Effect of the traffic-light system on nutrition labeling in processed food products in the Ecuadorian population

Abstract

The 2012 Ecuadorian National Nutrition and Health Survey found that a high proportion of the country's population is overweight or obese. To address this health issue, in 2013, the Ecuadorian government implemented the use of a mandatory traffic-light (TL) nutritional labeling to be displayed on all processed food and beverages for sale in the country. We aimed to evaluate Ecuadorian consumers’ preferences for the nutritional attributes represented in the TL nutritional labels. In this cross-sectional study, 1,152 Ecuadorian consumers aged 18 and older in charge of food purchases completed an online survey. Results of a choice experiment with yogurt products revealed strong preferences and willingness to pay for yellow and green labels reaching price premiums of $1.15 for fat and $1.34 for sugar relative to red labels. The study provides evidence that consumers have high understanding of the TL nutritional labeling and prefer products with colors representing healthier alternatives.

1. Introduction

Obesity and the condition of being overweight have become global concerns because of their association with chronic non-communicable diseases and its impact on the immune system (Kelishadi et al. 2017). Furthermore, overweight and obesity is a growing problem in most of Latin America countries consequence of shifts in diets and lack of physical activity (Popkin and Reardon 2018). According to the 2012 National Nutrition and Health Survey conducted in Ecuador (ENSANUT 2012-ECU), a large proportion of the population suffered from being overweight and obese: 29.9 per cent of children (5–11 years), 26.0 per cent of adolescents (12–19 years), and 62.8 per cent of adults (20–59 years) (Freire et al. 2015). Further, data from Ecuador's birth and death registry (2012) show that chronic non-communicable diseases were the leading causes of death. These diseases were estimated to be responsible for about one in five deaths in the country (12.4 per cent of ischemic heart diseases, 4.8 per cent of cerebrovascular diseases, 5.3 per cent of diabetes, and 3.9 per cent of hypertension (Freire et al. 2015; INEC 2022b).

Given the Ecuadorian health conditions context presented above, the Ecuadorian Ministry of Public Health (MSP) introduced the regulation of the labeling of processed food products and beverages in November 2013, and implemented it in 2014 (MSP 2013a), with the goal to improve the Ecuadorian population's diets. The regulation requires the use of a traffic-light (TL) nutritional labeling system on the packages of all processed food products. The system provides consumers with graphical nutritional information on high, medium, low, and no total fat, sugar, and salt (sodium) content in processed food products and beverages (Sandoval et al. 2019). The TL label used in Ecuador has three different colors that represents the levels of three nutritional components (sugar, salt, and fat) in one serving of a processed food product (Figure 1). For each nutrient, the red label indicates a high content of the nutritional component, yellow indicates a medium content, and green indicates a low content. Moreover, if no nutritional component is present, the phrase “does not contain” is used (Table 1). The Ecuadorian TL nutritional labeling can be located on the package's front, side, or back (MSP 2013a).

Most of the previous research on TL labels in Ecuador was conducted shortly after the policy implementation. This study aims to contribute to this literature by measuring the effects of the TL labeling policy on consumer choices, preferences, use, and knowledge almost 7 years after its implementation. Most generally, the study contributes to the TL labeling literature by analyzing the effect of TL label colors on consumer choices and preferences. This contrasts with previous studies on the subject that have mainly focused on evaluating the effectiveness of including TL labels in food products relative to products without labels (Song et al. 2021). Finally, as found in Song et al. (2021) there is also a limited number of studies evaluating the effect of TL labels on consumer purchasing behavior (stated or revealed). This study also helps to fill this research gap. Thus, this study's objectives are to: (1) evaluate consumers’ preferences for the nutritional attributes represented in the TL nutritional labels, and (2) assess the relation between consumers’ preferences for the nutritional attributes represented in the TL nutritional labels and their sociodemographic characteristics, knowledge, and use of TL nutritional labels.

Ecuador was the first Latin American country to implement a mandatory supplemental labeling policy. Therefore, this study also intends to guide public policy in Ecuador and other countries with similar label policies. This is important, as several countries in the region (e.g., Bolivia, Chile, and Peru) have recently introduced similar laws regulating the labeling of processed products. Still, the empirical evidence on its use and effectiveness is very limited (OMS 2017). Finally, local producers in Ecuador or producers in countries adopting similar supplemental information labels may benefit from an improved understanding of consumer preference and perceptions for the attributes in supplemental labeling programs in future products and formulations.

