The State of Food and Agriculture 2023

The environmental, SOCIAL and HEALTH hidden costs of agrifood systems

Methodology

Steven Lord at the University of Oxford Environmental Change Institute developed a model for the Food System Economic Commission (FSEC) to value the hidden costs of agrifood systems across three dimensions: environmental, social and health.¹ The model was paired with FAO’s Corporate Database for Substantive Statistical Data (FAOSTAT), and other global sources that had data available for multiple countries and time periods on the impacts of agrifood systems, including GHG and nitrogen emissions, land use, the burden of disease from dietary patterns and the incidence of moderate poverty and undernourishment. The model provides preliminary estimates of the annual quantified environmental, social and health hidden costs of national agrifood systems for 154 countries in 2016–2023. Referring to them as “quantified” acknowledges the data gaps in many countries that prevent the estimation of all hidden costs, such as those associated with pesticide exposure and land degradation. As the hidden costs are at country level and presented as a monetary measure, they can be aggregated at global, regional and income levels and compared with macroeconomic indicators such as GDP.

The annual hidden costs of agrifood systems are obtained by multiplying impact quantities (for example, of GHG emissions) by their respective marginal hidden costs at national level.

The hidden costs are measured in 2020 purchasing power parity (PPP) dollars, which represent the amount of a basic basket of goods and services that a single US dollar, once exchanged to local currency, would have purchased in a given country in 2020. In other words, PPP eliminates price level differences across countries and equalizes the purchasing power of currencies. The goods and services represent welfare through their consumption. Consequently, the measured hidden costs represent the reduction in welfare (welfare losses) due to a decline in purchasing power. An advantage of hidden costs measured as GDP PPP losses is their comparability with national accounts and other national spending measures. They also allow for the aggregation of results, across both cost categories (for example, between environmental and health costs) and countries. A disadvantage of measuring the hidden costs as GDP PPP losses is the inability to measure changes in income inequality. Another disadvantage is the assumption of perfect substitution between losses in natural, human and produced capital income flows. Finally, it is important to note that hidden costs differ from abatement costs (see Glossary), which are excluded from the analysis due to a lack of data and valuation factors.

To account for hidden costs that are borne by future generations, the model further employs as a reference a “middle-of-the-road” shared socioeconomic pathway (IPCC pathway SSP2) and assumes a Ramsey social discount rate with time preference of 0 and constant marginal expected utility of consumption of 1.5.² For a detailed description of how hidden costs are reconciled within the framework provided by SSP2, see Lord (2023).¹

Scope of the analysis

Figure 5 in Chapter 2 illustrates the scope of agrifood systems covered by the analysis, as well as the hidden costs considered. In a nutshell, the analysis covers costs from GHG emissions, nitrogen emissions, blue water use, land-use transitions, and poverty, as well as productivity losses from dietary patterns and undernourishment. Due to data gaps, pesticide exposure and land degradation are not considered. Forestry is also outside the scope of the analysis, as estimates of the hidden costs associated with forestry-related economic activities (for example, logging) were not available. Specifically, and as identified in Figure 5, the analysis includes hidden costs related to the following:

1. Environment – external costs (see Chapter 1) of externalities from GHGs emitted along the entire food value chain from food and fertilizer production and energy use; nitrogen emissions (volatilized and runoff) at primary production level and from sewerage; and water use and land-use change at farm level.
2. Social – as a result of productivity losses from undernourishment (as defined by FAO [2022]³) or through the contribution of agrifood systems to moderate poverty. The hidden costs related to social harm are assumed to be driven by policies and institutions failing to address the issues of poverty and food insecurity. The rationale is the following: first, sufficient calories are available worldwide to achieve zero hunger, so the prevalence of undernourishment indicates the failure of agrifood systems to distribute available supply; second, poverty among agrifood systems workers is also an indication of the failure of agrifood systems given the substantial downstream profits made by wholesalers, processors and retailers of food products.
3. Health – as a result of unhealthy dietary patterns that cause a burden of obesity and NCDs and, consequently, productivity losses. Specifically, unhealthy diets low in fruits, vegetables, nuts, whole grains, calcium and protective fats, and high in sodium, sugar-sweetened beverages, saturated fats and processed meat have been associated with preventable morbidity and mortality from neoplasms, cardiovascular disease and type-2 diabetes.⁴ A wide range of market, institutional and policy failures (see Chapter 1) drive these dietary patterns by making foods of high energy density and minimal nutritional value more available, cheap and convenient.

Lord (2023)¹ discusses the distinction, at national level, between the production of hidden costs (cost production), the bearing of hidden costs that may have been produced by that nation or another nation (cost bearing), and those actors receiving free benefits from the cost bearing of other actors (benefit receiving).

