public Policy and Marketing 16 (1), 228–241.

Variyam, J.N., 2008. Do nutrition labels improve dietary outcomes. HealthEconomics 17 (6), 695–708.

Wang, G., Fletcher, S.M., Carley, D.H., 1995. Consumer utilization of food labeling asa source of nutrition information. The Journal of Consumer Affairs 29, 368–380.

Wansink, B., 2006. Mindless Eating: Why We Eat More Than We Think. Bantan-Dell,New York.

Williams, P.G., 2005. Consumer understanding and use of health claims for foods.Nutrition Reviews 63 (7), 256–264.

World Health Organization (WHO), (2002). The world health report 2002. Reducingrisks, promoting healthy life. Geneva: World Health Organization, 2002.