1.1. Literature review

A body of literature has evaluated consumers’ understanding and perceptions of interpretative front-of-package labels (i.e., labels using symbols, figures, or cautionary text to indicate the nutrient content of a product), including TL labels (Song et al. 2021). Research has found that TL labels are generally easier to understand than other nutritional labels, such as NuVal, Facts-Up-Front, and the octagon nutritional labeling (Findling et al. 2018; Valverde-Aguilar et al. 2018; Song et al. 2021). Studies have also reported that the information presented in TL labels improves consumers’ understanding of nutritional product profiles (Grunert et al. 2010; Hieke and Wilczynski 2012; Arrúa et al. 2017; Findling et al. 2018). For example, a study of Uruguayan mothers reported that TL labels’ warnings improved their ability to identify high amounts of sodium in food products and enhanced awareness of the product's nutritional profile (Arrúa et al. 2017). Another study from Uruguay by Machín et al. (2018) found that information included in TL labels increase consumer perception of the product healthfulness compared with simple information food labels. There is also a branch of literature evaluating the effect of the presence of TL labels on consumer behavior (intended or actual) (Song et al. 2021). Song et al. (2021) conducted a meta-analysis on nutritional labeling literature and identified twenty-two studies focused on the effect of the TL on food consumer behavior (five conducted in the Latin American region). Overall, it is reported that the presence of TL labels increases the likelihood of consumers choice of healthier products (Song et al. 2021). Ares et al. (2014) found that TL labels enable Uruguay consumers to make selections for products with a low content of fat and sugar. Yoo et al. (2017) study in Uruguay found that TL labels increased consumer perception of the unhealthfulness of product attributes discouraging consumption. Finally, Talati et al. (2019) studied Argentina consumers food products selection when TL labels are introduced. Results indicated that the presence of TL labels influenced consumers to switch product selection to an option with a healthier profile.

Studies evaluating the effect of different TL label colors are less common. Balcombe et al. 2010 conducted a choice experiment for a basket of goods with different TL labels in the UK. Their findings showed that consumers are concerned about high levels of fat and salt and are willing to pay premium prices for products with green labels relative to those with red labels. Another study with German consumers reported that the use of TL labels that highlight the presence of low levels of nutrients like fat in pizza increased the probability of choosing the products (Drescher et al. 2014). Sonnenberg et al. (2013) conducted a study in a hospital cafeteria in the USA, in which they found that sales of food and beverages with red labels decreased, while sales of food and drinks with green labels increased after labels were placed on the products. We did not identify any study evaluating the effect of TL label colors on consumers in the Latin American region (Song et al. 2021).

Few studies have evaluated the use and effect of the TL nutritional labeling implemented in Ecuador, and most have used data gathered one or two years after the TL system was implemented (Souza Jaramillo 2015; Poveda Guevara 2016; Díaz et al. 2017; Freire et al. 2017; Peñaherrera et al. 2019). Several of these studies were qualitative or used samples that do not represent the Ecuadorian population (Maya Izurieta 2015; Souza Jaramillo 2015; Poveda Guevara 2016). Santos Collantes (2018) conducted a survey in the capital city, Quito, about consumers’ knowledge of information and use of TL labels. The author found that 76 per cent of respondents identified the information presented on the TL labels correctly, and 38 per cent used it for purchasing decisions. Ramos Padilla et al. (2017) explored the association between sex and age in attitudes toward the TL nutritional system. Their findings indicated that men are less likely to use TL labels than women, and consumers aged 18–40 years are more likely to use the labels to reduce fat, sugar, and salt consumption than consumers over 40 years. Another study Sandoval et al. (2019) evaluated TL labeling's effect on soft drink purchases using data from before (January 2013) and after the TL label policy was implemented (December 2015) in Ecuador. They found that the implementation of TL labels did not affect consumers’ purchases of soft drinks. Finally, a recent study by Cabrera et al. (2022) collected information about the TL labels use and knowledge by Ecuadorian teenagers. Results showed that less than 50 per cent of Ecuadorian teenagers do not use TL labels for purchasing decisions, while about 68 per cent were found to have a high to medium knowledge of the information presented in TL labels.

In summary, the majority of the literature on TL labels has focused on issues related to understanding the information presented on the labels. The literature on consumer preferences for TL label colors is more limited, particularly in the context of Ecuador.

2. Hypothesis

Based on the relevance of the TL label policy for healthy eating and other countries new adoption of similar policies, we can see that there is the opportunity to better understand consumers preference and use of SNL labels. We formulated two hypotheses for the study as follows:

• H1: The TL label colors affect consumers’ preferences for food products.

