Some like it healthy: Can socio-demographic characteristics serve as predictors for a healthy food choice?

Food Quality and Preference 46 (2015) 103–112

Contents lists available at ScienceDirect

Food Quality and Preference journal homepage: www.elsevier.com/locate/foodqual

Some like it healthy: Can socio-demographic characteristics serve as predictors for a healthy food choice? Caterina Contini a,⇑, Leonardo Casini a, Violeta Stefan b, Caterina Romano a, Hans Jørn Juhl b, Liisa Lähteenmäki b, Gabriele Scozzafava a, Klaus G. Grunert b a b

Università degli Studi di Firenze, Florence, Italy Aarhus Universitet, MAPP Centre, Aarhus, Denmark

a r t i c l e

i n f o

Article history: Received 27 February 2015 Received in revised form 23 June 2015 Accepted 13 July 2015 Available online 17 July 2015 Keywords: Health claims Food choice Extra virgin olive oil Denmark Italy Latent class clustering

a b s t r a c t Authorising new health claims in Europe will favour the diffusion on the market of a greater number of foods with health claims. This scenario presents new opportunities to promote healthy food choices and launches new challenges to define strategies aimed at promoting products on the market. The literature suggests that our understanding of consumers’ sensitivity to health claims is still fragmented and should be further investigated. Our objective is to study the relationship between choice behaviour, attitudes and socio-demographic characteristics in order to evaluate the effectiveness of consumer characteristics in predicting consumers’ choice of products with health claims. Towards this end, we have conducted a choice experiment for extra-virgin olive oil on a sample of Danish (n = 1024) and Italian (n = 1000) consumers. Applying the latent class approach has enabled us to identify a niche of individuals sensitive of health claims and to characterise them with respect to the rest of the population. The results supply insights for the development of more targeted health promotion campaigns, as well as for actions in food marketing. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction The consumer’s response to the supply of foods with health-related information on the label is a question that has attracted the interest of various authors these past years. Studies focusing on consumer-related factors point out the importance of three key aspects that concern the attitudinal sphere, which are personal relevance, interest in nutritionally healthy eating and trust in health-related messages. As far as the first element is concerned, the literature points out that the vulnerability to diseases is a decisive factor in the choice of products with claims indicating the food’s beneficial function. In particular, studies underline that the perception of a benefit on the personal level influences the intention to use the product more significantly than the relevance it may have for a relative or a friend. It is also stressed that a claim represents an effective choice driver for consumers at risk of certain diseases, especially when it refers to specific diseases and carries detailed information on function and health benefits (Dean et al., 2012; Wong et al., 2013). ⇑ Corresponding author at: Department of Agricultural, Food and Forestry Systems, Università degli Studi di Firenze, P.le delle Cascine, 18, 50144 Florence, Italy. E-mail address: caterina.contini@unifi.it (C. Contini). http://dx.doi.org/10.1016/j.foodqual.2015.07.009 0950-3293/Ó 2015 Elsevier Ltd. All rights reserved.

Another attitudinal factor that plays a role in choosing a health claim is the interest in nutritionally healthy eating. This aspect has been evaluated by means of the general health interest scale (GHI), which consists of eight statements related to an interest in eating healthily (Roininen, Lähteenmäki, & Tuorila, 1999). Studies have demonstrated the capability of the GHI to predict the choice of low fat foods, such as, for example, an apple instead of a chocolate snack, while the relationship between the GHI and the preference for foods with nutrition and health-related claims proves to be more uncertain (Roininen et al., 2001). Finally, moving on to consider the relationship between trust in health-related messages and food choice, the literature states that the trust in a health claim increases the intention to choose a food with that claim and the propensity to consume it. Trust is indeed the key element so that the consumer effectively uses the information in his decision-making process (Annunziata & Vecchio, 2011; Saba et al., 2010; Worsley & Lea, 2003). The aforementioned suggests that personal relevance, interest in nutritionally healthy eating and trust in health related messages can be expected to be predictors for food choices involving foods with health-related messages. On the other hand, the relationships between the interest in health claims and the socio-demographic variables are much less

