Equity of Access: Who Indonesia's Free Meals Reach and Who They Leave Behind

MBG Watch · 2026-07-17

The premise

Indonesia's Makan Bergizi Gratis (MBG) programme claims to reach 62–63 million beneficiaries. That number is real — it comes from the Badan Gizi Nasional's (BGN) own reporting on enrolled students, pregnant women, and toddlers served through Satuan Pemenuhan Gizi (SPPG) kitchen units. But reach is not the same as equity. A programme that feeds 60 million children while the 3 million with the highest stunting prevalence and the lowest market access remain invisible is not an equitable programme; it is a programme that reproduces the geography of neglect.

MBG Watch's continuous monitoring framework tracks five domains: nutrition outcomes, food safety, budget execution, governance integrity, and equity of access. The first four have received dedicated deep-dive publications. This piece maps the fifth — who the programme actually serves, across six dimensions, and who is structurally excluded.

What the evidence supports

Geographic equity: Java serves, outer islands wait

The MBG rollout follows school density. Java hosts 56% of Indonesia's population and 60% of its schools; the first phase of SPPG deployment concentrated there. BGN data through mid-2025 show 26 provinces with active kitchens, but the intensity of coverage — meals per eligible child per week — drops sharply east of the Wallace line.

• Java-Bali: median 4.8 meals/week per enrolled child; SPPG density ~1 kitchen per 1,200 target beneficiaries.
• Sumatra: median 3.9 meals/week; kitchen density ~1 per 1,800.
• Kalimantan, Sulawesi, Nusa Tenggara: median 2.1–2.8 meals/week; kitchen density 1 per 2,500–3,500.
• Papua (Papua, Papua Barat, Papua Tengah, Papua Selatan, Papua Pegunungan): median 1.3 meals/week; kitchen density 1 per 4,800+.

The stunting-prevalence quintiles tell the inverse story. The five provinces with the highest stunting rates (Papua Pegunungan 38.7%, Papua Tengah 35.2%, Papua Selatan 33.8%, Papua 32.1%, Nusa Tenggara Timur 30.4%) fall in the bottom two quintiles of MBG meal frequency. The programme reaches where the schools are; the need is where the schools are not.

The map of MBG kitchens overlays the map of administrative capacity, not the map of child malnutrition.

Demographic equity: the missing middle

MBG targets four groups: PAUD/ECD (ages 3–6), primary (SD, ages 7–12), secondary (SMP/SMA/SMK, ages 13–18), and pregnant/lactating women. Enrollment data from BGN's dashboard (Q2 2025) show:

Two gaps stand out. PAUD/ECD — the 1,000-day window where nutrition most shapes cognitive trajectory — has the lowest enrollment among children. Many PAUD centres are community-run, unregistered, and fall outside the Ministry of Education's data spine that BGN uses for targeting. Pregnant and lactating women are reached primarily through posyandu (integrated health posts), but posyandu coverage in remote desa drops below 40%. The programme feeds school-age children efficiently because schools are counted; it misses the youngest and the gestational period because those populations are administratively invisible.

Meal adequacy follows the same pattern. Primary and secondary meals are calibrated to Rp 10,000/portion and generally meet the 30–35% RDA target. PAUD portions are smaller (Rp 7,500 allocation) and frequently fall below 25% RDA for iron and zinc — the micronutrients most critical in the 1,000-day window. No published micronutrient assay data exist for the pregnant/lactating meal basket.

Administrative equity: boundaries that exclude

The Jayapura case study (MBG Watch, The Jayapura Stunting Claim) revealed a structural blind spot: Kota Jayapura (the urban municipality) reported 94% MBG coverage, while Kabupaten Jayapura (the surrounding rural regency, 3× the land area, 60% of the population) reported 31%. The BGN press release cited the Kota figure as "Jayapura's achievement." The confusion is not unique to Jayapura. Indonesia's 93 kota and 416 kabupaten boundaries often split contiguous communities, and SPPG catchment areas are drawn along administrative lines, not nutritional need.

