Care4Food: Choices and perceptions of food security and sustainability in a climate change context in Portugal

SCIENTIFIC REPORT Care4food CARE4FOOD: CHOICES AND PERCEPTIONS OF FOOD SECURITY AND SUSTAINABILITY IN A CLIMATE CHANGE CONTEXT IN PORTUGAL. Authors: Patrícia Abrantes (coord.), Cláudia Viana, Eduarda Marques da Costa, Eduardo Gomes 1 CARE4FOOD Choices and perceptions of food security and sustainability in a climate change context Escolhas e perceções de segurança alimentar e sustentabilidade num contexto de alterações climáticas em Portugal. Technical report Authors: Patrícia Abrantes (cood.) Cláudia Viana, Eduarda Marques da Costa, Eduardo Gomes Collaborators: Miguel Vilhena, Rafael Ramusga, Francisco Sousa MARCH 2025 How to cite: Abrantes, P., Viana, C., Marques da Costa, E., & Gomes, E. (2025). Care4Food: Choices and perceptions of food security and sustainability in a climate change context in Portugal – Tecnical report. Report Care4Food Project. DOI 10.5281/zenodo.15475486. Funding: Social Observatory of la Caixa Foundation under the call Social Impact of Climate Change (FP23-2B) 2 CONTENTS LIST OF FIGURES.......................................................................................................................... 4 LIST OF TABLES ........................................................................................................................... 5 ABBREVIATIONS .........................................................................................................................6 SUMMARY .................................................................................................................................. 7 1. INTRODUCTION....................................................................................................................... 8 1.2. Socio-ecological framework applied to food analysis .................................................... 11 1.3. Aim of the study ............................................................................................................. 14 2. DATA AND METHODS ........................................................................................................... 17 2.1. Case study presentation ................................................................................................. 17 2.2. Data collection ............................................................................................................... 19 2.2.1. Sample and survey structure .................................................................................... 19 2.2. Proposing quantitative indicators .................................................................................. 20 2.2.1. Food security indicator ............................................................................................. 21 2.2.2. Body mass index ...................................................................................................... 21 2.2.3. Food Sustainability ................................................................................................... 22 a) Indicator (food sustainability score) ............................................................................... 22 b) Sustainable meal option ................................................................................................ 25 2.3. Data analysis .................................................................................................................. 26 3. RESULTS ................................................................................................................................ 27 3.1. Household characterisation ........................................................................................... 27 3.2. Household food security ................................................................................................ 30 3.2.1. Food security indicator as a measure of economic accessibility ................................ 30 3.2.2. Physical accessibility to food .................................................................................... 31 2.2.3. Food utilization: dietary habits and practices ........................................................... 34 2.2.3.1. Gender and age in food utilization ......................................................................... 37 3.3. Barriers and enabler of food change .............................................................................. 39 3.4. Food sustainability awareness and the role of public policies........................................ 42 4. EXPLORATORY ANALYSIS OF PATTERNS AND DETERMINANTS OF FOOD CHOICES FOR FOOD SECURITY AND SUSTAINABILITY ............................................................................................... 46 4.1. Cluster Analysis of the Food and Contextual Environment in Portuguese NUTS III Regions ................................................................................................................................. 48 4.2 Individual-Level Food Behavioural changes ................................................................... 50 5. CONCLUSIONS AND RECOMMENDATIONS ........................................................................... 54 5.1. Main limitations ............................................................................................................. 54 3 5.2. Results in the broader conceptual framework ............................................................... 55 5.3. Policy Implications and Future Directions ...................................................................... 57 5.4. Building a National Food Sustainability Observatory ..................................................... 57 REFERENCES .............................................................................................................................. 59 4 LIST OF FIGURES Figure 1 - Greenhouse gas emissions per 100 grams of protein. ........................................................ 8 Figure 2: interrelations between food security and sustainability. ..................................................... 9 Figure 3: Social ecological model of behaviour change and the Capability, Motivation and Opportunity behavior (COM-B). ....................................................................................................... 12 Figure 4: Socio ecological model (SEM) of food behavior. ................................................................ 13 Figure 5: FAO’s holistic Food wheel including a SEM and COM-B perspective. ................................. 13 Figure 6: framework adopted from Mitchel etl. 2011 to analyse behavioural changes toward food security and sustainability ................................................................................................................ 15 Figure 7 – Workflow of the care4food study .....................................................................................16 Figure 8: Mainland Portugal NUTSII and III ....................................................................................... 18 Figure 9: level of education of the household representative .......................................................... 28 Figure 10: householder representative – professional status by region ............................................ 29 Figure 11: Food security composite indicator by region (%) ..............................................................30 Figure 12: Food security composite indicator by gender(%) ............................................................. 31 Figure 13 - Participant representation by average distance to a fresh food store (%) ......................... 31 Figure 14: Percentage of participants in the survey using each transport mode by region ................ 33 Figure 15: Type of diet by region, in percentage ............................................................................... 35 Figure 16:relation between the perception of a sustainable and healthy diet and meal option .. 37 Figure 17: relation between practices (food score) and perception of a sustainable diet ...................38 Figure 18: distribution of cooking responsibilities by gender ........................................................... 39 Figure 19: Changes made in eating habits ....................................................................................... 39 Figure 20: What is important to change (responses from 1 - not importante to 5 -very important) .. 40 Figure 21: What is the main difficulty to changing habits (responses from 5 - not important to 1 -very important) ....................................................................................................................................... 41 Figure 22: Perceived National Self-Sufficiency in the Context of Climate Change, by Region ..........43 Figure 23: Public policies support for nutritious and sustainable food ..............................................43 Figure 24: Perceived Importance of Public Policy Measures Supporting Nutritious and Sustainable eating (Scale: 1 – very important, 5 – not important) ....................................................................... 44 Figure 25: most frequently cited words in open-questions about suggestions. ................................ 45 Figure 26: NUTS III Clusters ............................................................................................................ 50 Figure 27: Cluster and variables B-plots ........................................................................................... 50 Figure 28 – main drivers of food habits change towards sustainability. ............................................ 