104

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

evident. This is a problem, as the latter are important indicators in order to develop marketing strategies and health-related policies, as they can be quantified and verified easily and thereby linked to issues of communicational accessibility. In the ambit of socio-demographic characteristics, the most significant predictor proves to be the country of residence (Saba et al., 2010; Van Trijp & Van der Lans, 2007). The literature interprets this phenomenon by associating it to various factors of the sociocultural environment, such as market exposure to health claims, trust in authorities, and familiarity with the carrier and the ingredient (Annunziata & Vecchio, 2011; Grunert et al., 2009; Saba et al., 2010). Conflicting results are recorded for gender, in relation to which several authors find a greater interest among women, while others observe no significant differences between the sexes (Ares & Gámbaro, 2007; Dean et al., 2007; Urala, Arvola, & Lähteenmäki, 2003; Urala & Lähteenmäki, 2007). The scenario for age also turns out somewhat unclear. Several studies indeed show that age has a positive influence on the attention to food healthiness and the interest in health claims (Siegrist, Stampfli, & Kastenholz, 2008), while others indicate that different age brackets evaluate the healthiness of foods with health claims in the same way (Ares, Giménez, & Gámbaro, 2009). Similar considerations can also be made for education, whose influence on consumer preferences has been studied in relation to functional foods. In this ambit, interest proves to be determined by product typology rather than by level of education. Consequently, generalisation does not appear to be justified (de Jong, Ocké, Branderhorst, & Friele, 2003; Verbeke, 2005). Finally, studies indicate that families with pre-adolescent children have a greater interest in information about nutrition on the label (Grunert & Wills, 2007), but the impact of the presence of children on the choice of foods with health claims remains to be further investigated. The current findings suggest that our understanding of consumers’ sensitivity to health claims is still fragmented and should be further investigated. The objective of our study is to clarify consumer behaviour regarding foods with health claims, analysing the relationship between the choice, the attitudinal factors and the socio-demographic characteristics in order to evaluate the effectiveness of consumer characteristics in predicting consumers’ choice of products with health claims. This analysis is fundamental for the European food sector, where the authorisation of new claims by the European institutions will favour the entry and diffusion on the market of a greater number of products with health claims. In this framework, knowing consumer preferences is of prime importance for public stakeholders, so that they can orient food choices towards a healthier diet and contribute to increasing public health and social wellbeing. Furthermore, a better knowledge of consumer behaviour represents a key element in defining strategies that target the promotion of products with health claims for marketers, too. Proposing to shed greater light on the relationship between consumer characteristics and the response to health claims, we have conducted a choice experiment on extra-virgin olive oil, which in 2012 obtained the authorisation to present a health claim on the label from the European Commission (European Commission, Reg. 432/2012). The recent acknowledgement of this product’s health-promoting properties on the European level makes the case study particularly interesting, as its results can provide significant practical implications for public and private operators of the agri-food system. The impact of the health claim on consumer preferences was evaluated in association with other attributes that contribute to the choice of the product. This approach enables us to analyse the importance of health claims in a context that includes important elements of the package information found in real purchase situations and to apply the analysis to a plausible context.

Another interesting aspect of the study lies in the fact that the experiment is conducted in two countries, Denmark and Italy, characterised by a different setting as far as the exposure to health claims, culinary tradition, and familiarity with the product are concerned. With the aim of finding out whether there is a segment of consumers responsive to health claims and how it differs from the rest of the population, we applied the latent class (LC) approach which permits an analysis of determinants of consumer choices, taking into account heterogeneity that may exist between different segments. In the result section, after describing the main segments that emerged, the paper will concentrate on the group most sensitive to health claims, analysing its attitudes and socio-demographic characteristics. The discussions propose new insights for the development of more targeted health consciousness campaigns and actions in food marketing. 2. Methodology 2.1. Overall approach Consumer preferences observed in both Italy and Denmark were merged into one dataset and analysed employing discrete choice models (McFadden, 1974). In addition, we adopted the latent class (LC) analysis approach to model estimation, which makes it possible to investigate heterogeneity by means of segmenting into groups with similar preferences (Greene & Hensher, 2003). The LC analysis outperforms the traditional segmentation techniques inasmuch as it is based on a likelihood model that permits statistical inference; furthermore, it permits us to evaluate the sufficient number of segments by comparing the information criteria as a function of the number of specified segments (Magidson & Vermunt, 2002; Train, 2003). In our study, the LC analysis was applied utilising the statistical software Latent Gold Choice 4.5 (Statistical Innovation Inc.). Then the predictors of preferences for health claims were identified, comparing the consumers most sensitive to health claims to the rest of the sample by means of Chi-squared Automatic Interaction Detection (CHAID) analysis. To do this we used SI-CHAID software, which is integrated with Latent Gold and makes it possible to grasp the degree of uncertainty associated with each individual’s belonging to a class. The CHAID analysis was also used to analyse the relationships between socio-demographic characteristics and the importance of attitudinal characteristics for the choice of olive oil in the sample. 2.2. Design of choice experiment The most important attributes of extra-virgin olive oil in consumer choice were selected by means of a literature review (Dekhili, Sirieix, & Cohen, 2011; Delgado, Santosa, & Guinard, 2013; García, Aragonés, & Poole, 2002; Manapace, Colson, Grebitus, & Facendola, 2011). The attributes identified were price, site of production and a health-related message. Regarding the choice of price level, 4 levels were considered. These were DKK 60, DKK 140, DKK 220, and DKK 300 for Denmark, and € 3, € 8, € 13, and € 18 for Italy (Table 1). They were identified starting from the 1st and the 99th percentiles of the price spread at which extra-virgin olive oil is purchased in the two countries (Nielsen data 2012). In particular, the two extreme levels were calculated increasing the 1st and 99th percentile by 50%. The other levels were defined by dividing the price interval defined by the extreme values into equal parts. Concerning the sites of production, the levels used in the analysis were Tuscany, Italy, EU, and Extra-EU (Table 1). Finally, the health related

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112 Table 1 Attributes and levels used in the experimental design. Attribute Price Levels

Site of production Levels

1 2 3 4

DKK 60 140 220 300

1 2 3 4

Extra EU EU IT Tuscany

Health related message Levels 1 2 3 4 5 6 7 8

Health claim Health claim Health claim Health claim Health claim Health claim Organic

€ 3 8 13 18

1 2 3 1 and Organic 2 and Organic 3 and Organic

Note: Health claim 1 ‘‘Olive oil polyphenols contribute to the protection of blood lipids from oxidative stress. The beneficial effect is obtained with a daily intake of 20 g of olive oil’’; Health claim 2 ‘‘2 tablespoons (ca. 20 g.) of olive oil per day may reduce the risk of coronary heart disease’’; Health claim 3 ‘‘20 g of this olive oil per day helps to protect blood lipids from harmful reactions’’.

