The true cost of Swedish food consumption A true cost accounting assessment of externalities associated with food consumption in Sweden Bachelor’s thesis in Global Systems Engineering Alicia Andersson Eddi Kusuran Klara Ihse Åkerström Alva Lanhed Sivertsson Lovisa Johansson Simon Leijon Department of Space, Earth and Environment CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2025 www.chalmers.se www.chalmers.se Bachelor’s thesis 2025 The true cost of Swedish food consumption A true cost accounting assessment of externalities associated with food consumption in Sweden Alicia Andersson Klara Ihse Åkerström Lovisa Johansson Eddi Kusuran Alva Lanhed Sivertsson Simon Leijon Department of Space, Earth and Environment SEEX16-25-04 Chalmers University of Technology Gothenburg, Sweden 2025 The true cost of Swedish food consumption An estimation of externalities associated with food consumption in Sweden ALICIA ANDERSSON EDDI KUSURAN KLARA IHSE ÅKERSTRÖM ALVA LANHED SIVERTSSON LOVISA JOHANSSON SIMON LEIJON © ALICIA ANDERSSON, KLARA IHSE ÅKERSTRÖM, LOVISA JOHANSSON, EDDI KUSURAN, ALVA LANHED SIVERTSSON, SIMON LEIJON 2025 Supervisor: Jörgen Larsson, Department of Space, Earth and Environment Examiner: Martin Persson, Department of Space, Earth and Environment Bachelor’s Thesis 2025 Department of Space, Earth and Environment SEEX16-25-04 Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Cover: AI-generated illustration depicting a dinner plate with Earth placed at its center Typeset in LATEX, template by Kyriaki Antoniadou-Plytaria Gothenburg, Sweden 2025 iv Abstract This thesis focuses on estimating the true cost of food consumption in Sweden by assessing the environmental and health related externalities associated with various food categories. The market fails to account for these costs, leading to unsustain- able consumption patterns and significant societal burdens. This study employs a True Cost Accounting approach to quantify and monetise various externalities, including global warming, eutrophication, degradation of ecosystems, biodiversity loss, ecotoxicity, human toxicity, and disease burdens. By utilising data from the Sustainability Assessment of Foods and Diets database and the Global Burden of Disease study, the research provides a comprehensive overview of the environmental and health impacts of food consumption in Sweden. The findings reveal a total external societal cost of SEK 301 billion, with 81% stem- ming from health-related externalities and the remaining 19% from environmental. When these costs are taken into account, the true cost of Swedish food consumption is nearly double the current national food expenditure. Externalities following di- etary risks and carbon emissions from food production are the largest contributors. Meat-based food categories contribute significantly to both environmental degrad- ation and health-related costs, while other categories such as wholegrain pasta and vegetables would generate a societal net benefit per kilogram due to their positive health impacts. The thesis also discusses the potential of market based interventions, such as taxes and subsidies, to internalise these costs and promote a more sustainable and healthy food consumption in Sweden. By incorporating these externalities into food pricing, this study aims to provide insights that guide policymakers, businesses, and consumers to make more informed decisions regarding the food we consume. v Sammandrag Denna uppsats fokuserar på att uppskatta den verkliga kostnaden för matkonsum- tion i Sverige genom att bedöma de miljö- och hälsorelaterade externaliteterna kopplade till olika livsmedelskategorier. Den nuvarande livsmedelsmarknaden tar inte hänsyn till dessa dolda kostnader, vilket leder till ohållbara konsumtionsmönster och betydande samhälleliga bördor. Studien använder True Cost Accounting för att kvantifiera och monetarisera olika externaliteter såsom global uppvärmning, över- gödning, ekosystemförstöring, förlust av biologisk mångfald, ekotoxicitet, mänsklig toxicitet och sjukdomsbördor. Genom att använda data från databaserna Sustainab- ility Assessment of Foods and Diets och Global Burden of Disease ger forskningen en omfattande översikt av de miljö- och hälsoe!ekter som matkonsumtion har i Sverige. Resultaten visar på en total extern samhällskostnad på 301 miljoner SEK, där 81% härstammar från hälsorelaterade externaliteter och 19% från miljörelaterade. När dessa kostnader beaktas är den verkliga kostnaden för svensk livsmedelskonsumtion nästan dubbelt så hög som den nuvarande nationella livsmedelsutgiften. External- iteter kopplade till kostrelaterade hälsorisker och koldioxidutsläpp från livsmedelspro- duktion är de största bidragande faktorerna. Livsmedelskategorier baserade på kött står för en stor del av både miljöförstöring och hälsorelaterade kostnader, medan andra kategorier såsom fullkornspasta och grönsaker kan ge en samhällsekonomisk nettofördel per kilogram tack vare sina positiva hälsoe!ekter. Uppsatsen diskuterar även möjligheterna att använda marknadsbaserade styrmedel, såsom skatter och subventioner, för att internalisera dessa kostnader och främja en mer hållbar och hälsosam livsmedelskonsumtion i Sverige. Genom att inkludera dessa externaliteter i livsmedelspriserna syftar studien till att ge insikter som kan vägleda beslutsfattare, företag och konsumenter mot mer informerade beslut om den mat vi konsumerar. Keywords: True Cost Accounting, Climate, Food consumption, Externalities, Health, Environment, Sustainability. Acknowledgements This thesis project began in January 2025 and was completed in May of the same year. It has been a rewarding and insightful experience, particularly given that the applied assessment method had not previously been used to evaluate the overall ex- ternal costs of food consumption in a Swedish context. We would like to express our sincere gratitude to all the experts and researchers who supported us throughout this journey. A special thank you to our supervisors Martin Persson and Jörgen Larsson at the institution of Physical Research Theory at Chalmers University of Technology, for their invaluable support over the course of this project. We are truly grateful for their guidance, critical insights and encouragement, all of which were instrumental in making this report possible. Gothenburg, May 2025 vii Glossary of Abbreviations Below is the list of abbreviations that have been used throughout this thesis listed in alphabetical order: Abbreviation Explanation a.i. Active Ingredient AMR Antimicrobial Resistance CO2e Carbon dioxide equivalents DALY Disability-Adjusted Life Year EEA European Environment Agency GBD Global Burden of Disease GDP Gross Domestic Product GHG Greenhouse Gases HDI Human Development Index IAM Integrated Assessment Model LCA Life-Cycle Assessment MSA Mean Species Abundance QALY Quality-Adjusted Life Year SAFAD Sustainability Assessment of Foods and Diets SCB Statistics Sweden (Statistiska Centralbyrån) SCC Social Cost of Carbon TCA True Cost Accounting TMREL Theoretical Minimum Risk Exposure Level WTP Willingness To Pay ix Contents Glossary of Abbreviations ix List of Figures xiii List of Tables xv 1 Introduction 1 1.1 Consequences of Swedish food consumption . . . . . . . . . . . . . . 2 1.2 The economic burdens of environmental impact and public health e!ects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Previous research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.4 Aim and research questions . . . . . . . . . . . . . . . . . . . . . . . 5 1.5 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2 Theory 7 2.1 Externalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 True Cost Accounting . . . . . . . . . . . . . . . . . . . . . . . . . . 7 3 Methods and materials 11 3.1 A!ected areas & assessment of externalities . . . . . . . . . . . . . . 11 3.2 Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.2.1 Sustainability Assessment of Foods and Diets . . . . . . . . . 13 3.2.2 Global Burden of Disease . . . . . . . . . . . . . . . . . . . . . 14 3.3 Assumptions and methodology limitations . . . . . . . . . . . . . . . 16 3.3.1 Data consistency . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.3.2 Extrapolation and generalisation . . . . . . . . . . . . . . . . 16 3.3.3 Avoidance of double counting . . . . . . . . . . . . . . . . . . 16 3.3.4 Linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.3.5 Temporal variations . . . . . . . . . . . . . . . . . . . . . . . . 16 3.3.6 Categorisation of food groups . . . . . . . . . . . . . . . . . . 17 3.4 Environmental impacts . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.4.1 Carbon footprint . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.4.2 New N input and new P input . . . . . . . . . . . . . . . . . . 20 3.4.3 Ammonia emissions . . . . . . . . . . . . . . . . . . . . . . . . 21 3.4.4 Blue water use . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.4.5 Pesticide use: Ecotoxicity . . . . . . . . . . . . . . . . . . . . 23 3.4.6 Cropland use . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 xi Contents 3.5 Animal welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.6 Health impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.6.1 Dietary risks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.6.2 Pesticide use: Human toxicity . . . . . . . . . . . . . . . . . . 31 3.6.3 Air pollution . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.6.4 Antibiotic use . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.6.5 Heavy metal exposure . . . . . . . . . . . . . . . . . . . . . . 35 3.7 Monetisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.8 National quantification . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3.9 Food group quantification . . . . . . . . . . . . . . . . . . . . . . . . 38 3.10 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 4 Results 41 4.1 Total true cost of Swedish food consumption . . . . . . . . . . . . . . 41 4.1.1 Environmental costs . . . . . . . . . . . . . . . . . . . . . . . 43 4.1.2 Health costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.2 True cost per food group . . . . . . . . . . . . . . . . . . . . . . . . . 45 4.3 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 5 Discussion 49 5.1 Interpretation and comparison of results . . . . . . . . . . . . . . . . 49 5.2 Uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 5.3 Implications for policy interventions . . . . . . . . . . . . . . . . . . . 54 6 Conclusion 59 A Appendix I xii List of Figures 1.1 Volume of food demand per capita in Sweden from statistics and expectancy from 2018–2030, figure reproduced from Statista in 2025 [8] 1 1.2 Milligrams of total antibiotic use per kilogram of livestock in 2020. Adjusted for di!erences in livestock numbers and species by stand- ardising to a population-corrected unit (PCU) [19] . . . . . . . . . . . 3 3.1 Social cost of CO2, CH4 and N2O (from left to right) calculated from Climate Impact Model developed by the Climate Impact Lab (CIL, Carleton et al., Rode et al.)), Greenhouse Gas Impact Value Estim- ator (GIVE) model [50] and Meta analysis global damage function estimation [51] over time. . . . . . . . . . . . . . . . . . . . . . . . . . 19 3.2 Worst seasonal water scarcity conditions (July-September) for European countries in 2022, measured for the WEI+ [58] with data from Joint Research Center (JRC), European Environment Agency (EEA) and Eurostat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 3.3 PAALA index for 1 kilogram of livestock products produced in Sweden 26 4.1 True cost of Swedish food consumption, including national expendit- ure along with external costs from environmental and health impacts 41 4.2 Share of di!erent environmental externalities (by indicator) contrib- uting to total environmental impact . . . . . . . . . . . . . . . . . . . 43 4.3 Share of di!erent health externalities (by indicator) contributing to total health impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.4 Market price, environmental and health-related external costs per kilogram for selected food categories . . . . . . . . . . . . . . . . . . 45 4.5 Sensitivity analysis of the total societal cost of food under alternative health and environmental valuation scenarios . . . . . . . . . . . . . . 