56 Figure 29 – The Care4Food Dashboard. .......................................................................................... 58 5 LIST OF TABLES Table 1 - Main indicators of the studied regions ............................................................................... 17 Table 2 - Sample size by region and gender ......................................................................................19 Table 3 – Sample size by region and gender ..................................................................................... 20 Table 4– Insecurity levels according to FEIS and Composite indicator ............................................... 21 Table 5– Nutritional status of the respondents according to WHO classification (in percentage) ...... 22 Table 6– Nutritional and sustainability features of suggested meals ................................................ 25 Table 7– Number and percentage of adults per household ............................................................... 27 Table 8– Percentage of adults and children by household ................................................................ 27 Table 9– Relation between the percentage of chronic diseases reported and the BMI ..................... 29 Table 10– Expenses on food by household with children and without children .................................30 Table 11– Diversity fresh vegetables and fruits supply within a 5 minute walking distance from home ........................................................................................................................................................ 32 Table 12– Mode of transport by distance time ................................................................................. 32 Table 13– types of stores to purchase fresh fruits and vegetables ..................................................... 33 Table 14 – Relation between the type of diet (b) and perception of a healthy and sustainable diet (a) ........................................................................................................................................................34 Table 15 – Statistical analysis between Food sustainability score and Healthy and sustainable perceptions ...................................................................................................................................... 35 Table 16– type of diet and preferred meal to eat or cook ................................................................ 36 Table 17– Sustainable food score by age ...........................................................................................38 Table 18– Pairwise comparison between perceived Barriers and Dietary Change Status* ................ 42 Table 19– Climate change and agriculture intensification awareness ............................................... 42 Table 20 - Overview of indicators by phase, framework, and data source used in cluster formation and in exploratory analysis .......................................................................................... 46 6 ABBREVIATIONS AMP Área Metropolitana do Porto AML Área Metropolitana de Lisboa BMI Body Mass Index CATI Computer-Assisted Telephone Interviewing COM-B Capability, Opportunity, Motivation – Behaviour model EU European Union FAO Food and Agriculture Organization (of the United Nations) FIES Food Insecurity Experience Scale GDP Gross Domestic Product GHG Greenhouse Gas (Emissions) GIS / ArcGIS Geographic Information System / ArcGIS Pro INE Instituto Nacional de Estatística (Portugal) NUTS Nomenclature of Territorial Units for Statistics SDGs Sustainable Development Goals SEM Socio-Ecological Model SPSS Statistical Package for the Social Sciences WHO World Health Organization 7 SUMMARY The Care4Food study investigates food choices and perceptions related to food security and sustainability within the context of climate change in mainland Portugal. Using a combined socio-ecological (SEM) and behavioral (COM-B) framework, the study analyzes both individuallevel behaviors and regional dynamics. It draws from a comprehensive survey of 1,813 respondents, alongside secondary data such as land use and regional statistics The study surveyed 1,813 individuals representing households, with a balanced gender distribution and an average respondent age of 50. Most households consist of two members, and 10% of participants were born outside Portugal. While 75% report no chronic illness, 62% of respondents are overweight or obese. Households with children spend less on food, likely due to prioritizing other expenses, and higher education levels are most prevalent in Grande Lisboa. 13% of households experience moderate to severe food insecurity, with the Algarve being most affected. Over 60% of respondents are overweight or obese, despite most believing they follow healthy diets, highlighting a clear gap between perception and behavior. A sustainable diet score was developed to address this misalignment. The score reveals that 75% of the sample has less healthy and sustainable practices which may lead to high health and environmental costs. Barriers to healthier eating include food cost, time constraints, and lack of knowledge. Public support is strong for policies that promote healthy and sustainable food systems, though taxation on meat remains unpopular. Women and older individuals are more likely to adopt sustainable practices. Findings also reveal significant territorial disparities, particularly between urbanized coastal areas and aging, low-density inland regions. Regions like Algarve and Alentejo Litoral stand out for their agricultural, touristic, and multicultural characteristics, while rural-urban transition zones show emerging accessibility issues, often car-dependent. This shows that policies should be better tailored to the local context, together with consumers, producers, policy-makers, land use planners, and nutrition experts. Overall, Care4Food underscores the need for systemic change that aligns food policy with health, equity, and environmental sustainability. The report advocates targeted policy interventions, integration of agroecology, and the creation of a Food Sustainability Observatory to support monitoring and civic engagement at local levels. A Geographic information system dasboarh is complementary to this report as a way of feeding and replicate food scores at a more administrative level (feeding municipalities public policies). 8 1. INTRODUCTION 1.1. Food security and food sustainability Global food systems—including production, transportation, processing, packaging, storage, retail, consumption, and food loss and waste—are both significantly contributing to and being affected by climate change. This dynamic threatens to undermine food security, defined as the condition in which “all people, at all times, have physical and economic access to sufficient, safe, and nutritious food that meets their dietary needs and food preferences for an active and healthy life” (FAO, 2019). This definition encompasses key dimensions: availability, accessibility, and utilization—in terms of both safety and nutritional value—as well as sociocultural acceptability and, critically, the stability of these conditions over time. While food is impacted by extreme events like floods or drought, causing food shortages and food insecurity (Ritchie (2024), food and intensive agriculture are also one of the main contributors to climate change, particularly through greenhouse gas emissions. According to Crippa, M. (2021) a third of GHG emissions comes from the agricultural sector. Animal-based foods, particularly red meat, dairy, and farmed shrimp, have the highest GHG emissions due to deforestation for grazing land, methane emissions from livestock digestion, nitrous oxide from animal waste and fertilizers, or the destruction of carbon-absorbing mangrove forests for shrimp farming. In contrast, plantbased foods, such as fruits, vegetables, whole grains, legumes, and nuts require less energy, land, and water, making them more environmentally sustainable (Babiker et al., 2022). Figure1 figure shows the GHG emissions per kilogram of food protein, with beef, followed by lamb and shellfish as highest contributors (Poore & Nemecek, 2018). Figure 1 - Greenhouse gas emissions per 100 grams of protein. Source: Poore, J., & Nemecek, T. (2018). Reducing food’s environmental impacts through producers and consumers. Science. Processed by Our World in Data [retrieved from https://ourworldindata.org/grapher/ghg-per-protein-poore on March 2025] On another angle, the huge consumption of meals containing high calories, meat protein, and dairy not only contributes to greenhouse gas emissions but also to the risk of disease (Willett, et 15 Figure 6: framework adopted from Mitchel etl. 2011 to analyse behavioural changes toward food security and sustainability The Care4Food study follows a three-step methodology to explore the relationship between individual food choices, sustainability, and territorial context. 