message is made up of two different elements: the health claim and organic certification. The decision to consider these two indications in a single attribute is based on the intent to analyse the interactions between the two elements that the consumers typically associate with food healthiness (Aschemann-Witzel, Maroscheck, & Hamm, 2013; Magnusson, Arvola, Hursti, Åberg, & Sjöden, 2003), and to obtain a more detailed picture of the response to health-related stimuli. In particular, the levels for this attribute were the result of all the possible combinations of organic certification with three different formulations of the health claim (Table 1). The first formulation was the claim authorised by the European Commission (Reg. 432/2012) ‘‘Olive oil polyphenols contribute to the protection of blood lipids from oxidative stress. The beneficial effect is obtained with a daily intake of 20 g of olive oil’’; the second was based on the other officially recognised claim for olive oil, which is the one authorised by the FDA in the USA ‘‘2 tablespoons (ca. 20 g) of olive oil per day may reduce the risk of coronary heart disease’’. Finally, the third was formulated in a more accessible language than the European claim without reference to a specific disease ‘‘20 g of this olive oil per day helps to protect blood lipids from harmful reactions’’. The decision to use different phrasings to convey the same message was motivated by the concern to comprise the possible differences in impact that the different formulations may have on the preferences of consumers. In addition to these three options there was a ‘‘no claim’’ level. The choice experiment design was then structured taking into consideration 2 attributes with 4 levels and 1 attribute with 8 levels. The choice experiment was a pairwise comparison between two options, which also provided for the possibility to not choose any of the proposed alternatives. Starting from 8128 possible pairs, we reduced their number with an orthogonal fractional factorial design, which satisfies attribute level balance and allows for an independent estimation of the influence of each attribute on choice. This proves to be the most effective solution in cases where there is no knowledge about the prior parameter values (Green & Srinivasan, 1990; Louviere,

105

Hensher & Swait, 2000; Bliemer & Rose, 2005). The design was obtained by means of Ngene software (ChoiceMetrics Ltd.). It produced 64 pairs divided into 4 blocks of 16 sets each. The choice sets obtained in this manner were presented to respondents in the form of images of labels differentiated by site of production, price, and health-related message in which the organic certification is expressed by the European logo and the wording, ‘‘organic’’ (Fig. 1). Respondents were asked to imagine that while shopping they come across two 1-l packages of extra-virgin olive oil and have to decide which product to purchase. In order to understand the no-choice behaviour, the respondents who had selected the no-choice option for more than 50% of the choice sets were asked to specify the reasons for this behaviour, choosing from a series of options that motivated their decision and providing information not included in our experimental design, such as the exclusive preference for a brand, a different area of origin or a different product format, the habit of purchasing directly from the producer, the need to know the type of crushing method, or other. 2.3. Other measures Other measures selected are divided into attitudinal and socio-demographic measures. The former include personal relevance of cholesterol and cardiovascular diseases, interest in nutritionally healthy eating, evaluated by means of the GHI and following the methodology proposed by Roininen et al., 1999, and trust in health claims evaluated by means of a seven-point Likert scale anchored with degrees of agreement and disagreement with the following statement ‘‘I believe that foods with health claims fulfil their promises’’ (Urala & Lähteenmäki, 2007). The socio-demographic characteristics concern the country of residence, gender, age, education, presence of children (younger than 12 years old), and family income. We also considered the familiarity with the product, expressed as the frequency of use of olive oil,

Fig. 1. Example of stimulus presented to respondents in the choice experiments.

106

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

Table 2 Sample composition.

3. Results

Denmark

* ** ***

Italy

Absolute figures

%

Education Low* Average** High***

210 534 280

Age Under 35 35–55 Over 55

3.1. The latent classes

Total sample

Absolute figures

%

Absolute figures

%

21 52 27

117 580 303

12 58 30

327 1114 583

16 55 29

278 470 275

27 46 27

366 530 104

37 53 10

644 1000 380

32 49 19

Gender Male Female

436 588

43 57

513 487

51 49

949 1075

47 53

Total

1024

100

1000

100

2024

100

Completion of grade 8 in Italy and of grade 10 in Denmark or a lower degree. Secondary education. Higher education level.

inasmuch as the literature indicates that this is an important factor to interpret the different attitude towards food in general (Lähteenmäki et al., 2010). Finally, the main reasons for purchasing olive oil were investigated.

2.4. The sample The questionnaires were administered on line in December 2013 to a sample representative for age, gender, and education (Table 2). In order to qualify respondents had to be consumers of olive oil. The analysis is based on all completed questionnaires, altogether 2024 (1024 in Denmark and 1000 in Italy). The sample shows a clear difference between Danish and Italian consumers in terms of familiarity with the product. Indeed most Italians (82%) use olive oil every day, while only a small minority (6 individuals out of 1000) uses it less than once per week. The situation is different for the Danish consumers: 13% consume olive oil every day, 54% consume it at least once per week, while the frequency of the remaining 34% is even less. We find similarities between the two countries, however, concerning the principal motivations tied to the consumption of extra-virgin olive oil, which are healthiness (stated as important by 89% of the Italians and 75% of the Danish consumers), taste (cited as important by 89% of the Italians and 75% of the Danes) and naturalness (indicated as important by 85% of the Italians and 70% of the Danes).