• H2: Ecuadorian consumers are WTP higher values food products with TL label colors representing healthier alternatives’.

• H3: WTP for TL label colors are associated with consumer sociodemographic characteristics.

3. Background on processed food regulations in Ecuador

In Ecuador, a valid sanitary certificate issued by the National Agency for Regulation, Control, and Health Surveillance (ARCSA) is required for the distribution and sale of processed food products in Ecuador (ARCSA 2016b). ARCSA will issue the certificate based on the classification of the food product's risk factors. The risk factor considers the food product's physical and chemical properties, preparation method, shelf life, and storage requirements. Therefore, to submit an application for the sanitary certificate and obtain a classification based on risk factors, the food manufacturer must first develop a technical product profile (ARCSA 2016a).

A technical product profile must include a declaration of the product ingredients, physical and chemical specifications of the product package, package design and labeling of the product, a description of the elaboration process, and certifications such as “organic certification” granted by the corresponding authority (MSP 2013a; ARCSA 2016b). Once ARCSA receives the technical product profile, supporting documentation (e.g., name of technical representative of the processing plant, operating permits, etc.), and required payments, the agency will evaluate the application and issue the product risk factor classification. After the review and classification of risk factors, ARCSA will issue the sanitary certification for the food product, which will be valid for 5 years (ARCSA 2016a).

The content and structure of food labels in Ecuador are regulated by the Ecuadorian Ministry of Health and the Ministry of Industry and Productivity (MSP 2013a; MIP 2014). For example, regulation prohibit the use of images of children, health professionals or celebrities or the inclusion of unproven health statements (MSP 2013). Regarding labeling of the nutritional content, food products in Ecuador needs to include two labels: a nutrient declaration label, and since 2014 the TL label. The nutrient declaration labels provide consumers with a summary of the nutrient composition of the food product (FAO and WHO 2001; Cabrera et al. 2022).

Ecuadorian laws also regulate the size and content of TL labels. The regulations, among others, include the color of the background (white or gray), the size (at least 15 per cent of the area where it is located if the area is larger than 19.4 cm²), its position (leftmost side in the front or back of the package), order of the colors and nutrients, and text font size, for example, the words ‘HIGH, MEDIUM, and LOW’ and ‘FAT, SUGAR, and SALT’ must use capital case letters. If color labels are displayed more than once, then the order of the attributes is sugar, fat, and salt; and in case an attribute is not present in the product, the ‘does not contain’ label must use Roman style and lower-case letters (MSP 2013; MIP 2014).

4. Methods

4.1. Data

This study uses cross-sectional data obtained from an online survey of 1,152 Ecuadorian consumers during December 2019. This sample size is appropriate for estimating the proportion of individuals using the TL label with 95 per cent confidence and a 3 per cent maximum error (Ott and Longnecker 2015). Data was collected by Universidad Técnica Particular de Loja's Foundation for Business and Social Development (FEDES). FEDES used SurveyMonkey to distribute the survey among an existing and active national research panel of consumers across the twenty-four provinces of Ecuador. The scope of the study was the population of Ecuadorian food shoppers, therefore, 384 surveys were collected for each region of the country (Costa, Sierra, and Amazon), where only consumers in charge of their own food purchase decisions and 18 years or older were surveyed. The study was approved by the university authors Institutional Review Board.

The survey instrument was developed by a research group that included Economists and Public Health professionals from Ecuador and the USA. Given the study objectives, the survey was organized into three parts: (1) responses to a choice experiment to assess their preferences for nutrient levels presented on the labels; (2) individuals’ knowledge, perceptions, and use of the TL nutritional labeling; and (3) socio-economic and demographic characteristics. A pilot test of the survey was carried out first with fifty-one respondents, and the information obtained in the test was used to refine the final survey instrument. The average completion time of the survey was 5.5 minutes.

4.2. Choice experiment

The choice experiment (CE) included in the survey had eight choice scenarios. Each scenario represented a situation, where the consumers observed two 150 g cups of yogurt with the same package and flavor that differed in price and the TL label colors for sugar, salt, and fat (i.e., the choice experiment included four product attributes). Yogurt was selected because it presents TL colors for the three non-price attributes (sugar, salt, and fat) and is consumed by a large proportion of consumers in the country (INEC 2015). Consumers were then asked to select between the two product profiles or a “none” option. Figure 2 shows an example of one of the choice scenarios presented to the consumers.

