Deforestation

In the 1990s, empirical studies of deforestation increasingly drew on economic models that assumed land was allocated across alternative uses to maximise economic returns. Tropical regions were the predominant focus (e.g., Reis and Guzman 1992, Nelson and Hellerstein 1997, Pfaff 1999, Cropper et al. 1999). Researchers commonly used static multinomial logit models to estimate probabilities of various land-use choices – retaining forest cover, converting to pasture, or cultivating crops – as functions of factors influencing agricultural profitability. This early research highlighted the importance of land characteristics (such as soil quality) and transportation costs in explaining deforestation.

Agricultural and Land-Use Policies

Building on this literature, Souza-Rodrigues (2019) developed a static structural model to assess the cost-effectiveness of alternative policies in the Brazilian Amazon in 2006. By exploiting regional variation in transportation costs to recover farmers’ responses to permanent changes in land-use returns, he finds that counterfactual payment programmes to avoid deforestation and land-use taxes on agricultural land can be highly effective in preserving the rainforest and substantially less costly than existing command-and-control policies that restrict agricultural area on private properties. This finding has important implications for conservation finance: market-based instruments can promote forest protection.

Yet, land-use change is inherently dynamic, with landowners making decisions that depend on current prices, expectations of future prices, and switching costs, such as forest clearing (Stavins and Jaffee 1990, Stavins 1999). These dynamic elements create differences between short- and long-run land-use elasticities, as permanent price increases can justify fixed costs of land conversion. In contrast, temporary price fluctuations may not (Scott 2013). This creates an external validity problem for static models: estimates based on short-term variation may not accurately predict the long-run effects of lasting policy changes.

Recent dynamic models address these issues by explicitly incorporating factors that evolve over time, allowing policy analysis to distinguish between transitional and long-term effects. Araujo et al. (2025b) examined deforestation and conservation policies in the Brazilian Amazon between 2008 and 2017, estimating the carbon-efficient level of forest cover – the amount preserved if farmers internalised the social cost of carbon. Compared to this efficient benchmark, business-as-usual results in massive deforestation, releasing 42 billion tonnes of CO2 due to an inefficient loss of 1.2 million km2 of forest cover – an area twice the size of France. A carbon tax aligned with the social cost of carbon would prevent this loss and generate over $1.6 trillion in social welfare gains. Importantly, a second-best policy, such as a tax on cattle ranching, captures up to 87% of the welfare gains from the carbon tax. This highlights that targeted sectoral policies, although second-best, may deliver significant environmental benefits.

Assunção et al. (2025) also developed a dynamic spatial model, focusing on ambiguity in location-specific agricultural productivity and carbon absorption capacity, to evaluate carbon pricing policies in the Brazilian Amazon. They also find that modest carbon prices could generate significant reductions in greenhouse gas emissions.

Scott et al. (2025) study cattle management dynamics, emphasising cattle’s dual role as both a consumption and a capital good. Temporary price spikes encourage ranchers to cull more cattle immediately, thereby shrinking future herd size and reducing incentives for deforestation. Persistent price increases lead ranchers to retain breeding cattle and expand herds. They show deforestation is largely unresponsive to temporary price shocks but highly elastic to persistent price changes. They also find that deforestation taxes can benefit farmers in the long run by raising beef prices and increasing profits per hectare.

Deforestation and Rainfall

Araujo (2024b) provides the first integration of climate feedback effects into structural land-use models. The Brazilian Amazon sustains rainfall by recycling moisture through evapotranspiration – the “flying rivers” mechanism (Nobre et al. 1991, Marengo et al. 2004). Deforestation can disrupt this cycle, reducing rainfall and agricultural productivity downwind (Spracklen et al. 2012). Araujo (2024b) develops a framework that integrates climate and economic land-use models and finds that having more forest upwind is associated with more rainfall. As an application, he simulates the rollback of protections for the Xingu Indigenous Territories in Mato Grosso. If unprotected, 47% of Xingu would be deforested, generating a climate externality by reducing rainfall by 20% in downwind regions, thus reducing crop yields and leading to the conversion of 60,000 km² of agricultural land to forest due to abandonment. This climate externality offsets 40% of the gains from expanding agricultural land into protected areas. This suggests traditional cost-benefit analyses may significantly underestimate the benefits of conservation by neglecting climate feedback.

Energy and Deforestation

Sant’Anna (2024) examines the deforestation footprint of biofuels by studying sugarcane ethanol supply in Brazil. He notes that sugarcane is a crop with declining yields over time until fields are replanted, motivating a dynamic model that disentangles the roles of acreage (extensive margin) and yields (intensive margin) on ethanol supply. He finds that 92% of new ethanol production comes from increases in planted area, 19% of which is direct deforestation. This highlights how biofuel policies intended to reduce carbon emissions may inadvertently increase them through land-use change effects.

Araujo (2024a) shows that deforestation, by reducing precipitation downwind, can have implications for hydroelectric generation. Using an econometric climate model that connects deforestation with rainfall patterns, he finds that Amazon deforestation since the 1980s reduced energy generation capacity of Teles Pires hydropower plant by 2.5 to 10%, representing an annual loss of approximately 10% of the plant’s revenue.

Market Design

The role of market design in achieving environmental goals is an important topic of growing empirical interest. Aronoff and Rafey (2025) study wetland conservation and offset markets in Florida, finding that offset markets generated substantial private gains from trade relative to direct conservation. At the same time, offset trading also produced unintended hydrological externalities by affecting flood risk. Aspelund and Russo (2025) investigated market design for the US Conservation Reserve Programme, linking auction bids to satellite-based land-use data. They find the programme generates $126 million per auction in welfare gains but captures only 15% of potential efficiency, partly because 75% of marginal winners would have conserved anyway. They propose alternative scoring rules accounting for expected additionality, closing 41% of the efficiency gap. Heilmayr et al. (2020) studied the carbon impacts of forest subsidies in Chile between 1986 and 2011. They find that while payments for afforestation increased tree cover, they also incentivised the plantation of exotic species at the expense of native ones. This illustrates that conservation programmes must be carefully designed to achieve intended environmental outcomes rather than simply maximising tree cover.

Future Work

The empirical literature on deforestation has progressed from documenting its key determinants to developing structural models that enable policy simulations and discussion of optimal policy design. Early studies relied on static, spatially explicit models, while more recent work incorporates dynamic frameworks and climate feedback loops. Future work can go further by developing new analytical tools for modelling and estimating dynamics, integrating scientific modelling of climate phenomena.Complementary advances in market design emphasise the role of carefully structured mechanisms, such as offsets and auctions, in improving conservation outcomes. Future work in this vein can inform ongoing efforts to establish voluntary and mandatory carbon markets, including on an international scale.

Policy Takeaways

Structural models of firms and firm incentives provide crucial insights for policy.

• Market Incentives. Market-based instruments like payments for conservation and land-use taxes can be very cost-effective and generate enormous welfare gains. Second-best sectoral policies targeting the major drivers of deforestation, such as cattle ranching, can capture most of the benefits.
• Dynamic Responses. Policy effectiveness depends critically on whether changes are perceived as temporary or permanent. Policies should account for these dynamic responses in their design.
• Programme Design. Conservation programmes must be carefully designed to achieve intended outcomes and additionality. Simple metrics like tree cover may miss important environmental goals like biodiversity preservation.
• Climate Feedbacks. Traditional cost-benefit analyses may underestimate conservation benefits by ignoring climate feedback. Protected areas may be more economically justified than previously recognised, particularly where forests provide crucial rainfall for agriculture.

For full reference list see the end of the conclusion chapter.