The analysis of the information criteria (Table 3), in combination with the significance of the parameter estimates and the meaningfulness of the parameter signs has led us to choose the 8-class model (Table 4). The importance of the attributes has been calculated for each of the classes on the basis of the relative maximum effect, normalised with respect to the sum of all the magnitudes of the estimated parameter values (Table 5). Through the combined analysis of the attribute importance and of the class specific attribute level part worths (Tables 5 and 4), the 8 classes can be grouped according to 4 principal tendencies: price sensitivity (52%), importance of the production area (23%), inclination towards the no-choice option (7%), and importance of health-oriented messages (18%). The first consideration concerns the importance of familiarity with the product. Our results show that based on the frequency of use of olive oil, it is possible to distinguish the consumers who choose oil for the qualitative attributes from those who do not choose or mostly choose on the basis of price. In particular, the price sensitive consumers are segmented into 3 different classes (classes 1, 2 and 3). These three classes are made up mainly of consumers who are not daily users of olive oil (that is to say, they use it less than 5 times per week) (Table 6). Moreover, as far as the socio-demographic characteristics are concerned, there is a prevalence of Danish consumers (64%) and, in particular, younger consumers (less than 35 years of age). Indeed, 79% of the young Danes belong to this segment. The main difference between class 1, 2 and 3 concerns the inclination to choose. Indeed, the individuals belonging to the first class choose one of the products in the choice set, while the no-choice option prevails for the other two classes. In this regard, the probability to choose oil is 47% for the consumers in the second class (cl. 2), and drops to 10% for the consumers in the third class (cl. 3). Analysing the parameters shows that this difference depends on the magnitude of the price interval at which individuals are willing to purchase oil: larger for the second class and limited to the lower price for the third class. Twenty-three percent of consumers (14 + 9%) are mostly influenced by the site of production, preferring European, Italian or Tuscan productions to non-European productions. These individuals habitually use olive oil (Table 6) and it comes as no surprise that the Italians (74%) prevail in this group, considering the greater importance of olive oil in Italian cooking habits. In this case too, two different behaviours can be distinguished. One type of consumer (class 4) tends to choose one of the products, while the

Table 3 Summary of latent class cluster models. Model 1-Cluster model 2-Cluster model 3-Cluster model 4-Cluster model 5-Cluster model 6-Cluster model 7-Cluster model 8-Cluster model 9-Cluster model 10-Cluster model

LL 28400.1 23749.8 22052.7 21011.8 20215 19757.9 19358.1 19086.6 18826.4 18600.7

BIC(LL)

AIC(LL)

AIC3(LL)

CAIC(LL)

Npar

56906.84 47720.38 44440.28 42472.72 40993.36 40193.43 39507.88 39079.07 38672.86 38335.63

56828.27 47557.61 44193.31 42141.56 40578.01 39693.89 38924.14 38411.14 37920.74 37499.32

56842.27 47586.61 44237.31 42200.56 40652.01 39782.89 39028.14 38530.14 38054.74 37648.32

56920.84 47749.38 44484.28 42531.72 41067.36 40282.43 39611.88 39198.07 38806.86 38484.63

14 29 44 59 74 89 104 119 134 149

Note: LL = Log-likelihood; BIC(LL) = Bayesian information criterion based on the log-likelihood; AIC(LL) = Akaike information criterion based on the log-likelihood; AIC3(LL) = Akaike information criterion, with 3 as penalising factor, based on the log-likelihood; CAIC (LL) = Consistent Akaike information criterion, based on the loglikelihood; Npar = number of estimated parameters.

107

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112 Table 4 Estimates of parameters of the latent class model with eight segments. Attributes

Price Level Level Level Level

1 2 3 4

Price sensitive

Interested in the production area

Inclined towards the no-choice option

Interested in health-oriented messages

Class1

Class2

Class3

Class4

Class5

Class6

Class7

Class8

0

0

0

0

0 1.38⁄⁄ 0.45⁄⁄ 0,14

0 0.07 1.09⁄⁄ 1.41⁄⁄

0 0.71⁄⁄ 0.61⁄⁄ 0.42⁄⁄

0

2.44⁄⁄ 4.95⁄⁄ 7.09⁄⁄

0.81⁄⁄ 3.23⁄⁄ 4.12⁄⁄

4.87⁄⁄ 7.20⁄⁄ 7.09⁄⁄

0.67⁄⁄ 1.63⁄⁄ 2.49⁄⁄

0.31⁄ 1.04⁄⁄ 1.91⁄⁄

Site of production Extra EU EU IT Tuscany

0 0.47⁄⁄ 0.60⁄⁄ 1.13⁄⁄

0 0.76⁄⁄ 1.29⁄⁄ 1.47⁄⁄

0 1.17⁄⁄ 1.57⁄⁄ 1.57⁄⁄

0 3.09⁄⁄ 4.13⁄⁄ 4.95⁄⁄

0 0.98⁄⁄ 4.08⁄⁄ 4.30⁄⁄

0 0.81⁄⁄ 0.93⁄⁄ 0.57⁄⁄

0 0.07 0.77⁄⁄ 0.99⁄⁄

0 0.95⁄⁄ 1.27⁄⁄ 1.41⁄⁄

Health oriented messages No message Health claim 1 Health claim 2 Health claim 3 Health claim 1 and organic Health claim 2 and organic Health claim 3 and organic Organic