The three non-price attributes considered in the CE were TL label colors for sugar, fat, and salt, the visible attributes in the TL labels. The TL label colors for sugar and fat had three levels: red, yellow, and green, while the TL label color for salt included only two levels: yellow and green. These attribute levels were defined after reviewing the product database from the Kantar World Panel Company and a nutrient composition database of 177 yogurt products of the ARCSA (Kantar World Panel 2019). None of the yogurt products in the ARCSA database had a red label for salt.

The price attribute had four levels: $0.50/cup, $0.65/cup, $0.75/cup, and $0.90/cup. These prices were defined based on a sample of prices from four retail locations (supermarkets and convenience stores) and seven brands of yogurts collected in two cities (Loja and Quito) between October and November 2019. The mean value of prices for a 150 g cup of yogurt was $0.71 (range = $0.50–$1.50). The price levels in the experiment were set using 7 per cent and 30 per cent values above and below the mean price observed in those cities ($0.71).

SAS version 9.4 was used to design the choice experiment (SAS Institute Inc. 2012). The combination of all labels for sugar, salt, and fat and price levels resulted in a total of seventy-two (3 × 3 × 2 × 4) possible product profiles and 2,556 possible choice scenarios |$( {{\rm{C}}_{72}^2} )$|⁠, in which |${\rm{C}}_{\rm{n}}^{\rm{r}}$| denotes the number of unordered subsets (i.e., combinations) of n objects taken r at a time (Wackerly et al. 2014). Finally, a fractional factorial design selection procedure was used to create sixteen possible choice scenarios. The scenarios were blocked into two different versions of the questionnaires. One version of the questionnaire was randomly presented to respondents.

The presentation of the choice scenarios was preceded by an information section. The section included an explanation of the attributes and levels presented in the yogurt product, and instructions on how to do the selection of the preferred product. The information section also asked respondents to make choices as if they were in the regular shopping location and reminded them about their budget constraint, which is one of the ex-ante approaches suggested in the literature to reduce hypothetical bias (i.e., respondents are urged to be honest and behave as if they really had to pay) (Loomis 2014).

4.3. Individuals’ knowledge and use of TL nutritional labeling

Individuals' knowledge of the TL nutritional labeling was evaluated using a set of two questions with four items. (Cabrera et al. 2022) The first question asked, ‘What components are included in the traffic light nutritional labels?’ This was a multiple-choice question with four options and one correct answer (see Table 2). The second question asked about the level of the nutritional content (high, medium, and low) of the foods associated with each color on the traffic-light nutritional labels (red, yellow, and green); thus, it had three parts. Respondents were required to match the colors presented in the TL labels with the concentration level they represent (red-high; yellow-medium; and green-low) (Cabrera et al. 2022). Each correct answer was assigned 1 point, for a maximum knowledge score of 4 points and a minimum of 0 points. Scores with 0 to 1 points were classified as low knowledge scores, 2 to 3 points as medium knowledge scores, and 4 points a perfect or high knowledge scores.(Cabrera et al. 2022) The use of the TL label was evaluated with the question: Do you shop based upon the traffic light nutritional labeling? The respondents selected the frequency with which they use TL in purchase decisions (see Table 2).

4.4. Theoretical and econometric model

The econometric model was specified using Lancaster's (1966) demand theory based upon the products’ characteristics or attributes and the random utility model (McFadden 1973). According to Lancaster (1966), consumers have preferences for products’ characteristics, rather than preferences for the products per se; thus, each good represents a set of attributes. Moreover, consumers choose the goods with a set of attributes that maximizes their utility under a budget constraint. In this context, the attributes of interest were the colors for sugar, salt, and fat displayed in the TL labels.

Consumers’ product selections can then be analyzed using the random utility model. In this model, the indirect utility of each choice depends upon product characteristics (non-price attributes) and price; thus, the model is consistent with Lancaster's demand model, and utility maximization subject to a budget constraint (Train and Weeks 2005). Therefore, for consumer n (i.e., respondents in the choice experiments) choosing product l in choice scenario t, the utility of choice l (U_nlt) is as follows:

in which |$n{\rm{\ }} = {\rm{\ }}1, \ldots ,{\rm{\ N}};{\rm{\ l\ }} = {\rm{\ }}1, \ldots ,{\rm{\ L}};{\rm{\ t\ }} = {\rm{\ }}1, \ldots ,{\rm{\ T}}.{\rm{\ }}{V}_{{\rm{nlt}}}$| is the measurable portion of utility related to the observed product attributes and price, and |${e}_{{\rm{nlt}}}$| represents unobserved consumer and product characteristics affecting utility not included in |${V}_{{\rm{nlt}}}$|⁠.