47 4.6 Sensitivity analysis of external costs by food category. Bars represent baseline estimates of external costs, while whiskers illustrate the range resulting from applying low and high DALY valuations . . . . . . . . 48 A.1 Estimated externality costs (by indicator) per kilogram for aggregated food categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II A.2 Estimated annual externality costs (by indicator) for aggregated food categories. Categories marked with asterisk (*) make up a total cost of SEK 666,203,600,00 due to the dietary risk of underconsumption of wholegrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III xiii List of Figures xiv List of Tables 3.1 Considered environmental, social and health externalities, categorised by indicator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.2 Sustainability footprints and corresponding units, based on SAFAD [37] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 3.3 Dietary risk factors, associated disease outcomes, and the optimal intake levels of foods associated with risk (TMREL), based on GBD 2021 [42] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.4 Methods for economic valuation of environmental externalities by in- dicators for Swedish food consumption . . . . . . . . . . . . . . . . . 17 3.5 Carbon pricing in 2024 price level . . . . . . . . . . . . . . . . . . . . 20 3.6 Estimated external costs of pesticides in SEK (2024 price level), de- rived from Leach & Mumford [45] . . . . . . . . . . . . . . . . . . . . 24 3.7 Methods for economic valuation of health-related externalities by in- dicators for Swedish food consumption . . . . . . . . . . . . . . . . . 28 3.8 Mapping of GBD dietary risk factors to food categories . . . . . . . . 29 3.9 Subpopulations a!ected by specific chemical exposures in the study by Thomsen et al. [73], including population size and assumed aver- age body weight used in the calculations . . . . . . . . . . . . . . . . 36 3.10 Monetisation of environmental and health externalities by their in- dicator, at 2024 price level . . . . . . . . . . . . . . . . . . . . . . . . 37 3.11 Valuation of a DALY from various sources, converted to SEK (2024 price level) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.12 Estimated societal costs of carbon emissions for yearly Swedish food consumption at varying discount rates and the respective share of Total National Expenditure (TNE) . . . . . . . . . . . . . . . . . . . 39 3.13 Valuation of nitrogen emissions from various sources, converted to SEK (2024 price level) . . . . . . . . . . . . . . . . . . . . . . . . . . 39 4.1 National market expenditure and estimated externality costs by en- vironmental and health-related indicators . . . . . . . . . . . . . . . . 42 4.2 Net true costs for selected food items (rounded) . . . . . . . . . . . . 45 4.3 Estimated annual national costs of food categories . . . . . . . . . . . 46 xv List of Tables xvi 1 Introduction In recent decades, shifts in global food production and consumption have fundament- ally transformed the way we eat, with significant implications for both sustainability and public health [1]. As the global food system is increasingly recognised as a ma- jor contributor of environmental degradation and disease burden, there is a growing need to reassess how we value food. Rising populations and incomes have led to an overall increase in food demand, particularly for energy-dense products such as high-fat and animal-based foods. The production and consumption of these foods often come at the expense of nutritional quality and environmental resilience, pla- cing significant strain on ecosystems and contributing to greenhouse gas emissions [2], biodiversity loss [3], land degradation [4], excessive pesticide use [5], and high water consumption [6]. In addition to environmental impacts, the modern food sys- tem undermines e!orts to promote healthy, sustainable diets, exacerbating global health challenges [7]. Sweden is no exception to these trends. Figure 1.1 illustrates the volume of food demand in Sweden between 2018 and projections for 2030 [8]. This growing de- mand mirrors global patterns, with a shift toward more resource-intensive and less nutritionally beneficial foods. Figure 1.1: Volume of food demand per capita in Sweden from statistics and expectancy from 2018–2030, figure reproduced from Statista in 2025 [8] 1 1. Introduction In Sweden, the gap between the price consumers pay for food and the broader so- cietal impacts remains substantial, largely due to unaccounted externalities. An externality is a cost or benefit of an activity that a!ects an uninvolved third party, without being reflected in the market price [9]. This report applies a True Cost Accounting (TCA) approach to explore how environmental and health-related ex- ternalities from Swedish food consumption can be quantified and internalised. 1.1 Consequences of Swedish food consumption In 2022, Swedish agriculture was responsible for emitting 6.4 million tonnes of car- bon dioxide equivalents (CO2e), accounting for approximately 14% of the country’s total territorial emissions [2]. However, when emissions are instead calculated based on food consumption, the figure equates to 15 million tonnes of CO2e [10]. The climate e!ects of Swedish food consumption occur both within Swedish territory and in other countries through food imports and it is estimated that about 60% of all greenhouse gas emissions from Swedish food consumption occur abroad [11]. Furthermore, Swedish agriculture accounts for approximately 90% of the national ammonia emissions. Along with nitrogen leakage and other pollutants, these emis- sions contribute to deterioration of air quality, acidification, and eutrophication of water and soil [10]. Similarly, food consumption places an increasing strain on public health. In 2019, dietary habits was identified as the third largest contributor to disease burden in the country [7]. This raises urgent questions about how to balance food security, environmental sustainability, and population health in the years ahead. The poor nutritional content and high energy density of the food we consume pose an in- creased risk of malnutrition and diseases such as obesity, cardiovascular diseases, metabolic disorders as well as negatively impacting mental health and perceived well-being [12]. The prevalence of obesity or overweight has risen to more than 50% of the population, from Body Mass Index (BMI) measurements [13], following a global trend, while over 2 million people are su!ering from cardiovascular diseases [14]. The Swedish Public Health Agency (Folkhälsomyndigheten) emphasises that high intake of sugar, salt, and fats, along with a low intake of fibres and nutritional legumes, are a few of the dietary factors contributing to the large disease burden in Sweden. An example of this is the intake of red and processed meat, which typically contains high levels of salt and fat and can increase the risk of the diseases previously men- tioned. The Swedish Food Agency recommends limiting prepared red meat intake to less than 350 grams of prepared product per week [15]. Precise data on actual intake is unavailable because total consumption is based on import and sales fig- ures, and the reported weight refers to raw meat, which is heavier and often includes bones. However, the Swedish Board of Agriculture could estimate an average pre- pared red meat consumption of 511 grams per week [16] based on a national dietary survey conducted in 2010–2011 [17]. It is important to note that this average masks substantial variation within the population. While some individuals consume well 2 1. Introduction above the recommended limit, others eat little or no meat. As a result, a certain proportion of the population faces elevated disease risks, even though the national average meat consumption remains below the EU average. The usage of pesticides, herbicides, and other chemicals and toxins used to enhance crop yields adversely a!ects both human and animal health [18]. In addition, the excessive use of pharmaceutical products, such as antibiotics in animal production to prevent diseases and loss of livestock, can cause antimicrobial resistance [19]. This constitutes a global health risk that could heavily impact medical treatments and endanger both human and animal life. Although antibiotic usage in Swedish animal production remains relatively low, a portion of imported meat comes from countries such as Denmark, Ireland, the Netherlands, Germany, Poland, and Spain, where antibiotic usage is significantly higher, as illustrated in Figure 1.2. Figure 1.2: Milligrams of total antibiotic use per kilogram of livestock in 2020. Adjusted for di!erences in livestock numbers and species by standardising to a population-corrected unit (PCU) [19] 1.2 The economic burdens of environmental im- pact and public health e!ects The current agri-food systems have negative impacts both locally and globally, af- fecting areas such as environmental, health and social well-being [20]. These impacts a!ect both current and future generations. For example, the current externalised environmental costs are estimated at USD 7 trillion and the costs to human life at USD 11 trillion [21]. In comparison, the total global food consumption is valued at USD 9 trillion. The negative consequences of Swedish food consumption include considerable health- related costs, as well as expenses linked to work absence, such as sickness benefits 3 1. Introduction and reduced productivity. The societal cost of the 1.35 million Swedes that su!ered from obesity in 2022 amounted to SEK 125 billion [2]. Meanwhile, the total cost of cardiovascular diseases was estimated at SEK 60 billion in 2019 [14]. Environmental externalities are challenging to quantify because the estimation of mitigation and adaptation costs does not fully capture the intrinsic value of the loss of environ- mental quality. As a result, di!erent types of cost analyses are required to estimate these damages accurately. For example, to illustrate the scale of the costs, climate- related adaptation of infrastructure in Sweden was estimated in 2021 price levels to range between SEK 148 and 221 billion. [22]. A fundamental reason for the current unsustainable and unhealthy food system is that market prices fail to reflect the external costs. These externalities include the burden on healthcare, productivity losses, and environmental degradation. Since they are not captured in food prices, these costs are often overlooked by both con- sumers and policymakers. Utilising a TCA approach enables the identification, quantification, and monetisation of these externalities. This provides greater trans- parency regarding the real cost of food and supports more informed decisions. In recent years, there has been an increased focus on health related taxes on food products. Multiple health organisations, such as the World Health Organisation (WHO) and the American Heart Association, recommend a tax on sugar-sweetened beverages as a part of government policy to decrease the risk of diseases such as diabetes and cardiovascular diseases [23]. In 2022, the WHO released a manual for such a tax as a guide for policy makers to promote healthy diets [24]. As of 2022, taxes on at least one type of sugar-sweetened beverage had been applied in at least 108 countries worldwide, Sweden not included, although the coverage of these taxes varies [25]. A notable example is in the UK, where a sugar tax has contributed to the reduction of sugar levels in soft drinks. This was done by introducing a producer fee that in- creases with the sugar content [26]. Another case where market based interventions have been implemented was in Denmark, where a tax was introduced on saturated fats above a certain threshold [27]. 1.3 Previous research Despite TCA being a fairly new approach, several studies have explored its applic- ation in the food system, highlighting the societal costs of food production and consumption. Two such reports by Perotti [28] in Switzerland and The Rockefeller Foundation in the US [29] conducted TCA assessments, quantifying the external- ities and their impacts. Their findings demonstrated significant di!erences in cost between food categories. Various policy instruments, such as environmental taxes and fiscal proposals, have been explored to encourage healthier and more sustainable food choices in Sweden. 