Nutrient adequacy for poor households in Africa would improve with higher income but not necessarily with lower food prices

The study addresses how changes in food prices and household incomes affect dietary intake and nutrient adequacy among African households. Despite historical improvements in global food security through lower food prices and higher incomes, sub-Saharan Africa still has the lowest share of its population meeting diet quality standards, with high burdens of energy, protein, and micronutrient deficiencies and substantial child undernutrition. Nutritious diets remain unaffordable for most poor consumers, partly due to high relative prices of nutrient-rich foods and low purchasing power. The authors aim to model consumer preferences and quantify how dietary energy and key macro- and micronutrients respond to total expenditures and food prices, comparing poor and wealthier consumers. They simulate two policy levers—a cash transfer to raise expenditures and targeted price discounts—to evaluate potential impacts on diet quality, including intake sufficiency, macronutrient balance, and nutrient density.

Demand system modeling since the 1970s (e.g., Deaton and Muellbauer) provided foundational estimates of income and price elasticities in low-income countries but often excluded Africa and relied on cross-sectional data. Early findings may not fully reflect today’s nutrition context of dietary transitions, fortification, and processed foods. Newer flexible models (e.g., EASI) and panel data can better capture substitution patterns and heterogeneity. Meta-analyses suggest income-inelastic demand for staples and more elastic demand for aspirational foods but with wide cross-country variation, often driven by cross-sectional methods. Experimental and quasi-experimental studies (RCTs, choice experiments) confirm that poor consumers increase energy and dietary diversity when incomes rise, though model-based predictions can differ from experimental estimates due to model specification, endogeneity, or program effects on preferences. Evidence on price elasticities in Africa is limited; panel-based estimates allowing elasticities to vary with expenditure are rare, with some studies showing strong sensitivity of energy and nutrient intake to staple prices (e.g., maize in Malawi) and lower price responsiveness in panel versus cross-sectional estimates. Evidence on subsidies/discounts largely comes from middle- and high-income countries; limited cases from Asia and South Africa show modest or mixed effects on demand and diet quality. This motivates comprehensive, multi-country panel-based demand modeling for Africa that can generate consistent expenditure and price elasticities and simulate policy impacts.

Data: The study uses LSMS-ISA nationally representative panel datasets from Malawi, Niger, Uganda, Tanzania, and Nigeria (various rounds spanning 2005–2019; details in Table 2). Households report 7-day at-home food consumption by item and source; unit values for non-purchased items are imputed. Food items are aggregated into 18–19 food groups per country, mapped to nutrient composition tables to derive household-level dietary energy and nutrient intakes. Total household expenditures (7-day) include food at home and non-food categories; food away from home is treated as part of the numéraire due to insufficient nutrient detail. Households are partitioned into four expenditure groups (Q1–Q4) using international poverty lines (PPP-adjusted). Prices: Unit values from purchasing households are used and imputed for non-purchasers to reflect opportunity costs. Modeling: An Exact Affine Stone Index (EASI) incomplete demand system is estimated for food groups and a composite numéraire, allowing flexible Engel curves and price responses, with price–expenditure interactions and household covariates. Community-level correlated random effects control for time-invariant unobserved heterogeneity. Censoring in 7-day consumption is addressed via Tobit for latent demand; cross-equation restrictions (homogeneity, symmetry, adding-up) are imposed. Endogeneity: (1) The Stone Price Index within real total expenditures is instrumented using a modified index based on sample-average budget shares. (2) Quality heterogeneity and price search are handled via food group-level Fisher Ideal Price Indices and price instruments constructed from neighbor households; first-stage relevance and exclusion are examined in supplementary materials. Elasticities: Budget shares, expenditure and own/cross-price elasticities are predicted at each household–year; standard errors are simulated from the parameter covariance matrix. Nutrient demand elasticities are derived combining food elasticities with nutrient composition. Diet quality measures: (i) intake sufficiency (household consumption vs household EAR/EER for DE, carbohydrates, protein, fat, iron, zinc, vitamin A, folate); (ii) macronutrient balance (shares of carbs, fat, protein in DE against WHO ranges); (iii) Nutrient-Rich Food Index (NRFI), rewarding densities of 9 beneficial nutrients and penalizing 3 moderation components. Policy simulations: A cash transfer (CT) equal to 20% of the median Q1 household’s total expenditures (country-specific) and 25% price discounts (PDs) applied separately to five categories—staple grains (SG), starchy staples (SS), pulses/nuts/seeds (PN), fresh fruits and vegetables (FFV), and animal-source foods (ASF)—are simulated. Costs: CT costs are fixed per household; PD costs equal the redeemed value based on predicted post-discount quantities, ignoring administrative costs.

