Indonesia Says It Has Too Much Rice. Three Official Numbers Say Otherwise

How 31.3 million tonnes, 34.77 million tonnes and 38.6 million tonnes describe one 2026 harvest, and why a self-sufficiency ratio that jumped 23 points in a year cannot survive its own arithmetic

SeaPRwire – On 30 September 2026, workers at a Bulog warehouse in Balikpapan, East Kalimantan, were photographed checking rice stocks. The picture ran with a story that has become routine in Jakarta this year: state procurement had passed 4.11 million tonnes of farmers’ rice, beating the 4 million tonne target for the whole of 2026 with three months to spare, and the national reserve had touched about 5.4 million tonnes, the highest since independence.

The same week, the government said something larger. Indonesia had reached rice self-sufficiency, the 2026 import order was cancelled, a programme to clear forest for new rice fields was being wound down, and officials were exploring premium rice exports to Malaysia. The reserve, readers were told, covers demand into the first quarter of 2027.

All of that may be true. The problem is narrower and more awkward. The four numbers doing the work in those two paragraphs, production, self-sufficiency, reserve and consumption, come from different agencies, use different denominators, and in at least one case cannot all be true at the same time. This piece takes them one at a time and checks whether they add up. Where they do not, the gap is not a scandal. It is a measurement problem that the policy has already spent money against.

The Claim, Quoted Precisely

Hasil Sembiring, special adviser to the agriculture minister for production enhancement, put the case on the record at a Jakarta briefing on 10 August 2026. Output had risen, reserves were stronger, incentives had improved, and the country was on a path to a more productive and self-sufficient food system.

The numbers behind that sentence, as published by the statistics agency and relayed by state media, are these. Paddy output reached 34.69 million tonnes in 2025, against 30.62 million tonnes in 2024, a gain of 13.29%. Harvested area rose to 11.32 million hectares from 10.05 million. The food self-sufficiency ratio moved from 87.14% in 2024 to 110.25% in 2025.

Those are not forecasts. They are presented as measured outcomes. Two of them, the production total and the ratio, are the load-bearing claims for everything that follows: the cancelled import, the export trial, the wind-down of land clearing.

The 13.29% Is a Land Story, Not a Yield Story

Start with the easiest check, because it is pure division.

Divide 34.69 million tonnes by 11.32 million hectares and output per hectare comes to 3.06 tonnes. Do the same for 2024: 30.62 divided by 10.05 gives 3.05 tonnes. The difference is under one percent.

Harvested area, meanwhile, rose 12.6% in a single year. That is a very large movement in a country where new paddy land is scarce and where the government has been trying to stop converting forest into fields. It is possible. It is also the kind of number that usually comes from a change in how area is counted rather than from 1.27 million new hectares appearing between two harvests.

The library decomposition used for this piece puts 94.8% of the 2025 output gain on area and 5.2% on yield. The published totals support that split. They do not support a story about farmers getting more rice from the same ground, which is the version usually told, because yield did not move.

This matters for the policy, not just for the ledger. Area-led growth has a hard ceiling and a political cost, since the cheapest remaining land is forest. Yield-led growth does not. If the 13.29% came from area, then the 2026 target inherits a problem: the same trick cannot be run twice without more land, and the land programme is being closed.

Three Forecasts for One Harvest, 23.3% Apart

The 2026 numbers get worse before they get better. Three published forecasts exist for the same crop year, and they do not describe the same country.

Table 1: What three published sources project for Indonesia’s 2026 paddy crop

Source: ANTARA , August 2026; agency projections as reported.

The gap between the low and high figure is 23.3%. On a base of roughly 35 million tonnes that is about 7.3 million tonnes, or more than the entire national reserve at its record peak. A country that cannot say within 7 million tonnes how much rice it will grow this year is not in a position to cancel imports on the strength of the forecast.

The low end deserves particular attention because of who produced it. A 31.3 million tonne crop would be below the 2025 outturn of 34.69 million, which means the food agency’s own projection implies a 9.8% decline in the year the government declared self-sufficiency and stopped importing. Either the projection is conservative, or the declaration rests on a crop the agency does not expect.

The high end, 38.6 million tonnes, would be a 11.3% increase on 2025 and would require another area jump on top of the one that just happened. Nothing in the procurement data suggests that is under way: Bulog’s 4.11 million tonnes bought from farmers is strong, but it is a purchase number, not a production number, and it can rise while output falls if farmers choose to sell to the state instead of to traders.

The Ratio That Cannot Hold

Now the self-sufficiency ratio, which is the number most quoted and the least examined.

The series runs 91.04% to 98.88% across 2020 to 2023, drops to 87.14% in 2024, then leaps to 110.25% in 2025. A 23.1 point move in twelve months.

A ratio of production to consumption is only as good as its denominator, and the denominator here is not published with the ratio. But it can be recovered by division. At 110.25%, a 34.69 million tonne crop implies national consumption of 31.47 million tonnes. At 87.14%, a 30.62 million tonne crop implies consumption of 35.14 million tonnes.