Impact quantities data sources and coverage

Impact quantities refer to by-products of activities in agrifood systems, such as GHG emissions, that result in hidden costs. Data on impact quantities were obtained over 2014–2020 for 154 countries. Missing data were interpolated using moving average or regional change rates. Data for 2021–2023 – including GDP and other macroeconomic indicators – were then extrapolated using specific statistical methods or projections provided by FAO and the World Bank. The following sections present the data sources and coverage across the three hidden cost categories (environmental, social and health). For a detailed description of the methods of interpolation and extrapolation, and data sources, see Lord (2023).¹

Environmental impact quantities

Starting with GHG emissions, country-level data for (direct and indirect) tier-1 CO², methane (CH⁴) and nitrous oxide (N²O) emissions were retrieved from FAOSTAT for 2014–2020.⁵ Blue agricultural water-use data at country level were taken from FAO’s Global Information System on Water and Agriculture (AQUASTAT) between 2014 and 2020.⁶ Land-use conversion data – that is, the conversion of forest and unmanaged grassland (a broad category including shrubland, grassland and unmanaged rangeland) to cropland and pasture, and cropland and pasture to forest and unmanaged grassland – from 2014 to 2019 were obtained from the HIstoric Land Dynamics Assessment+ (HILDA+) dataset.⁷ Nitrogen emissions from volatilized ammonia (NH³) and nitrogen oxides (NOx) to air from agricultural production and energy use in 2015 were obtained from the European Commission’s Emissions Database for Global Atmospheric Research (EDGAR) v5.0 dataset.^8–10 Amounts of nitrogen runoff to surface waters and leaching to deep waters were calculated from Integrated Model to Assess the Global Environment – Global Nutrient Model (IMAGE–GNM) spatial datasets.^{11, 12}

Social impact quantities

Country-level data on the prevalence of undernourishment and the number of undernourished for the years 2014–2020 were obtained from FAOSTAT.¹³ Data on poverty gaps and the number of people in moderate poverty at the 3.65 per day 2017 PPP dollar income poverty line were obtained from the World Bank.¹⁴ The share of agrifood systems workers in total employment is used as a proxy for the share of agrifood systems workers in moderate poverty.¹⁵ For most countries with high levels of moderate poverty, this proxy is likely an underestimation, as most agrifood systems workers are in agriculture, which has a higher prevalence of poverty.¹⁶

Health impact quantities

For dietary patterns, the burden of preventable morbidity and mortality on human capital is measured by DALYs lost for each country between 2014 and 2019.¹⁷ DALYs also estimate high BMI for each country in the same period.¹⁷ Mediation factors are used to avoid double attribution of DALYs to both high BMI and dietary factors.^{18, 19} This interdependence means that DALYs represent one impact quantity per country per year and that the burden of disease from obesity and NCDs attributable to unhealthy dietary patterns are not treated as two separate quantities. Another complication is attributing the burden of disease to the activities of agrifood systems actors, as poverty and genetics can be co-factors in obesity and NCD prevalence. In this study, 75 percent of DALYs are attributed to the failure of agrifood systems activities. This attribution level is varied in uncertainty analysis.²⁰

Lord (2023) discusses the limitations in data and the costing methods in more detail. It presents the breakdowns of national hidden cost production and bearing not included in this report.¹

Marginal hidden cost data sources and methods

The marginal hidden costs are calculated in 2020 PPP dollars using the SPIQ-FS version 0 marginal damage cost model developed for the FSEC,ⁿ and are provided with uncertainty estimates in the form of parameterized probability distributions.^22–26 Damage to future economies is estimated based on business-as-usual future projections (SSP2).² Poverty is the exception, as it was costed directly using data from the World Bank; it was not modelled with uncertainty. As with the impact quantities, the following sections describe the data sources and method for valuing marginal hidden costs across the three dimensions.

Environmental marginal costs

For GHG emissions, SPIQ-FS resamples the simulations of the social cost^o of GHGs in 2020 by the Interagency Working Group on Social Cost of Greenhouse Gases (IWG-SCGHG) in 2020.^{28, 29} IWG-SCGHG simulations are provided for three discount rates (2.5 percent, 3 percent and 5 percent) and five socioeconomic scenarios. Using national GDP growth projections for SSP2 to 2100,² global rates matched a discount rate of 3 percent. Given this discount rate, the social costs of carbon under the five scenarios were sampled uniformly for additional uncertainty estimates of economic futures under SSP2. Social costs are given separately for CO², CH⁴ and N²O. Costs of a GHG emission in a country are borne globally through climatic changes. To attribute the cost of an emission as a cost to the country that caused the emission, it is assumed that economic actors in that country are required to pay an amount per emission equal to the social cost of the respective GHG. In principle, such funds would go to compensate the cost bearers of the emission inside or outside the country.

