Data • does EU policy work • case 1

The ban that moved the barrels

The EU's RED II measure phases palm oil out of European biofuel by 2030[1], a step often called a palm-oil ban though it touches only fuel, not the palm in food or chemicals. The intent was to take the deforestation out of Europe's energy. This page scores the measure against what physically happened, which turns out to be a different thing: European palm-biofuel demand fell by about half, but roughly 98 percent of that displaced to other markets rather than disappearing, and the deforestation it targeted followed its own drivers. This is the first of a series scoring EU environmental measures against the measured reality, with CBAM and the deforestation regulation to follow.

How to read this page: ● measured sourced data · ◐ inferred analyst reading, basis linked · ○ projected anchored to a real starting point. Bracketed citations link to the sources at the foot of the page.

EU palm biofuel

−52%[6]

4.5 to 2.17 Mt, 2019 to 2023

Displaced, not cut

~98%[7]

of the withdrawn demand reappeared elsewhere

Deforestation avoided

~1.6%[12]

the most a full ban could achieve

Indonesian production

still rising[9]

45 to 47.5 Mt across the phase-out

The claim, scored

Four links from intent to outcome, each scored against the evidence.

The EU removed its palm-biofuel demand.

● True

EU palm burned for fuel fell steeply from its 4.5 Mt peak in 2019: down 27% between 2021 and 2022, and from 2023 palm biofuel no longer counts toward EU renewable mandates. Total EU palm imports fell about 46% (7.1 to 3.8 Mt, 2019 to 2023).[6,5] ● measured

Palm production fell as a result.

● False

A 75% cut in EU palm imports is modelled to produce only about 1.8% less production in Indonesia and Malaysia: most of the withdrawn demand is displaced rather than eliminated. Output stayed broadly resilient (about 45 to 47.5 Mt), with year-to-year dips.[7,9] ◐ inferred

Deforestation fell because of the EU measure.

● False

Even a full EU ban is modeled to avoid only about 1.6% of Indonesian deforestation; an EU-only freeze cuts global land-use-change CO2 by under 1%. The actual decline began in 2017, before RED II existed, and tracks the palm price crash and the moratorium.[12,13,14] ◐ inferred

The displaced palm quietly returns to the EU.

● Plausible, flagged

Palm was replaced in EU biodiesel by used cooking oil, much imported from China (+39% in 2024); some of that 'UCO' may be mislabelled virgin palm, which is why the EU imposed anti-dumping duties on Chinese biodiesel in 2024. The HS 1518 monitor below is the live watch.[23,24] ◐ inferred

The waterbed

Push palm down in one market, it rises in another. ● measured

The palm the EU stopped buying did not stop being grown. Indonesia's domestic biodiesel mandate ramped from B20 to B40 over exactly the phase-out window, and the palm it burns rose from 3.2 to 13.3 million tonnes a year, more than three times the entire EU biofuel appetite the measure removed[8]. Production kept climbing, and exports rerouted to China, Africa and India, each of which absorbed more than the EU gave up[10]. A general-equilibrium model puts a precise figure on it: a 75 percent cut in EU palm imports yields only a 1.8 percent fall in production, roughly 98 percent displaced[7].

EU demand down, domestic demand up (Mt/yr)

EU palm for biofuel (2023 approx.)2019 to 20234.5 Mt2.2 Mt ▼Indonesian domestic biodiesel palm uptake2016 to 20253.2 Mt13.3 Mt ▲Indonesian palm production2019 to 202645 Mt47.5 Mt ▲

Export shift, 2022 to 2023

EU-11.6%China+23%Africa+33%India+8%

■ EU · ■ other markets. Indonesian palm exports by destination[10].

The forest followed its own clock

Indonesian primary forest loss, and why the timing rules the EU out. ● measured with the reading ◐ inferred

Deforestation did fall, steeply, from about 930,000 hectares in 2016 to a 20-year low around 2021[14]. But the fall began in 2017, a full year before RED II was even adopted, and it tracks the palm price crash and Indonesia's forest moratorium, not European demand. The modeling agrees: even an outright EU ban would avoid only about 1.6 percent of Indonesian deforestation, because half the restricted supply leaks to other buyers and palm is only a third of the clearing[12,13]. Then it rebounded, up 66 percent in 2025 to an 8-year high, driven by the domestic biodiesel push, food estates and nickel mining[16].