It is assumed further that |${V}_{{\rm{nlt}}}{\rm{\ }}$|is a linear function of price |${p}_{{\rm{nlt}}}$|⁠, and non-price attributes|$\ {X}_{{\rm{nlt}}}$|⁠,; thus, equation (1) can be written as:

in which |${\alpha }_{\rm{n}}$| and |$\beta _{\rm{n}}^{\rm{^{\prime}}}{\rm{\ }}$|represent the parameters for price and observed product characteristics, respectively, for individual n.

To estimate the parameters of the utility function in equation (2), the error |${e}_{{\rm{nlt}}}$| is assumed to have an extreme valued distribution; thus, it can be shown that the probability that consumer n will choose product |${\rm{l}}$| in the choice scenario t, conditional on the coefficient vector |${\theta }_{\rm{n}} = [ {{\gamma }_{\rm{n}}{\rm{\ }}\beta _{\rm{n}}^{\rm{^{\prime}}}} ],$| is (Revelt and Train 1998):

in which |${V}_{{\rm{nlt}}} = - {\alpha }_{\rm{n}}{{\rm{p}}}_{{\rm{nlt}}} + \beta _{\rm{n}}^{\rm{^{\prime}}}{x}_{{\rm{nlt}}}$|⁠. Moreover, conditional on |${\theta }_{\rm{n}}$|⁠, the probability of consumer |$n{\rm{^{\prime}s}}$| observed sequence of L choices is expressed as (Train 1998; 2009):

in which |$t( {n,{\rm{\ l}}} )$| denotes a specific product, |${\rm{l}}$|⁠, that consumer n selects in choice scenario t. The coefficient vector |${\theta }_{\rm{n}}$| is unobserved for every consumer, n, and varies through the population with density |$g( {{\theta }_{\rm{n}}\vert {\rm{\Gamma }}} ),$| in which the parameters of the distribution of |${\theta }_{\rm{n}}$| are |${\rm{\Gamma }}$|⁠. Therefore, the unconditional probability of the observed choice sequence (i.e., the mixed logit choice probability) is:

The log-likelihood function for all n consumers is:

Estimation of the parameters |${\rm{\Gamma }}$| ws conducted using simulated maximum likelihood procedures in STATA software (Train 1998; 2009; Rigby and Burton 2006). With respect to the distribution of the coefficients in |${\theta }_{\rm{n}}$|⁠, the price coefficients were specified as lognormal with standard deviation (SD) constrained to 0, and the non-price attributes as normal (Carson and Czajkowski 2019), so as to obtain well defined distributions of WTP estimates for the attributes (i.e., with finite second moments). As shown in Carson and Czajkowski (2019), the mean willingness to pay (⁠|${\rm{WT}}{{\rm{P}}}_{\rm{c}}$|⁠) for attribute c is then the ratio of the attribute's mean coefficient, |$\widehat {{{\rm{\beta }}}_{\rm{c}}},{\rm{\ }}$|to the natural exponential function of the estimated price coefficient, |${\rm{WT}}{{\rm{P}}}_{\rm{c}} = \widehat {{{\rm{\beta }}}_{\rm{c}}}/{\rm{exp}}( {{\rm{\hat{\alpha }}}} )$|⁠.

4.5. Factors that affect WTP for TL attributes

The relation between consumers’ characteristics and the WTP values for levels in the TL labels was analyzed using a two-step approach. The first estimated individual consumers’ WTP values for the color labels of the attributes (salt, sugar, and fat) presented in the TL labels. The second step used linear regression models with WTP values as dependent variables and consumer characteristics as explanatory variables.

The estimation of the individual consumers’ WTP values relies on the Baye's rule. As the density of |${\theta }_n$| is conditional on the parameter vector for each individual's sequence of choices, it can be rewritten as (Revelt and Train 1998; Train 2009):

thus, the expected value of any function of |${\theta }_n$|⁠, such as the WTP value, is given by:

which can be estimated using simulations:

in which |${\theta }^d$| corresponds to the d-th draw from the population density |$g( {{\theta }_n\vert \Gamma } )$|⁠, and |${R}_n( {{\theta }^d} )$| is the probability of an individual n’s sequence of choices. Estimated parameters |$\hat{\Gamma }$| are used rather than the parameters |$\Gamma $| (Hess 2007) using 1000 Halton draws.

