Behavioural, Nutritional, and Health Effects of Food Carbon Taxes: Evidence from a QUAIDS Demand System for Pakistan

Food pricing policies are increasingly discussed as instruments to internalise environmental externalities and improve diets. The motivation is clear: animal-based foods account for a disproportionate share of food-system greenhouse gas (GHG) emissions (Clark et al., 2019; Poore & Nemecek, 2018), while dietary patterns high in red and processed meat are strongly associated with cardiovascular disease (CVD) risks (Afshin et al., 2019; Willett et al., 2019). Yet the case for food carbon taxes cannot be made on emissions alone. In low- and middle-income countries, food expenditure shares are high and nutritional adequacy is often fragile, so price shocks can have first-order welfare and nutrition consequences (FAO, 2022). Whether carbon pricing in food delivers health co-benefits, or instead worsens diet quality for vulnerable households, is fundamentally an empirical question about substitution behaviour. This paper develops a behavioural microsimulation framework that links food carbon taxation → prices → consumption substitution → protein intake → predicted CVD risk. The core behavioural engine is a Quadratic Almost Ideal Demand System (QUAIDS), estimated on household microdata from Pakistan. QUAIDS is well-suited here because it accommodates flexible Engel-curve shapes and non-linear expenditure effects (Banks et al., 1997), which matter in settings with substantial heterogeneity in food budgets and diet composition. The demand system follows the standard AIDS/QUAIDS tradition (Deaton & Muellbauer, 1980) and is estimated with demographic controls and survey weights. Food is grouped into nine aggregates that reflect both emissions and nutrition channels: cereals; red meat; poultry; fish; dairy; fruit and vegetables; beverages; processed foods; and plant-based protein (legumes/peas/beans). Prices are constructed from unit values and total expenditure is equivalised using a square-root scale. From the estimated parameters, we recover own- and cross-price elasticities and propagate policy-induced price changes through predicted budget shares and implied quantities. Nutritional outcomes focus on protein (given its policy relevance for affordability and adequacy), derived from group-specific nutrient coefficients. Health outcomes are evaluated using comparative risk assessment logic by mapping dietary changes into shifts in cardiovascular disease (CVD) risk factors, drawing on established evidence on diet–disease relationships from large-scale comparative risk assessments (Lim et al., 2012; Micha et al., 2017) and widely used dietary benchmarks for healthy and sustainable diets (Springmann et al., 2016; Tilman & Clark, 2014). We simulate three carbon-tax scenarios applied to all food: no revenue recycling, full revenue recycling via an equal per-capita lump-sum transfer, and targeted partial recycling, where 20% of revenues subsidise plant-based protein (legumes/peas/beans). The tax is calibrated at €5 per tCO₂e, consistent with low-end carbon prices observed across many existing instruments, particularly in low- and middle-income settings (World Bank, 2023). This design captures a key tension: while carbon pricing can reduce diet-related emissions, substitution toward cheaper calories or processed foods may dilute nutritional and health co-benefits (Smed et al., 2016). The partial-recycling scenario tests whether modest earmarking can redirect substitution toward lower-emission, nutritionally favourable foods relative to the pure-tax and lump-sum benchmarks. Methodologically, the paper embeds a demand system within a fiscal microsimulation framework to jointly evaluate climate, nutrition, and health outcomes, and shows how recycling design governs whether behavioural responses translate into co-benefits.