0250500750100020162017201920212022202320242025RED II adopted (Dec 2018)the decline started a year earlierkha primary forest loss (GFW/UMD)

Even the reality has competing official versions: for 2024 the satellite record (GFW/UMD) shows loss falling 11 percent while Indonesia's own SIMONTANA system shows it rising 61 percent, because the two count differently and use different years[15]. This page cites the satellite series and names it; the disagreement is itself part of the story.

The world's forests on a globe, three ways ● measured

Layer·Zoom to
Forest (evergreen / broadleaf)
Forest (needleleaf / mixed, boreal)
Woody savanna / shrub
Savanna / grassland
Cropland
Barren
Snow & ice
Water
every forest class in green, boreal Finland included; MODIS IGBP land cover 500 m, 2024

Loading satellite forest layer…

Three satellite layers, because forest cover, forest condition, and forest loss are three different questions. Forest cover is where forest simply is, every forest on Earth in green including boreal Finland, from ESA WorldCover[35]. Tropical condition is the European Commission's Tropical Moist Forest monitoring, humid tropics only, in the literature's classes: undisturbed, degraded, deforested (plantations included), and regrowth[26]. Global loss is the Hansen record, whole planet, red by year of clearing[15]. The green layers sit on a light globe; the red loss flips it to dark. One honesty the three layers make plain: regrowth to new forest is a small, slow signal next to loss, so the world regrows far less than it clears, which is the next panel's point.

Regrowth is not the forest that was cleared ● measured

The green regrowth on the map is not the dark green of undisturbed forest, and the difference is the reason replanting does not offset clearing. A young secondary forest is a fast annual carbon sink, taking up carbon about eleven times faster per hectare than mature forest, which is why growing forests, not old ones, are the net oxygen source[27,31]. But it holds only a fraction of the stock that clearing released, and the functions that make old forest irreplaceable, its species composition, its soil and peat carbon, recover over more than a century, if at all[28,29].

FunctionAfter 20 yearsTo ~90% of old-growth
Annual carbon uptake (the sink rate) — ~3 tonnes C per hectare a year, about 11 times the net rate of mature forest[27,30]already faster than old-growthflux, not a stock
Above-ground carbon stock (what clearing releases) — 11-fold variation with rainfall; drier and repeatedly cleared sites far slower[27]roughly a third to a half~66 years (median)
Canopy structure[28]partial25 to 60 years
Species richness (how many species)[29]~80%~50 years
Species composition (which species)[29,28]~34%over 120 years, centuries for old-growth specialists
Soil and peat carbon[34,33]little restored; drained peat still emittingdecades to centuries; peat may be irreversible
Aerosol and cloud chemistry (BVOCs)[32]different, not quantifiednot established at stand scale
Net oxygen — the 'lungs of the Earth' line is a myth for mature forest[31]tracks carbon uptakemature forest is net-zero at equilibrium

Boreal forest inverts the picture: 80 to 95 percent of its carbon is in soil and peat, not trees, so recovery is governed by the slowest pool. Burned or cleared boreal stands are net carbon sources for years, and permafrost and peat loss can be effectively irreversible on any management timescale.[33]

Recovery times are medians to 90 percent of old-growth values from the tropical-forest recovery literature; site variation is large (biomass recovery spans an eleven-fold range with rainfall). Two honesties the table keeps: the fast soil recovery is for physical and nutrient attributes under light land use, not soil-carbon stock after intensive clearing or on peat; and the aerosol chemistry is a real effect of forest age and species that is not yet quantified at whole-stand scale, so it is marked unquantified rather than given a false number[28,32].

The carrot and the stick

If the trade ban was displaced, did the market channel do better? ◐ inferred

The import ban was the stick. Running alongside it was a carrot: schemes that gave producers a financial reason to comply with better practice. The honest read is that the carrot, specifically the buyer-driven no-deforestation commitments, is the more plausible driver of the actual decline than the trade measure, with the clearest evidence being that clearing stayed at a 20-year low in 2021 even as palm prices spiked, breaking the usual link. But certification badges as such show little net effect once leakage is counted, and both channels miss the roughly 40 percent of production that comes from untraceable smallholders.

RSPO certification

weak evidence

Voluntary certified-sustainable-palm premium market, founded 2004; about 19-20% of global production certified.

