JMP 1: Costly Regulation, Minimal Results: The EU’s Deforestation Regulation Effect on Global Soy Trade (under R&R in the European Review of Agricultural Economics)

One-sentence takeaway

The EU’s new deforestation regulation for soy shifts trade away from Europe and toward China, raises prices for EU consumers, and does little in the short run to reduce deforestation pressure in major South American producers.

Abstract

The European Union Deforestation Regulation (EUDR) aims to curb tropical deforestation by restricting market access to soy and other “forest-risk” commodities that are not verified as deforestation-free. However, the EU is a relatively small buyer in the global soy market, raising concerns that its import restrictions may simply redirect trade to other destinations rather than reduce overall deforestation.

I use a structural gravity model of global soy trade—covering soybeans, soybean oil, and soybean cake—to quantify how EUDR compliance costs for exports to the EU would reshape trade flows, prices, and welfare. I treat compliance requirements as additional trade costs on EU-bound imports and simulate two scenarios: (i) all exporting countries comply and continue to trade with the EU, and (ii) Brazil, Argentina, and Paraguay (BAP)—the main deforestation-risk exporters—do not comply and are effectively shut out of the EU soy market.

Three main results emerge. First, restricting access to the EU market reallocates soy exports from South America to unregulated markets, especially China, rather than significantly reducing South American exports. Second, EU importers face substantial price increases and welfare losses, and substitute toward soy supplied by the United States and Canada. Third, South American exporters see lower producer prices but little change in terms of trade or total exports in the short run, indicating limited leverage of EU policy to induce pro-forest land use decisions.

Overall, the EUDR in its current unilateral form is likely to rearrange who buys soy rather than meaningfully reduce deforestation linked to soy production, highlighting the need for broader, coordinated, and supply-side complementary policies.

Motivation

Soy expansion is one of the major drivers of tropical deforestation, particularly in Brazil and other South American countries. The EU has responded by adopting the EUDR, which:

  • bans imports of soy and other forest-risk commodities unless they are:
    • deforestation-free,
    • compliant with local laws, and
    • fully traceable through due diligence; and
  • places the legal responsibility on importers and traders operating in the EU.

This policy is intended to use market access to incentivize cleaner production. However:

  • the EU accounts for only about 8% of soybean-related exports from Brazil, Argentina, and Paraguay, while
  • China alone absorbs over 60% of global soybean imports.

This raises a central question for policymakers interested in trade and the environment:

Can a single (relatively small) buyer, acting unilaterally through trade restrictions, significantly reduce global deforestation, or will trade simply be diverted to other markets?

My job market paper provides a quantitative answer for the soy sector.

Research Question

I address the following core question:

What are the trade, price, and welfare effects of the EU Deforestation Regulation in the global soy market, and how much leverage does the EU actually have to reduce deforestation pressure in major exporting countries?

More specifically, I ask:

  1. How does EUDR-style compliance cost on EU imports reallocate soy trade across origins (South America, North America, rest of the world) and destinations (EU, China, rest of the world)?
  2. How much do EU prices and expenditures increase as a result of the regulation?
  3. Do deforestation-risk exporters such as Brazil, Argentina, and Paraguay face enough economic pressure—via lower prices, worse terms of trade, or reduced export potential—to change land-use decisions?

Empirical Approach

Data

I build a panel dataset for 90 countries from 2007–2022, combining:

  • Bilateral trade flows for soybeans, soybean oil, and soy cake (CEPII–BACI),
  • Intranational trade flows constructed from FAOSTAT Food Balance Sheets,
  • Effectively applied tariffs (ITC MacMap, including preferential tariffs), and
  • Standard gravity controls (distance, contiguity, common language, colonial ties from CEPII).

Structural gravity framework

I estimate separate structural gravity models for:

  • soybeans,
  • soybean oil, and
  • soybean cake,

using PPML with exporter-year, importer-year, and pair fixed effects. From tariff elasticities, I recover Armington elasticities (elasticity of substitution across origins).

