A country's exports are a rough inventory of what it has learned to do. Indonesia's are heavy on things dug up or grown: coal, palm oil, and — since it stopped shipping nickel ore — stainless steel. The idea behind the plot above is simple and quite old: the next thing a country manages to make is usually near something it already makes, because the skills, suppliers and habits carry over. So you can list the candidates. This is that list, and its shape is the news.
What this gives you: the list, and each product's position on it. There are 976 products Indonesia does not currently sell competitively; the plot holds the 400 it has most to build on. For each one: how many similar things it already makes, how hard the thing is to make, and how much the world buys of it every year.
What it does not give you: a plan, or a ranking you can quote. Split each measure at its upper quarter and 98 products fall in the closest quarter, 98 in the hardest-to-make quarter, and 2 in both. Those 2 are vulcanised rubber thread and hat linings. The whole world buys about $1.0bn of them a year between them — roughly 0.3% of what Indonesia already exports.
Both halves are the finding. There is somewhere to go, and the short steps are the small ones. A list that showed only the first half is how a country ends up with a policy aimed at an empty corner.
The corner that is empty
Rank the 400 products by how close Indonesia already is, then rank them again by how hard they are to make, and the two orders come out close to mirror images of each other. Measured on a scale where −1 would be a perfect mirror and 0 would mean no relationship at all, they sit at −0.699. In words: the nearer a product is, the less there is to learn from making it.
You can see it without the arithmetic. Here are the five products Indonesia is closest to selling competitively, in order, named exactly as the trade tables name them:
- Vegetable roots and tubers
- Fruit, edible — bananas, (including plantains), fresh or dried
- Cocoa beans — whole or broken, raw or roasted
- T-shirts, singlets and other vests
- Cereals — rice in the husk (paddy or rough)
Cassava, bananas, raw cocoa beans, cotton T-shirts, unmilled rice. These are real opportunities and they are worth having. None of them teaches a country anything it does not already know, which is the whole reason a government reaches for downstream industry in the first place. and the shape is unmistakable: the corner where a product is both within reach and worth learning holds 2 of 400.
"Sells competitively" is doing precise work here. A product counts as one Indonesia already sells if it makes up at least as large a share of Indonesia's exports as it does of all the world's exports put together. So it is a measure of specialisation, not of volume: Indonesia imports and re-sells plenty of things that never qualify, and a product it ships a little of but everyone else ships a lot of counts as one it does not have.
Indonesia has already run this experiment once
In 2013–2015 Indonesia was shipping nickel ore. Then it stopped: by 2015 ore shipments were under a hundredth of their earlier level, and smelters at home turned the ore into ferro-nickel, nickel mattes and stainless steel instead. It is the textbook version of moving up, and it is the most studied thing in this case.
The dollar figure went up 14.6 times over. The tonnage went up 34.6 times over. Both are true, and only the first one is ever quoted. The whole ledger, as this case totals it: $26.5bn a year more processed exports, $783m a year of ore sales given up, and an extra $3.6bn a year of imported inputs for the smelters — coal, coke, chromium ore, lime. So 13.6% of the gain goes straight back out on things bought abroad, and the trade Indonesia gave up to get it was the small side of the bargain.
And Indonesia's overall standing barely moved. On this case's own reckoning it sat 76th of 156 countries a decade ago and 78th of 156 now. Over the same decade Vietnam moved up 12 places, India 7 and the Philippines 3.
Does this match the official ranking?
Closely. Rebuilt from the raw trade data, this case's ranking agrees with the Growth Lab's published one at 0.85 on a 0-to-1 scale, and lands on average 2.4 places away from it. For a reconstruction from scratch that is a clean replication, and it is the weakest kind of good news available.
Matching a published number proves the arithmetic, not the meaning. Two figures from this case say why. Indonesia's rank moved 2 places across a decade — while the typical year-to-year move is 4.3 places and the largest is 14. The decade-long story is smaller than the annual wobble around it.
And in this part of the table the ranking is packed tight: moving ten places is worth about 0.09 of an index point. The measure was designed to separate the top of the world from the bottom, not to order the countries in the middle — and Indonesia is in the middle. So a rank that replicates to two places is still a rank that should not be read as a score.
What this cannot tell you
The list is real. Almost everything a reader would want to do with it is not supported, and the case published three checks that say so — it set the pass marks before it computed the answers, and two of the three failed.
| Check fixed in advance | Result | What it came out at |
|---|---|---|
| Do the export totals match Indonesia's own published figure? | failed | 17.9% above, against a 5% tolerance |
| Do the import totals match? | passed | −0.87% off, against a 10% tolerance |
| Did building the smelters show up as machinery imports? | failed | machinery up 29.4% while all imports rose 27.1% — invisible |
- The dollar totals are not Indonesia's own. The source used here is reconciled from what every country says it bought, and for 2023 it puts Indonesia's goods exports 17.9% above the figure Indonesia itself publishes — against a tolerance of 5% set before anyone looked. Imports reconcile almost exactly (−0.87%, inside a 10% tolerance), so the mismatch is on the export side. Shares and structure survive it, because they are ratios and a roughly proportional gap cancels; the levels do not.