To cost agricultural blue water use, SSP2 discount rates were used for the impacts of future water scarcity. Marginal hidden costs are, however, underestimated due to a lack of cost data on the loss of environmental flows. Cost of land-use changes in terms of lost, retained or returned ecosystem services per hectare per year are derived from the Ecosystem Services Valuation database.^{30, 31} To the degree possible, carbon sequestration services were excluded from the valuation to avoid double counting with the costing of GHGs. National-level discount rates under SSP2 were used to discount lost ecosystem services from deforestation from 2020 to 2100 to obtain cumulative values per hectare of land-use change. For land being returned to its natural habitat, 14 years of gained ecosystem services were used to obtain cumulative value per hectare of land-use change. This was varied in uncertainty analysis. Costing nitrogen emissions relies on SPIQ-FS for the volatilization of NH³ (ammonia) and NOx (nitrogen oxides) to air and the runoff of reactive nitrogen into surface waters and soil leaching, predominately soluble NO₃- (nitrate).

Social marginal costs

SPIQ cost modelling includes a model from the number of undernourished and DALYs from protein–energy malnutrition based on data from WHO. The productivity losses of protein–energy malnutrition are costed using historical labour productivity data from the International Labour Organization (ILO).³² As for moderate poverty, data on the daily 3.65 2017 PPP dollar national poverty gap in 2014–2020 are from the World Bank,¹⁴ and adjusted by inflation in PPP terms to 2020 PPP. Poverty gaps were converted into income shortfall per annum. The total attributable cost of poverty is defined as the income-equivalent welfare required to eliminate moderate poverty that is attributable to distributional failures in agrifood systems. It is calculated by multiplying the relevant moderate poverty headcount by the average income shortfall in PPP terms.

Health marginal costs

The productivity losses of diseases attributable to diets and high BMI are costed using historical labour productivity data from ILO.³²

Intensity indicators of the environmental, social and health hidden costs of agrifood systems

Measuring the hidden costs of agrifood systems at national level in GDP PPP allows the costs to be compared to national indicators, such as agricultural gross value added (GVA) in PPP terms. This report, therefore, proposes three intensity indicators calculated as ratios between different types of cost (environmental, social or health) and different macroeconomic indicators.

The higher the value of these indicators, the more damaging the hidden costs being considered relative to the benefits brought about by the agrifood activities causing those costs. A zero value denotes zero net cost bearing, while a negative value represents net benefits. An example of the latter would be a gain in ecosystem services from the contraction of agricultural land and restoration of habitat.

Agricultural externalities impact ratio

The first indicator is the agricultural externalities impact ratio (AEIR), which is obtained by dividing the present value of hidden costs from agricultural production and land-use change in GDP PPP by the GVA of agriculture, forestry and fisheries (AFF). GVA AFF data are retrieved from the World Bank for all 154 countries as a percentage of GDP and then multiplied by GDP PPP.³³ GVA AFF is averaged over 2016–2020, and the average is converted to 2020 PPP dollars for consistency with the numerator. And as hidden costs can be aggregated at the global, regional or national level, so can the indicator. The following formula shows how the AEIR indicator is calculated and how it is derived from other two indicators:

where,

ALEC is the per hectare present value of hidden costs from agricultural production and land-use change, including agricultural water use, land-use changes (from forests to crops or pastures and vice versa), farm-level nitrogen emissions, and farm-level GHG emissions as an intensity measure of these hidden costs per unit of agricultural land (land being the primary factor of production in agriculture), and

ALEB is the per hectare GVA AFF, as an intensity measure of agricultural (primary phase) productivity.

Social distribution impact ratio

The third indicator is the social distribution impact ratio (SDIR), which is obtained by dividing the sum of (i) the income shortfall of agrifood workers from the moderate international poverty line (at 3.65 per day in 2017 PPP dollars) and (ii) the present value of productivity losses driven by undernourishment, by the average income of the moderate poor. It is calculated using the following formula:

where,

SDPOVA denotes the income shortfall from the moderate poverty line of agrifood systems workers,

SDPOUC denotes the annual total productivity losses driven by undernourishment (assumed, for simplicity purposes, to be experienced by the moderate poor) using historical labour productivity data from ILO,³² and

SDINC denotes the total annual income of the moderate poor.

SDIR is calculated as the average over 2016–2020. Income of the moderate poor is obtained from World Bank data and averaged over 2016–2020.

Dietary patterns impact ratio

The second indicator is the dietary patterns impact ratio (DPIR), which is obtained by dividing the present value of productivity losses from obesity and NCDs driven by dietary patterns (in GDP PPP) by GDP PPP. The following formula shows how the DPIR indicator is calculated and how it is derived from other two indicators:

where,

DPPCAP represents the per capita productivity losses from dietary patterns costed using historical labour productivity data from ILO,³² and

GDPCAP represents the per capita GDP PPP.^p