Most forest carbon credits did some real conservation work. They just claimed far more than they delivered. A new Nature Communications paper looked at six independent evaluations covering 44 first-generation REDD+ projects. The result is pretty brutal.... Most projects did reduce deforestation. But across the portfolio, they claimed 10.7 times more avoided deforestation than independent estimates justify. That means the core problem wasn’t that these projects did nothing. It’s that the credit numbers were massively inflated. The usual defence has been that independent researchers used coarse global forest-loss data, while project developers used better bespoke maps. The paper, though, actually finds the opposite. In project areas, the global ACC dataset often detected similar or even higher deforestation than the certified project estimates. So the gap can’t be explained away as a remote sensing artefact. The bigger problem was the counterfactual. REDD+ credits were issued by comparing observed forest loss in project areas against a modelled baseline of what would have happened otherwise. That sounds reasonable until you see how those baselines were built. Projects were allowed to use expert-selected “reference areas” and flexible ex ante modelling approaches. Both choices created room for inflation. Across 17 projects where the authors could reconstruct the reference areas, those reference areas were systematically more exposed to deforestation pressures than the projects themselves. • They were more accessible. • They were less forested. • They often looked like places more likely to be cut down anyway. So when project outcomes were compared against those reference areas, the project appeared to be avoiding more deforestation than it really was. The paper estimates that reference area choice explains about 57% of the over-crediting. Then comes the second problem. Even after accounting for reference area bias and remote sensing differences, the authors attribute most of the remaining gap to ex ante modelling. In plain English, the future was modelled far too aggressively. Within the 17-project subset, certified assessments reported median avoided deforestation of 0.95% per year. The quasi-experimental estimate was 0.18%. Swap in the certified project-side deforestation rates and it barely changes, rising to 0.20%. Swap in the certified reference areas and it jumps to 0.64%. That tells you where the inflation was coming from: from baseline design. The most important part of this paper is that it doesn’t argue all these projects were useless. It argues many did create real conservation benefits. But climate claims depend on quantity, not intention. If one credit is supposed to represent one unit of avoided emissions, then overstating avoided deforestation by a factor of ten breaks the system. Basically, this means carbon markets can’t rely on project-selected reference areas plus flexible ex ante models and still claim high integrity. The authors argue for ex post crediting against credible counterfactuals using quasi-experimental methods. That would mean fewer credits, but more honest ones... Link to paper: nature.com/articles/s41467-0…