Expenditure elasticities: Demand is most expenditure elastic for animal-source foods and beverages, less so for fruits/vegetables, nuts, fats, cereals, and tubers. Relative to prior meta-analyses, average expenditure elasticities are larger: 80–140% higher than Colen et al. for many groups (SG, SS, PN, ASF), 28–58% higher than Muhammad et al., close in Tanzania and larger by 24% (Malawi) and 21% (Uganda) compared with careful cross-sectional models. Nutrient intake elasticities: Poor households (Q1) generally have higher expenditure elasticities of intake than wealthier groups across macro- and micronutrients. Poor consumers have expenditure-elastic demand for zinc (everywhere except Malawi), iron (Nigeria only), vitamin A (Niger, Uganda, Tanzania), and folate (everywhere except Nigeria). Energy (DE) demand is typically expenditure inelastic except for Q1 in Niger, Uganda, and Tanzania. As expenditures grow, diets shift toward fat and protein and away from carbohydrates. DE sufficiency and CT impacts: The probability a household’s DE intake meets its EER rises steeply with expenditure; increases exceed 35 percentage points in Malawi, Niger, and Nigeria. A CT targeted to Q1 increases the share meeting DE requirements from 75%→82% (Malawi), 59%→76% (Niger), 34%→42% (Uganda), 19%→37% (Tanzania), and 50%→61% (Nigeria). Protein sufficiency rises notably (e.g., Tanzania 45%→61%; increases also in Malawi +7 pp, Niger +11 pp, Uganda +9 pp, Nigeria +9 pp). CTs also raise sufficiency for iron, zinc, and folate, with exceptions (iron in Niger; vitamin A in Nigeria already sufficient for most). NRFI responses to CTs vary: improved for all but the poorest in Malawi and Niger; improved for all but the wealthiest in Tanzania; decreased for all in Nigeria and all but the wealthiest in Uganda. Macronutrient balance: CTs do not exacerbate overconsumption; they reduce carbohydrate-heavy diets among Q1 in Malawi and Tanzania, with minimal effects on low fat/protein shares. Price effects: In single-staple countries (Malawi–maize; Niger–millet; Tanzania–maize), poor consumers’ intake across many nutrients is highly sensitive to the dominant staple’s price, producing large negative intake elasticities that cut across nutrients, including those not concentrated in the staple. In multi-staple countries (Uganda, Nigeria), no single price dominates; intake responds to a wider array of prices (SG, SS, PN). Cross-price effects are substantial: declines in complementary foods account for about half of the protein decrease after millet price increases in Niger and one-third after maize price increases in Malawi; similar patterns for rice (Niger, Uganda), roots/tubers (Tanzania), and cassava (Nigeria). PD impacts: For Q1, most PDs increase DE sufficiency, with exceptions—SS discounts in Niger and ASF discounts in Nigeria reduce DE. SG discounts come closest to CT effects on DE in Niger and Malawi. Protein sufficiency improves with SG and PN discounts in Malawi, Niger, Tanzania; PN in Niger and Uganda; ASF in Malawi, Tanzania, Nigeria. No single category universally improves all micronutrients. Notable adverse effects include: SG discounts lowering vitamin A in Niger and Uganda; SS lowering zinc/folate in Niger; PN lowering vitamin A in Uganda; FFV lowering vitamin A in Uganda; ASF lowering iron (Uganda) and vitamin A/folate (Uganda, Tanzania). Dietary balance under PDs: SG discounts generally worsen carbohydrate overconsumption (except Malawi, where it is already high). PN and ASF discounts reduce carbohydrate-heavy diets in Malawi. Among wealthier consumers, fat overconsumption in Uganda worsens under PN/ASF discounts; in Niger, SG discounts increase carb overconsumption and fat underconsumption, whereas SS discounts improve balance (Q4). Costs: For Q1, all PDs cost less per household than CTs (CT: US$30–47 per month, lowest Uganda, highest Nigeria). SG discounts are most expensive (up to US$24 per Q1 HH/month in Niger; US$58 for Q4). In Uganda, SS is the most expensive PD due to staple mix. PN discounts are relatively low-cost (median ≈US$4 per Q1 HH/month in Uganda, the highest among countries), and not excessively costly for wealthier groups. FFV discounts are also low-cost for Q1 (up to US$3.98 in Malawi). ASF discounts are small for Q1 (up to US$4.83 in Nigeria) but costly for Q4 (up to US$24.86 in Malawi). Considering dietary impacts and costs, PN and FFV discounts appear more promising than SG or ASF discounts.