1. Regional Food System Analysis: In the first step, after literature review, we collect and analyze which key regional indicators of the mainland Portugal better reflect a food system analysis. This analysis is based on indirect data, including census statistics and land use/land cover maps, to understand how structural and geographic factors shape the food environment (NUTS III are represented). Food security dimensions of availability and accessibility are analyzed here, for instance through indicators such as food deserts or fresh veg&fruits accessibility 1 1 Food Desert Methodology The analysis was based on COS 1.1 and 1.2, combined with a 500-meter buffer around fresh food markets, using spatial overlay to extract urban areas located outside these buffer zones. Example – Barrancos: Total urban area: 660,000 m²; Urban area outside a 500-meter radius of a market: 77,500 m²; Resulting percentage classified as a food desert: 11.74%. To estimate how much population resides in these areas, BGRI and subsection data were used, considering the Modifiable Areal Unit Problem (MAUP). Population proportions were estimated based on the area of each subsection that falls inside and outside the buffer. Results: Total population by subsection: 9,855,909; Population intersecting the 500-meter buffer: 6,908,689; Estimated population in urban food desert areas: 3,183,274 (i.e., living more than 500 meters from a fresh food outlet) In addition to this calculation, further analyses were conducted: The fresh-to-bad food ratio, The percentage of "bad food" outlets near schools (by municipality). The percentage of urban area classified as food desert 16 2. Population Survey: The second step involves a direct survey of the mainland Portuguese population to gain insights into daily food choices and perceptions of sustainable diets. The focus is on two dimensions of food security: accessibility and utilization. The survey explores how people make decisions about what to eat, their awareness of environmental impacts, and how they perceive the sustainability of their diets. A Food score is built under certain responses. 3. Exploratory quantitative analysis: In the final step, a quantitative model—combining statistical analysis and machine learning techniques—is applied to identify key drivers of behavioral change. This model integrates individual-level factors and environmental/contextual patterns to better understand the complex dynamics influencing food-related behaviors. The results are synthesized and shared through a dashboard and a story map, designed to facilitate public understanding and support broader dissemination to both scientific and nonscientific audiences (Figure 7). Figure 7 – Workflow of the care4food study (Image generated from chatGPT, 2025) The Care4Food study aligns with Sustainable Development Goals 2 (Zero Hunger), 10 (Reduced Inequalities), and 11 (Sustainable Cities and Communities), as well as the European Union’s Farm to Fork Strategy, by promoting equitable access to sustainable and healthy diets while addressing regional disparities and supporting the transformation of food systems at the territorial level. 17 2. DATA AND METHODS 2.1. Case study presentation The study overviews Portugal Mainland referring to the continental part of the Portuguese Republic, excluding the autonomous regions of the Azores and Madeira 2 . It is bordered by Spain to the east and north, and the Atlantic Ocean to the west and south. In terms of administrative and statistical divisions, Portugal follows the European Union's Nomenclature of Territorial Units for Statistics (NUTS) system. This system is used for regional statistics, economic analysis, and European funding allocation. The NUTS system is divided into three levels: - NUTS I: this includes Mainland Portugal, Azores, and Madeira; - NUTS II: it comprises seven regions on the mainland (Norte, Centro, Grande Lisboa, Península de Setúbal, Oeste e Vale do Tejo, Alentejo, Algarve) and the two island regions. - NUTS III: these are subregions within the NUTS II areas, offering a more detailed level of regional analysis. Mainland Portugal is divided into 26 NUTS III subregions, and these will be used for the aggregated analysis 3 . The following table highlights their total population, gender and age distribution, and GDP per capita, in 2023. All regions exhibit a slightly higher female population percentage, consistent with national trends. Regarding GDP, the AML, particularly Grande Lisboa, stands out with the highest value, reflecting its status as the country's economic hub. Other regions like Algarve and Alentejo Litoral also show relatively high GDP per capita, indicative of their economic activities (tourism and agriculture). Table 1 - Main indicators of the studied regions NUTS III Region Total Population Female Population (%) GDP per Capita (€) Age 0-14 (%) Age 65+ (%) Norte Alto Minho 234 215 52,6 63 11,0 32,3 Cávado 429 833 51,7 67,2 12,9 33,1 Ave 422 464 51,5 64,6 12,1 32,3 Área Metropolitana do Porto 1 802 664 52,4 76,7 12,4 35,0 Alto Tâmega 83 669 52,2 52,8 8,9 28,1 Tâmega e Sousa 409 348 51,5 50,5 12,2 30,5 Douro 184 195 52,2 60,5 10,2 24,1 Terras de Trás-os-Montes 107 473 51,9 59,4 9,5 24,0 Centro Região de Aveiro 384 689 51,9 77,6 12,5 34,0 2 The Azores and Madeira regions were excluded from this analysis due to time and budget constraints. Their inclusion would require theoretical and methodological adjustments to account for their insular and peripheral characteristics, which differ significantly from the mainland context in terms of geography, food systems, and socio-economic dynamics. 3 Some variables of the INE are only available for the Area Metropolitana de Lisboa. That is the aggregation of Peninsula de Setubal and Grande Lisboa. 18 Região de Coimbra 446 982 52,5 69,9 11,3 31,0 Região de Leiria 297 222 51,7 75,8 12,4 30,5 Viseu Dão Lafões 256 810 52,4 61,8 11,3 29,1 Beira Baixa 81 694 52,1 65 10,4 28,0 Beiras e Serra da Estrela 209 896 52,3 55,1 9,7 25,7 Oeste e Vale do Tejo Oeste 388 396 51,6 59,8 13,0 32,6 Médio Tejo 234 765 52,3 62,2 11,3 28,9 Lezíria do Tejo 247 764 51,9 66,8 12,8 30,8 Grande Lisboa + Peninsula de Setúbal (former NUTS II Área metropolitana de Lisboa )* Grande Lisboa + Peninsula de Setúbal 2 961 177 52,8 127,3 + 54,4 14,5 36,4 Alentejo Alentejo Litoral 101 388 48,5 101 11,9 29,3 Baixo Alentejo 115 757 50,9 79,3 12,9 30,4 Alto Alentejo 104 081 52,4 60,9 11,8 29,4 Alentejo Central 153 475 51,9 67,9 12,2 31,0 Algarve Algarve 484 122 51,4 87 13,7 32,7 * This region is composed of Nuts 3 Peninsula de setúbal and Nuts 3 grande lisboa, but for the purpose of this study where we aim to compare both metropolitan regions (Porto and Lisbon) we have aggregated the two subregions in one. Source: INE, Estimativas da População, 2023 Figure 8: Mainland Portugal NUTSII and III Source: DGT, SNIG, 2024 19 2.2. Data collection The data acquisition process combines both direct and indirect methods. Direct data is collected through a survey conducted by GfK, which provides insights into individual behaviors and perceptions regarding food security and sustainability. Indirect data is gathered from sources such as the National Institute of Statistics, OpenStreetMap, and the National Geographic Information System (SNIG), particularly for land use and land cover information. While the survey data captures individual-level decision-making, the statistical and spatial data help contextualize those behaviors within broader environmental and socio-economic settings. All collected data are systematically organized and stored in a spatial database, both at the individual level and aggregated at the NUTSIII regional level ensuring a structured foundation for integrated analysis (will be available in Arcgis online, Zenodo and University of Lisbon repository). A dashboard for visualization and interactive analysis can be found at Available at https://www.arcgis.com/apps/dashboards/9088fa801b454f32afa39d5c84e10c7b 2.2.1. Sample and survey structure The universe of the survey, conducted by GfK for the Care4Food project, consisted of individuals representing their household, of both genders, aged 18 and older, residing in mainland Portugal. Data collection was carried out through telephone interviews, using the CATI (Computer-Assisted Telephone Interviewing) system, between July 8 and August 24, 2024. Interviews were conducted between 5 PM and 10 PM on weekdays and between 11 AM and 10 PM on weekends. Respondents were selected using the quota sampling method, based on a matrix that crossreferenced Gender, Age (4 groups: 18 to 24, 24 to 44, 45 to 64 and 65 and more), and Region. There are 8 regions comprising the Norte, where the Metropolitan Area of Porto was separated from it because of its urban characteristics; the other regions are Centro, Oeste e Vale do Tejo, Grande Lisboa e Peninsula de Setubal (both forming the Metropolitan area of Lisbon, Alentejo and Algarve). These Regions (NUTSII) served as the starting point for selecting households through the random generation of fixed and mobile phone numbers, following the prefixes of each region and operator, while applying the quotas (table 1 and 2). The final sample comprised 1,813 valid interviews, with an acceptable margin of error of 2% within a 95% confidence interval. Table 2 - Sample size by region and gender Region Man Woman Nonbinary NR Sample Norte 154 167 2 323 Área Metropolitana do Porto (AMP) 154 172 326 Centro 148 160 308 Grande Lisboa 170 197 3 370 Oeste e Vale do Tejo 71 83 1 155 Península de Setúbal 69 82 151 Alentejo 44 47 91 Algarve 41 48 89 Sample 851 956 5 1 1813 Source: Care4Food, 2025; GfK, 2024 20 Table 3 – Sample size by region and gender Region 18-24 years 25-44 years 45-64 years 65 >= Norte 8% 30% 36% 25% Área Metropolitana do Porto 8% 30% 36% 26% Centro 7% 27% 35% 31% Oeste e Vale do Tejo 7% 28% 34% 30% Grande Lisboa 9% 33% 32% 26% Península de Setúbal 9% 30% 32% 28% Alentejo 4% 25% 36% 34% Algarve 4% 29% 37% 29% Total 8% 30% 35% 28% Source: Care4Food, 2025; GfK, 2024 The survey started after agreeing on the informed consent about the study and it had 35 questions, and it has the following structure: individual and household 4 socio-economic, demographic, and health situation characterization; its food security, focusing on accessibility; dietary habits and practices; factors of changes in dietary behaviour toward sustainability and healthy food choices; perceptions of policy measures and governmental responsibilities. Most part of the questions were closed responses with a nominal and ordinal scale. The survey was built to address the factors that determine behaviour changes toward healthier and more sustainable food choices, namely Motivation, Capability and Opportunity. It also followed 3 questions of the the food insecurity experience scale (fies) of FAO (Ballard, 2013) and key questions adapted from the National Food, Nutrition and Physical Activity Survey of 20152016 (Lopes at al., 2018), as some comparison could be expected. A pre-test was conducted with the objective of establishing a minimum threshold below which interviews would be considered low quality, evaluating question comprehension, identifying difficulties encountered by respondents, testing the CAPI script to validate inconsistencies between responses, and assessing the total duration of the interview which took about 15 to 18 minutes. 