Fewer Credits, More Credibility: jurisdictional baselines in REDD+ provide more conservative, and appropriate, estimates of avoided deforestation.

Abstract

Reducing emissions from deforestation and forest degradation (REDD+) projects can
mobilise crucial finance for tropical forest conservation through the sale of carbon credits.
The first generation of REDD+ methodologies used historical deforestation rates from
reference areas to predict deforestation in the absence of the REDD+ intervention,
establishing the project’s baseline scenario ex-ante. Carbon credits were issued based on
the difference between the baseline scenario and the deforestation observed in the project
area. However, independent academic studies using ex-post quasi-experimental methods
reported that many REDD+ projects had used baseline scenarios that overestimated
business-as-usual deforestation, leading to substantial over-issuance of carbon credits. The
research contributed to a scandal which wiped $1bn off the value of Voluntary Carbon
Markets. Verra, the world’s largest accreditor of carbon credits, has since launched VM0048,
a REDD+ methodology which uses jurisdictional rather than project-level data to calculate
ex-ante baselines. However, to date, there are no studies that have assessed whether
VM0048 produces more appropriate baselines (judged by agreement with ex-post quasiexperimental methods) compared to the now inactive first-generation methodologies. Here,
we retrospectively apply the VM0048 methodology to a set of REDD+ projects in Colombia
which have been assessed using ex-post quasi-experimental methods. Results show
VM0048 baselines are on average 80% (ranging from 59% to 94%) more conservative than
those from the now inactive first-generation REDD+ methodologies, and forecasts of
avoided deforestation closely align with independent ex-post quasi-experimental
assessments of project effectiveness. Nonetheless, important caveats remain. Most
importantly, VM0048 ex-ante baselines fundamentally cannot account for unexpected
deforestation rates not reflected in historical trends. Whilst divergence between methods is
inevitable, VM0048 ultimately remains constrained by the limitations of any ex-ante
framework. Our results, although specific to the Colombian context, therefore represent
cautious optimism for the next generation of REDD+ projects: we find that the
methodological refinements in VM0048 produce more conservative baselines, yet the exante design could continue to facilitate systematic over- and under- estimated baselines in
REDD+ globally.

Date of Award	15 Sept 2025
Original language	English
Supervisor	Julia Patricia Gordon Jones (Supervisor), Eleanor Warren-Thomas (Supervisor) & Katie Devenish (Supervisor)