I then follow Anderson and Yotov (2018) to construct:

  • baseline trade costs and multilateral resistance terms, and
  • conditional counterfactuals that:
    • hold each country’s total soy supply and expenditure fixed, and
    • allow bilateral trade flows and price indices to adjust when trade costs change.

Policy experiments

I simulate the EUDR as a destination-specific increase in trade costs for EU-bound imports:

  1. Compliance scenario (All comply)
    • All exporters face a moderate compliance cost (modeled as a ~6% ad valorem trade cost) on exports to the EU.
    • The compliance cost estimates are sourced from from EU impact assessments and industry estimates of procurement cost increases.
  2. Non-compliance scenario (BAP excluded)
    • Brazil, Argentina, and Paraguay (BAP) face prohibitive trade costs and are effectively shut out of the EU soy market.
    • Other suppliers still face the 6% compliance wedge.
    • This reflects a plausible situation where major deforestation-risk countries choose not to comply if compliance costs exceed the “European premium.”

For each scenario, I compute:

  • changes in bilateral exports,
  • price indices,
  • producer prices, and
  • terms of trade and welfare (through changes in price indices).

Key Findings

1. Trade is diverted from the EU to China, not eliminated

Under both scenarios, the EUDR leads to large trade reallocations but small reductions in exports from Brazil, Argentina, and Paraguay:

  • When BAP are restricted from the EU market, their exports to China increase sharply, especially for soybean cake:
    • Brazil’s soybean cake exports to China rise by more than 1,000%,
    • Argentina’s and Paraguay’s exports also surge.
  • The EU’s reduced demand is effectively offset by expanding demand from China and other non-EU markets.

Implication:
The regulation mainly shifts deforestation-embodied soy from the EU market to China, rather than reducing the overall volume exported by deforestation-risk producers.

2. The EU pays higher prices and loses welfare

The EU faces significant price and welfare impacts, particularly when BAP do not comply:

  • Price effects:
    • Soybean prices in the EU can increase by up to 50–60%.
    • Prices for soybean cake and soybean oil also rise.
  • Trade diversion:
    • The EU increasingly imports soy products from the United States and Canada:
      • US soybean exports to the EU nearly triple in the non-compliance scenario.
      • US and Canadian soybean cake exports to the EU increase by several hundred percent.
  • Welfare losses:
    • EU expenditure losses on soy products are substantial, with the largest absolute losses in:
      • Spain, the Netherlands, Italy, Germany, France, Poland, and Denmark.

Implication:
The EU bears sizable adjustment costs—higher prices and reduced terms of trade—without proportionate environmental gains in the short run.

3. South American exporters see lower prices but limited pressure to change land use

For Brazil, Argentina, and Paraguay (BAP):

  • Producer prices for soy products tend to fall when they face restrictions in the EU.
  • Price indices for soy often decline, especially for soybeans, reflecting loss of a high-paying market.
  • However, in the short-run conditional equilibrium:
    • Total exports and terms of trade change very little.
    • BAP can redirect volumes to China and the rest of the world.

When I relax the constraint on expenditures (allowing global demand to adjust):

  • BAP total exports can actually rise, especially for soybean cake, as lower producer prices make them more competitive in non-EU markets.

Implication:
The EU’s share in BAP’s export portfolios is too small to generate strong economic pressure for land-use change. The EUDR alone is unlikely to be a sufficient lever to reduce deforestation in major soy-exporting countries.

4. Segregated supply chains and leakage undermine environmental effectiveness

The simulations suggest the emergence of segregated supply chains:

  1. A relatively small “compliant” chain:
    • Verified deforestation-free soy shipped to the EU, primarily from:
      • North America (US, Canada), and
      • a subset of South American suppliers and regions already under strict zero-deforestation commitments.
  2. A much larger “unregulated” chain:
    • Soy shipped to China and other non-EU markets without equivalent deforestation or traceability constraints.

This pattern is consistent with:

  • Leakage documented in the climate and trade literature: policies in one market push emissions or deforestation to other markets rather than eliminate them.
  • Evidence from Brazil of within-country displacement of deforestation when regulations target specific regions or buyers.