2.2. Proposing quantitative indicators The survey has mostly nominal and ordinal questions, and for purposes of greater analytical power we have created some quantitative variables enabling a more robust statistical analysis of relations between variables, namely of food security, health and food scores. 4 Household composition referred to the number of people living in the same house, regardless of family relationships. 21 2.2.1. Food security indicator Food security, or the perception of food security is analysed three questions adapted from the FAO - Food Insecurity Experience Scale (FIES) (Ballard, et al. 2013): Were you worried about not having enough food due to a lack of money or other means (e.g., lack of equipment, time, difficulty accessing the supermarket, etc.)? Were you unable to have a healthy and nutritious diet due to a lack of money or other means? Did you run out of food at home due to a lack of money or other means? The scale ranges from 1 (food secure) to 4 (severe food insecurity). As we only had 3 questions out of the 8 normally used by FAO, we propose a composite indicator based on 5 variables of the survey (having children, being unemployed, having a lower income, and expenses on food). According to literature review, there are some key factors that contributes to food insecurity related with socio-economic factors (Grimaccia & Naccarato, 2019). As variables were mostly nominal, Chi-square helped us to identify which variables were related with the FEIS scale. They were “having teenage children” (<0.01), “income” (<0.01), “food expenses” (<0.05), and “professional status” (<0.01). The composite indicator is a weighted mean of the variables and FEIS as follows:.5) 𝐼𝑛𝑠𝑒𝑐𝑢𝑟𝑖𝑡𝑦 𝑖𝑛𝑑𝑖𝑐𝑎𝑡𝑜𝑟 = (𝑐ℎ𝑖𝑙𝑑𝑟𝑒𝑛17 ∗0.15)+(𝑖𝑛𝑐𝑜𝑚𝑒∗0.15)+(𝑝𝑟𝑜𝑓𝑒𝑠𝑠𝑖𝑜𝑛∗0.15)+(𝐹𝐸𝐼𝑆 ∗0.5) The results are classified through an adaptation of the FIES scale < 0,5 - Food security; between 0,5 to 0,9 – marginal or potential food security; between 1 to 1,5 – Moderate insecurity and > 1,5 – severe insecurity As we can see in table 4 both indicators are very close to each other, being security overestimated by the FEIS indicator in 7%. Otherwise, other values are close. Table 4– Insecurity levels according to FEIS and Composite indicator FEIS indicator (FAO) Composite indicator Security 1338 74% 1213 67% Potential/marginal security 248 14% 309 17% Moderate insecurity 143 8% 177 10% Severe insecurity 84 5% 114 6% Source: Care4Food, 2025; GfK, 2024 2.2.2. Body mass index Self-reported weight and height in the survey were used to calculate the participants’ body mass index (BMI), classified according to the criteria of the World Health Organization (WHO, 2010) 5 5 From WHO (2010) “A healthy lifestyle - WHO recommendations at https://www.who.int/europe/news-room/factsheets/item/a-healthy-lifestyle---who-recommendations” [retrieved on 28/03/2025) 22 Table 5– Nutritional status of the respondents according to WHO classification (in percentage) BMI Nutritional status Below 18.5 Underweight 1,66% 18.5–24.9 Normal weight 36,18% 25.0–29.9 Pre-obesity 49,74% > 30.0 Obesity 12,42% Source: Care4Food, 2025; GfK, 2024, adapted from WHO, 2010 2.2.3. Food Sustainability a) Indicator (food sustainability score) To better understand perceptions of sustainable dietary habits and practices, our analysis focuses on responses to the question: “Do you consider that you follow a healthy and sustainable diet?”, using a Likert scale from 1 (not at all healthy and sustainable) to 5 (very healthy and sustainable). While this self-assessment provides insight into individual perceptions, it is important to acknowledge that such responses may be overestimated due to social desirability bias— individuals tend to report behaviors they perceive as socially acceptable or desirable (de Ridder et al., 2012). To address the gap between perceived and actual practices, we propose a Food Sustainability Score, constructed from the frequency of responses to a series of dietary and sustainable practices behavior questions. This allows for a objective comparison between perceived sustainability and actual reported habits. - “Do you consider that you follow a healthy and sustainable diet...? 5 - Never; 4Occasionally in the month; 3 - Once a week; 2 - A few times a week; 1 - Every day”; - “How often do you consume each of the following foods…? 6 - Never; 5Occasionally in the month; 4 - Once a week; 3 - A few times a week; 2Every day; 1 - Several times a day. The groups of products are: White meat and fish; Red meat; Milk and dairy products; Eggs; Cereals, nuts, and legumes (includes bread, excludes ultra-processed breakfast cereals, for example); Vegetables; Fruits; Sweet and savoury snacks; cured meats and ultra-processed foods such as pre-cooked meals; Takeaway food. - “How often do you have lunch and/or dinner outsider home? 1 - Every day, 2 - A few times a week, 3 - Once a week, 4 - Occasionally in the month, 5 – Never” - “How often do you buy fresh food products (i.e., fruit and vegetables)?: 1 - Every day, 2 - A few times a week, 3 - Once a week, 4 - Occasionally in the month, 5 – Never” - What transport mode do you use? car, walking, other. - Do you waste food? 1 - Every day, 2 - A few times a week, 3 - Once a week, 4 - Occasionally in the month, 5 – Never” The final food score is calculated from four sub-scores: 1. Protein Score (score_proteina) 2. Fruit and Vegetable Score (score_fruta) 3. Processed Foods Score (score_processados) 4. Sustainability Score (score_sustentabilidade) 23 Each sub-score ranges from 1 (least sustainable) to 5 (most sustainable). The final score is the arithmetic mean of the four sub-scores, i.e. a composite score between 1 and 5, representing the participant’s general diet quality, considering nutritional balance and sustainable behaviors (see box 1). This sustainable food score is alligned with academic research operationalizing sustainable diets through composite indices, namelly, the Sustainable Diet Index (SDI) (Seconda et al., 2020) combining dietary quality, environmental footprint, economic cost, and socio-cultural practices; the Dietary Pattern Sustainability Index (DIPASI) (Bôto et al., 2024) measured deviation from the mediterranean diet as a reference sustainable pattern, or the Planetary Health Diet Index (PHDI) (Bui et al., 2024) quantified adherence to the EAT-Lancet diet, linking higher scores to reduced mortality and environmental impact. Box 1 Sustainable food Scoring System – Formula Explanation The set of formulas presented here compute a score based on questionnaire responses. The calculation is structured into four sub-scores: - Protein Score (score_proteina) - Fruit and Vegetable Score (score_fruta) - Processed Foods Score (score_processados) - Sustainability Score (score_sustentabilidade) Each sub-score ranges from 1 (least sustainable) to 5 (most sustainable). The final score is the arithmetic mean of the four sub-scores. 1. Common Formula Structure All scoring functions follow the same mathematical pattern: number(min(5, max(1, base_score + adjustments))) The min(5, …) function ensures that no score exceeds 5. The max(1, …) function ensures that the score does not fall below 1. This guarantees that all sub-scores are normalized within the 1–5 range. 2. Protein Score (score_proteina) Variables (questions) considered: cereais – frequency of cereal consumption carne_branca – white meat (fish/poultry) carne_vermelha – red meat leite – milk and dairy products leguminosas – legumes (beans, lentils, etc.) Scoring logic: The protein score begins with a base value of 3, then applies several conditional adjustments: Cereals and balanced protein sources: increased score if cereals are consumed daily or several times per week, combined with moderate consumption of white meat or dairy. Red meat: low or occasional intake (nunca, ocasionalmente, 1x por semana) → +1 Moderate (algumas vezes por semana) → –2 Frequent or daily (1x por dia, várias vezes) → –3 White meat and dairy: Moderate weekly consumption increases the score; excessive or daily intake decreases it. 24 Legumes: Regular consumption (weekly or more) adds +1. 3. Fruit and Vegetable Score (score_fruta) Variables considered: frutas – fruit intake frequency vegetais – vegetable intake frequency Scoring logic The score evaluates how regularly fruits and vegetables are consumed: Frequency pattern Score Both “várias vezes por dia” 5 Both daily or one daily and the other multiple times 4 Both several times per week 3 Both once per week 2 Less frequent consumption 1 An additional penalty (–1) is applied if one food group is consumed very frequently while the other is consumed rarely or never (to discourage imbalance). 