4 1. Introduction One such proposal is the report Matskatteväxling by Larsson et al. [30], exploring how altered pricing, based on environmental taxation and subsidies, could incentiv- ise such shifts in food consumption. 1.4 Aim and research questions The aim of this thesis is to estimate the true cost of Swedish food consumption by monetising key environmental, health-related externalities as well as animal welfare impact. The report applies a TCA approach inspired by previous assessments from other countries, using Swedish data on agriculture, food imports, and consumption patterns. This thesis will also quantify and compare the external costs of the main food categories, such as meat, dairy, vegetables, and fruit, and examine how these costs di!er. The findings of the thesis are intended to serve as a tool for guiding market interventions, such as taxes or subsidies, that support a shift toward a more sustainable and health-oriented food consumption in Sweden. Since estimations of external costs often involve significant underlying uncertainties and methodological di!erences, valuations may vary significantly between sources and are often presented as a range (e.g., low, base, and high estimates). Therefore, a sensitivity analysis will be conducted to assess how the choice of valuation influences the results. The study addresses the following research questions: • What is the true cost of Swedish food consumption, considering environmental and health-related externalities, as well as animal welfare? • How do the external costs of di!erent food categories compare? • How does the choice of valuation method influence the estimated cost of Swedish food consumption? 1.5 Scope This study assesses the true cost of food consumption in Sweden, primarily focusing on environmental and health-related externalities. Although numerous externalities could theoretically be included, the analysis prioritises those that are measurable, quantifiable, and supported by available data. Other potential externalities, such as those related to socioeconomic disparities and labour conditions in food production, are excluded due to data limitations and methodological constraints. Among the social externalities, only animal welfare is included, since reliable data is available. The selection of externalities is further constrained to those that are both relevant to the Swedish context and supported by su"cient data. In cases where Sweden- specific data is unavailable, international sources are used. 5 1. Introduction The analysis is conducted for both the total annual external costs of Swedish food consumption and the cost per kilogram for aggregated food categories. Within each category, the externality values reflect weighted averages across multiple products. As a result, di!erences between individual products, brands, and production meth- ods, such as conventional and organic farming, are not explicitly captured in the analysis. 6 2 Theory This chapter outlines the theoretical foundations and key concepts relevant to as- sessing the true cost of food consumption in Sweden, and aims to lay a foundation for understanding the need for market-based interventions for more sustainable con- sumption patterns. 2.1 Externalities Externalities represent a form of market failure as they lead to an ine"cient al- location of resources [31]. Since these external e!ects are not reflected in the market price, overproduction and overconsumption of some products are possible consequences. When this leads to pressure on natural resources, it is a negative ex- ternality, which means that the production of goods leads to costs for society. There are also examples of positive externalities that benefit society, such as the consump- tion of certain products contributing to better public health. These externalities can lead to underproduction and underconsumption of a product, as their broader societal values are not reflected in the market price [32]. The United Nations food summit 2021, with the aim to develop strategies to achieve the UN’s 17 sustainable development goals, addressed the issue of externalised costs in the food system [21]. It emphasises how unsustainable, una!ordable and un- healthy food production and consumption arises because of externalities making these types of food more profitable to produce. As a result, healthier and more sus- tainable food options are often more expensive for consumers and less profitable for businesses. This market failure arises from an imbalance, as the actors who profit from the food system are not the ones bearing the full costs it imposes on society and the environment. To address these externalities and to transition the food sys- tem toward greater sustainability and health, many economists agree on the need to internalise the externalities of food products in their prices through true pricing [21], [31]. 2.2 True Cost Accounting Although the need to internalise externalities has been acknowledged for more than a century, there have been di"culties in quantifying and valuing them [31]. The internalisation of externalities refers to the process of including the total costs and benefits in the pricing of goods and services. This can involve measures to reduce or 7 2. Theory prevent negative externalities or create positive externalities, and can be done either through business models or regulatory policies. TCA is a relatively new alternat- ive that provides frameworks for identifying, quantifying and valuing externalities in order to internalise them [21]. The method utilises a holistic approach used to assess the societal costs related to production and consumption [33]. In contrast to traditional economic accounting, TCA frameworks provide a more comprehensive understanding of the actual impact. Quantifying externalities can be done with a variety of methods, one common tech- nique being Life Cycle Assessment (LCA) [33]. LCA is a standardised methodology used to evaluate the environmental and social impacts of a product throughout its entire value chain [34]. This framework enhances transparency by quantifying emissions and resource use, identifying critical areas where interventions can reduce negative e!ects on ecosystems and human health. To ensure consistency and credib- ility, these analyses follow international guidelines such as the ISO 14040 and 14044 standards, which are described in the International Reference Life Cycle Data Sys- tem developed by the European Commission. While the results of an LCA is typically expressed in natural units, i.e. the units the externalities were originally measured in, many quantitative TCA frameworks translate the environmental and social impact into monetary units [33]. Monetary valuation can be done with a variety of means, depending on the nature of the ex- ternality and data availability [32]. These valuation techniques include using market prices and other empirical methods such as: • Production factor: Measuring how an externality a!ects production, e.g. crop yield or forest growth. • Contingent valuation: Measuring the social costs of an externality through surveys. • Averting expenditures: Estimating environmental externalities based on the costs of avoiding negative environmental impacts, e.g. the cost of installing water filters to reduce the amount of nitrates in the drinking water. A key concept in environmental economics is an individual’s willingness to pay (WTP) for a certain service or good [32]. It is often assessed through the contingent valuation method, which can be used to estimate the value that individuals place on non-market goods and services such as cultural, environmental, and health issues [35]. Contingent valuation lets economists simulate a market by giving respondents a hypothetical scenario and asking them how much they are willing to pay for a specified level of a good, or a change in its quality [36]. The method provides a rel- atively inexpensive way of collecting data directly from consumers instead of relying on secondary data. However, this comes with a few uncertainties. Since they are not faced with a real financial commitment, respondents might over- or underestimate their actual willingness to pay in a certain scenario. Lack of information, as well as other kinds of bias, can also a!ect the reliability of the results. 8 2. Theory TCA frameworks are particularly useful in the food and agriculture sector in order to measure and value both the positive and negative environmental, health and so- cial costs [33]. The increased use of TCA has led to the development of numerous methodologies. Although this has contributed to greater transparency in assessing the true costs of food systems, the diversity of approaches has also made it more challenging to directly compare results across the di!erent frameworks. Within the frameworks, a variation of system boundaries, units, and monetisation methods cre- ates inconsistency, consequently limiting qualitative comparisons of the outcomes. 9 2. Theory 10 3 Methods and materials This chapter describes the methodology used to quantify the externalities associated with food consumption in Sweden. Both environmental and health related external costs have been considered and translated into monetary values. The methods cover national level estimates as well as pricing for di!erent food categories where data availability allowed for further breakdown. 3.1 A!ected areas & assessment of externalities A range of potential externalities were initially identified to assess the impacts of Swedish food consumption. This early selection was guided both by the relevance of the indicators and the expected availability and quality of the data. The aim was to establish a comprehensive overview of possible areas of impact before narrowing the scope for the final quantitative analysis. For clarity and comparability, the ex- ternalities were grouped into three major impact categories: environmental, health and social. This categorisation allows for a clearer understanding of how di!erent types of costs are distributed and enables comparison between them. Among these, animal welfare is the only social externality considered. It is therefore treated as a distinct category due to its ethical dimension, separate from health and environ- mental valuations. 11 3. Methods and materials Table 3.1: Considered environmental, social and health externalities, categorised by indicator Indicator Externality Description Carbon footprint Climate change Emissions of greenhouse gases contrib- uting to global warming New N input Eutrophication Nutrient overload in water bodies lead- ing to algal blooms, oxygen depletion, and loss of aquatic biodiversity New P input Eutrophication Nutrient overload in water bodies lead- ing to algal blooms, oxygen depletion, and loss of aquatic biodiversity Blue water use Water scarcity Excessive freshwater withdrawal from groundwater and surface water bodies like rivers and lakes Pesticide use Ecosystem toxicity Chemical exposure harming biod- iversity and ecosystems Ammonia emissions Eutrophication Airborne nitrogen contributing to nu- trient overload, leading to algal blooms, oxygen depletion, and loss of aquatic biodiversity Cropland use Biodiversity loss and habitat conversion Occupation of land for agriculture pre- venting other ecological functions, dis- placing natural habitats and ecosys- tems Animal welfare index Animal su!ering Impact on animal health and well-being Dietary risks Health impact of diet related diseases Disease burden from under- or overcon- sumption of foods and nutrients Pesticide use Human toxicity Health e!ects from occupational expos- ure, local environmental exposure, and dietary residues Air pollution Human toxicity Disease burden due to decreased air quality Antibiotic use Health impact due to antimicrobial resist- ance Health burden due to reduced antibiotic e"cacy and increased treatment fail- ures Heavy metal exposure Human toxicity Toxicity from foodborne heavy metals Table 3.1 presents the range of externalities initially considered, with environmental, social, and health-related impacts highlighted in green, blue, and pink, respectively. Not all of these were included in the final calculations, as some lacked su"cient data or valuation methodologies. 