Implication:
As long as large alternative markets remain unregulated, the EUDR unilateral policies risk creating a “clean” EU supply chain without substantially changing global deforestation outcomes.

Policy Relevance

For policymakers, international organizations, and firms, the paper highlights that:

  • Unilateral, demand-side measures like the EUDR:
    • are effective at “cleaning” the EU’s own consumption footprint,
    • but have limited leverage over global deforestation when the EU is not the dominant buyer.
  • More effective strategies likely require:
    • Broader coalitions of importers (e.g., EU + US and other major markets) with coordinated standards,
    • Supply-side instruments such as:
      • forest conservation payments,
      • conditional trade agreements linked to land-use outcomes,
      • support for local governance and traceability systems, and
    • Territorial approaches that align with domestic policies in producer countries.

The results therefore speak directly to ongoing debates about:

  • how to design trade–environment linkages,
  • the role of the EU in shaping global value chains,
  • and how to avoid creating fragmented markets that weaken environmental leverage.

JEL Codes and Keywords

  • JEL Codes: Q17, Q56, F18
  • Keywords: European Union Deforestation Regulation (EUDR); structural gravity model; deforestation; soy; trade policy; tariffs

Full Paper

JMP 2: Foreign Demand, Export Sales, and Deforestation: A Commodity-Level Analysis

One-sentence takeaway

International demand for agricultural commodities significantly increases deforestation, with the strongest effects for land‑intensive perennial export crops (stimulants and spices, palm, cocoa, rubber, coffee, nuts) including soybeans and cattle in South America and Southeast Asia—accounting for roughly one‑quarter to one‑third of forest‑to‑agriculture conversion over 2001 to 2022.

Abstract

International agricultural trade is widely viewed as a key driver of tropical deforestation, yet there is limited causal evidence on how strongly commodity‑specific foreign demand translates into deforestation across countries and regions. Most existing work either uses accounting approaches that allocate forest loss to trade ex post, or focuses on a narrow set of countries or crops.

I quantify the causal effect of international demand on deforestation using a global country–commodity–year panel covering 138 countries, 18 commodity groups, and 2001–2022. I measure international demand in two ways: (i) a shift–share foreign‑demand that aggregates plausibly exogenous importer–product demand shocks, and (ii) export sales by country and commodity. To address simultaneity between exports, prices, and deforestation, I construct a foreign‑demand shifter based on supply‑corrected imports from a gravity model, and use it as an instrument for export sales in a control‑function PPML framework saturated with country–year and commodity–year fixed effects.

Three main results emerge. First, higher foreign demand leads to significantly more deforestation: a 1% increase in foreign demand raises commodity‑level deforestation by about 0.11%, and a 10% increase in export sales increases deforestation by 2–4% on average. Second, these effects are highly heterogeneous across commodities. Elasticities are largest for land‑intensive perennial export crops—palm, cocoa, rubber, stimulants/spices, nuts, coffee, sugar crops, soybeans, and cattle—while staple grains and many horticultural crops show weak or no responses. Third, impacts are geographically concentrated in South America and Southeast Asia, with important but smaller effects in parts of Africa and the rest of Asia, and near‑zero responses in Europe and North Asia.

Back‑of‑the‑envelope calculations suggest that growth in international demand and export sales can account for roughly 28–41 million hectares of agricultural expansion into forests between 2001 and 2022—about 25–37% of observed forest‑to‑agriculture conversion. These elasticities can be used as behavioral parameters in quantitative trade and land‑use models and helps to conduct policy experiments for trade‑linked environmental regulation/ policies.

Motivation

Agricultural expansion is now one of the main drivers of tropical deforestation, with far‑reaching consequences:

  • Biodiversity loss, as tropical forests host a disproportionate share of global species richness.
  • Climate and hydrological impacts, including local and regional warming, altered rainfall patterns, and teleconnected effects on distant regions.
  • Human health costs, such as increased respiratory and cardiovascular morbidity from smoke exposure and changes in vector‑borne disease patterns.