0 0.77⁄⁄ 0.78⁄⁄ 0.96⁄⁄ 2.39⁄⁄ 1.91⁄⁄ 1.91⁄⁄ 1.09⁄⁄

0

0 0.45⁄ 0.71⁄⁄ 1.34⁄⁄ 1.11⁄⁄ 0.66⁄⁄ 0.93⁄⁄ 1.13⁄⁄

0 0.15 0.09 0.19 1.07⁄⁄ 1.18⁄⁄ 1.10⁄⁄ 0.80⁄⁄

0 0.62⁄⁄ 0.91⁄⁄ 0.74⁄⁄ 1.32⁄⁄ 1.53⁄⁄ 1.30⁄⁄ 1.16⁄⁄

0

0.01 0.40⁄⁄ 0.19 0.78⁄⁄ 0.65⁄⁄ 0.37⁄⁄ 0.30⁄⁄⁄

1.23⁄⁄ 0.59⁄⁄ 0.96⁄⁄ 0.83⁄⁄ 0.45⁄ 0.13 0.29

0 1.08⁄⁄ 0.95⁄⁄ 0.70⁄⁄ 1.12⁄⁄ 1.33⁄⁄ 1.18⁄⁄ 0.36⁄⁄

0 0.16 0.31⁄ 0.28 3.05⁄⁄ 3.33⁄⁄ 3.00⁄⁄ 3.11⁄⁄

Asc No choice Choice

0 5.52⁄⁄

0

0

0 2.99⁄⁄

0

0 2.75⁄⁄

0 3.00⁄⁄

0 2.29⁄⁄

Class size

23%

18%

14%

9%

10%

8%

2.21⁄⁄

0.11

12%

0.57⁄⁄

7%

Note: ⁄ and ⁄⁄ denote significance at the 5% and 1% level, respectively. Health claim 1 ‘‘2 tablespoons (ca. 20 g.) of olive oil per day may reduce the risk of coronary heart disease’’; Health claim 2 ‘‘Olive oil polyphenols contribute to the protection of blood lipids from oxidative stress. The beneficial effect is obtained with a daily intake of 20 g of olive oil’’. Health claim 3 ‘‘20 g of this olive oil per day helps to protect blood lipids from harmful reactions’’.

Table 5 Relative importance of the attributes for the eight classes identified.

Price (%) Site of production (%) Health oriented messages (%)

Class1

Class2

Class3

Class4

Class5

Class6

Class7

Class8

67 11 23

65 23 12

71 16 13

29 57 14

21 59 21

41 26 34

24 33 44

29 21 50

Table 6 Frequency of use of extra-virgin olive oil in the eight classes identified.

Every day (%) 5–6 times a week (%) 3–4 times a week (%) 1–2 times a week (%) 2–3 times a month (%) Once a month (%)

Class 1

Class 2

Class 3

Class 4

Class 5

Class 6

Class 7

Class 8

32 13 15 16 11 13

46 11 13 13 9 8

30 8 16 18 15 14

58 15 8 9 6 4

85 6 3 3 2 1

29 11 7 16 12 26

69 14 8 4 2 2

42 17 19 11 6 6

other (class 5) prefers not to choose. The latter tendency prevails among the consumers who purchase oil in the area of origin, directly from the producer, rather than in a shop. In this case, oil is generally purchased in bulk without any label. Consequently, the behaviour of the purchaser who buys directly from the producer is mainly influenced by the relationship with the producer himself, who becomes the guarantor of quality. Therefore, we can reasonably assume that the preference for the no-choice option is related to the fact that the indications on the label used in our choice experiment were not enough for the respondent to make his choice. A third tendency (class 6) refers to those who do not purchase oil if the label does not carry a brand, which they feel indispensable

for their choice. For these consumers, olive oil carries little importance in their diet, and more than a quarter of them use this food no more than once per month (Table 6). This group mainly features Danish consumers (70%), more than 34 years of age (79%), and with a middle-low level of education (80%). Finally, the choices of 18% of consumers prove to be notably influenced by health-oriented messages. These consumers have a marked familiarity with the product, and most of them consume oil more than 4 times per week (Table 6). The analysis of the class specific attribute level part worths show that a part of these (class 8) (8%) prefers to purchase products with an organic certification, while the other (class 7) (10%) mainly chooses on the basis of the presence of a health claim without showing a substantial

108

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

difference between the different proposed phrasings. Class 7 will be the object of further analyses and will be termed ‘‘consumers sensitive to health claims’’.