3.2 Data collection To assess the externalities related to food consumption, the study relies on data from multiple sources covering both environmental and health impacts. Two main databases are used: the Sustainability Assessment of Foods and Diets (SAFAD), which provides data on sustainability footprints, and the Global Burden of Disease database (GBD), which o!ers estimates of health impacts linked to various risk factors, including dietary risks. 12 3. Methods and materials 3.2.1 Sustainability Assessment of Foods and Diets One of the main sources for data collection on externalities is the SAFAD database, an open access tool for assessing the impact of foods and diets [37]. The tool includes eight environmental indicators along with indices for animal welfare and antibiotic use. The footprint of each product is representative of the Swedish market, as calcu- lations are based on import shares for the raw commodities and the amount required of each related ingredient [38]. The impact of each product is calculated through- out the whole value chain and includes emissions from raw primary commodities, transportation of the commodities, processing, packaging, and waste in production, retail, and consumption. When available, o"cial statistics from e.g. the Food and Agriculture Organization Statistics (FAOSTAT) and the European Statistical O"ce (Eurostat), as well as data from trade organisations and scientific literature such as published LCAs, were used in the calculations. In cases where there were gaps in the available data, extrapolation and approximations were applied. The resulting impacts are presented per kg of product for each food item. To connect external costs to specific foods, this study uses food categories rather than individual food items. This approach simplifies the analysis while still reflect- ing consumption patterns and impact. The categorisation is based on 38 food groups developed by Larsson et al. [30], which is based on data from the SAFAD database. Each category represents a weighted average of the 20 most sold food products by weight within that group, ensuring that the values reflect real consumption beha- viour. These categories collectively cover the entire food consumption in Sweden, making them suitable for national-level estimations. The food categories developed by Larsson et al. [30] are used throughout the study as a common structure to organise and present the externalities. The SAFAD data- set contains relevant data linked to these categories, particularly for aspects related to production. Additional sources were used to complement this information, espe- cially for health-related impacts where SAFAD did not provide su"cient coverage. Wherever possible, externalities were matched to the same category structure to en- sure coherence across the analysis. The footprints from SAFAD used in this report can be seen in Table 3.2. Table 3.2: Sustainability footprints and corresponding units, based on SAFAD [37] Footprint Category Unit Carbon footprint kg CO2e New N input kg N New P input kg P Blue water use m3 Pesticide use g a.i. Ammonia emissions kg NH3 Cropland use m2·year Animal welfare Index Antibiotic use Index 13 https://safad.se/ 3. Methods and materials 3.2.2 Global Burden of Disease The GBD study is the most extensive and detailed scientific initiative developed to quantify health trends and risks [39]. Led by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington, and supported by a global network of over 12,000 researchers, the GBD framework provides comprehensive estimates of disease burden and risk factors across 204 countries. GBD combines diverse data sources to produce standardised estimates of disease burden, allowing for comparisons across time, geography, and health conditions. A key metric developed by the GBD study is the Disability-Adjusted Life Year (DALY), which captures the total burden of disease by combining the years lost due to premature mortality (YLL) and years lived with disability (YLD) [40]. This relationship can be expressed as: DALY = YLL + YLD (3.1) One DALY represents one year of healthy life lost. This metric is particularly useful in identifying diseases that may not be fatal but still impose a significant burden through long-term disability. Quality-Adjusted Life Year (QALY), on the other hand, is more commonly used in health economics and clinical decision making [41]. It reflects both the quality and the quantity of life lived. One QALY equates to one year in perfect health. If a person lives for a year in a health state valued at 0.5 (on a scale from 0 = death to 1 = perfect health), this would be equivalent to 0.5 QALYs. While both DALY and QALY aim to quantify the burden of disease, they di!er in perspective: DALYs focus on lost health and are often used to assess population- level health burdens [40], whereas QALYs emphasise gained health and are primarily applied in evaluating healthcare interventions at the individual level [41]. To attribute disease burden to specific risk factors, such as diet or air pollution, GBD applies a comparative risk assessment framework [42]. This method estim- ates the theoretical reduction in disease burden if the population were exposed to an optimal exposure level, called the Theoretical Minimum Risk Exposure Level (TMREL). Risk relationships are modelled using established relative risks from meta-analyses, applied across population distributions stratified by age, sex, and geography. Non-linear dose–response functions are often used, meaning the health risk does not necessarily increase proportionally with exposure. Importantly for this study, the GBD estimates used are those specifically calculated for Swedish popula- tion characteristics, including national patterns of dietary intake, disease rates, and demographic profiles. 14 3. Methods and materials Table 3.3: Dietary risk factors, associated disease outcomes, and the optimal intake levels of foods associated with risk (TMREL), based on GBD 2021 [42] Dietary Risk Factor Associated Diseases TMREL Diet low in fruits Cardiovascular diseases, type 2 diabetes, chronic kidney disease, tuberculosis, respirat- ory cancers 340–350 g/day Diet low in vegetables Cardiovascular diseases, type 2 diabetes, chronic kidney disease, tuberculosis, esopha- geal cancer 306–372 g/day Diet low in legumes Ischemic heart disease 100–110 g/day Diet low in whole grains Cardiovascular diseases, type 2 diabetes, chronic kidney disease, tuberculosis, colon and rectum cancer 160–210 g/day Diet low in nuts & seeds Ischemic heart disease 19–24 g/day Diet low in milk Prostate cancer, colon and rectum cancer 280-340 g/day (m)a 500-610 g/day (f)a Diet high in red meat Cardiovascular diseases, type 2 diabetes, chronic kidney disease, tuberculosis, colon and rectum cancer 0–200 g/day Diet high in processed meat Cardiovascular diseases, type 2 diabetes, chronic kidney disease, tuberculosis, colon and rectum cancer 0 g/day Diet high in sugar- sweetened beverages Cardiovascular diseases, type 2 diabetes, chronic kidney disease, tuberculosis 0 g/day Diet low in fiber Cardiovascular diseases, type 2 diabetes, colon and rectum cancer 22–25 g/day Diet low in calcium Prostate cancer, colon and rectum cancer 0.72–0.86 g/day (m)a 1.1–1.2 g/day (f)a Diet low in seafood omega-3 Ischemic heart disease 470–660 mg/day Diet low in polyunsatur- ated fatty acids Ischemic heart disease 9–10% of total daily energy Diet high in trans fatty acids Ischemic heart disease 0–1.1% of total daily energy Diet high in sodium Cardiovascular diseases, chronic kidney dis- ease, stomach cancer 1–5 g/day a m = males, f = females Table 3.3 illustrates the links between various dietary risk factors and disease out- comes. However, most dietary risk factors do not directly cause diseases, but instead exert their e!ects through intermediate biological mediators such as elevated blood pressure, high blood glucose, or low nutrient levels [42]. This structure is reflected in the GBD 2021 model, where associations are mapped through these mediators before leading to specific health outcomes. For example, a diet low in milk increases the risk of colon and rectum cancer primarily by contributing to low calcium in- take, while high sodium intake elevates the risk of cardiovascular diseases through its impact on systolic blood pressure. 15 3. Methods and materials 3.3 Assumptions and methodology limitations Throughout this study, several assumptions and limitations were made to enable analysis despite constraints in data availability, accuracy, and scope. These were necessary to ensure feasibility but also to introduce uncertainties that should be considered throughout the report. 3.3.1 Data consistency It is assumed that the data sources used are comprehensive and su"cient repres- entations of real-world conditions. However, potential quality gaps can a!ect the accuracy of the results. Variations and missing data could introduce bias, which leads to sensitive estimations of the findings. Moreover, data from di!erent sources may have been composed using di!erent assumptions and limitations. 3.3.2 Extrapolation and generalisation In cases where specific data was unavailable, values from similar studies or contexts were used as proxies. This allowed for broader extrapolation, but also introduced uncertainty, as some estimates may not fully reflect Swedish conditions. While some data rely on national or regional averages, local variations in environmental and health factors may be overlooked. As a result, applying these values across di!erent contexts may a!ect the accuracy of the findings. 3.3.3 Avoidance of double counting E!orts were made to prevent overlap in impact calculations. However, due to in- terconnected factors and unclear descriptions for some data sources, there remains a risk that certain parts have been counted twice. This could alter the final results, potentially leading to an overestimation of the true impacts. 3.3.4 Linearity Parts of the methodology is based on the assumption of linear relationships between input variables and their associated impacts. This means that changes in one factor are assumed to result in proportionally equal changes in outcomes. While this sim- plification facilitates analysis and comparability, it may not fully capture complex, non-linear dynamics present in real-world systems. 3.3.5 Temporal variations Variations in time could introduce uncertainty of the results. Data used for monet- isation terms and production numbers was collected over varying years. This tem- poral variation can lead do reduced consistency and weaker estimations. Moreover, food production systems and dietary patterns are dynamic and constantly evolving, consequently failing to accurately capture future trends or shifts in consumer be- haviour. Additionally, since some of the externalities accumulate progressively over 16 3. Methods and materials time, it complicates the estimate predictions over time. Annual estimates may fail to fully represent the long-term impacts for some of these externalities. 3.3.6 Categorisation of food groups In this study, food items have been categorised into broader groups such as processed meat and vegetables rather than the individual products. While this has simplified the calculation process, it may also mask significant variations in environmental and health impacts within those groups. This is not only relevant for the individual products itself, but also that di!erent production methods may have been used. The distinction between organic and conventional production becomes less apparent once the products are categorised. 