4. Processed Foods Score (score_processados) Variable considered: snacks – frequency of processed snack consumption Scoring logic Frequency of consumption Nunca (never) 5 Ocasionalmente no mês 4 Uma vez por semana 3 Algumas vezes por semana / Uma vez por dia 2 Várias vezes 1 This score inversely reflects the frequency of processed food intake: less frequent consumption yields a higher score. 5. Sustainability Score (score_sustentabilidade) Model starts at 2 expecting more possible positive actions than negative ones Variables considered: take_away – frequency of take-away meals jantar_fora – frequency of eating out onde_compra – type of shop (e.g., supermarket, market, local producer) como_desloca – mode of transport for food shopping sobras – frequency of food waste compra_frutas – frequency of purchasing fruit and vegetables perto_longe – proximity preference when buying products and if it’s part of the daily commuting Scoring logic: Rare or no take-away / eating out +1 Buying from local markets or producers +1 Using sustainable transport (walking/bicycle) +1 Using private car –2 31 According to figure 12, women are more affected by food insecurity than men, with 19% of women experiencing moderate to severe food insecurity, compared to 13% of men with a chisquare association between food security status and gender of p < 0.05). Figure 12: Food security composite indicator by gender(%) Source: Care4Food, 2025; GfK, 2024. 3.2.2. Physical accessibility to food Food access can be both economic and physical—including time-related factors. Figure 13 illustrates the average travel time to access a store that sells fresh fruits and vegetables, which is 11 minutes for most parto f the participants, with a standard deviation of 7 minutes. Most regions report travel times between 10 and 14 minutes. The Area Metropolitana do Porto (AMP) has the shortest average distance, while in Alentejo, the average travel time is above the national average. Figure 13 - Participant representation by average distance to a fresh food store (%) Source: Care4Food, 2025; GfK, 2024. 72% 15% 8% 5% 62% 19% 11% 8% 0% 10% 20% 30% 40% 50% 60% 70% 80% Man Woman 32 Most participants (68%) buy fresh fruits and vegetables near their homes, while 8% purchase them closer to their workplace, and 24% report buying in both locations. Most participants are satisfied with their access to fresh products within a 5-minute walk from home (69%), with a statistically significant association of p < 0.01. Table 11– Diversity fresh vegetables and fruits supply within a 5 minute walking distance from home Is there a diversity of fresh food supply near home? Are you satisfied with the diversity of fresh food supply directly from the producer? No Yes No Yes Norte 34% 66% 19% 81% Área Metropolitana do Porto 27% 73% 18% 82% Centro 39% 61% 16% 84% Oeste e Vale do Tejo 38% 62% 21% 79% Grande Lisboa 26% 74% 24% 76% Península de Setúbal 27% 73% 25% 75% Alentejo 32% 68% 20% 80% Algarve 27% 73% 24% 76% Total 31% 69% 20% 80% Source: Care4Food, 2025; GfK, 2024. Of the 1,795 participants who responded to the question about the modes of transport that they use to purchase fresh fruits and vegetables, some use multiple transport modes, while the majority (63%) reported using a private car, and 34% reported walking. The car use is significantly associated with the average distance to the store (p < 0.01). Table 14 shows that for short distances (within 5 minutes), participants are more likely to walk (37%) when compared to the private car use (32%). Table 12– Mode of transport by distance time Soft mode Public transport Car Total Soft mode Public transport Car Total 5 minutes 268 8 438 714 37% 10% 32% 33% 6 to 10 274 18 524 816 38% 23% 39% 38% 11 to 15 109 20 247 376 15% 26% 18% 17% 16 or more 74 31 150 255 10% 40% 11% 12% 725 77 1359 2161 100% 34% 4% 63% 100% Source: Care4Food, 2025; GfK, 2024. Differences in the use of transports modes by region are statistically evident (<0.01), confirming that the transportation modes vary depending on the region in which people live. Some regions have a significantly higher proportion of people walking to acquire fresh food, while others rely less on walking. 33 The percentage-based heatmap shows where walking is most common and where it is less used. AML and AMP have higher values (42% and 35%) possibly meaning that urban areas with high accessibility may see more walking, while rural areas might depend on motorized transport. This can also show that in more urbanized regions, car use might be lower due to the availability of nearby stores, and that metropolitan regions may have better public transport infrastructure, leading to a higher percentage of residents using walking and transport mode. Figure 14: Percentage of participants in the survey using each transport mode by region Source: Care4Food, 2025; GfK, 2024. There is a statistically significant relationship between walking and buying from local or smallcircuit markets, while the use of a private car is strongly associated with global retail purchases (p < 0.01). When asked about the type of stores where they purchase fresh products, 20% of participants reported buying within small circuits, including direct purchases from local producers and street markets. This percentage is slightly higher (21%) in the Norte, AML, and OVT regions, where local agricultural production is more common. There is also a statistically significant regional variation in small circuit purchasing behavior (p < 0.05). Most participants purchase fresh products from local supermarkets, followed by global retail chains. Notably, 102 individuals reported buying from all types of outlets, while 387 participants shop in both local supermarkets and small-circuit markets. Table 13– types of stores to purchase fresh fruits and vegetables Small chains Local Supermarket and grocery stores National/ global supermarkets chains All types Local and Small circuits Norte 25% 42% 27% 1% 4% AMP 21% 43% 31% 1% 4% Centro 21% 40% 25% 3% 11% OVT 21% 29% 22% 6% 21% Grande Lisboa (AML) 17% 43% 27% 3% 11% P. Setúbal (AML) 14% 34% 22% 5% 24% Alentejo 18% 40% 16% 5% 21% 34 Algarve 17% 31% 29% 6% 17% Total 20% 39% 26% 3% 12% Source: Care4Food, 2025; GfK, 2024. 2.2.3. Food utilization: dietary habits and practices Participants largely follow a protein-rich diet (66%), while 31% adhere to a Mediterranean diet, and only 3% identify a vegan or vegetarian. Although 81% of participants report following a healthy and sustainable diet either daily or several times a week, this percentage drops among those following protein-heavy (67%) and high-fat diets (55%), and is higher among those following Mediterranean (81%) and flexitarian diets (40%). The relation between both variables are statistically significant at p < 0.01. Table 14 – Relation between the type of diet (b) and perception of a healthy and sustainable diet (a) Do you consider that you follow a healthy and sustainable diet…(a) What type of diet do you consider you follow? (b) Never Occasionally in the month Once a week A few times a week Every day Total (b) Mainly based on meat, fish, milk, and dairy products 86% 83% 83% 67% 55% 66% Mainly based on fruit, vegetables, fish, and olive oil as main fat (e.g. Mediterranean, Flexitarian) 13% 14% 16% 31% 40% 31% Special diet 0% 0% 0% 0% 1% 0% Vegan & Vegetarian 1% 3% 1% 2% 5% 3% Total (a) 5% 9% 5% 47% 34% 100% Source: Care4Food, 2025; GfK, 2024 The regional distribution of diets follows a similar pattern across the country, with the majority of participants adhering to diets rich in fat, dairy, and protein. The Mediterranean and flexitarian diets are more prevalent in the Algarve (39%), while special diets—such as vegetarian, vegan, or other specific dietary patterns—are more common in metropolitan areas. 35 Figure 15: Type of diet by region, in percentage Source: Care4Food, 2025; GfK, 2024. To assess whether participants perceive a connection between their diet and healthy, sustainable habits, we correlated indicators of perceived healthy and sustainable diets with actual healthy and sustainable practices (i.e. food score). The Pearson correlation coefficient was 0.28 (p < 0.01), indicating a very weak positive relationship between the two variables. This suggests a slight gap between perceived and real sustainable habits. Participants generally believe they are quite sustainable (4/5), but their actual behaviors reflect only moderate sustainability (3/5). The data show reasonable consistency, but also highlight a perception–practice gap that could be addressed through education and behavioral reinforcement. Table 15 – Statistical analysis between Food sustainability score and Healthy and sustainable perceptions Indicators Score Perceptions Mean 3 4,0 Median 3 4,0 Mode 3.25 4,0 Standard deviation 0,8 1,1 Curtosis -0,6 0,9 Assimetry 0,1 -1,2 Pearson correlation = 0,24 Source: Care4Food, 2025; GfK, 2024. Additionally, we assessed whether there was an association between diet perception, type of diet, and the type of meal participants would choose for lunch or dinner. The results revealed a statistically significant relationship (p < 0.01). 