Put those two implied figures side by side and the problem appears. Consumption would have fallen 10.4% in one year, from 35.14 to 31.47 million tonnes, in a country whose population grew, whose free meal programme expanded to 83 million beneficiaries, and whose government was simultaneously worried about feeding people.

A 10.4% drop in rice consumption in a single year has no plausible mechanism. Households do not halve their rice. The free nutritious meals programme, budgeted at IDR 335 trillion for 2026 and covering 83 million people, adds demand rather than removing it, at roughly IDR 304 per person per day on the programme’s own arithmetic.

So one of the two ratios is wrong, or the two are computed on different bases. Since both are attributed to the same agency and the same methodology family, the likeliest explanation is a denominator that changed definition between years, most probably a shift between “production available for food” and gross paddy output, or a change in the conversion rate from paddy to milled rice.

There is a third witness. The government’s own working figure for national rice consumption is 2.5 million tonnes a month, which is 30 million tonnes a year. Against 34.69 million tonnes of paddy that is a ratio of about 115.6%, not 110.25%. Against 31.3 million tonnes of forecast 2026 output it is 104.3%, and against a milled-rice conversion, which is how consumption is actually measured, it falls below 100%.

Three official anchors, 110.25%, 115.6% and 104.3%, none of which is the number in the press release.

Reserves, and the Difference Between a Stock and a Buffer

The reserve numbers are the strongest part of the government’s case, and they are internally consistent in a way the ratios are not.

Table 2: Government rice reserve by release date, with the stated target for comparison

Source: Shangbao Indonesia , 9 October 2026; earlier figures from state media as compiled.

The build from 3.25 to 5.4 million tonnes in nine months is real, and it was bought at IDR 6,500 per kilogram, a floor price that protects farmers and costs the budget money on every tonne. The 2026 food security budget is IDR 210.4 trillion, up from IDR 139.4 trillion in 2025.

Here is the catch. A 5.4 million tonne reserve against 2.5 million tonnes of monthly consumption is 2.2 months of cover, not the “into the first quarter of 2027” that the coverage language implies. To reach the first quarter of 2027 from October 2026 requires about five months. The national figure that does reach five months, 12.53 million tonnes total stock on 7 January 2026, includes household and trader inventories, which the government does not control and cannot release.

Government reserve coverage at that January date, on the same arithmetic, was 1.3 months. That is the number that governs what happens in a bad year. It has improved since, to roughly 2.2 months, and it is still not five.

Self-Sufficiency in One Commodity, Imports in the Rest

The word in the announcement is food, but the number behind it is rice. That distinction is where the policy gets expensive.

Indonesia tracks 11 strategic food commodities. Eight of them are reported as running a production surplus in 2026. Rice is one. The three that are not are the ones that show up in the import bill every year, and they are not small: soybean imports are projected at 26.4 million tonnes for 2026, garlic at 679,895 tonnes, beef at 321,660 tonnes. Those are forecasts, and they should be read as forecasts, but the direction has been stable for a decade.

Soybean is the awkward one, because it is not a luxury. Tempeh and tofu are the cheapest protein in the country and they are made from imported beans. A rice surplus of 3 million tonnes and a soybean deficit of 26 million tonnes are not the same kind of achievement, and no ratio that averages them is useful.

The free nutritious meals programme sharpens the point. Budgeted at IDR 335 trillion for 2026 and reaching 83 million people, it is the largest demand-side intervention in Indonesian food since the 1998 crisis. Its per capita allocation works out to about IDR 304 a day, which buys roughly a fifth of a kilogram of rice at the procurement floor price, or a portion of eggs, or a small piece of fish. The programme’s kitchen-level sourcing decisions, not the paddy harvest, will determine whether it lifts protein consumption or simply adds volume to the rice channel.

There is also a labour question hidden in the area story. If 1.27 million additional hectares were harvested in 2025, someone planted and cut them. Indonesian agriculture has been losing workers to manufacturing and services for a decade, and the harvest labour market is already tight in the main producing provinces during the peak season. An area-led expansion is therefore not just a land constraint. It is a labour constraint that shows up as a wage bill two seasons later.

None of this is an argument against the reserve. Buying 4.11 million tonnes at IDR 6,500 protects farmers from the price collapse that a big harvest normally brings, and that is a legitimate use of a state buyer. The argument is about which number gets used to justify the next decision. A rice surplus is not food security. It is a rice surplus, in one commodity, in one year, produced mostly by putting more land under the crop.

Where the Numbers Cannot Coexist

Three specific incompatibilities, stated as plainly as possible.

First, a 2025 self-sufficiency ratio of 110.25% and a 2026 production forecast of 31.3 million tonnes cannot both be right unless consumption collapses by a further 9% in 2026. The ratio says the country has a surplus. The forecast says next year’s crop is smaller than this year’s consumption.

Second, the decision to cancel the 2026 rice import sits on top of that forecast, not on top of the 2025 outturn. Imports are bought against the coming year. If the food agency’s own projection is 31.3 million tonnes and consumption is 31.47 million tonnes at the 110.25% denominator, there is no surplus to export and very little to hold.