3.4 Environmental impacts The assessment of environmental externalities is primarily based on a meta-analytical approach, drawing on comparisons across existing literature. Using the environ- mental footprint categories from the SAFAD database as a foundation, each cat- egory’s impact was examined and corresponding monetary values were assigned where possible. Certain categories were excluded due to data limitations and high levels of uncertainty. A summary can be seen in Table 3.4. The methodology for each category will be further explained in the following sections. Table 3.4: Methods for economic valuation of environmental externalities by in- dicators for Swedish food consumption Indicator Calculation Carbon footprint Estimated global damage costs using the SCC [43], applied to CO2e emissions New N input WTP-based estimate using regional valuation of ni- trogen reduction to the Baltic Sea [44], adjusted for excess nutrient use New P input Excluded Blue water use Excluded Pesticide use: Ecotoxicity Estimated using the PEA model, GDP-adjusted for Sweden based on data from the UK, the US, and Germany covering cost of remediation of damaged habitats[45]. Ammonia emissions A WTP-based estimate using regional valuation of nitrogen reduction to the Baltic Sea, adapted and scaled for Swedish ammonia emissions [44] Cropland use Compensation costs based on the value of ecosystem services lost due to land use [46] 17 3. Methods and materials 3.4.1 Carbon footprint Carbon footprints are typically measured in carbon dioxide equivalents (CO2e), a standardised metric that expresses the global warming potential (GWP) of various greenhouse gases (GHG) relative to the impact of 1 kilogram of carbon dioxide (CO2) [47]. This allows emissions of gases like methane or nitrous oxide to be compared and aggregated in a common unit. To assign a monetary value to carbon emissions, two primary approaches are com- monly used: the Social Cost of Carbon (SCC) and the Marginal Abatement Cost (MAC). A practical baseline for MAC values can be seen in carbon pricing mechan- isms like the European Union Emissions Trading System (EU ETS) [48]. The EU ETS operates as a cap and trade system where emitters must purchase allowances for their emissions, e!ectively creating a market price for carbon. In this case, the MAC represents the cost of reducing one additional metric ton of CO2 emissions and are therefore reflecting market based mitigation costs [49]. SCC, on the other hand, refers to the estimated economic damage resulting from emitting one additional metric of CO2 into the atmosphere [43]. It’s a policy driven metric capturing a wide range of external costs globally using Integrated Assess- ment Models (IAMs). The SCC is highly sensitive to the discount rate, which reflects how we value future costs and benefits relative to those in the present. The discount rate determines how much weight we place on the future damages caused by GHG emissions today. A higher discount rate means that we place less value on future damages, making today’s emissions seem less costly in economic terms. This results in a lower SCC. The discount rate is highly political because it reflects eth- ical judgments about future interests to present economic priorities. Governments and institutions therefore debate what rate is appropriate. For example, the US government has used discount rates ranging from 1.5% to 2.5% in estimating the SCC. Table 3.1 presents the estimated social costs for three major GHGs: carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O) [43]. These costs are projected from 2020 to 2080, under di!erent discount rate scenarios. The wide variation in cost across gases reflects their di!ering GWPs and atmospheric lifetimes, which translate into distinct levels of environmental harm per unit emitted. In this analysis, the values for the year 2020 are used, based on the availability and reliability of published estimates as well as the intention to align the valuation with current economic conditions. However, future SCC estimates are expected to project higher costs as a results of increasing climate impacts [43]. 18 3. Methods and materials Figure 3.1: Social cost of CO2, CH4 and N2O (from left to right) calculated from Climate Impact Model developed by the Climate Impact Lab (CIL, Carleton et al., Rode et al.)), Greenhouse Gas Impact Value Estimator (GIVE) model [50] and Meta analysis global damage function estimation [51] over time. In Sweden, a fixed carbon price has been set by policymakers as a tool to reduce emissions by influencing consumer and producer behaviour [52]. Through the im- plementation of carbon taxes, the cost of emitting CO2 is internalised into the price of fossil fuels, e!ectively reducing the purchasing power for carbon intensive goods and services. The tax rates vary depending on the type of fuel, reflecting di!erences in carbon content. The tax per ton of CO2 is derived from a conversion formula with standardised emission factors [53], and corresponding fuel tax [52]. For example, for petrol: Carbon tax per kg CO 2 = 3.14 2.36 = 1.33 SEK/kg CO 2 (3.2) MAC pricing is based on WTP, as determined through the auction mechanism of the EU ETS. Accordingly, it is presented in Table 3.5 using an average value, as well as high and low auction price level from 2024. The SCC is shown with corresponding discount rates of 1.5%, 2%, and 2.5%, representing a range of high, medium, and low price respectively. Lastly, the carbon tax price is fixed per kilogram of CO2 and presented in Table 3.5 for 2024 medium price rate. 19 3. Methods and materials Table 3.5: Carbon pricing in 2024 price level Approach Scope Price (SEK/kg) Source Avg/mid High Low Policy Carbon taxes in Sweden 1.3 * - - Swedish Tax Agency, 2024 [52] MAC Auction price EU ETS 0.74 0.86 0.57 EEX, 2025 [54] SCC Estimation of damage costs for CO2 2.2 3.9 1.4 EPA, 2023 [43] * Carbon tax rate for petrol (motor gasoline) with environmental class 1 (miljöklass 1) Although SCC and MAC originate from di!erent frameworks, they can converge under optimal climate policy conditions. In theory, if a government sets a carbon price equal to the SCC, then market actors would reduce emissions up to the point where the MAC of the last unit reduced equals the SCC [43]. In practice, however, most existing carbon prices fall short of the estimated SCC. For example, although Sweden’s carbon tax is relatively high compared to the MAC implied by the EU ETS, it remains significantly below the SCC estimates derived from scientific as- sessments of global climate damages. In this assessment, the SCC is applied as the valuation metric for carbon emis- sions, as it better reflects the full societal cost of climate related damages. Using a lower price, such as the existing carbon tax, would risk underestimating the true external cost of emissions and preserve the gap between market signals and actual environmental harm. 3.4.2 New N input and new P input The new nitrogen (N) and phosphorus (P) input categories from SAFAD quantify the added nutrients from mineral fertilisers and biological fixation [38]. However, these indicators do not account for the impact on terrestrial or aquatic ecosystems, since the e!ects are highly dependent on local conditions, such as soil type, topo- graphy, climate, crop type, and the chemical form of the nutrient [55]. Since these factors are di"cult to assess, a simplified approach was used, as follows: Nutrient input to the production system → Nutrient output via harvested products = Excess nutrients (3.3) It was assumed that all excess nutrients have the potential to contribute to eutroph- ication. To get the share of excess nutrient use compared to total nutrient input, the following formula was used: Share of excess nutrients of total input = Excess nutrients Nutrient input to the production system (3.4) This was calculated for both nitrogen and phosphorus using data on Swedish ag- 20 3. Methods and materials riculture from Statistics Sweden (Statistiska centralbyrån, SCB) [56]. The excess was assumed to be consistent across food categories, resulting in a single average value for each nutrient. The resulting excess nutrients per year for phosphorus was zero. Given this rough estimate, the associated cost of eutrophication is therefore negligible. As a result, a decision was made not to include the cost of phosphorus in the assessment. For nitrogen, Equation 3.4 yielded a value of approximately 29%. This percentage was then applied to the original SAFAD data for new nitrogen input to estimate the amount of excess nitrogen use. The cost of eutrophication was based on the Swedish Environmental Protection Agency’s (Naturvårdsverket) price database which provides an estimated price for reduced nitrogen emissions to the Baltic Sea [44]. The price was based on the total WTP of all nine countries bordering the Baltic Sea to reduce eutrophication and achieve the goals of the Baltic Sea Action Plan, divided by their combined nitrogen reduction commitments. However, since the emissions do not need to be reduced to zero to maintain good environmental status, applying this price directly to the emissions would overestimate the environmental impact. Therefore, an adjustment of the price was necessary to reflect the cost per kilogram of nitrogen emissions. To estimate the cost per kilogram of nitrogen emissions, the WTP was instead di- vided by the total nitrogen emissions, using values obtained from the Baltic Marine Environment Protection Commission (HELCOM) [57]. The inflation-adjusted WTP of SEK 43 billion per year and the total nitrogen input of 977 kilotonnes per year were used, resulting in the following estimate: Price/kg N emissions = WTP Total N emissions ↑ 44.28 SEK/kg N emissions (3.5) The cost of food consumption was then calculated as: Cost = Price/kg N emissions ↓ (share of excess nutrient input) ↓ New N input (3.6) 3.4.3 Ammonia emissions Ammonia emissions included in the SAFAD data originate from fertiliser and ma- nure use [38]. While ammonia can contribute to several environmental issues, such as acidification, the price estimate will consider only its e!ects related to eutroph- ication. 21 3. Methods and materials The Swedish Environmental Protection Agency has estimated the price of reduced ammonia (NH3) emissions by calculating the proportion of national emissions that reach the Baltic Sea [44]. The deposition of nitrogen compounds in lower oxidation states, such as ammonia, ammonium (NH4 +), and organic nitrogen, to the Baltic Sea was divided by the total Swedish ammonia emissions. This resulted in a frac- tion of 0.153. This fraction was then multiplied by their nitrogen reduction price, resulting in the estimated price for reduced emissions of ammonia. However, as the objective in this report is to determine the cost per kilogram of am- monia emitted rather than reduced, the fraction was instead multiplied by the cost of nitrogen emissions, as estimated in Equation 3.5. This resulted in the following calculation: Price/kg NH 3 emissions = 44.28 ↓ 0.153 ↑ 6.77 SEK/kg NH 3 emissions (3.7) 3.4.4 Blue water use Data provided by SAFAD presents the amount of freshwater used during production [38]. To estimate the cost, it is necessary to determine what portion of this water is scarce, as the environmental impacts depend on it. A restoration cost for scarce blue water use from True Price Foundation, which was set at 1.330 EUR/m3, was studied [46]. However, it is di"cult to determine the exact details required, such as the precise location and local conditions at the time of production. Therefore, this category was excluded, as a reliable estimate could not be made. Water scarcity is generally low in Sweden with a Water Exploitation Index plus (WEI+) of 0.4%, which can be seen in Figure 3.2. The WEI+ is the percentage of available renewable freshwater resources that is being consumed during the worst quarterly water scarcity conditions of Q3 from July to September [58]. This suggests that domestic water-related environmental costs are minimal. However, Sweden im- ports a significant share of its food from countries with higher water stress. The largest sources of food imports include the Netherlands with 15% of total imports, Denmark with 14% and Germany with 13%, all of which report higher WEI+ values of 6.0%, 18.6%, and 4.5% respectively, with data from 2024 [59], which can be seen in Figure 3.2. These figures indicate that imported food may be associated with higher levels of water exploitation. As a result, the exclusion of blue water costs may lead to an underestimation of the total environmental burden, particularly for food categories with high water foot- prints sourced from water-stressed regions. If these costs were to be fully accounted for, it is likely that significant di!erences between food categories would emerge in terms of their water-related externalities. 