36 In terms of dietary habits, most participants (66% or 1,175 individuals) follow a diet rich in fat, protein, and dairy, with 62% of those selecting meat and fish-based dishes. Approximately 38% of participants tend to choose less sustainable and less nutritious options, often favoring meatbased meals—though some of these choices may be influenced by perceived health benefits of salads or grilled items, without considering the full nutritional context of the dish. Meanwhile, 40% of participants prefer moderately sustainable and healthy options, with codfish emerging as the most frequently chosen dish. This reflects strong cultural food preferences, as codfish is common across all diet types. Notably, some individuals identifying as vegetarian or vegan selected animal-based dishes, which may indicate a gap in diet literacy—with 18 out of 48 respondents in these groups making such choices. Interestingly, only 22% of participants opted for healthier and more sustainable meal options. Table 16– type of diet and preferred meal to eat or cook Meat, fish, milk, and dairy products Fruit, vegetables, fish, and olive oil as the main fat Special diet Vegan & Vegetarian Total Avocado, Algarve mango, and shrimp salad (1) 15% 22% 0% 13% 17% Grilled beef steak, rice, and salad (1) 26% 11% 40% 6% 21% Goat meatballs with seasonal vegetables (2) 3% 2% 0% 2% 3% Boiled codfish with patato and vegetables (2) 36% 41% 60% 17% 37% Chickpeas with poached eggs (3) 16% 18% 0% 21% 17% Tofu jardinière (3) 3% 6% 0% 42% 5% Total 66% 31% 0% 3% 100% Source: Care4Food, 2025; GfK, 2024. Figure 16 highlights an almost inverse relationship between dietary perceptions and actual meal choices, suggesting that individuals may place more importance on their perception of eating healthily than on the reality of their dietary behavior. 37 Figure 16:relation between the perception of a sustainable and healthy diet and meal option 8 Source: Care4Food, 2025; GfK, 2024. 2.2.3.1. Gender and age in food utilization The gender dimension, illustrated in Figure 17, shows that women are more likely to engage in healthier and more sustainable practices (77% compared to 23% of men). Conversely, men are more likely to adopt less healthy practices (57% compared to 43% of women). However, these differences are not statistically significant. 8 Classification: perception < 3 less healthy and sustainable practices; 3 to 4 moderate Healthy and sustainable practices and > 4 healthy and sustainable practices. Meal: less healthy and sustainable: Avocado, Algarve mango, and shrimp salad (1), Grilled beef steak, rice, and salad (1); moderate healthy and sustainable: Goat meatballs with seasonal vegetables (2) Boiled codfish with patato and vegetables (2) Healthy and sustainable: Chickpeas with poached eggs (3), Tofu jardinière (3) 0% 20% 40% 60% 80% 100% Less healthy and sustainable Moderate healthy and sustainable Healthy and sustainable Perception Meal option 38 Figure 17: relation between practices (food score) and perception of a sustainable diet Source: Care4Food, 2025; GfK, 2024. Although the majority of participants have healthy practices (47%), individuals aged 45 and older tend to follow a healthier and more sustainable diet compared to younger age groups. This association with age is statistically significant (p < 0.01). Table 17– Sustainable food score by age 18-24 25-44 45-64 65 and + Total Less healthy and sustainable practices 9% 35% 31% 25% 47% Some Healthy and sustainable pratices 7% 26% 38% 28% 42% Healthy and sustainable pratices 4% 21% 36% 39% 11% Total 8% 31% 37% 30% 1813 Source: Care4Food, 2025; GfK, 2024. The distribution of cooking responsibilities within households is uneven between men and women (p < 0.01). 70% of women report being primarily responsible for cooking, compared to 32% of men. Meanwhile, 27% of men state that they share the task with another household member, compared to 19% of women. Shared cooking responsibilities are more common among individuals aged 25 to 64, occurring in 73% of those cases. Men perception Women perception Men score Women score less sustainable 57.3 42.7 57.0 43.0 Moderate 45.1 54.9 43.5 56.5 Highly sustainable 46.0 54.0 23.0 77.0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 % 39 Figure 18: distribution of cooking responsibilities by gender Source: Care4Food, 2025; GfK, 2024 3.3. Barriers and enabler of food change When asked about making changes toward a healthier and more sustainable diet ("Have you already changed or would you consider changing your food choices to contribute to environmental sustainability?"), 32% (589 participants) reported that they have already changed their eating habits. 35% (635) said they have not made any changes, while the remaining 33% indicated they are considering making changes. Figure 19: Changes made in eating habits Source: Care4Food, 2025; GfK, 2024 When asked what changes participants made, many reported making multiple changes (113 participants). The most common change was reducing red meat consumption (197 responses), followed by buying local and seasonal products or growing their own food, which together accounted for 170 responses. Regarding motivation for these changes, 321 responses cited health reasons, while 272 responses were driven by environmental sustainability concerns. Another Didn't change 35% Considering changing 33% Changed 32% Didn't change Considering changing Changed 40 aspect explored among those who have changed or considered changing their diet is what they consider important to change. Overall, participants rated most aspects as very important, with the exception of buying alternative products and switching exclusively to a vegetarian or vegan diet, which received lower importance ratings (median scores of 1 and 2, respectively). This may reflect cultural influences that warrant further analysis (Figure 20). Figure 20: What is important to change (responses from 1 - not importante to 5 -very important) Source: Care4Food, 2025; GfK, 2024 Regarding barriers to dietary change (Figure 20), participants generally agreed that the price of food is a significant obstacle. Among those who have already changed their eating habits, additional barriers included the lack of clear food labeling and the belief that unhealthy food helps manage stress. Participants who are considering a change cited several barriers: high food prices, limited availability of fresh products, lack of time to cook, insufficient knowledge, and again, the perception that unhealthy food helps cope with stress. For those who are not planning to change, the main obstacles were food cost and lack of knowledge. 47 1 OF_TECH_MUNICI PAL Environmental context Physical opportunity Use of technologies in the governement sector INE, Stats yearbook 20212023 1 OF_TECH_POP Environmental context Physical opportunity Proportion of population using digital technologies INE, Stats yearbook 20212023 1 OF_VARPOP Environmental context Physical opportunity Variation in population size over time Population census, 2021 2 MR_CHANGE_HA BITS (dependent Z) Individual Reflexive motivation Has changed toward sustainable consumption: 1 = Yes; 0 = No survey, 2024 2 CPP_GENDER Individual Social Psychological capacity / Physical Gender of the respondent: 1 = Male; 2 = Female; 0 = Non-binary; 3 = Prefer not to answer survey, 2024 2 CPP_AGE Individual Psychological / Physical capacity Age of the respondent survey, 2024 2 OS_ADULTS Social context Social opportunity Number of adults in the household survey, 2024 2 OS_CHILDREN Social context Social opportunity Number of children in the household survey, 2024 2 CF_PROFESSION Individual Physical capacity Professional status: 1 = Retired; 2 = Employed; 3 = Student; 4 = Unemployed; 5 = Domestic work survey, 2024 2 MR_INSECURITY_ FAO Individual Reflexive motivation Food Insecurity Experience Scale (FIES) – FAO food security index: 0 = secure; 1 = Moderately insecure (or at risk); 3 severely insecure survey, 2024 2 CF_COOKING Individual Physical capacity Main person responsible for cooking: 1 = Respondent; 2 = Spouse; 3 = Shared; 4 = Other; 5 = No one cooks survey, 2024 2 CF_SHOPPING Individual Physical capacity Main person responsible for shopping: 1 = Respondent; 2 = Spouse; 3 = Shared; 4 = Other; 5 = No one shops survey, 2024 2 OF_DISTANCE Environmental context Physical opportunity Time to walk to nearest fresh food store: 1 = 5 minutes; 2 = 10 minutes; 3 = 15 minutes or more survey, 2024 2 OF_DIVERSITY Environmental context Physical opportunity Is there a diversity of fresh food available within a 5-minute walk? 1 = Yes; 2 = No survey, 2024 2 CP_CLIMATECHA NGE Individual Psychological capacity Awareness of climate change: 2 = Yes; 1 = No; 0 = Don’t know survey, 2024 2 CP_AGRICULTURE Individual Psychological capacity Awareness of intensive agriculture issues: 2 = Yes; 1 = No; 0 = Don’t know survey, 2024 2 OF_P24_1_PRICE OF FOOD Environmental context Physical opportunity Price of food as a barrier to changing habits (1 = Very important to 5 = Not important) survey, 2024 2 CP_FOOD_LACK_K NOWLEDGE Individual Psychological capacity Lack of knowledge as a barrier (1 = Very important to 5 = Not important) survey, 2024 2 MA_STRESS Individual Automatic motivation Stress as a barrier (1 = Very important to 5 = Not important) survey, 2024 2 OS_NORMS_VALU ES Social context Social opportunity Cultural and social norms as a barrier (1 = Very important to 5 = Not important) survey, 2024 2 MR_FEAR_ADEQU ATE_NUTRITION Individual Reflexive motivation Fear of inadequate nutrition as a barrier (1 = Very important to 5 = Not important) survey, 2024 2 OF_OFFER OF FOOD Environmental context Physical opportunity Lack of fresh food availability as a barrier (1 = Very important to 5 = Not important) survey, 2024 2 MR_LACK_OF_TA STE Individual Reflexive motivation Lack of taste or cooking skills as a barrier (1 = Very important to 5 = Not important) survey, 2024 48 2 CF_LACK_OF_TIM E Individual Physical capacity Lack of time to cook as a barrier (1 = Very important to 5 = Not important) survey, 2024 2 CP_AUTOSUFFICIE NCY Individual Psychological capacity Awareness of self-sufficiency: 2 = Yes; 1 = No; 0 = Don’t know survey, 2024 2 OS_STATE_INTER VENTION Social context Social opportunity Belief in the need for state intervention: 2 = Yes; 1 = No; 0 = Don’t know survey, 2024 2 CP_BMI Individual Physical capacity Body Mass Index (BMI) survey, 2024 2 CP_EDUCATION Individual Psychological capacity Education level: 0 = Illiterate; 1 = Primary; 2 = Basic School; 3 = High School; 4 = Graduate; 5 = Doctorate survey, 2024 2 CF_DISEASES Individual Physical capacity Presence of chronic disease: 1 = Yes; 0 = No survey, 2024 2 MR_DIET_PERCEP TION Individual Reflexive motivation Self-perceived diet quality: 5 = Very good to 1 = Very poor survey, 2024 A two-step exploratory analysis was conducted. First, environmental and social context variables were used (see table 20 column - step) to identify types of regions. Secondly, individual context variables were used to analyze what best predicts behavioral changes. In this second stage, sustainable diet score and habits change are the variables that were more likely to be used as dependent variables. 