3.2. Predictors of the preferences for foods with a health claim The class of consumers sensitive to health claims obtained by applying the LC model was subjected to further analyses. The results of a set of Chi-squared Automatic Interaction Detection analysis are illustrated in Figs. 2–5 and discussed below. Among the attitudinal characteristics, trust in health claims (LR chi square = 85.44; df = 3; p < 0.0001) proves to be a predictor of class membership, with individuals sensitive to health claims manifesting a marked trust compared to the rest of the population (Fig. 2). It is interesting to underline, in this regard, that the trust in health claims also represents the major difference between the consumers who choose oil with an organic certification and those who prefer the health claim (LR chi square = 30.56; df = 2; p < 0.001). In particular, the individuals sensitive to health claims prove to have greater trust than those who prefer organic products (class 8). Our analysis confirms the importance of the two other attitudinal characteristics as well – interest in nutritionally healthy eating (LR chi square = 17.72; df = 3; p = 0.0034) and personal relevance (LR chi square = 10.11; df = 1; p = 0.0015) – and sheds light on how all 3 variables examined can be considered as valid predictors of the preferences of consumers sensitive to health claims. We have thus verified the importance of the three key attitudinal factors and confirmed the results of the other studies on the topic. Moving on to consider the influence of socio-demographic characteristics on being sensitive to health claims, the results show how the country of residence (LR chi square = 139.72; df = 1; p < 0.0001), gender (LR chi square = 5.78; df = 1; p = 0.016) and the presence of children (LR chi square = 7.63; df = 1; p = 0.005) significantly influence the likelihood of being in the class sensitive to health claims (Fig. 2), while age (LR chi square = 3.94; df = 1; p = 0.092) does not achieve standard levels of statistical significance and education (LR chi square = 0; df = 0; p = 1) appears to be insignificant. It therefore seems necessary to investigate the relationships between socio-demographic variables and attitudinal factors for a better understanding of the results. Our findings show the presence of a significant relationship between interest in nutritionally healthy eating and age (LR chi square = 49.22; df = 12; p < 0.0001), gender (LR chi square = 37.29; df = 6; p < 0.0001), presence of children (LR chi square = 23.02; df = 6; p = 0.0008), and level of education (LR chi square = 19.73; df = 6; p = 0.0060) (Fig. 3). In particular, the elderly (older than 65), women, families with children, and persons with a higher education (equal to or higher than secondary school) show a greater interest in nutritionally healthy eating. A significant difference is also noted between countries (LR chi square = 77.69; df = 6; p < 0.0001), with a prevalence of Italian consumers among those who pay greater attention to the health aspects of foods. The trust in health claims proves to significantly differ for country (LR chi square = 450.33; df = 6; p < 0.0001), age (LR chi square = 48.36; df = 12; p < 0.0001) and presence of children (LR chi square = 59.50; df = 6; p < 0.0001), while gender and education are not significant statistically (Fig. 4). We observe, in particular, a greater trust among Italian consumers, younger individuals, and families with children. As far as the country is concerned, our results show a low trust index in health claims among the Danish consumers. Finally, personal relevance proves to be significant for age (LR chi square = 45.44; df = 2; p < 0.0001) and gender (LR chi square = 14.08; df = 1; p = 0.00018), with a greater personal relevance in men and in the elderly (Fig. 5).

Fig. 2. Profile of consumers sensitive to health claims with respect to the attitudinal and socio-demographic variables that the CHAID analysis has shown as significantly different from the rest of the population. Education is not included in the picture, since it does not distinguish significantly between consumers sensitive to health claims and the rest of the sample. Numbers in brackets indicate which categories have been merged by the CHAID algorithm. The broken line histogram indicates the profile of the whole sample.

4. Discussion and conclusions The attention for the health-related characteristics of foods is a phenomenon that is spreading in contemporary society (Grunert & Wills, 2007), but it is one that still concerns a niche of the population (Casini, Contini, Romano, & Scozzafava, 2015). The interest in health-related properties therefore risks remaining latent the moment that consumers’ preferences are analysed as a whole. Applying the LC approach to the declared preferences of a significant sample of the Danish and Italian populations has permitted us to observe the importance that health claims assume in choice behaviour, identifying the consumer segment most sensitive to health messages and characterising it on the basis of attitudinal and socio-demographic aspects. The results obtained have permitted us to propose an interpretative model of the relationships between health claim, socio-demographic characteristics, attitudes, and choice behaviour. Our model described in Fig. 6 points out that the relationship between the health claim and choice behaviour depends on a combination of factors of an attitudinal nature. Coherently with the findings in the literature, the choice decision originates in the presence of a specific interest in the information on the label induced by a general interest in nutritionally healthy eating and/or personal relevance (Dean et al., 2012; Lähteenmäki, 2013; Wong et al.,

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

109

Fig. 3. Consumer profile defined by means of the relationship between the socio-demographic variables and the different levels of the general health interest scale (GHI).

Fig. 4. Consumer profile defined by means of the relationship between the socio-demographic variables and the different levels of trust in health claims.

110

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

Fig. 5. Consumer profile defined by means of the relationship between the socio-demographic variables and personal relevance.

2013). On the other hand, the use the consumer makes of this information depends on his trust in the message, which becomes responsible for the information effectiveness (Worsley & Lea, 2003). Our analysis integrates this framework with the addition of the role of the socio-demographic characteristics, showing how these are tied significantly to one or more attitudinal factors. In particular, some socio-demographic characteristics have impacts of the same sign on attitudinal variables. Such is the case of the presence of children, which favours two attitudes and consequently a higher preference for health claims. This finding is coherent with the literature (Grunert & Wills, 2007), but it also adds another element to our knowledge in the ambit of consumer behaviour, directly tying the presence of children to the choice of the health claim. The same reasoning is valid for the country of residence, for which our

results confirm the literature that reports that the differences between countries represents a valid predictor of the choice of health claims (Saba et al., 2010; Van Trijp & Van der Lans, 2007). In particular, our results show a lower GHI and a lack of trust in health claims among the Danish consumers, which can be related to this country’s recent market exposure to health claims (Grunert et al., 2009; Lähteenmäki et al., 2010). In other cases though, the same socio-demographic variable has a different influence on the attitudinal factors, favouring one attitude and limiting another at the same time. These different effects determine consequences on the final choice that are not always predictable, and do not always permit identifying a univocal relationship between the socio-demographic characteristics and choice behaviour. This result enables us to interpret the different choice behaviours observed in literature by gender (Ares &

Fig. 6. Interpretative model of the relationships between health claim, socio-demographic characteristics, attitudes, and choice behaviour. Non-significant relationships between socio-demographics and attitudes are omitted.