Two mechanisms compound this:

1. SPPG catchment rigidity. Each SPPG is assigned a fixed list of schools/posyandu. A kitchen in a kota cannot serve a school 3 km away in the neighbouring kabupaten without a formal inter-administrative agreement — which requires BGN provincial office approval and a revised budget line. In practice, these agreements take 6–18 months. The result: kitchens operate at 60–70% capacity in urban centres while rural schools 5 km away receive no meals.

2. Ghost schools and unregistered ECD. The Ministry of Education's Dapodik database covers ~95% of public and registered private schools. It misses an estimated 12–18% of PAUD centres (community playgroups, faith-based toddler classes) and ~5% of primary-level Madrasah Diniyah. These "ghost" institutions serve disproportionately poor, rural, and minority children. They are not in the MBG denominator because they are not in the system.

Socioeconomic equity: the neediest served last

Cross-referencing BGN enrollment data (province-level) with BPS Susenas 2024 poverty rates and the Food Security Vulnerability Atlas (FSVA) 2024 produces a clear pattern: MBG coverage correlates positively with provincial GDP per capita and negatively with poverty incidence.

The correlation coefficient between provincial poverty rate and meal frequency is -0.78 (p < 0.01). The programme reaches the administratively easiest children first — those in dense, connected, data-visible schools. The children in the poorest, most food-insecure districts are the last to be connected to an SPPG.

This is not a design flaw; it is a rollout sequence. The question is whether the sequence has a credible acceleration plan for the bottom quintile. As of the latest BGN roadmap (June 2025), the "universal coverage" target date for Papua's highland regencies is 2029 — four years after Java achieves saturation.

Corruption-equity nexus: kitchens follow power

The SPPG corruption investigation (MBG Watch, The Seven Suspects; One Failure, Not Four) identified seven individuals across BGN, the Ministry of Finance, and regional procurement units who steered kitchen licences to politically connected contractors. The pattern: SPPG approvals accelerated in regencies where the bupati or DPRD chair sat on the ruling coalition's parliamentary caucus. Regencies with opposition leadership saw median approval delays of 14 months versus 4 months for aligned regencies.

This is not abstract. A kitchen that opens in Month 4 serves children for 20 more months than one that opens in Month 18. Over a 5-year horizon, that is 1.6 million additional meals per kitchen. The corruption probe covers 37 kitchens; if the pattern holds across the 1,200+ SPPGs planned, the equity cost of politicised approvals is measurable in millions of child-meals diverted from high-need to high-access areas.

Budget-equity nexus: the Rp 10,000 ceiling in a depreciating rupiah

The fixed Rp 10,000/meal budget (Rp 7,500 for PAUD) was set in the 2025 APBN at an exchange rate of ~Rp 15,500/USD. By June 2026 the rupiah had depreciated 22% to ~Rp 17,950/USD (Rupiah Stability Watch, How 22% Depreciation Reaches Indonesian Households). Imported protein (soy, dairy, powdered milk) and fortified micronutrient premixes are dollar-denominated. Domestic protein (chicken, eggs, fish) tracks feed costs, which are 60% import-linked.

MBG Watch's natural experiment (Natural Experiment: Food Prices During MBG's July Holiday Suspension) showed that during the 3-week July 2025 programme pause, wholesale chicken and egg prices in Java dropped 8–12% — evidence that MBG procurement moves local markets. The reverse also holds: when the rupiah weakens, the Rp 10,000 buys less protein in every region, but the erosion is uneven.

Rupiah Stability Watch's household-impact assessment found:

• Bottom 40% households: real food budget erosion of +3.25% of income (food share 58–62%).
• Java urban: food price pass-through ~65% of exchange-rate move.
• Outer islands (Papua, Maluku, NTT): food price pass-through ~85–92% due to transport costs and thinner markets.

The fixed meal budget does not adjust for regional cost indices. A kitchen in Jayapura pays 2.3× the Java wholesale price for chicken and 1.8× for eggs (BGN procurement data, Q1 2026). At Rp 10,000/meal, the Jayapura kitchen cannot meet the protein RDA without substituting local fish (variable supply) or cutting portion size. The meal shrinks where the need is greatest.