Findings support the centrality of poverty reduction for improving diet quality: increases in total expenditures are strongly associated with higher intake of energy and multiple nutrients, especially among the poorest. CTs substantially raise the share of households meeting energy and key nutrient requirements without worsening macronutrient imbalances, though they do not eliminate all intake gaps. Food prices also matter: in single-staple settings, the dominant staple’s price can drive intake across nutrients; in multi-staple settings, responses to prices are more diffuse across food groups. Complex cross-price substitutions mean a given price change can have counterintuitive and heterogeneous effects on nutrient density and balance, varying by country and expenditure group. No universal food price subsidy unambiguously improves all diet quality dimensions in all contexts. While SG discounts can improve several outcomes, they are expensive; discounts on nutrient-dense foods (FFV, ASF) can help but often involve tradeoffs and smaller effects. PN discounts stand out as offering meaningful improvements across nutrients for poor households at relatively low cost. Overall, the structural demand approach provides policy-relevant insights on how consumer preferences shape dietary responses to income and price changes and can guide prioritization of interventions and agricultural R&D targeting staples with the greatest potential nutrition benefits for the poor.

Using flexible, panel-based demand system estimation across five African countries, the study shows that raising poor households’ total expenditures yields robust improvements in diet quality and intake adequacy, while lowering specific food prices does not necessarily close nutrient gaps due to complex substitution patterns. Staple prices are especially influential in single-staple contexts; in multi-staple contexts, multiple prices matter. Policy simulations indicate that CTs broadly improve sufficiency without exacerbating imbalance, while PDs have mixed effects, with PN and FFV discounts offering favorable impact–cost profiles compared to SG and ASF discounts. The work underscores that healthy diets remain costly relative to poor households’ purchasing power; thus, poverty reduction and attention to staple price dynamics are critical. Future research should utilize enhanced panel surveys with longer food lists, better within-household consumption measurement, and richer data on food away from home and processing/fortification, enabling refined modeling and more precise program design.

Key limitations include: (1) Household-level intake is modeled rather than individual-level consumption, so within-household allocation is not observed; sufficiency at the household level does not guarantee individual adequacy. (2) Consumption is measured over a 7-day recall period and subject to measurement error. (3) Insufficient detail on food consumed away from home prevents modeling its nutrient contribution. (4) Lack of data on processing (e.g., milling, fortification) may affect nutrient content assumptions. Simulation assumptions add further limitations: fixed shares of CT used to raise expenditures (no modeled returns from investments), and no general equilibrium effects on producer prices, wages, or consumer prices (beyond the simulated PD). These choices likely yield conservative CT impacts and may omit market feedbacks that could dampen or alter results.