4.1. Cluster Analysis of the Food and Contextual Environment in Portuguese NUTS III Regions In this analysis we aimed to identify types of territories (regions NUTSIII) based on variables reflecting the food environment (demographic, socioeconomic, and infrastructural factors) and the social context. That is a set of Indicators were selected to reflect dimensions such as: - Food environment and Infrastructure: fast food availability, proximity to retail, horticulture production, car ownership, water consumption. - Socioeconomic and demographic context: unemployment, purchasing power, education, urban pressure, foreign population, aging index All variables were standardized (z-scores) to ensure comparability across regions. K-Means clustering was applied using Euclidean distance. The Elbow method and silhouette scores supported choosing 4 clusters: - Cluster 1: Rural and Aging Context Low access to food stores, low fast food and car ownership. High aging index, lower education, and purchasing power. Indicates a low-access, low-choice food environment, possibly at risk of food insecurity among older populations. - Cluster 2: Agricultural and multicultural Area High water use, foreign population, high agricultural exports and and accessible food environment. Moderate socioeconomic indicators and lower urban pressure. Represents a semirural and specialized environment. - Cluster 3: Transitional/Developing 49 Mixed profile with some urban traits and moderate scores across most indicators. May represent semi-urban or periurban environment. - Cluster 4: Urban, Diverse & Consumer-Oriented High densities of stores where it is possible to access fast and ultra-processed food, proximity to fresh food stores, higher levels of foreign population, purchasing power, young population and education. Strong infrastructure (internet, water access). Reflects a high-access, high-choice food environment typical of metropolitan areas. Analyzing these regions through the lens of the COM-B model allows us to infer how regional differences may shape sustainable and secure food behaviors. In this model, Opportunity refers to environmental and social-structural factors that either enable or constrain food behavior at a broader level. Proximity to food stores, car ownership, and the quality of urban infrastructure serve as useful proxies for physical opportunity. Clusters 2, 3, and 4 benefit from more favorable opportunity structures, while Cluster 1 faces more significant limitations. These opportunities are closely interlinked with Capability and Motivation. For instance, environments characterized by higher educational levels or a more diverse, younger population may enhance capability and foster motivational factors such as habits, norms, and cultural openness toward sustainable food behaviors. Clusters 2 and 4 exhibit the strongest motivational potential, driven by economic vitality and cultural diversity. In contrast, Cluster 1 may encounter motivational barriers, largely due to an aging population and more static social structures. The regional clusters reflect observed or inferred food-related behaviors, shaped by varying levels of food accessibility, infrastructure, and socioeconomic and cultural conditions. These behaviors influence both individual food choices and broader outcomes such as food sustainability and food security. Notably, Cluster 3 appears most vulnerable, potentially lacking both opportunity and capability. On the other hand, Cluster 1 aligns with high-choice, high-consumption environments, which present different kinds of challenges for sustainable food behavior. This clustering approach provides a valuable framework for contextualizing territorial differences and guiding region-specific strategies in public health, food policy, and spatial planning. 50 Figure 26: NUTS III Clusters Source: Care4Food Figure 27: Cluster and variables B-plots Source: Care4Food 4.2 Individual-Level Food Behavioural changes 51 To complement the contextual analysis, an individual-level behavioral model was developed using a binary logistic regression applied to the full national sample and subsequently disaggregated by Clusters 1, 2, 3, and 4. The model estimated the probability of reported changes in food habits as a function of selected predictors. The dependent variable, MR_CHANGE_HABITS, was a binary indicator denoting whether respondents reported a shift toward more sustainable food practices (1 = Yes; 0 = No). To preserve statistical power and avoid listwise deletion, “Don’t know” responses were not treated as missing values but were instead recoded into dummy variables and retained as valid categories. This approach recognizes that non-response may capture meaningful constructs (e.g., uncertainty or low awareness) and allows for a more inclusive estimation of behavioral predictors. All predictors exhibited acceptable multicollinearity levels, with Variance Inflation Factors (VIFs) below 2—well within conventional thresholds (typically VIF < 5 or 7). The general model was estimated using a nationally representative sample of 1,170 individuals, comprising 599 respondents who reported no change and 571 who reported a change. Analyses were conducted in SPSS (version 28). Following preliminary testing, the final model was estimated using the stepwise forward conditional method, which sequentially introduces predictors based on their statistical significance. Key configuration parameters included a significance level of 0.05 for entry, 0.10 for removal, and a classification cut-off of 0.50 for predicted probabilities. Model diagnostics assessed overall fit, classification accuracy, and residual patterns. The logistic regression model follows the standard form (1) log(𝑃 1−𝑃)= β0+β1𝑋1+⋯+β𝑘𝑋𝑘 where P is the probability of reporting a change in food habits, β0 is the intercept, and β1 through βk are the estimated coefficients for each predictor variable x1through Xk The predicted probabilities can be obtained by transforming the logit score Z using the logistic function of equation (2): (2) 𝑃 = 1 1+𝑒−𝑍 The national-level logistic regression model was statistically significant at step 11, χ²(11) = 236.108, p < .001, explaining approximately 24.4% of the variance in the outcome variable (Nagelkerke R² = .244). The Hosmer–Lemeshow test indicated a good model fit (χ² = 4.649, df = 8, p = .794). The model achieved an overall classification accuracy of 69.0%, correctly predicting 68.4% of cases with no behavioral change and 69.7% of cases reporting a change. Model performance was further evaluated using Receiver Operating Characteristic (ROC) curve analysis. The area under the curve (AUC) was 0.748 (SE = 0.014, 95% CI [0.720, 0.775]), indicating good discriminative ability. This suggests that the model correctly distinguishes between individuals who changed their behavior and those who did not approximately 74.8% of the time. The result was statistically significant (p < .001), confirming that the model performs significantly better than chance. 52 Of the 26 predictors included in the initial model, 11 were statistically significant (p < .05) at the national level. When disaggregated by food environment clusters, the number of significant predictors varied: 4 for Cluster 1, 8 for Cluster 2, 3 for Cluster 3, and 11 for Cluster 4 (see Table 21). Table 21 - Logistic Regression Coefficients (B) and Odds Ratios (Exp(B)) for General Model and Clusters 1–4 of Dietary Change (MR_CHANGE) Variable code General model Cluster 1 Cluster 2 cluster 3 cluster 4 Z MR_CHANGE B Exp(B) B Exp(B) B Exp(B) B Exp(B) B Exp(B) x1 CPP_GENDER 0.368 1.445 0.714 2.042 x3 CP_CLIMATECHANGE 1.008 2.739 1.074 2.926 x5 OF_PRICE OF FOOD - 0.848 0.428 -0.176 0.839 x6 CP_LACK_KNOWLEDGE -0.18 0.835 - 0.262 0.769 -0.139 0.87 x8 OS_NORMS_VALUES - 0.307 0.735 x9 MR_FEAR INADEQUATE_NUTRITION - 0.227 0.797 -0.232 0.793 x10 OF_OFFER OF FOOD -0.402 0.669 x11 MR_LACK_OF_TASTE 0.276 1.318 x15 OS_STATE 0.423 1.526 0.404 1.497 x16 CP_BMI -0.056 0.945 x17 CP_EDUCATION 0.326 1.385 0.541 1.717 0.376 1.457 1.118 3.058 0.229 1.257 x19 MR_DIET_PERCEPTION 0.308 1.36 0.361 1.435 0.267 1.306 x23 MR_INSECURITY_FAO 0.516 1.676 x24 CF_COOKING -0.391 0.677 x26 OF_DISTANCE 0.206 1.229 Dummy (don't know answers) CP_CLIMATECHANGE 1.206 3.339 0.026 MR_FEAR INADEQUATE_NUTRITION - 1.618 0.198 - 21.03 0 0.004 MR_LACK_OF_TASTE - 0.999 0.368 OS_NORMS_VALUES - 21.053 0 - 1.411 0.244 CF_LACK_OF_TIME - 20.693 0 Some statistically significant predictors were consistent with theoretical expectations, such as gender, education level, and perceived diet quality, thereby reinforcing the validity of the analytical framework. In contrast, variables including age, employment status, and household composition did not emerge as significant in the general model. 