22 3. Methods and materials Figure 3.2: Worst seasonal water scarcity conditions (July-September) for European countries in 2022, measured for the WEI+ [58] with data from Joint Research Center (JRC), European Environment Agency (EEA) and Eurostat 3.4.5 Pesticide use: Ecotoxicity Pesticides such as herbicides, insecticides, nematicides, and fungicides are commonly used in agriculture to improve e"ciency and increase yields by targeting fungi, harm- ful bacteria and insect pests [60]. While e!ective in boosting production, pesticide use can have substantial negative impacts on ecosystems. This report employs the Pesticide Environmental Accounting (PEA) model to estimate the external environ- mental costs associated with pesticide applications [45]. The model integrates the Environmental Impact Quotient (EIQ), which quantifies the relative ecotoxicological risk of various pesticide active ingredients (a.i.). EIQ scores are based on toxicity data and environmental persistence, and they are used to weight pesticide impacts by active substance and application rate. Within the ecological scope of the model, two primary categories of environmental e!ects are assessed: aquatic and terrestrial [45]. Aquatic e!ects include pesticide 23 3. Methods and materials runo! and toxicity to aquatic life such as fish, while terrestrial e!ects include harm to birds, bees, beneficial insects and chemical persistence in soils and plants. The external costs incorporated in the model are derived from studies conducted in the United Kingdom (UK), United States (US) and Germany (DE) [45]. To be able to apply the model to Sweden (SE) which is di!erent in terms of both pesticide usage intensity and Gross Domestic Product (GDP) per capita, it is necessary to adjust the cost estimates accordingly. The adjustment uses a GDP-weighted trans- formation as follows: CSE = ( CUK GDPUK + CUS GDPUS + CDE GDPDE ) ↓ 1 3 ↓ GDPSE (3.8) where C = Mean cost The PEA model is based on data from three high-income countries where the costs of environmental remediation and monitoring are relatively high. Consequently, the model may overestimate the external costs for countries with lower service costs and weaker enforcement systems and potentially underestimate them, since the costs are based on a WTP. Furthermore, the extent of environmental damage from pesticides depends not only on toxicity but also on usage intensity, landscape characteristics and ecosystem sensitivity [45]. Although these country specific variables are not directly included in the model, the PEA allows for indirect adjustment by scaling costs according to GDP per capita, which serve as proxies for remediation capacity and population exposure, respectively. Table 3.6: Estimated external costs of pesticides in SEK (2024 price level), derived from Leach & Mumford [45] Ecological category Mean cost per kg a.i. (SEK) UK US DE Pollution incidents, fish deaths and monitoring costs 9.19 3.9 19.9 Biodiversity/wildlife losses 13.7 5.0 2.4 Bee colony losses 1.08 3.55 0.54 Using this method, the estimated average external environmental cost for pesticide use in Sweden is calculated to be approximately 19.5 SEK/kg a.i. This is in com- parison with the highest country specific cost, from the UK at 23.9 SEK/kg a.i., and the lowest, from the US at 12.4 SEK/kg a.i. [45]. The adjustment ensures the model reflects Sweden’s economic context and agricultural profile more accurately. 3.4.6 Cropland use The cropland use metric from SAFAD quantifies the land area (m2) required annu- ally to produce food, as defined by Röös et al. [38]. Cropland use has a significant 24 3. Methods and materials impact on ecosystem services and biodiversity. The occupation of land for agricul- tural activities limits its availability for other ecological functions, leading to the displacement of natural habitats and ecosystems. This results in losses in both biodiversity and ecosystem services [61]. The True Price Foundation assigns monetary values to ecosystem services lost due to land use for di!erent biome types, based on a study by De Groot et al. [62]. These values represent compensation costs, which are the estimated value of the ecosystem services no longer provided by the land when it is used for agriculture [46]. To apply these values, specific biome cover data is required to apply the appropriate valuation categories. A biome refers to a large ecological unit defined by dominant vegetation in terrestrial regions or biogeochemical properties in marine areas, within which ecosystems function in broadly similar ways [61]. Since approximately 70% of Sweden is covered by forest, and detailed land cover data is limited, forest was as- sumed to be the dominant biome nationwide [63]. Therefore, the Other forest biome category from True Price was used to represent Sweden’s costs due to cropland use. The data provided by SAFAD is expressed in m2·years. In order to apply the True Price monetisation value, this was converted to MSA·ha·years. Mean Spe- cies Abundance (MSA) is the proportion of species populations remaining under the current land use compared to its original, undisturbed natural state [61]. An MSA-value of 0.6 for Sweden, sourced from the Global biodiversity model for policy support, was used [64]. The result was then divided by 10,000 to convert square metres into hectares. All other biome types presented by the True Price Foundation have higher costs than the other forest category. Inland wetlands, which exist in Sweden, are valued at more than ten times higher [46]. This suggests that the estimation may be underestimated due to the absence of detailed biome cover data for Sweden. Furthermore, the impact of imported food is not accounted for in this estimation, meaning that costs associated with more sensitive ecosystems may have been overlooked. 3.5 Animal welfare The Animal Welfare Index in this study is based on the Perception Adjusted Animal Lives A!ected (PAALA) metric, as described by Röös et al [38]. PAALA quantifies perceived animal su!ering per kilogram of commodity, making it particularly suit- able for dietary assessments. The PAALA index is constructed from three components: • Number of animals a!ected per kilogram of food produced • Perceptual ability of the animal species to experience su!ering 25 3. Methods and materials • Degree of su!ering, including factors such as exposure to disease and stress during the production system, as well as the impact of slaughter The formula for calculating PAALA is as follows: PAALA = number a!ected ↓ ability to perceive su!ering↓ su!ering (including slaughter) ↓100(3.9) Although similar to the more conventional metric Animal Life Years Su!ered (ALYS), PAALA di!ers by expressing perceived su!ering per kilogram of food, thus o!ering a more consumption-oriented perspective. The indices are shown in Figure 3.3. Figure 3.3: PAALA index for 1 kilogram of livestock products produced in Sweden In terms of valuation, the ALYS metric has been compared to the human DALYs framework, which allows for a monetisation of welfare impacts for animals based on the value of life quality. Including the value of animals in cost–benefit analyses is significant, as it acknowledges the intrinsic worth of non-human lives. Based on a WTP estimation for data regarding poultry, the value of animal welfare per a!ected unit has been approximated at USD 0.10 to USD 0.37 per kilogram [65]. However, relying solely on human WTP of other species may severely underestimate the true intrinsic value since WTP reflects human preferences, awareness and biases. This implies that any attempt to construct social welfare functions for valuing animal welfare, necessarily reflects underlying ethical and philosophical assumptions, high- 26 3. Methods and materials lighting the normative choices embedded in welfare aggregation across species [66]. Kuruc and McFadden [67] present an example of a generalised totalist utilitarian welfare function, following traditional economist frameworks. In their analysis, the value of animal welfare impact is calculated to USD 122,789.19, which is about SEK 1,2 million for annual meat consumption of an average non-vegetarian diet for an American. This is not representative for Sweden since dietary patterns di!erentiate between countries, but it can act as a perception of magnitude. The model assigns animals a fixed negative utility value across species, based on the ethical assump- tion that lives in industrial farming conditions are worse than non-existence. This is operationalised in the model by setting animal welfare at a human-equivalent util- ity level of USD 1.00 per day, below the international human poverty line of USD 1.90 per day. These utility losses are then aggregated across the number of animal life-years required to sustain typical meat consumption, yielding a total estimate of the external welfare cost. An alternative monetisation method involves assessing the additional costs required to improve animal welfare, such as increased living space, better nutrition, and im- proved health, thereby reduce su!ering. However, such cost estimates have not yet been quantified in detail and are therefore not monetised in this thesis. A formal model addressing this issue has been developed, providing a theoretical basis for understanding the economic impact of animal welfare improvements at slaughter [68]. While valuing animal welfare strictly through damage cost approaches o!ers practical advantages for integration into economic analyses, it also raises important ethical concerns. Reducing the intrinsic value of animal life to market prices risks marginalising the moral worth of sentient beings. These considerations led to the conclusion that the available monetisation method- ologies are not su"ciently reliable for this analysis. Consequently, this impact area was excluded from the results presented in this thesis. 3.6 Health impacts This section presents the methods used to quantify health-related impacts associ- ated with the Swedish food consumption patterns. It includes five distinct health indicators: dietary risks, pesticide exposure, air pollution, heavy metal exposure, and antibiotic use. Each indicator required a tailored methodological approach, de- pending on the type and availability of the data. An overview of the included indicators and their methodological basis is provided in Table 3.7, with further detail in the subsections that follow. 27 3. Methods and materials Table 3.7: Methods for economic valuation of health-related externalities by indic- ators for Swedish food consumption Indicator Calculation Dietary risks DALY estimates from GBD 2021 [69], based on associations between dietary intake and disease outcomes Pesticide use Average cost per kg a.i. derived from Zandonella et al. [70], applied to pesticide use in food consumed in Sweden Air pollution DALY estimates from GBD 2021 [69] for PM and ambient ozone, and characterisation factors (µDALY/kg emitted) for NH3 from Humbert [71] Antibiotic use DALYs lost due to infections with antibiotic-resistant bac- teria from the European Centre for Disease Prevention and Control [72] Heavy metal exposure DALY estimates per contaminant (lead, inorganic arsenic, cadmium, and methylmercury) from Thomsen et al. [73] and monitoring data from Petersen et. al. [74] 3.6.1 Dietary risks To estimate the disease burden linked to diet, data from GBD 2021 [69] was used. For dietary risks, GBD provides point estimates of DALYs attributable to each di- etary risk in Sweden for the year 2021. These estimates represent the best central value based on 1,000 simulations and reflect Swedish dietary patterns, disease rates, and demographics [42]. To allocate these burdens to specific foods for the calculation of annual and per- kilogram externality costs, GBD risk factors were matched to food categories in the modified SAFAD dataset by Larsson et al. [30]. As shown in Table 3.8, not all GBD dietary risks could be clearly mapped to the food categories. Only dietary risks that could be directly linked to specific food groups were included in the category-level calculation, while others were excluded to avoid double-counting or ambiguity. 