Third, the export initiative to Malaysia has no volume behind it that survives the reserve math. Every tonne exported comes out of a 5.4 million tonne stock that represents 2.2 months of cover. A country with 2.2 months of buffer does not have exportable surplus, whatever its ratio says.

None of this means Indonesia is short of rice today. Bulog bought 4.11 million tonnes, the reserve is at a record, prices were defended at the farm gate, and the harvest was large. It means the four numbers used to describe that success are drawn from four different frames, and the frame that gets quoted, the ratio, is the one that moves the most and is explained the least.

The Fraud Cases That Make the Optimism Expensive

There is a reason to be strict about agricultural statistics in Indonesia that goes beyond bookkeeping, and it is sitting in the court records.

eFishery, once valued at USD 1.4 billion, collapsed after revenue irregularities surfaced at the end of 2024. Its founder was sentenced to nine years in April 2026; prosecutors had sought ten. Cumulative funding was about USD 415 million across four rounds, and investor losses are put at roughly USD 300 million. A fine of IDR 1 billion was imposed, a sum that rounds to nothing against the losses.

TaniHub is the second case, and the more instructive one. Six defendants were sentenced in June 2026: the founder to nine years, the finance director to seven, a former MDI Ventures chief executive to five, and a former BRI Ventures chief executive to three. State losses were assessed at USD 25 million, against USD 20 million from MDI and USD 5 million from BRI Ventures. The company had been valued at USD 218 million in May 2021. Its lending arm reached a non-performing loan ratio of 63.9% in 2023, was restricted by the regulator in April 2023, and lost its licence in May 2024. Bad loans of IDR 693 billion, operating losses of IDR 437 billion, and overstated earnings of IDR 359 billion are all in the file.

The pattern is the one that matters for statistics: a lending book, a valuation and a growth story that all depended on numbers nobody outside the company could check. Agricultural data in Indonesia is harder to falsify than a fintech loan book, because the statistics agency does sample surveys. But the same reflex, treating an unverified number as a settled one because the policy needs it, is what produced a 23 point jump in a ratio that no one has published a denominator for.

What Would Prove This Wrong

The argument here is that the 110.25% ratio is a measurement artefact, that the 2026 crop is smaller than the 2025 crop, and that the reserve is thinner than the coverage language suggests. Four observations would settle it against me.

If the statistics agency publishes, before the end of 2026, a methodology note that fixes the consumption denominator and shows both years on the same basis, with 2024 also above 100%, then the jump is a revision and not an artefact. Watch for a restated 2024 figure.

If Bulog’s procurement for the full 2026 season exceeds 5 million tonnes without a further price increase above IDR 6,500, then the crop is genuinely larger than the low forecast and the area expansion was real.

If the government publishes monthly rice stock including household and trader inventories for two consecutive quarters and the government share of that total rises above 45%, the reserve is deeper than 2.2 months of cover implies.

And if an actual export contract to Malaysia is signed and shipped, with a volume and a date, then there is exportable surplus and this piece is wrong about the arithmetic. Not the announcement of intent. The bill of lading.

Methodology

Data as of October 2026. Three denominators are in play and they are not interchangeable. The self-sufficiency ratio is treated here as production divided by consumption; the consumption figure is never published, so it is recovered by dividing published output by the published ratio, which is exact arithmetic on two stated numbers and inherits any rounding in the ratio. Per hectare output divides paddy output by harvested area, both from the statistics agency as relayed by state media, and is a gross paddy measure, not milled rice. The paddy to milled conversion is not applied anywhere in this piece, because the conversion rate is not stated in the sources; where a milled comparison would change a conclusion, the text says so. Reserve coverage divides the government-held stock by the stated monthly consumption of 2.5 million tonnes, and excludes household and trader inventories, which is why it differs from the national stock coverage figure. The 94.8% / 5.2% decomposition is a library calculation from the area and output totals, and the direct per hectare arithmetic is shown so a reader can check it independently. Cost figures are nominal rupiah, not inflation adjusted.

Limitations

The three 2026 forecasts are attributed to different intermediaries and may not share a cut-off date, so part of the 23.3% spread could be timing. The consumption denominator is inferred, not published, which means the 10.4% implied drop is a property of the ratios rather than a measured fact. And the reserve series mixes target, year-end and intra-year figures from different releases, so the month-to-month path is indicative.

Sources

• ANTARA , Indonesia confident on food security outlook as farm output rises, 10 August 2026
• Shangbao Indonesia , Bulog procures 4.11 million tonnes, reserve peak 5.4 million, 9 October 2026
• CNA , TaniHub case verdict and investor losses, June 2026
• The Straits Times , eFishery founder sentenced to nine years, April 2026
• Agency production and area statistics for 2024 and 2025 as relayed by ANTARA, no public dataset link (no public link)

This article was prepared by the expert team at Global Infinity Labs, drawing on insights and data provided through its MCP data services.

About Global Infinity Labs

Global Infinity Labs is an AI innovation company specializing in MCP data services.

Email: support@globalinfinitylabs.com