Cut deforestation about 33% inside certified plantations, but those held under 1% of remaining forest (selection bias), and the pooled net effect is statistically insignificant after leakage. The premium is tiny and often below the cost of certifying.[17,18,19]

NDPE buyer commitments

stronger evidence

No Deforestation No Peat No Exploitation policies, covering 83% of Indonesian and Malaysian refining capacity by 2020, driven by Western consumer-goods buyers and lenders.

The most plausible driver of the actual decline: large-scale corporate clearing fell, and 2021 deforestation stayed at a 20-year low even as palm prices hit their highest since 2010, breaking the usual price-drives-clearing link. But it leaks to non-NDPE mills and the ~40% independent-smallholder segment, and has plateaued.[21,22,14]

RED's low-ILUC exemption

loophole

The regulation's own carrot: palm certified 'low indirect-land-use-change risk' is exempt from the phase-out and keeps EU market access.

The stick quietly re-created a carrot. Interpreted loosely, about 8 Mt/yr could qualify by 2030 with no real ILUC reduction, roughly double the palm used in EU biofuel in 2017, which the ICCT flags as a loophole.[20,1]

The leakage watch, live

China's share of EU used-cooking-oil imports, monthly, from this site's HS 1518 monitor. ● measured

What replaced palm in EU biodiesel is used cooking oil, and much of it now comes from China[23]. The catch: some of that "UCO" may be mislabelled virgin palm, which would mean part of the phased-out palm is quietly returning to the EU through a category that is harder to verify, the reason the EU imposed anti-dumping duties on Chinese biodiesel in 2024. This panel is the live watch: China ran 27 percent of EU HS 1518 imports by weight in 2026-02, against 25 percent in 2021-07[24].

0%20%40%60%2021-072022-012022-072023-012023-072024-012024-072025-012025-072026-01

China share of EU HS 1518 (used cooking oil and residues) imports by weight, monthly to 2026-02[24]. Rising Chinese UCO is not proof of fraud, but it is the channel to watch, and it is already litigated.

What this is, and is not

This is not an argument that the EU should do nothing, nor that palm oil is harmless. Naively replacing palm with rapeseed or soy can be worse, because they yield far less oil per hectare, so a blunt ban can move emissions the wrong way. The point is narrower and it is measured: as a deforestation instrument, an import restriction that a producer can neutralize by redirecting trade does very little, and the honest arithmetic says so. Policy that bites has to reach the producer, through conservation, finance and buyer pressure, not just the customs line. That is the general lesson, and the next two cases test it: does CBAM cut emissions or just reshuffle who ships the carbon, and does the deforestation regulation cut clearing or reroute the dirty supply to non-EU buyers.

Every figure here carries its source and its tier. The model outputs (the 98 percent, the 1.6 percent) are projections from peer-reviewed equilibrium models, labelled as such; the trade and deforestation series are measured, with the data system named. Where sources disagree, the disagreement is shown, not resolved.