28 3. Methods and materials Table 3.8: Mapping of GBD dietary risk factors to food categories GBD Dietary risk factor Mapped to food category Reason if not mapped Diet low in fruits Fruits - Diet low in vegetables Vegetables - Diet low in legumes Legumes - Diet low in whole grains Whole grain categories - Diet low in nuts & seeds Nuts & seeds - Diet low in milk Milk - Diet low in fiber Not mapped Di"cult to isolate source, over- laps with multiple food categor- ies Diet low in calcium Not mapped Already indirectly captured via “low milk”; risk of double- counting Diet low in seafood omega-3 Fish & shellfish - Diet low in polyunsatur- ated fatty acids Not mapped Spread across multiple oils/fats, unclear mapping Diet high in red meat Two categories: Beef & lamb, Pork - Diet high in processed meat Processed meat - Diet high in sugar- sweetened beverages Sugar-sweetened softdrinks - Diet high in sodium Not mapped A!ects many foods; cannot alloc- ate to single category Diet high in trans fatty acids Not mapped Trans fats are ingredient-level; di"cult to assign to specific cat- egory The DALY values obtained from GBD was used to estimate the annual health- related costs for each food category. For red meat such as beef & lamb and pork, consumption data from the modified SAFAD dataset was used to proportionally allocate the burden between the two categories. This was possible because the disease burden results from overconsumption, allowing the cost to be distributed according to the relative intake of each type of meat. In the case of whole grain products, the burden originates from insu"cient intake. Since the health benefit is associated with increased consumption of the entire food group, it is not mean- ingful to separate the e!ect of separate whole grain categories such as bread and pasta. The deficiency reflects an overall lack of whole grain consumption, regardless of which specific product is underconsumed. Therefore, the burden was not divided across subcategories but instead attributed to whole grain products as an aggregate. To estimate DALYs per kilogram of food, dietary intake data from GBD (g/day by age group 25+) [75] was combined with population data from SCB [76] to produce a population-weighted average daily intake. This was then scaled up to the entire Swedish population to calculate annual national intake in kilograms. Although GBD provides data only for individuals aged 25 and above, the burden was distributed across the entire population. This decision was based on evidence that diet-related diseases develop over time [77] and that early-life dietary patterns are closely linked 29 3. Methods and materials to later health outcomes [78]. The TMREL values defined by the GBD were used as benchmarks for optimal intake levels, as presented in Table 3.3. For foods with excessive intake, the lowest value in the TMREL range was used to calculate the deviation. This allocated the DALYs across the largest possible intake surplus, resulting in a conservative per-kilogram cost estimate: DALY/kg (excess) = Total DALYs Actual intake → Lower TMREL bound (3.10) For foods with insu"cient intake, the highest value in the TMREL range was used. In the case of diet low in milk, which had separate TMRELs for males and females, the average of the two upper bounds was applied. This allocated the DALYs across the largest possible deficiency, again resulting in a conservative cost: DALY/kg (deficiency) = Total DALYs Upper TMREL bound → Actual intake (3.11) Each DALY-per-kilogram value was then assigned to the corresponding food group in SAFAD. A positive value indicates a health cost due to overconsumption, while a negative value represents a health benefit from addressing underconsumption. This method assumes a linear relationship between intake deviation and health impact, meaning that each additional kilogram above or below the optimal intake contributes proportionally to the total DALY burden. The resulting DALY/kg value can be interpreted as the marginal health burden of consuming one extra, or one less, kilogram of a given food. This assumption is made to be able to produce a single value, even though GBD applies non-linear dose response functions [42]. It is therefore of importance to note that the method does not account for the diminish- ing or accelerating risk e!ects that may exist at di!erent intake levels. Moreover, while GBD estimates DALYs using the full intake distribution, the per- kilogram analysis here was limited to risks where the national average intake lies outside the TMREL range. Consequently, the risk factor diet low in seafood omega- 3 which could be mapped to fish & shellfish, was excluded from the per-kilogram analysis since the mean Swedish intake already falls within the optimal range. In- cluding it could overstate the societal cost of average consumption, as the remaining burden likely arises from specific subpopulations. Finally, to estimate the total national burden from diet-related diseases, the DALY value reported by GBD for all dietary risks combined was used, rather than limiting 30 3. Methods and materials the sum to only those that could be mapped to SAFAD. This avoids double-counting and ensures inclusion of all the dietary risks in GBD. 3.6.2 Pesticide use: Human toxicity To estimate the health-related costs of pesticide exposure, this study uses a Swiss analysis by Zandonella et al. [70], which provides both consumption data and mon- etised health impact estimates. A comparison with Swedish data from Cederberg et al. [79] indicates that the proportional use of herbicides, fungicides, and insecticides is relatively similar in the two countries. Since these pesticide types di!er in their potential health impacts [60], this similarity supports the use of the Swiss estimates as a reasonable proxy for Swedish conditions. A consumption-based approach was used, meaning the estimates reflect the health impact of pesticides used in the production of food consumed in Sweden, regard- less of where the pesticides were applied geographically. It should be noted that applying the Swiss health cost estimate to pesticide use associated with Swedish food consumption assumes that the health impacts per kilogram of pesticide are similar across countries. This introduces uncertainty. Since the consumption-based approach includes both domestically produced and imported food, the actual health costs might di!er depending on where the pesticide exposure occurs. In some cases, health-related costs in production countries could be lower due to weaker healthcare systems, regulatory environments, or lower valuation of health outcomes, whereas in other cases they could be higher due to less protection for agricultural workers or greater exposure risks. Both Sweden and Switzerland have high levels of human development [80] and per capita health spending [81], supporting the assumption that the health impact per unit of pesticide use could be similar in both contexts. However, Switzerland has a slightly higher Human Development Index (HDI) [80] and marginally higher health expenditures per capita [81], which could lead to an overestimation of the costs. Furthermore, the analysis assumes a linear relationship between pesticide use and associated health costs, although this is likely a simpli- fication that does not fully reflect the complexity of real-world exposure-response dynamics. Zandonella et al. [70] provides several cost estimates derived from di!erent meth- odological approaches. To account for this variation, the average of these estimates was used to calculate a representative cost per kilogram of a.i. Cost/kg a.i. = Total Swiss health cost (CHF) Total Swiss pesticide use (kg a.i.) (3.12) 31 3. Methods and materials This unit cost was then used in a general formula to estimate health related costs: Cost = Cost/kg a.i. ↓ Pesticide use (kg a.i.) (3.13) This calculation was performed both for total pesticide use linked to Swedish food consumption, and by food group. The pesticide consumption data was obtained from the SAFAD dataset. 3.6.3 Air pollution Air pollution is complex and often di"cult to trace or quantify. A major limita- tion in assessing its health impact from agriculture is the lack of detailed data on emissions, dispersion, and exposure pathways. In particular, the formation of par- ticulate matter (PM) further complicates matters, as its dynamics make it di"cult to establish clear links between emissions and health outcomes. Air pollution, and especially PM2.5, is linked to respiratory and cardiovascular diseases, as well as pre- mature death [82]. Agricultural emissions, such as ammonia, contribute significantly to the formation of secondary PM, making them a notable public health concern [83]. The calculations used to estimate air pollution impacts from agriculture is simplified and generalised, focusing on broader assumptions rather than specific geographical data. These models may overlook complexities in atmospheric chemistry, pollutant dispersion, and long-range transport of secondary pollutants as stated above. The following method was used to estimate the total DALYs associated with air pollution from agriculture in Sweden: Firstly, data from GBD regarding the total health burden in Sweden due to air pollution was collected [69]. The data included estimates of DALYs attributed to PM, ambient ozone pollution and nitrogen dioxide pollution. The GBD dataset provided the total DALYs from air pollution in Sweden, allowing for an overview of the health impacts at a national level. The share of each spe- cific pollutant related to agriculture was identified by analysing available sectoral emission data from the EEA [84]. The total DALYs from agriculture-related air pol- lution were estimated by multiplying the proportion of each pollutant attributable to agriculture by the total DALYs associated with that pollutant, as shown in the equation below: Total DALYsagriculture = n∑ i=1 (Sharei ↓ Pollutanti) (3.14) 32 3. Methods and materials However, for ambient ozone and nitrogen dioxide pollution, certain considerations and limitations were applied. For ambient ozone formation, a simplified 1:1 ratio between non-methane volatile organic compounds (NMVOCs) and NOx was applied, deviating from the more typical 70:30 ratio commonly used for urban environments [85]. Since agriculture is predominantly rural, where ozone formation tends to be NOx-limited rather than NMVOC-limited, this simplification was made to better reflect the atmospheric chemistry in those areas. The share of pollutants was then calculated using a newly weighted average, which was later used in the health im- pact calculations. Nitrogen dioxide was not included in the report primarily due to a lack of specific data related to its agricultural emissions. Additionally, its health impacts are more relevant for urban sectors such as tra"c and industry, rather than for agriculture. Furthermore, since nitrogen dioxide is included in NOx and acts as a precursor to ozone rather than a direct pollutant, its inclusion would not significantly a!ect the analysis. Since ammonia emissions were not included as a separate category in the GBD data, the health impacts from ammonia emissions was calculated separately. This was done by taking the total national emissions and multiplying them by a factor representing the health damage per kilogram emitted in rural areas, expressed as µDALY/kg emitted pollutant [71]. This approach allowed for an estimation of the impact of ammonia using emission-specific health impact factors appropriate for rural settings. Due to limited data availability regarding emissions per food item, the calculations for the individual food categories are mostly based on ammonia emissions, and to a lesser extent NMVOCs. Data for emissions of the other pollutants per kilogram of food was not found, which limited the analysis. The annual cost per capita for each category was calculated using the following equation: Cost per capita/year = ∑ ( kg Pollutant kg food ↓ kg food/person/year ↓ DALY kg Pollutant ↓ SEK/DALY ) (3.15) These results were also multiplied by the total population for the, giving the total annual cost from these emissions. Moreover, since Sweden imports approximately 60% more agricultural products and food than it exports [86], it is assumed that the emissions per unit of food are 33 3. Methods and materials similar for imported and domestically produced products. To reflect that Sweden consumes more food than what is produced domestically, the territorial external costs are adjusted using a factor of 1.6. This factor represents the ratio between total food consumption in Sweden, using both domestic production and net imports, but also the domestic production alone. This adjustment for import and exports is only done for the total cost calculation and not for the food categories, since those are consumption-based and already adjusted. 