International demand for agricultural commodities—especially cattle, soy, palm oil, cocoa, coffee, and rubber—has been repeatedly implicated in this process. Recent studies estimate that roughly one‑third of tropical deforestation is embodied in agricultural trade, with hotspots in South America, Southeast Asia, and parts of Africa. Yet much of this evidence is based on:

  • Accounting frameworks that trace deforestation through supply chains without identifying causal elasticities, or
  • Aggregate analyses that link trade liberalization or crop prices to total forest loss in specific countries.

As a result, we lack comparable, causal estimates of how strongly commodity‑specific foreign demand and export sales translate into deforestation across multiple commodities and continents.

At the same time, policymakers are increasingly turning to trade‑linked regulations—such as the EU Deforestation Regulation (EUDR) and similar initiatives in other high‑income markets—to curb deforestation by conditioning market access on environmental performance. These policies implicitly assume that reducing demand for “forest‑risk” commodities in regulated markets will slow forest conversion in exporting countries.

This paper asks how strong that demand–deforestation linkage actually is, and for which commodities and (sub-)continents it is most consequential.

Research Question

I address the following core question:

How strongly does international demand for specific agricultural commodities translate into deforestation at the country–commodity level, and how does this relationship vary across commodities and continents?

More specifically, I ask:

  1. Pooled elasticities:
    What is the average elasticity of deforestation with respect to:

      1. a shift–share measure of foreign demand, and
      1. export sales, across all countries and commodities?

  2. Commodity heterogeneity:
    How do these elasticities differ across 18 commodity groups, including land‑intensive perennial crops (palm, cocoa, rubber, coffee, nuts, stimulants/spices), soybeans, cattle/pasture, staple grains, and horticultural crops?

  3. Geographic heterogeneity:
    How large are demand‑driven deforestation responses in different regions—South America, Southeast Asia, Africa, North & Central America, Europe, North Asia, Oceania, and the rest of Asia?

  4. Contribution to global deforestation:
    How much of observed forest‑to‑agriculture conversion between 2001 and 2022 can be attributed, in a causal sense, to growth in foreign demand and export sales?

Empirical Approach

Data

I assemble a global panel dataset at the country–commodity–year level:

  • Coverage:
    • 138 countries
    • 18 commodity groups (aggregated from FAOSTAT items and HS trade codes)
    • 2001–2022
    • 41,829 country–commodity–year observations
  • Deforestation (outcome):
    • Commodity‑attributed deforestation from the DeDuCE model (Singh 2024), which spatially attributes tree‑cover loss to specific agricultural commodities.
    • I use unamortized deforestation (hectares) allocated to 18 aggregated commodity groups (e.g., cocoa, coffee, palm, rubber, soybeans, cattle/pasture, sugar crops, cereals, cassava, fruits, vegetables, other oilseeds).
  • Trade (treatment):
    • Bilateral trade flows from CEPII–BACI, mapped from HS codes to the same 18 commodity groups.
    • I construct country–commodity–year export values (USD) and quantities.

The panel is unbalanced because not all countries produce all commodities (e.g., rubber and cocoa are absent from most temperate countries). I retain country–commodity pairs that consistently appear in the BACI data and exhibit agricultural deforestation over the sample.

Empirical framework

I focus on two related measures of international demand:

  1. Export sales for country \(i\), commodity \(k\), year \(t\), denoted \((X_{ikt}\).
  2. A shift–share foreign‑demand index, \(FD_{ikt}\), that aggregates importer–product‑year demand shocks faced by country–commodity pair \((i,k)\).

I estimate two main sets of models:

  • Reduced‑form PPML models that relate deforestation directly to foreign demand: \[\mathbb{E}[C_{ikt} \mid FD_{ikt}, \text{FE}] = \exp(\beta\, FD_{ikt} + \alpha_{it} + \alpha_{kt}),\] where \(C_{ikt}\) is deforestation, \(\alpha_{it}\) are country–year fixed effects, and \(\alpha_{kt}\) are commodity–year fixed effects.