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

Gámbaro, 2007; Dean et al., 2007; Urala & Lähteenmäki, 2007; Urala et al., 2003), age (Ares & Gámbaro, 2007; Ares et al., 2009) and education (de Jong et al., 2003; Verbeke, 2005). As for gender, it is pointed out how this variable has a predictive capability that can mainly be attributed to the type of health benefit associated with the product. Indeed, for olive oil, the greater relevance for males corresponds to a clear and positive response to health claims. And this occurs despite the fact that being a woman involves greater attention to healthy eating. Bearing these results in mind, we can reasonably suppose that the effectiveness of gender as a predictor can further increase in the case in which gender acts in the same manner on personal relevance and on interest in healthy eating. As far as age is concerned, we observe only a weakly significant relationship with choice, with a prevalence of young people among those who choose the health claim. This phenomenon can be explained by the different effects of age on the attitudinal variables. Indeed, these effects, on one hand, would determine an increased preference for a health claim (personal relevance and interest in nutritionally healthy eating), while on the other hand, it would determine a reduction (trust in health claims). Therefore, the combination of these effects is not manifested in a clearly distinct behaviour of young people compared to the elderly, and does not enable us to univocally associate age to a given behaviour. The education variable does not prove to influence choice significantly, despite the fact that it has a positive effect on the interest in nutritionally healthy eating. This phenomenon can be interpreted with the fact that the more educated also have a good knowledge of the product’s nutraceutical properties and, consequently, do not orient their preferences on the basis of the health claim presence (Verbeke, Vermeir, & Brunsø, 2007). A final reflection is warranted for the differences in the choice behaviours of the two countries considered in our study, which differ substantially with respect to two factors. The first concerns the trust in claims and, consequently, the effectiveness of health messages, while the second factor concerns the product’s importance in food habits. In particular, familiarity appears to be principally responsible for the greater attention to quality. Danish consumers, for whom the product plays a more marginal role in their food habits, are indeed principally influenced by price, while the Italians devote greater attention to the quality attributes, the most important of which is the area of origin. The importance of origin for the Italian consumers may furthermore be motivated by Italy’s role as an oil producing country, which can translate into associating the product with ‘‘something of our own’’. The results expounded here can represent a useful information basis for the development of more targeted health consciousness campaigns, as well as for actions in food marketing. The forthcoming diffusion on the European market of a greater number of products with health claims can represent an interesting challenge in the ambit of improving the quality of people’s diet, especially in countries where consumers have had the opportunity to acquire a certain familiarity with health claims. In countries where there has been less exposure to health claims, communication campaigns aimed at familiarising consumers with this type of information will prove fundamental, fostering message comprehension and trust. In a scenario of this type, those who could most benefit from the diffusion of products with health claims are represented by families with children, which display a marked sensitivity to health-related information on labels. The situation is a bit more complex for the elderly, inasmuch as on one hand, they show a great interest in health-related information but, on the other hand, they have little trust in it. Transforming the potential interest in products with health claims into effective choice behaviours must

111

necessarily involve communication campaigns aimed at guaranteeing the truthfulness of health-related messages. For the operators of the agri-food sector too, the diffusion of foods with health claims can represent an interesting opportunity to grab by means of implementing marketing strategies aimed at the different consumer segments. A key element in this ambit consists in the specific health benefit that can be associated to the food product considered. Benefits more targeted at specific socio-demographic segments will require quite differentiated marketing strategies, capable of supporting the weaker attitudinal variables of the different segments. Our study has contributed to drawing a clearer picture of the relationships between socio-demographic and attitudinal characteristics and choice behaviours, which can be of great help in developing new products and, especially, in implementing specific marketing strategies. This type of study appears even more decisive in the hypothesis of opening up to new markets, even in very different sociocultural contexts, where the relationships we have described could take on new facets and shed light on new elements that could contribute to further defining the overall picture. A challenging object of further study could concern integrating analysis with a closer look at the aspects tied to understanding the health message and the role of communication in determining choice behaviour. These further studies could contribute to shed light on several aspects that our paper targets less, such as those concerning the differences between age brackets for which a different understanding could play an important role. The implications of these further analyses would also be important for the food sector and to promote healthy food choices, considering the importance that the choice of language plays in communication campaigns.

References Annunziata, A., & Vecchio, R. (2011). Functional foods development in the European market: A consumer perspective. Journal of Functional Foods, 3, 223–228. Ares, G., & Gámbaro, A. (2007). Influence of gender, age and motives underlying food choice on perceived healthiness and willingness to try functional foods. Appetite, 49(1), 148–158. Ares, G., Giménez, A., & Gámbaro, A. (2009). Consumer perceived healthiness and willingness to try functional milk desserts. Influence of ingredient, ingredient name and health claim. Food Quality and Preference, 20(1), 50–56. Aschemann-Witzel, J., Maroscheck, N., & Hamm, U. (2013). Are organic consumers preferring or avoiding foods with nutrition and health claims? Food Quality and Preference, 30, 68–76. Bliemer, M.C.J. & Rose, J.M. (2005). Efficiency and sample size requirements for stated choice studies, Working Paper ITLS-WP-05-08, Institute of Transport and Logistics Studies. Casini, L., Contini, C., Romano, C., & Scozzafava, G. (2015). Trends in food consumptions: What is happening to Generation X? British Food Journal, 117(2), 705–718. Commission Regulation (EU) No 432/2012, of 16 May 2012 establishing a list of permitted health claims made on foods, other than those referring to the reduction of disease risk and to children’s development and health. Official Journal of the European Union, L 136/1. (). de Jong, N., Ocké, M. C., Branderhorst, H. A. C., & Friele, R. (2003). Demographic and lifestyle characteristics of functional food consumers and dietary supplement users. British Journal of Nutrition, 89, 273–281. Dean, M., Lampila, P., Shepherd, R., Arvola, A., Saba, A., Vassallo, M., et al. (2012). Perceived relevance and foods with health-related claims. Food Quality and Preference, 24(1), 129–135. Dean, M., Shepherd, R., Arvola, A., Vassallo, M., Winkelmann, M., Claupein, E., et al. (2007). Consumer perceptions of healthy cereal products and production methods. Journal of Cereal Science, 46(3), 188–196. Delgado, C., Santosa, M., & Guinard, J. X. (2013). Olive oil consumer research: Methods and key learnings. In R. Aparicio & J. Harwood (Eds.), Handbook of Olive Oil: Analysis and Proprieties (pp. 561–587). New York: Springer. Dekhili, S., Sirieix, L., & Cohen, E. (2011). How consumers choose olive oil: The importance of origin cues. Food quality and preference, 22(8), 757–762. García, M., Aragonés, Z., & Poole, N. (2002). A repositioning strategy for olive oil in the UK market. Agribusiness, 18(2), 163–180. Green, P. E., & Srinivasan, V. (1990). Conjoint analysis in marketing: new developments with implications for research and practice. Journal of Marketing, 54(4), 3–19.