In the second step, we used a random-effects panel linear regression model (Campbell 2007; Barrowclough et al. 2019):

in which |$WT{P}_{nc}$| is the nth household WTP for attribute c (i.e., |$c\ $|= color of nutritional label), |${t}_c$| and b are coefficients, |${y}_n$| is a vector of household-related characteristics, |${h}_i$| is a household-specific random error, and |${u}_{nc}$| is the error term. This approach estimates the marginal effects of household-related characteristics (knowledge, use, perceptions, and sociodemographic characteristics) on the attributes’ average WTP values (Campbell 2007).

5. Results

5.1. Sample characteristics

The sample consisted in 1,132 Ecuadorian respondents, 18 years or older, and in charge of food purchasing decisions. The average survey respondent was 39 years old and had a median monthly household income between $521 and $686 (Table 2). Most survey respondents were female (55.64 per cent) and had a college degree or more (70.66 per cent). The sample was distributed equally in the three regions of Ecuador (33.3 per cent). Compared to the population in Ecuador, the sample is older than the national average (30 years in population) and more educated (6.3 per cent of population have a college degree) (INEC 2015; ONU 2022). The sample also had a slightly larger share of female respondents (49.6 per cent in population), The distribution of income in the sample was very similar to that reported by the Servicio de Rentas Internas (SRI) in 2020 (SRI 2020; Lucero, 2021). Some of the differences are expected given the study's focus on adult's individuals, which are older and more educated than the entire population and the use of an online survey.

TL labels seem to be important for purchasing decisions since about 85 per cent of respondents indicated that they use the TL labels to make food shopping decisions (15 per cent indicated they never use the labels) (Figure 3). However, results also show that only about 39 per cent of respondents indicated they use it most of the time (25 per cent) or always (14 per cent). Most respondents (87.33 per cent) identified fat, sugar, and salt as the nutritional components included in the TL labels (Figure 4). Moreover, with respect to overall knowledge about the TL label, 81.77 per cent of respondents were found to have a high level of knowledge (4 points), 7.30 per cent a medium level (2–3 points), and 10.94 per cent a low level of knowledge (0—1 points) (Figure 5).

5.2. Choice experiment

The mixed logit model estimation results are presented in Table 3. The estimated coefficients correspond to the parameters of the distribution of the random coefficients (⁠|${\alpha }_n,\ {{\boldsymbol{\beta }}}_{\boldsymbol{n}}$|⁠). Two columns are included, one for the mean parameters and another for the SD of the distributions. All estimated mean parameters and most SDs were statistically significant at the 1 per cent level.

All estimated mean parameters but that the one for Price can be interpreted as the mean effect of the attributes on the indirect utility function. For the TL color attributes represented with dummy variables, the red color for fat and sugar and yellow for sugar were selected as the baseline attributes. Therefore, positive estimated mean parameters for TL labels in the model indicate that consumers obtain higher levels of utility with yellow or green TL color labels relative to the red label (i.e., healthier alternatives are preferred to the less healthy ones). The positive sign in the alternative specific constant (ASC) coefficient indicates, on average, respondents preferred the “none” option. This seem to reflect the fact that the “none” option was selected in about 43 per cent of choice occasions. Finally, the positive Price coefficient does not have a direct interpretation as an effect on the indirect utility function. Since its distribution is lognormal, by design, the mean price effect is given by -exp(⁠|$\hat{\alpha }$|⁠), where the |$\hat{\alpha }$| represents the estimated mean price parameter.

With respect to the estimated SDs of the coefficients’ distributions, model results indicate Ecuadorian consumers have heterogeneous preferences for some but not all the label options. Heterogeneous preferences were found for green labels in sugar and fat and the yellow label for sugar. Heterogeneous preferences were also identified for the ASC.

The mean WTP values estimated in Table 4 are the monetary value consumers are willing to pay for a label color relative to a baseline label for a 150 g cup of yogurt. All five non-price attributes had statistically significant and positive mean WTP with respect to the baseline red labels for sugar and fat, and the yellow label for salt in yogurt products; thus, the WTP values can be interpreted as the premium consumers were willing to pay to avoid the red label for sugar and fat, and the yellow label for salt. The results showed that consumers were willing to pay, on average, $1.07 and $1.15 more for yellow and green labels, respectively, relative to the red label for fat. For sugar, consumers’ mean WTP was $1.22 and $1.34 for yellow and green labels, respectively, in relation to a red label. Finally, the mean WTP for salt was $0.19 for a green label compared to a yellow label.