53 When linking individual behavior change to regional clusters, preliminary exploration suggests: - Ageing regions (Cluster 1) demonstrate lower change intention, consistent with barriers such as low education and strong ingrained habits. - Agricultural and multicultural regions (Cluster 2) exhibit mixed responses, maybe influenced by habits and food preferences. This Cluster offers strong potential for developing short food supply chains and fostering local food resilience. - Transitional areas (Cluster 3) fall in between, showing variability that may reflect evolving infrastructure and awareness. These regions should be prioritized for infrastructure improvements and proactive food environment planning - Urban/affluent regions (Cluster 4) show higher proportions of individuals intending to change behavior, possibly due to higher education and label awareness. Urban clusters (Cluster 4) may require policies promoting healthy food choices and urban design that mitigate overexposure to ultra-processed foods. 54 5. CONCLUSIONS AND RECOMMENDATIONS This study aimed to provide an in-depth understanding of food choices and perceptions among the population of mainland Portugal, analyzed at both regional and individual levels. In addition to the collection of secondary indicators (such as statistics and land use data), a comprehensive survey of 1,813 individuals was conducted. This allowed the identification of territorial patterns and individual drivers of dietary change toward more sustainable food behaviors. The study employed a combined socio-ecological and behavioral framework (SEM and COM-B) to examine how individual and environmental factors influence food behavior, with the additional goal of informing public policy and supporting systemic changes in food systems. This study confirms that dietary behavior change toward sustainability is not merely a matter of individual choice, but the result of complex, interacting factors spanning biological, psychological, environmental, social, and territorial dimensions. In line with socio-ecological models (Story et al., 2008; HLPE, 2017), our findings show that personal capability and motivation are deeply influenced by contextual opportunity structures. These insights are especially relevant in the context of ongoing global efforts to transform food systems in alignment with the Sustainable Development Goals and European policy frameworks. 5.1. Main limitations This study also faced some limitations. They are related with the sample itself. Although quota sampling ensured demographic balance across age, gender, and region, certain vulnerable or atrisk groups may still be underrepresented—such as undocumented migrants or individuals living in extreme poverty or in food deserts. These populations are often more severely food insecure and face more complex barriers to adopting sustainable eating habits, potentially skewing the results toward more favorable averages. The study also relies heavily on self-reported data, including dietary habits, health status (e.g., self-reported weight and height for BMI), and perceptions of sustainable practices. While this approach is practical and cost-effective, it introduces potential biases, such as recall bias and social desirability bias. Participants may overstate the healthiness or sustainability of their diets, particularly when these behaviors are viewed as socially desirable. This tendency is reflected in the weak correlation (Pearson = 0.23) between perceived and actual sustainable food practices. To address this, a sustainable diet score was calculated to better assess real behaviors. Another limitation involves the measurement of food insecurity. The study employed only 3 of the 8 standard questions from the FAO’s Food Insecurity Experience Scale (FIES), which reduces the capacity to capture the full spectrum of food insecurity experiences. The full FIES is designed to assess not only food access, but also the psychological and behavioral impacts of food insecurity. As such, using a shortened version may have underestimated moderate or severe food insecurity, or failed to reflect coping strategies. To compensate, a composite food security 55 indicator was developed, incorporating variables such as the presence of children, employment status, income level, and food expenses. Furthermore, the study is cross-sectional, offering a snapshot in time. While this design is suitable for identifying associations and describing current behaviors, it does not support causal inference. For instance, although the analysis shows a relationship between high BMI and chronic disease, it cannot determine whether poor dietary habits caused obesity, or if pre-existing health conditions influenced eating behaviors. Longitudinal studies would be required to clarify such causal relationships. Lastly, caution is needed when interpreting regional data. While the study aggregates results at the NUTS III level to allow for regional comparison, the sampling frame was designed at the NUTS II level. As a result, intra-regional disparities—particularly in larger or more heterogeneous regions—may be obscured. Interpretations at the sub-regional level should therefore be approached with care, recognizing potential sampling limitations and variability within regions. 5.2. Results in the broader conceptual framework At the regional level, the data highlight that mainland Portugal is far from homogeneous. Clear distinctions emerge between coastal and inland areas, with metropolitan regions showing different dynamics compared to aging, low-density interior regions. Certain territories—such as Alentejo Litoral and Algarve—stand out due to the influence of tourism, agriculture, and multiculturalism, while rural-urban transition zones show increasing levels of accessibility, albeit still heavily reliant on car use. At the individual level, the results reveal diverse perceptions and behaviors: ● 13% of Portuguese households face moderate to severe food insecurity, with the highest rates in the Algarve region. ● 66% of diets are rich in meat and dairy, and 3% of the participants follow vegetarian or vegan diets. Algarve report following a mediterranean diet (39% against 31% national average) ● Only 22% chose highly sustainable meals when presented with several options—showing a significant gap between knowledge and action, even among those who identify with sustainable dietary practices. ● The average food sustainability score is 3 out of 5, placing 52% of the population above average. 12% of the sample as a score above 2. Coimbra Region as the higher average score. In contrast the perception of having a sustainable diet with overrated as 81% believe following and sustainable and healthy diet. ● 32% have changed their eating habits, mainly reducing red meat; 33% are considering changes and 35% do not consider change rating barriers like cost and knowledge has perceived obstacles. ● People who changed their diet have changed because of health issues (61%) ● 40% believe their diet has impact on climate change, but the other 40% don’t believe in that impact ● Cooking responsibilities remain gendered, with 70% of women reporting primary responsibility versus 32% of men. 56 ● Most participants are satisfied with their access to fresh products within a 5-minute walk from home (69%), but 63% of participants use a private car for food shopping, highlighting reliance on non-sustainable transport in many regions. ● 85% support stronger government action, specially making fruits and vegetables more affordable and accessible, support nutrition education and public awareness campaigns Then, a comprehensive analysis of regions and individuals was explored with machine learning and cluster analysis identified territorial patterns influencing food behaviors, showing that both individual and regional factors like urbanization, income, and access, can have influence on dietary practices and readiness to change. The five most influential individual variables of driving behavioral change were: Lack of taste in healthy food, Cultural habits, norms and values, habits, difficulties in understanding food labels, Lack of time, Educational level, sustainability awareness and in a small portion age and gender. When linking individual behavior change to regional clusters, preliminary exploration suggests: Ageing regions (Cluster 1) demonstrate lower change intention. Agricultural and multicultural regions (Cluster 2) exhibit mixed responses, maybe influenced by habits and food preferences. Transitional rural to urban areas (Cluster 3) fall in between, showing variability that may reflect evolving infrastructure and awareness. Urban regions (Cluster 4) show higher proportions of individuals intending to change behavior, possibly due to higher education and label awareness. Figure 28 highlights main drivers towards sustainable food habits change. Although all the components of the COM-B are represented, most of them reflect psychological capacities and physical opportunities acting as barriers or as enablers (also see figure 6 – conceptual framework). The results may suggest that public policy interventions not only should target Groups differently, but also territories. Figure 28 – main drivers of food habits change towards sustainability. Therefore, what are the most effective ways to engage individuals and territories in long-term food sustainability and climate change resilience?