Working papers, publications, and the thesis, grouped by status. Drafts of unpublished work are available on request.
Working papers
The Net Climate Value of Carbon Credits: Pricing Leakage, Time, and Reversal
with Jens Abildtrup, Philippe Delacote, Andreas Kontoleon, Ben Filewod, Hannes Böttcher, Tommaso Chiti, Ana Rey, Jurij Diaci, Aleksi Lehtonen, Helga Pülzl, Andreas Schindlbacher, and Miguel A. Zavala · Working paper · Replication code to be made accessible upon publication
A forest carbon credit is issued for one tonne of CO₂ kept from the atmosphere, but rarely delivers the full tonne. After on-site verification, three channels — market leakage, finite storage duration, and reversal risk — erode credited carbon. We calibrate each into one issuance rule across sixteen forest practices and three European biomes, propagating parameter uncertainty. Only 12–56% survives as net climate value (median 24%), and five European schemes over-credit by 29–59%. Storage duration is the largest shortfall: schemes credit 30- and 100-year contracts identically. Reaching 100 MtCO₂ yr⁻¹ of genuine removals needs ~121–265 Mha — approaching or exceeding the EU’s 160 Mha forest estate — while crediting the shortfall at face value over-credits by ~85 MtCO₂e yearly, a $4–28 billion social cost. Rankings and every scheme’s over-crediting hold under plausible parameters. The EU’s newly adopted carbon-farming methodology leaves all three channels unpriced; the forest-management methodology now in preparation should price them.
Audited Alike, Valued Alike? Household Preferences for Public or Private Delivery of EU and Tropical Forest Carbon
Choice experiment in France, Germany and Denmark · Pre-registration on OSF before the pilot
Who should close the EU’s 2030 land-sector gap, and how much do households weigh the damage that leaving it open causes others? In a pre-registered choice experiment in France, Germany and Denmark, respondents choose between emissions cuts, inaction and forest programmes varying in organiser, location and recipient. The EU must soon decide how far certified markets and international credits should stand in for public action at home. With quality audited, incomplete-contract theory gives no efficiency reason to prefer either provider, and carbon is a global public good. Two hypotheses test whether households nonetheless care who delivers a tonne, and whether tropical credits, the focus of over-crediting concerns, change this. A randomised treatment shows half the households their share of worldwide damage from inaction. Two treatment hypotheses test whether this changes support for action and whether households weigh it like money paid, as a global social cost of carbon assumes.
Two Countries, One Divide: Fear of Climate Change Splits Support for Forest Conservation
with Jens Abildtrup, David W. Shanafelt, Antonello Lobianco, and Harald Vacik · Highly Commended Paper Award, 8th Workshop on Non-Market Valuation (Leeds, 2024)
Who would pay for forest conservation, and why? Public programmes depend on the answer, yet studies report average willingness to pay, which hides the divides beneath it, and opinions on climate change have become politically polarised. Discrete choice experiments in Austria (N = 595) and France (N = 1,874) value publicly funded programmes that restore damaged private forests with climate-adapted stands. We model preference heterogeneity with a stated selection rule and report every model we estimate. The same divide appears in both countries: three quarters of respondents support the programmes, one quarter mostly declines to pay. Fear of climate change draws the line, most sharply in Austria; whether people believe that climate change can still be stopped makes no difference. Supporters would pay three to four times as much as the reluctant for climate-adapted forests, similar amounts in both countries, and in Austria amounts of the order of the planting costs. Climate concern divides public support for forest conservation; local party politics does not.
Classes or Continuum? A Pre-Registrable Specification Rule for Latent Class Logit Models
Companion R package: klue
Latent class multinomial logit (LCMNL) models describe preference heterogeneity as discrete classes, mixed multinomial logit (MMNL) models as a continuous distribution. Fitting an LCMNL requires three choices: starting values for a likelihood with many local maxima, the number of classes, and whether classes fit better than a continuous distribution. This paper proposes a rule for all three that can be pre-registered. Starting values come from clustering respondents by their choices and fitting a multinomial logit (MNL) to each cluster. The Bayesian information criterion (BIC) sets the number of classes, with an entropy check on the certainty of class assignment. An MMNL is the benchmark: if its BIC is lower, the heterogeneity is treated as continuous. A Monte Carlo study evaluates the rule under a blocked D-efficient design, as used in practice, and an idealised randomised design; five public datasets illustrate it. Under the blocked design, 93% of clustering starts reach the highest log-likelihood found, against 78–81% of starts from random partitions or the pooled MNL; under the randomised design clustering has no advantage. Six clustering starts reach it about as often as Apollo’s start-value search, in a fraction of the time. BIC selects the true number of classes in 94.5% of blocked-design datasets. The benchmark always picks the LCMNL when simulated preferences are discrete and never when they are continuous, even where BIC alone finds spurious classes. On the public datasets the MMNL fits best on the four with real choices and ties with the LCMNL on the synthetic one.
A Hybrid Machine Learning Approach for the Prediction of Forest Stand Growth Under Different Climate Scenarios
with Antonello Lobianco, Nikola Besic, Jean-Daniel Bontemps, and Laurent Hertzog
Publications
Do Individual PES Buyers Care About Additionality and Free-Riding? A Choice Experiment
with Jens Abildtrup, Claire Montagné-Huck, Salomé Gorel, and Anne Stenger · Ecological Economics, 213, 107944 (2023) · DOI
A Review of Forest Management Practices Potentially Suitable for Carbon Farming in European Forests
Chiti, T., Rey, A., Abildtrup, J., Böttcher, H., Diaci, J., Frings, O., Lehtonen, A., Pülzl, H., Schindlbacher, A., & Zavala, M. A. · Journal of Environmental Management, 398, 128391 (2026) · DOI
Carbon Farming in the European Forestry Sector
Chiti, T., Rey, A., Abildtrup, J., Böttcher, H., Diaci, J., Frings, O., Lehtonen, A., & Zavala, M. A. · From Science to Policy 17, European Forest Institute (2024) · DOI
Maximising Forest Carbon Storage while Aligning Sustainability, Value Chains and Policy
Chiti, T., Rey, A., Zavala, M. A., Manrique, S., & Frings, O. · Chapter 4 in N. Grima, L. Walther, & F. Krumm (Eds.), Productive and Biodiverse: Managing Europe’s Forests for Economic and Ecological Outcomes, pp. 69–99 · IUFRO World Series 46, International Union of Forest Research Organizations, Vienna (2026) · Report
PhD thesis
Pricing Credibility: Supply-Side Integrity and Public Support in Nature Markets
AgroParisTech · defence 10 December 2026
The thesis asks what makes nature markets credible, in three parts. The first builds the measurement instrument: a specification workflow for latent class choice models. The second studies demand-side legitimacy: who supports forest conservation programmes and carbon offsetting, and how that support depends on the institutions delivering them. The third turns to supply-side integrity: an issuance rule that prices leakage, storage duration, and reversal risk into forest carbon credits.