5.3. Effects of sociodemographic characteristics

Table 5 presents the results of regression analyses that explored the relations between WTP values for TL labels, consumers’ characteristics, and their knowledge and use of the TL labels on yogurt products. Two random-effects regression models were estimated, one for WTP for green and yellow labels for fat, and another for WTP values for green and yellow labels for sugar. In each regression model, in addition to consumer sociodemographic characteristics, the explanatory variables included dummy variables for the green labels; therefore, the yellow labels were the baseline. A regression model for salt labels was not estimated as there was no evidence of heterogeneity in the WTP for green labels in salt relative to the yellow label.

The panel-regression model results show that the WTP for green labels for fat and sugar was $0.09 and $0.12, respectively, higher than the corresponding WTP for yellow labels (P < 0.01). The effects of consumers’ characteristics, location, age, income, knowledge, and use were also statistically significant (P < 0.10 or less). Consumers located in the Amazon region were WTP $0.09 and $0.06 less for green and yellow fat and sugar labels, respectively, relative to consumers located in Sierra region. An additional year in age increased consumers’ WTP for green and yellow fat labels by $0.0006. In contrast, an additional thousand dollars in income decreased the WTP for sugar labels by $0.01. Moreover, the WTP for green and yellow fat and sugar labels was associated positively with consumers’ knowledge of the information on the TL labels and the labels’ use in purchase decisions (see Table 5).

6. Discussion

Ecuador was one of the first countries in Latin America to mandate the TL label to improve the population's nutrition. This study's goal was to evaluate Ecuadorian consumers’ preferences for TL label colors. The results of the study are relevant for other countries in the region that have adopted TL labels and different types of supplemental nutritional labels.

The survey results showed that a large proportion of the respondents (85 per cent) indicated that they use the label for purchase decisions. This result is similar to that reported in ENSANUT-2018 (INEC 2022a), which found that 77.47 per cent of the Ecuadorian population used TL labels for their purchase decisions. However, our level of use is considerably higher compared to the results of Santos Collates (2018) and Teran et al. (2019), who reported levels of use of only 38 per cent and 27 per cent, respectively. This study and ENSANUT-2018 use national-level samples, while collates and Teran et al.’s studies were limited to a single city (Quito).

With respect to consumers’ level of knowledge about the labels, 82 per cent of respondents were found to have a high level of knowledge (all 82 per cent had a perfect score). These results are similar to those in previous studies, although the methods used differed. Santos Collantes (2018) used a single self-reported TL knowledge question and found that 76 per cent of respondents indicated that they understood the TL labels. Teran et al. (2019) used different questions and a different scale to measure knowledge and found that 88.7 per cent knew about the TL labels. The high level of use and knowledge about the TL labels provides a good indication of the TL policy's relative success. Both knowledge and use of the TL labels are necessary, although not always sufficient, to promote diet change.

The choice experiment results revealed strong preferences for green and yellow labels for fat and sugar relative to red labels. These results are similar to those of Scarborough et al. (2015) (Scarborough et al. 2015), who found that food products with red fat and sugar labels were less likely to be selected than those with green labels. In addition, the results that showed preferences for green labels are similar to those in Thorndike et al.’s (2014) intervention study. They found that the presence of TL labels on beverages decreased consumers’ selection of red-labeled drinks and increased their consumption of beverages with green TL labels.

Given that the average market price for yogurt products is $0.71, the estimated WTP values for yellow and green labels were very high. These results are consistent with those obtained in Balcombe et al. (2010) study, in which the authors found that consumers were willing to pay high premiums to acquire goods with yellow relative to red nutrient labels. They reported that consumers were willing to pay £12.08 more to purchase a basket of goods with a green label for fat relative to one with a red label, and £14.22 to acquire a basket of goods with a green label for sugar relative to one with a red label.

The results of the WTP values for the yellow and green labels indicated that consumers valuation of the sequence of labels is not linear (i.e., yellow relative to red and green relative to yellow). For example, in the case of the labels for fat, consumers were willing to $1.07 for the yellow label relative to the red, but only approximately $0.08 to $0.09 more for the green label relative to the yellow. A similar pattern was also observed for the WTP values for sugar labels. Thus, consumers appear to be concerned largely with avoiding red labels.

The significance of the SDs of the distribution of coefficients suggested heterogeneity in the preferences for TL label colors, except for salt. However, the results of the regression analyses (Table 5) identified only a few consumer-related variables that affected the WTP values for yellow fat and sugar labels, and only location of residence was found to have an economically significant effect. Thus, there appear to be certain cultural aspects related to location that affect consumers’ WTP for products with the various labels.