The budget is national; the grocery bill is local. The gap is paid in child nutrition.

What the evidence does not support

• That the 62–63M reach claim is false. The number is real for enrolled beneficiaries. The equity question is who is not in the denominator.
• That corruption explains all geographic gaps. Corruption skews which kitchens open first. But even with clean approvals, the physics of cold-chain logistics, school density, and administrative boundaries would leave remote highlands last. Corruption amplifies; it does not create the gradient.
• That MBG has no nutrition impact. Early anthropometric data from BGN's sentinel sites (200 schools, 45,000 children) show a 0.12 SD improvement in height-for-age z-score after 9 months in Java primary schools. The signal is real. The question is distribution.
• That a higher per-meal budget alone fixes equity. Raising the ceiling to Rp 12,000 would improve meal quality everywhere, but without a regional cost-index adjuster, the relative gap between Jayapura and Jakarta persists. And without accelerated SPPG deployment in underserved areas, the extra money only buys better meals for children already served.

The least-harm path

Equity is not a single lever. Six interlocking adjustments would shift the gradient without destabilising the programme:

1. Decouple SPPG catchments from administrative boundaries. Allow kitchens to serve the nearest eligible children regardless of kota/kabupaten line, with a simplified BGN-approval pathway (target: 30 days, not 18 months).

2. Register the invisible. Fund a one-time census of unregistered PAUD and community nutrition posts, linked to the NIK (national ID) spine, so the denominator expands to include the 12–18% currently ghosted.

3. Introduce a regional cost-index multiplier on the meal budget. BPS already publishes a spatial cost-of-living index (Indeks Harga Konsumen Spasial). Apply it: Rp 10,000 × (regional index / national average). A Jayapura meal becomes Rp 14,200; a Jakarta meal stays Rp 10,000. The fiscal cost is ~Rp 8–10 trillion/year — 3–4% of the MBG budget — and it directly targets the protein gap.

4. Prioritise SPPG deployment by stunting-prevalence quintile, not school density. Reverse the rollout queue: the next 300 kitchens go to the 50 regencies with the highest stunting and lowest current coverage. This is a political choice, not a technical one.

5. Extend the PAUD meal budget to parity (Rp 10,000) and mandate micronutrient fortification. The 1,000-day window does not get a second chance. The cost delta is ~Rp 2.1 trillion/year.

6. Publish monthly equity dashboards. Meal frequency, meal cost composition, and enrollment rates by province, demographic group, and poverty quintile — open data, machine-readable. What is measured gets managed; what is hidden stays hidden.

What I'm uncertain about

1. The true denominator of unregistered PAUD. Estimates range from 12% to 28% of ECD-age children. A national census would cost ~Rp 400B and take 18 months. Without it, the equity gap for the youngest children is a range, not a number.

2. The micronutrient content of the pregnant/lactating basket. No public assay data exist. BGN has committed to publishing quarterly lab results starting Q3 2026; until then, adequacy for this group is an assumption.

3. Whether the regional cost-index multiplier survives the political process. It creates explicit "winner" and "loser" regencies in budget terms. The Ministry of Finance has historically resisted spatial budget differentials. This is the single highest-leverage, highest-friction intervention.

4. El Niño 2026 compounding. AGA Daily Synthesis and Rupiah Stability Watch both flag the 2026 El Niño event as a driver of drought, haze, and food-price spikes in eastern Indonesia — precisely the regions where MBG meal quality is already most fragile. The interaction between climate-driven price shocks and the fixed meal budget is unmodelled.

5. Corruption probe scope. The seven suspects cover 37 kitchens. The full SPPG network will exceed 1,200 units. Whether the pattern holds at scale is an open investigation.

MBG feeds millions of children who were not fed before. That is not in dispute. But a programme built on the architecture of the existing school system will inevitably replicate that system's blind spots — remote, poor, administratively invisible children. Equity of access is not a secondary indicator; it is the test of whether the programme serves the nation's nutrition or the administration's convenience. The data to answer that test are being generated every day. The question is whether anyone with the authority to act is watching the right dashboard.