Sources and method (35)
  1. [1] Directive (EU) 2018/2001 (RED II), Article 26, and Delegated Regulation (EU) 2019/807 (13 March 2019): palm oil is the only feedstock classified high indirect-land-use-change risk; biofuel use frozen at 2019 levels then phased to 0% by 2030. Biofuel-only scope; food/feed/oleochemical palm is untouched
  2. [2] Directive (EU) 2023/2413 (RED III): raised the renewables target but left the palm high-ILUC phase-out unchanged; Parliament's push for an immediate palm/soy phase-out was rejected
  3. [3] WTO DS593 (Indonesia v EU): panel report circulated 10 Jan 2025, adopted 24 Feb 2025; found the EU failed to properly review the ILUC data and faulted the low-ILUC certification design. Escalated to Article 22.6 retaliation arbitration, March 2026
  4. [4] WTO DS600 (Malaysia v EU): panel report circulated 5 March 2024, adopted 26 April 2024; partial win, the framework upheld but specific design/implementation faulted
  5. [5] USDA FAS (via IndexMundi), EU-27 palm oil imports: 7.1 Mt (2019 peak) to 3.8 Mt (2023), about -46%
  6. [6] Transport & Environment and USDA FAS: EU palm oil used for biofuel, 4.5 Mt (2019) to 2.17 Mt (2023), about -52%; in 2018 about two-thirds of EU palm imports were burned for energy
  7. [7] Heimann et al. (2024), 'Phasing out palm and soy oil biodiesel in the EU', GCB Bioenergy 16:e13115 (Kiel Institute, DART-BIO general-equilibrium model): a ~75% cut in EU palm imports from Malaysia/Indonesia produces only ~1.8% less palm production there, ~98% displaced, with vegetable-oil exports to other Asian countries rising ~12%
  8. [8] USDA FAS GAIN, Indonesia Biofuels Annual (ID2025-0029): crude palm oil absorbed by the domestic biodiesel mandate rose from 3.2 Mt (2016) to 13.3 Mt (2025), a +10 Mt/yr increase; mandate ramp B20 (2018) to B30 (2020) to B35 (2023) to B40 (2025)
  9. [9] USDA PSD Online: Indonesian palm oil production rose across the EU phase-out window (about 45 to 47.5 Mt), world production 76.7 to 81.4 Mt; it did not fall
  10. [10] GAPKI (Indonesian Palm Oil Association), exports by destination 2022 to 2023: EU -11.6%, China +23%, Africa +33%, India +8%; each single-market gain exceeded the EU decline
  11. [11] Trase / Stockholm Environment Institute: Indonesian domestic palm use rose from 32% of production (2018) to 44% (2022), driven by the biodiesel and oleochemical push
  12. [12] Busch et al. (2022), Environmental Research Letters 17:014035: a modeled EU restriction on high-deforestation palm would avoid only ~1.6% of Indonesian deforestation, because ~52% of restricted imports shift to non-participating markets, supply elasticity is 0.13, and palm is only ~32% of deforestation
  13. [13] Yarlagadda et al. (2025), Nature Communications 16:1520 (GCAM model): an EU-only palm freeze cuts global land-use-change CO2 by a negligible amount; even extending to the top five rich importers cuts it by ~0.8%
  14. [14] Gaveau et al. (2022), PLOS ONE: Indonesian primary forest loss fell from ~930,000 ha (2016) to ~203,000 ha (2021); a 1% palm-price fall is associated with 0.68% less forest loss. Forest-to-industrial-palm conversion fell from ~224,000 ha (2012) to ~15,500-22,000 ha (2021), a ~90% drop
  15. [15] Global Forest Watch / University of Maryland (Hansen): Indonesian primary forest loss by year; for 2024 satellite data shows -11% while Indonesia's official SIMONTANA shows +61%, the two systems pointing in opposite directions
  16. [16] Auriga Nusantara / TheTreeMap via Mongabay: Indonesian deforestation surged 66% in 2025 to an 8-year high, tied to biodiesel expansion, the food-estate programme and nickel mining
  17. [17] Roundtable on Sustainable Palm Oil (RSPO, founded 2004): about 19-20% of global palm production is certified; only ~64% of certified oil is actually sold as certified, and the Book & Claim premium is small (about USD 2.50-3.50/t, since fallen), often below the cost of certification
  18. [18] Carlson et al. (2018), PNAS 115(1):121-126: RSPO certification cut deforestation ~33% inside certified plantations, but with strong selection bias, certified areas held less than 1% of the forest remaining inside Indonesian oil-palm plantations, and there was no effect on peatlands or fire
  19. [19] Sibhatu & Qaim (2026), Environmental Research Letters, systematic review and meta-analysis: the within-concession deforestation cut is largely offset by leakage to non-certified land, so the pooled net effect of RSPO certification on deforestation is statistically insignificant
  20. [20] ICCT: RED's 'low-ILUC-risk' exemption lets certified palm keep EU market access; interpreted loosely, about 8 Mt/yr could qualify by 2030 with no actual ILUC reduction, roughly double the ~4 Mt used in EU biofuel in 2017