The results of the regression analyses provided evidence that label knowledge and use are associated with individuals’ WTP for yogurt products with yellow or green relative to red TL labels. These results are supported by Teran et al. (2019) and Drichoutis et al. (2005) results, who found that knowledge of the labels’ information increased the consumption (a reflection of preferences) of healthy (green label) products. Although the levels of knowledge and reported use of TL are already very high in the country, these results suggest that additional efforts to improve knowledge and promote use can still influence consumers’ preferences for the labels. Efforts to educate the population about TL use in Ecuador were very prominent after the regulation was implemented, but are less common now (Díaz et al. 2017). As location of residence was found to have an important effect on WTP for the labels, additional educational efforts can use these results to target those locations.

The results of the study have important policy implications. While we found that the Ecuador adult population knows the information presented in the TL label and reports using it for purchasing decisions, a significant share uses it infrequently; thus, more work is needed to promote its use. This study's results complement those Cabrera et al. (2022) found exploring TL label knowledge and use among Ecuadorian children. The study found low levels of TL knowledge and use among this group. Future educational campaigns should be designed considering these findings to target different populations according to their needs. Educational efforts for adults should emphasize the promotion of TL label use and motivate parents to teach their children about nutrition. Educational campaigns focusing on kids need to inform and promote TL label use.

Study findings provide evidence that TL labels in food products help achieve the final policy objective of reducing the purchase and consumption of high levels of sugar, fat and salt. However, something to remember is that the TL label information presented to study participants during the choice experiments was in the front of the package. Current TL label regulations in Ecuador do not require these labels to be present in the front package. Thus, the government should consider requiring TL labels on the front of the food package.

Study results also suggest that consumers are mainly avoiding products with red color labels, as there was no difference in WTP values for green and yellow TL colors. This limits the reduction in sugar, fat, and salt consumption that could be achieved with current labels. Alternative label schemes could be considered to better inform consumers about various nutrient concentration levels. Another alternative is the change in the nutrient concentration levels that define the TL colors (i.e., reduce maximum levels that define a red TL label).

Moreover, national and international food manufacturers can use the results of this study for future product formulation. Study results found that consumers prefer products with TL label colors reflecting healthier alternatives and are willing to pay premiums for these attributes. This suggests potential opportunities for developing and marketing new products better aligned with consumers' preferences. In fact, according to the results, decreasing nutrition concentrations from red TL label to yellow TL label levels might suffice to increase the demand for their products or at least to remain competitive in the market.

The study has some limitations. First, although the research found strong preferences and positive willingness to pay for yellow and green labels on yogurt products, WTP values can differ from product to product (Balcombe et al. 2010; De-Magistris and Lopéz-Galán 2016; Sandoval et al. 2019). Second, TL labels are displayed on all beverages and food products and are available for consumers of all ages; however, our study population was limited to consumers 18 years or older. Third, the sample is not totally representative of the Ecuadorian population. Fourth, the study used stated preferences, which might be subject to hypothetical biases, but even if they were present, the qualitative results still provide important information. Future research could use real product selection scenarios, although the hypothetical research design used in this study offered more flexibility to use a wider range of label values (i.e., colors), which in turn allowed us to acquire a more detailed understanding of TL labor colors’ preferences (Sacks et al. 2009). Future research on the WTP for TL labels could adjust the procedures to incorporate not only choices for one unit of a product but the total amount demanded.

7. Conclusion

Overweight and obesity are a problem of public concern given the potential future health problems and presence of NCDs. Through the implementation of SNL, governments aim to improve population dietary habits. Our study provided evidence that TL label colors help Ecuadorian consumers’ make choices consistent with preferences for products with lower sugar and fat content. Our results also provide a better understanding on the longer-term effects that TL labels have on purchasing decisions. In future, the results of this study can be used as a basis to monitor consumer preferences for different products and color labels as the policy evolves. Ideally, the evaluation of TL label policies should use actual purchase data from households or supermarkets, but these data are unavailable in many developing countries. Data obtained using stated preferences methods (e.g., choice experiments) provide a low-cost alternative to evaluate the effect of nutritional labeling policies, but the studies need to be implemented before a policy is implemented to collect baseline data.

Acknowledgement

Not applicable.

Funding

This research received support from the Universidad Nacional de Loja.

Data Availability

The dataset used and/or analyzed during the current study are available from the corresponding author on reasonable request.

References