3.6.4 Antibiotic use The health burden of antimicrobial resistance (AMR) due to antibiotic use was assessed using the antibiotic index from the SAFAD database. National data on antibiotic sales for veterinary purposes in di!erent EU member states is collected annually through the ESVAC project [87]. However, data on the actual use of these products, as well as species-specific data, is still lacking, as member states are cur- rently not required to report antibiotic use data by animal species for food-producing animals. Because of this, the calculations of the antibiotic index are based on the assumption that the amount of sold antibiotics mirrors the amount of antibiotics used [88]. The SAFAD antibiotic use index was calculated by comparing the total sales of an- tibiotics in veterinary medicinal products across di!erent EU countries with the ex- pected sales based on the proportion of livestock species and their biomass [88]. For each country, the expected sales were derived by multiplying the proportion of each species biomass in that country by the standardised antibiotic usage values (mg/kg biomass) for each species. These standardised values reflect the average amount of antibiotics used for each species, based on available data from six European coun- tries. The resulting ratio of total sales to expected sales was then multiplied by the standardised value for each species, as seen in Equation 3.16, producing the index. This approach enables for estimating antibiotic use based on national di!erences in species distribution, even when direct usage data is unavailable. In this report, the antibiotic index is used as an estimate of the amount of antibiotics used (mg) to produce one kilogram of each food category. Total sales Expected sales ↓ Standardised usage value = Antibiotic index (3.16) The yearly health cost of infections caused by antibiotic-resistant bacteria, meas- ured in DALYs, for all EU member states was collected from the European Centre for Disease Prevention and Control [72]. An estimate of the health burden directly related to food consumption was then made based on the assumption that 22% of all AMR is attributable to the food system [89]. This number was then divided by the amount of antibiotics sold for veterinary purposes in the EU to give an estimate of the DALYs lost to AMR per milligram of antibiotics, using data on antibiotic 34 3. Methods and materials sales collected from the European Medicines Agency [90]. The calculations are based on the assumption that all use of antibiotics contribute equally to AMR, which is a simplification. In reality, the spread of AMR depends on numerous di!erent factors, such as the type of antibiotics and the way it is used, which exceeds the scope of this report. 3.6.5 Heavy metal exposure To estimate the disease burden from foodborne heavy metals, two di!erent methods were used to calculate the total disease burden and the burden per food category. These methods relied on studies conducted in Denmark. Due to the lack of Swedish data and given the geographical and cultural similarities between the two countries, it was assumed that the Danish data was a reasonable proxy for Swedish conditions. For the total disease burden, a study by Thomsen et al. [73] was used. It estimated DALYs attributable to four chemical contaminants commonly found in food: inor- ganic arsenic (i-As), lead (Pb), methylmercury (MeHg), and cadmium (Cd). The associated health outcomes include intellectual disability, chronic kidney disease, and various forms of cancer, such as lung, bladder, and skin cancer. The burden was calculated using monitoring data and expressed as DALYs per 100,000 individu- als. In the case of Pb, the study provided two estimates based on uncertainty in the con- version factor between dietary intake and blood lead levels [73]. The lower bound resulted in 6.6 DALYs/100,000, while the upper bound was 22 DALYs/100,000. These di!erences arise due to variation in how e"ciently dietary lead is assumed to transfer to blood lead levels. To derive a single estimate, the mean value between the lower and upper bound was used in this analysis. These DALY rates were scaled up to the Swedish population to estimate the total disease burden from heavy metal exposure via food. For disease burden by food category, both the study by Thomsen et al. [73] and a report by Petersen et al. [74] were used. The latter monitored food contaminants between 2004-2011, presenting these in micrograms of contaminant per kilogram of food (µg/kg) for the same four heavy metals. The DALY per microgram of contaminant (DALY/µg) was calculated using the fol- lowing formula: Total DALY/year Average exposure/day ↓ Average weight ↓ Population ↓ 365 = DALY/µg (3.17) Both the total DALYs per year and the average daily exposure for the four chem- 35 3. Methods and materials icals were sourced directly from the report by Thomsen et al. [73]. The average exposure was defined as the amount of contaminant per kilogram bodyweight per day. However, only Cd and i-As exposure was calculated for the whole population, Pb exposure data was limited to 5-year olds and MeHg exposure was calculated only for women in fertile age (15-49 years old) [73]. Accordingly, the Population and Average weight parameters varied for each chemical. For MeHg, Cd and i-As an average weight of 70 kilogram was assumed, while for Pb, a weight of 21.5 kilo- gram was assumed. The subpopulation sizes for the year 2019 were obtained from Statistics Denmark [91]. These values are shown in Table 3.9. Table 3.9: Subpopulations a!ected by specific chemical exposures in the study by Thomsen et al. [73], including population size and assumed average body weight used in the calculations Chemical Subpopulation Size Assumed weight Pb 5-year olds 58,352 21.5 kg MeHg Women 15-49 1,267,296 70 kg Cd Entire population 5,806,081 70 kg i-As Entire population 5,806,081 70 kg These DALY values were then compared to the numbers in the study by Petersen et al. [74]. By multiplying the amount of DALYs associated with each chemical with the amount of chemical found in di!erent food groups, a number of DALYs per kilogram was generated for each available food category. 3.7 Monetisation In order to compare and aggregate the various external costs, it was necessary to assign monetary values to each externalities. To ensure consistency across monetary values, all foreign currencies were converted to Swedish kronor (SEK). Historical av- erage exchange rates from the year in which the original amount was reported were used for the conversion. If the specific year was not stated, the average rate from the year the report was published was applied instead. All exchange rates were obtained from the o"cial database of Sweden’s central bank (Sveriges Riksbank) [92]. To account for inflation, all amounts were subsequently adjusted to 2024 values us- ing the price conversion function of SCB (Prisomräknaren) [93]. This tool adjusts for inflation based on the Consumer Price Index (CPI), allowing for a consistent comparison of monetary values across di!erent years. Table 3.10 summarises the final monetised values applied in this study for each environmental and health-related externality by indicator. All values presented in the table reflect the amount after currency conversion and inflation adjustment. 36 3. Methods and materials Table 3.10: Monetisation of environmental and health externalities by their indic- ator, at 2024 price level Externality indicator Unit SEK/unit Source Carbon footprint kg CO2e 2.16 EPA, 2023 [43] New N input kg N 44.28 Swedish Environmental Protec- tion Agency, 2018 [44] Pesticide use kg a.i. 19.5 Leach and Mumford, 2008 [45] Ammonia emissions kg NH3 6.77 Swedish Environmental Protec- tion Agency, 2018 [44] Cropland use MSA·ha·year 12,761 True Price, 2023 [46] Dietary risks DALYs 1,250,800 True Price, 2023 [46] Pesticide use kg a.i. 225 Zandonella et al., 2014 [70] Air pollution DALYs 1,250,800 True Price, 2023 [46] Antibiotic use DALYs 1,250,800 True Price, 2023 [46] Heavy metal exposure DALYs 1,250,800 True Price, 2023 [46] 3.8 National quantification National quantification of the true cost of food consumption in this thesis is achieved by summing current national food expenditure with the aggregated monetised ex- ternal costs. True Cost = National Food Expenditure + n∑ i=1 Externalityi (3.18) When necessary, the Swedish population from 2021 (10,452,326 individuals) [76] was used to scale externalities to the national level. This aligns with GBD, while the SAFAD dataset reflects conditions for 2024. Although this means that data sources refer to slightly di!erent years, the di!erence in population size between 2021 and 2024 is marginal [76]. Given the approximate nature of the estimates used through- out this report, the impact of this simplification on the final results is negligible. The national food expenditure was estimated by multiplying the average annual food spending per household by the total number of households in Sweden, based on data from SCB [94]. While this provides insight into current household-level spending, these expenditures include VAT. Since VAT is a transfer payment between con- sumers and the public sector, and therefore not a net cost to society, it should not be considered a societal cost in an economic sense. Therefore, the VAT portion of the national expenditure was excluded. In Sweden, the VAT on food is 12% [95], meaning that the reported national expenditure should be adjusted as in the equa- tion below. 37 3. Methods and materials Net expenditure (excl. VAT) = Total expenditure (incl. VAT) 1 + VAT rate (3.19) To determine the average annual per capita food expenditure, the VAT-adjusted household food expenditure was divided by the average number of persons per house- hold, as shown in the equation below: Per Capita Food Expenditure = E P (3.20) where E = average annual food expenditure per household (excl. VAT) P = average number of persons per household 3.9 Food group quantification To calculate the true price for individual food categories, average market prices were combined with externality prices. These average market prices were primarily derived from data provided by Larsson et al. [30]. For some categories, however, no specific prices could be extracted. In those cases, current market prices were obtained from three major Swedish retail chains: ICA [96], Willys [97], and Coop [98]. 3.10 Sensitivity Analysis To determine how variations in the valuations of both environmental and health externalities a!ect the results, a sensitivity analysis was conducted. Externalities are often di"cult to monetise due to their complexity, resulting in di!erent valu- ations. This can significantly impact both the final outcomes and the reliability of the estimated total cost. In this report, the sensitivity analysis focuses on DALYs, since they are used for most health-related externalities, as well as on the environ- mental externalities with the highest total values, as these factors have the strongest influence on the overall result. The di!erent valuations for DALYs are presented in Table 3.11. The valuation used in the results is based on estimates from True Price [46], while the alternative values used in the sensitivity analysis originate from the Swedish Environmental Protec- tion Agency [44]. These were initially expressed as QALYs and later converted to DALYs, assuming that the monetary valuation of a QALY is relatively similar to that of a DALY. This approach was adopted to allow the use of Swedish data for 38 3. Methods and materials monetising DALYs, despite the considerable uncertainty associated with assuming that the va