  21. [21] Chain Reaction Research: No Deforestation No Peat No Exploitation (NDPE) policies covered 83% of Indonesian and Malaysian palm refining capacity by April 2020 (up from 74% in 2017), implementation about 78%; driven by Western consumer-goods buyers and their lenders
  22. [22] Nature Sustainability (2026), 1,541-household survey: zero-deforestation commitments reach some tied smallholders but awareness is low, there is almost no price differentiation, and independent smallholders (about 40% of production) are largely bypassed
  23. [23] Transport & Environment: used cooking oil (much imported from China, EU purchases +39% in 2024) is the main palm substitute in EU biodiesel; some 'UCO'/POME may be mislabelled virgin palm, and the EU imposed anti-dumping duties on Chinese biodiesel in August 2024
  24. [24] A1AYN, HS 1518 EU import monitor (UN Comtrade snapshot): monthly EU imports of HS 1518 (used cooking oil and residues) by partner, the live leakage-watch series
  25. [25] A1AYN, vegetable-oil and biofuel monitor: palm and substitute-oil prices, production and blending mandates
  26. [26] Vancutsem et al. (2021), 'Long-term monitoring of forest cover changes in the humid tropics', Science Advances 7:eabe1603; JRC Tropical Moist Forest product (© European Union), the 30 m classification of undisturbed, degraded, deforested and regrowth forest shown on the map
  27. [27] Poorter et al. (2016), 'Biomass resilience of Neotropical secondary forests', Nature 530:211-214: median 66 years to recover 90% of old-growth above-ground biomass; young secondary forest takes up ~3.05 tC/ha/yr, about 11 times the net rate of old-growth; 11-fold site variation driven by rainfall
  28. [28] Poorter et al. (2021), 'Multidimensional tropical forest recovery', Science 374:1370-1376: soil attributes recover in under a decade, structure and richness in 2.5-6 decades, but biomass and species composition take over 120 years; recovery is three largely independent processes
  29. [29] Rozendaal et al. (2019), 'Biodiversity recovery of Neotropical secondary forests', Science Advances 5:eaau3114: species richness recovers to ~80% in 20 years (~5 decades to full), but species composition only ~34% in 20 years and takes centuries; secondary forest does not substitute for old-growth
  30. [30] Heinrich et al. (2021), Nature Communications 12:1785: Amazon young secondary forests accumulate ~3.0 tC/ha/yr (western) vs 1.3 (eastern); disturbance cuts the regrowth rate 8-55%; per-hectare flux is high but total landscape drawdown is small
  31. [31] Y. Malhi (Oxford) and National Geographic: a mature forest at equilibrium adds essentially zero net atmospheric oxygen (photosynthesis balanced by respiration and decomposition); net oxygen tracks net carbon uptake, so the 'forests produce 20% of the world's oxygen / lungs of the Earth' framing is misleading for mature forest
  32. [32] Biogenic-VOC and secondary-organic-aerosol literature (Nat. Commun. 2025 s41467-025-59953-2; ACP 2026): forest age and species composition demonstrably change isoprene and monoterpene emissions and aerosol formation, but the evidence is leaf- and species-level, and a clean secondary-versus-old-growth stand-level aerosol figure is not established
  33. [33] Boreal carbon: roughly 80-95% of boreal carbon is in soil and peat, not living biomass; post-fire stands are net carbon sources for years, soil respiration can show no recovery even after several years, and permafrost/peat loss may be effectively irreversible (Forests 2021 10.3390/f12070880; GCB 2021 10.1111/gcb.15721; Sci. Rep. 2021 s41598-021-87343-3)
  34. [34] Tropical soil and peat carbon: more than half of tropical forest carbon is below ground; secondary-forest age does not reliably rebuild soil carbon, and drained tropical peat keeps subsiding and emitting under secondary-forest and fern regrowth (Nature 2022 s41586-022-05679-w; Catena 2021 S0341816221005774)
  35. [35] ESA WorldCover 2021, 10 m global land cover (CC BY 4.0, contains modified Copernicus Sentinel data); the forest-cover layer that shows every forest on Earth, including boreal Finland, in green

Verified 2026-07-04 across seven research threads. This is a curated, fully-cited module: the annual trade, production and deforestation series are small and each figure carries its citation; the peer-reviewed model outputs are tagged as projections, not observations. The one live series is the HS 1518 used-cooking-oil leakage panel, read from this site's UN Comtrade snapshot. Advocacy and industry sources are named where used; the load-bearing quantitative claims (Heimann 2024, Busch 2022, Yarlagadda 2025, Carlson 2018, Sibhatu & Qaim 2026) are peer-reviewed. Deforestation figures name their data system (GFW/UMD satellite vs Indonesia's SIMONTANA), which disagree by year and direction.

← All dataVegetable oil monitorFeedstock trade (HS 1518)CBAM (case 2, coming)