112

C. Contini et al. / Food Quality and Preference 46 (2015) 103–112

Greene, W. H., & Hensher, D. A. (2003). A latent class model for discrete choice analysis: Contrasts with mixed logit. Transportation Research, 37(8), 681–698. Grunert, K. G., & Wills, J. M. (2007). A review of European research on consumer response to nutrition information on food labels. Journal of Public Health, 15(5), 385–399. Grunert, K. G., Lähteenmäki, L., Boztug, Y., Martinsdóttir, E., Ueland, Ø., Åström, A., et al. (2009). Perception of health claims among Nordic consumers. Journal of Consumer Policy, 32, 269–287. Lähteenmäki, L., Lampila, P., Grunert, K. G., Boztug, Y., Ueland, Ø., Åström, A., et al. (2010). Impact of health-related claims on the perception of other product attributes. Food Policy, 35, 230–239. Lähteenmäki, L. (2013). Claiming health in food products. Food Quality and Preference, 27, 196–201. Louviere, J. J., Hensher, D. A., & Swait, J. D. (2000). Stated choice methods: analysis and applications. Cambridge University Press. Magidson, J., & Vermunt, J. (2002). Latent class models for clustering: A comparison with K-means. Canadian Journal of Marketing Research, 20(1), 36–43. Manapace, L., Colson, G., Grebitus, C., & Facendola, M. (2011). Consumers’ preferences for geographical origin labels: Evidence from the Canadian olive oil market. European Review of Agricultural Economics, 38(2), 193–212. Magnusson, M. K., Arvola, A., Hursti, U. K. K., Åberg, L., & Sjöden, P. O. (2003). Choice of organic foods is related to perceived consequences for human health and to environmentally friendly behaviour. Appetite, 40(2), 109–117. McFadden, D. (1974). Conditional logit analysis of qualitative choice behaviour. In P. Zarembka (Ed.), Frontiers in Econometrics IV. London: Academic Press. Roininen, K., Lähteenmäki, L., & Tuorila, H. (1999). Quantification of consumer attitudes to health and hedonic characteristics of foods. Appetite, 33, 71–88. Roininen, K., Tuorila, H., Zandstra, E. H., De Graaf, C., Vehkalahti, K., Stubenitsky, K., et al. (2001). Differences in health and taste attitudes and reported behavior

among Finnish, Dutch and British consumers: A cross-national validation of the health and taste attitude scales (HTAS). Appetite, 37(1), 33–45. Saba, A., Vassallo, M., Shepherd, R., Lampila, P., Arvola, A., & Dean, M. (2010). Country-wise differences in perception of health-related messages in cerealbased food products. Food Quality and Preference, 21, 385–393. Siegrist, M., Stampfli, N., & Kastenholz, H. (2008). Consumers’ willingness to buy functional foods. The influence of carrier, benefit and trust. Appetite, 51(3), 526–529. Train, K. (2003). Discrete Choice Methods with Simulation. New York, NY: Cambridge University Press. Urala, N., & Lähteenmäki, L. (2007). Consumers’ changing attitudes towards functional foods. Food Quality and Preference, 18(1), 1–12. Urala, N., Arvola, A., & Lähteenmäki, L. (2003). Strength of health-related claims and their perceived advantage. International Journal of Food and Technology, 38, 815–826. Van Trijp, H. C., & Van der Lans, I. A. (2007). Consumer perceptions of nutrition and health claims. Appetite, 48(3), 305–324. Verbeke, W. (2005). Consumer acceptance of functional foods: Socio-demographic, cognitive and attitudinal developments. Food Quality and Preference, 16, 45–57. Verbeke, W., Vermeir, I., & Brunsø, K. (2007). Consumer evaluation of fish quality as basis for fish market segmentation. Food Quality and Preference, 18(4), 651–661. Wong, C. L., Arcand, J., Mendoza, J., Henson, S. J., Qi, Y., Lou, W., et al. (2013). Consumer attitudes and understanding of low-sodium claims on food: An analysis of healthy and hypertensive individuals. The American Journal of Clinical Nutrition, 97(6), 1288–1298. Worsley, A., & Lea, E. (2003). Consumers’ personal values and sources of nutrition information. Ecology of Food and Nutrition, 42, 129–151.