How Were India's Carbon Market Targets Set?

The claim being tested

India's Carbon Credit Trading Scheme (CCTS) assigns each obligated industrial plant a target: a ceiling on how much carbon it may emit per unit of what it produces. The Bureau of Energy Efficiency has said, in general terms, that these targets are set relative to each sub-sector's best performer the cleanest plant becomes the reference point, and every other plant is pulled toward it.

That claim makes several distinct predictions, not just one. It predicts that the best-performing plant should itself face little or no cut. It predicts that plants further from the frontier should face proportionally larger cuts, applied on average across the whole group. And it implies the targets were genuinely calculated for each plant's specific position, not assigned from a short list of round numbers.

We tested all three predictions separately, because they can and do pull apart.

Three ways of measuring the same claim

1. Was the frontier plant spared?

In each sub-sector we identified the single plant with the lowest baseline emission intensity ie the frontier. We measured the cut it was actually given, and expressed it as a ratio against its sub-sector's average cut. A ratio near 0 means the cleanest plant paid almost nothing, as the stated rule predicts. A ratio near 1 means it was cut by roughly the same percentage as everyone else, which looks more like a flat, uniform reduction than benchmarking.

2. How much pressure fell on the typical plant?

The frontier ratio only tells you about one plant. A second measure, which we call lambda, asks a broader question: across every plant in a sub-sector, what fraction of the distance to the frontier does the average target close? A sub-sector can, in principle, apply real pressure to its typical plant (a high lambda) while still not sparing its single best performer (a high frontier ratio) - these are genuinely different facts, and several sub-sectors show exactly this split.

3. Was each target calculated, or inspired from a short list?

The number of distinct cut percentages within a sub-sector, and how much they vary, tells you something about process rather than outcome. If 46 plants share only 5 distinct cut rates with almost no variation between them, that sub-sector was likely not solving an equation for each plant but it was applying a small number of administrative rates. If 75 plants each carry their own unique cut percentage, real plant-level calculation was clearly happening, whatever its result.

Put together, these three measures separate what happened to the best plant, how hard the average plant was pushed, and whether the number was calculated or copied. A sub-sector's position on all three tells a fuller story than any one alone.

Four patterns emerge

Genuine, calculated benchmarking. Six sub-sectors combine a near-zero frontier ratio with near-total individualization - almost every plant carries its own distinct rate, and the cleanest plant is the one exception, paying little or nothing. Chlor-alkali (30 plants, all 30 with distinct rates) exempted its frontier entirely against a 6.1% group average. Cement grinding units (57 plants, 51 distinct rates) did the same against a 4.8% average. Paper's RCF-based mills (28 plants, 27 distinct rates), and textile's spinning (60 plants, 59 distinct rates) and processing (34 plants, all 34 distinct) sub-sectors show the identical signature: individually calculated targets that specifically protect the best performer. This is the cleanest evidence in the whole dataset that the stated rule was actually applied as described.

Calculated, but not generous to the leader. A second group is just as individualized -plant-level calculation, not flat rates -but does not spare its frontier plant. Cement's two largest sub-sectors sit here: Ordinary Portland Cement (45 plants, 44 distinct rates) cut its cleanest plant to 65% of the group average; Portland Pozzolana Cement, the single largest sub-sector in the entire scheme at 75 plants, gave every one of its 75 plants a unique rate, yet still cut its frontier plant to 43% of average. Paper's integrated plants (19 plants, all 19 distinct) show the same pattern, at 42%. What makes this group interesting is that their average pressure, measured by lambda, is not weak -PPC's lambda of 0.120 is among the highest in the entire dataset, higher than several of the "exempted" sub-sectors above. So these sub-sectors push real, calculated pressure across their whole plant population; they simply do not carve out the special exception for the very best performer that the stated rule implies.

Flat administrative rates. A third group shows the opposite fingerprint: high frontier ratios paired with almost no individualization at all. Textile composite mills (46 plants) share just 5 distinct cut rates, varying by only hundredths of a percentage point, and cut their cleanest plant to 99% of the group average -despite that plant emitting a small fraction of what its dirtiest neighbours do. Textile fibre (33 plants, 12 distinct rates) shows the same thing, with the frontier plant cut slightly more than average. Petrochemicals (11 plants, 8 distinct rates, essentially zero variation) applied what amounts to a flat 5% to nearly everyone. These are not calculations that happened to produce uniform results -the near-total absence of distinct rates indicates a small number of pre-set percentages applied regardless of where a plant actually stood.

Individualized, but the weakest pressure of all -iron and steel. The scheme's largest sector by both plant count (255) and emissions does not fit cleanly into either calculated group. It is fully individualized -247 of 255 plants carry distinct cut rates, ruling out a flat administrative rate. But its frontier plant was still cut to 81% of the group average, and its lambda of 0.063 is the lowest of any sub-sector with a comparable number of plants -lower than every group above, including the flat-rate ones. In plain terms: steel's targets were genuinely calculated per plant, but on average they pull the typical plant only about 6% of the way toward its cleanest peer, and even that peer was not meaningfully spared.

This figure needs one important caveat. Steel carries no further breakdown by production route in the available data, despite spanning a nearly 70-fold range of baseline intensities -a spread almost certainly driven by fundamentally different processes (integrated blast-furnace steelmaking versus scrap-based mini-mills) being measured on one common scale. A single "frontier plant" drawn from this pooled range may not be a meaningful benchmark for the rest of the sector at all. Steel's low lambda may reflect weak convergence pressure, or it may reflect the fact that no single frontier can meaningfully anchor a sector this technologically varied. Both readings point to the same conclusion: steel's targets cannot be fully understood without knowing which plants share a production route, and that information is not publicly available.

This figure also required a correction along the way. An earlier version of this analysis showed steel's frontier plant being cut harder than the sector average -the opposite of any coherent rule, and itself a strong signal of an error upstream. That turned out to be a mismatch between an updated 2026-27 target and an outdated 2023 baseline for 41 plants. Once corrected, steel settled into the pattern described above.

A quick reference

(Aluminium refining and secondary aluminium, cement clinkerization, composite and white cement, and paper's agro-based and speciality plants are excluded here: each has fewer than eight plants, too few to draw a reliable conclusion.)

The overall picture

No single method describes how India's carbon market targets were set. Roughly a quarter of sub-sectors show individually calculated targets that genuinely spare their best performer, consistent with the stated rule. A similar-sized group shows individually calculated targets that apply real average pressure but do not meaningfully protect the frontier. A third group shows almost no individualization at all -a small number of flat rates applied regardless of a plant's starting position, most visibly across textile's two largest sub-sectors. And the scheme's single largest sector, iron and steel, is fully individualized yet applies the weakest average convergence pressure of any comparably sized group -a result that may reflect genuinely weak benchmarking, or may simply reflect that steel's plants are too technologically different from one another for a single frontier to mean anything.

What the three measures make visible, taken together, is that the gap between the CCTS's stated methodology and its realised targets is not a single, uniform gap. It varies by sub-sector, along at least two independent dimensions -how much the best performer was protected, and how much of the calculation was genuinely individualized versus administratively flat -and the scheme's public notifications do not explain why.

That so many sub-sectors — chlor-alkali, cement grinding, textile spinning and processing among them — show fully individualized, plant-by-plant calculation is itself worth noting: the analytical rigour to do this well clearly already exists inside the scheme. Extending that same rigour, and explaining it publicly, to every sub-sector would let India's carbon market claim a level of transparency few first-generation compliance schemes anywhere have managed this early.