- The position in the table is a band, not a place. Around Indonesia's 78th place, 11 economies sit close enough together that this method cannot order them. The published series jumps an average of 4.3 places a year on its own, and the whole decade's change is 1.9% of one standard deviation. Anyone quoting a single place is quoting noise.
- It does not agree with the published benchmark on level. Against Harvard's Growth Lab rankings, on an identical list of countries, this pipeline puts Indonesia an average of 11.1 places lower — in 30 of 30 years, so it is a systematic offset and not noise. What does agree is the order: between 0.95 and 0.99 on a scale where 1 would mean the two rankings sort every country identically. The machinery is sound; its zero point is not.
- Building the smelters is invisible in the machinery bill. The case predicted that a smelter build-out would show up as a surge of imported equipment. It does not: machinery imports rose 29.4% while all imports rose 27.1%. What did surge was the feedstock — chromium ore up 12,348%, coal up 1,325% in a country that is the world's largest coal exporter — while scrap steel, which the case expected to rise, fell 23%. That check is recorded as a failure and the prediction behind it as wrong.
- Goods only. The source has no services in it at all. Software, tourism, shipping, business services — none of it is on the plot, and for a country wondering what to do next that is a large hole.
- The plot is a shortlist, not the whole field. It holds the 400 nearest of 976. The ones left off are further away still, so leaving them off cannot rescue the empty corner — but the horizontal axis is truncated, and it is not a wide axis: the closest of the 400 has only about 1.7 times as much of the surrounding know-how already in place as the 400th does. Read a product's horizontal position as an ordering, not as a distance.
What a government or an NGO can actually do with this
Three things follow, and the first is uncomfortable.
- Read the shortlist as a description, not a strategy. Indonesia sat at rank 76 of 156 on this measure a decade ago and rank 78 of 156 now. If the products nearest to what it already exports are agricultural staples and basic garments, that is a finding about the current basket rather than a recommendation to pursue them. The method answers "what is close", and close is not the same as worth doing.
- Use the distance, not the ranking. The rank itself wanders about ±1.9 places a year with nothing behind the movement, so a place in a league table is mostly noise. The useful number is how far a desired product sits from existing capability, because that prices what has to be built to get there — training, suppliers, logistics, standards. A ranking hides that; a distance costs it.
- Argue with the objective before using the output. Relatedness optimises for feasibility, and feasibility is not value. Nickel is the worked example on this page: $1.8bn of ore against $32.5bn of processed exports at the top of the series. A policy that wants value added has to accept longer jumps and say so openly, rather than borrowing the authority of a model that was optimising for something else.
What you must not conclude. This is not an investment list. A product being close says the surrounding know-how exists; it says nothing about whether anyone would make money, whether the electricity or the workforce or the port capacity is there, what it would cost, what it would emit, or whether world demand is growing or shrinking. Nor does the empty corner mean Indonesia cannot make hard things — Vietnam moved 12 places in a decade doing exactly that, and this method would not have predicted it, because a country that decides to jump is not constrained to what is next door. Treat the plot as a map of what is cheap to try, not as a verdict on what is possible.
How many things does Indonesia sell, and is that number moving?
One more figure worth sitting with. The count of products Indonesia sells competitively was 257 in 1995, peaked at 304 in 2000, and is 241 today. Over the same span its goods exports went from $53.5bn a year to $302.6bn. So the country sells far more, and it does not sell a wider range of things than it did at the start. The nickel move was real and it changed what is in the basket; it did not change how many things are in it.
Could this run in daily operations?
FEASIBILITY ONLY — none
Every pipeline here is a batch backfill: it fetches a season or an archive, not a live feed. Anything operational means rebuilding the ingest for near-real-time arrival, whatever the verdict below says.
What stops it. The export reconciliation against UN Comtrade failed, and so did the nickel capital-goods test.
What it would need. Reconciliation inside tolerance first. Trade statistics are revised for years after publication, so anything operational also needs a revision policy.
If you want the detail
On the technical article the vertical measure is called product complexity and the horizontal one density. They mean what the axes on the plot say they mean: how much know-how a product embodies, judged by which countries manage to export it at all, and how much of that know-how a country already has to hand. The same page carries every check, the failed ones included, and the corrections made to this case's own earlier claims. You can turn the network in three dimensions yourself on the dashboard. Every figure above is read from one record, so the two versions cannot drift apart.
Words on this page, in plain language 2 terms
- a look
- One pass of the satellite over a field. More looks means the crop's cycle is sampled more often; too few and a short stage — flooding, heading — can happen entirely between two looks and never be seen.
- standard deviation
- The usual distance of values from their average — a measure of spread rather than of level. Two rankings with the same mean can differ enormously in how far apart the places are.
Data vintage 2024; 30 years from 1995. Goods trade only. Written to be read without a background in economics; nothing was rounded to make a point, and the two failed checks are on the page rather than in a footnote.
Found something wrong on this page? Report a correction — it opens a pre-filled issue — or email taufik.adi@openstudy.id. Corrections are credited by name in the errata, and one that changes a finding says so on the page.