  • Control‑function PPML models that estimate the causal effect of export sales on deforestation, instrumenting \(\log X_{ikt}\) with foreign demand:

    • First stage (linear with fixed effects):
      \[\log X_{ikt} = \pi\, FD_{ikt} + \alpha_{it} + \alpha_{kt} + \nu_{ikt}.\]
    • Second stage (PPML with residual inclusion):
      \[\mathbb{E}[C_{ikt} \mid \log X_{ikt}, \hat{\nu}_{ikt}, \text{FE}] = \exp\big(\alpha\, \log X_{ikt} + \delta\, \hat{\nu}_{ikt} + \alpha_{it} + \alpha_{kt}\big).\]

This approach:

  • Retains zero-deforestation observations,
  • Is robust to heteroskedasticity, and
  • Yields elasticities of deforestation with respect to foreign demand and export sales.

I also estimate heterogeneous elasticities by:

  • Commodity group,
  • Continent, and
  • Commodity–continent pairs,

and conduct extensive robustness checks using alternative transformations (log, log(1 + C), asinh, levels) and 2SLS.

Identification and construction of foreign demand

The key identification challenge is simultaneity and reverse causality:

  • Deforestation and exports are jointly determined: expanding agricultural land can raise export capacity.
  • Global price shocks affect both exports and deforestation.
  • Country‑specific shocks (infrastructure, governance, policies) can move both exports and forest outcomes.

To address this, I construct a shift–share foreign‑demand instrument:

  1. Supply‑corrected imports:
    • Estimate a multiplicative gravity model at the bilateral product level: \[ X_{ijst} = \exp(\gamma_{ist} + \gamma_{jst} + \gamma_{ijs}) \varepsilon_{ijst}, \] where \(i\) is exporter, \(j\) importer, \(s\) product, \(t\) year.
    • Use exporter–product–year and pair–product fixed effects to purge exporter‑side supply shocks and isolate importer–product demand.
    • Construct supply‑corrected imports for each \((j,s,t)\) and then apply a “leave‑one‑out” correction to exclude imports from country \(i\).
  2. Shift–share aggregation:
    • For each country–commodity pair \((i,k)\), compute fixed export shares \(\omega_{ijs}^{k}\) over 2001–2022 that capture the importance of each destination–product pair \((j,s)\) in its export portfolio.
    • Aggregate lagged importer–product demand shocks using these shares: \[ FD_{ikt} = \sum_{j,s} \omega_{ijs}^{k}\, \log(\text{Import}_{js,t-1,-i}). \]
  3. Fixed effects and timing:
    • Use lagged foreign demand to further mitigate reverse causality.
    • Include country–year fixed effects to absorb domestic policies, macro shocks, and governance changes.
    • Include commodity–year fixed effects to absorb global price and technology shocks.

Under standard shift–share assumptions (many small destination–product shocks, exogenous importer‑side demand), \(FD_{ikt}\) provides plausibly exogenous variation in international demand faced by each country–commodity pair.

I also construct a second, more conservative foreign‑demand measure directly from importer–product–year fixed effects, which capture inward multilateral resistance and demand but strip out more exporter‑specific variation.

Key Findings

1. International demand significantly increases commodity-level deforestation

  • In pooled PPML specifications, the elasticity of deforestation with respect to foreign demand is about 0.11:
    • A 1% increase in foreign demand raises deforestation by roughly 0.11%.
  • In control‑function PPML models that instrument export sales with foreign demand, the export elasticity lies between 0.24 and 0.38:
    • A 10% increase in export sales leads to about a 2.4–3.8% increase in deforestation, on average.
  • These estimates are somewhat smaller than some single‑country, single‑commodity studies (which often focus on the most extreme frontiers), but they are broadly consistent with the idea that a substantial share of deforestation is induced by export demand.

Back‑of‑the‑envelope calculations suggest that growth in international demand and export sales across the 18 commodity groups account for approximately:

  • 28–41 million hectares of agricultural expansion into forests,
  • Roughly 25–37% of observed forest‑to‑agriculture conversion between 2001 and 2022.

2. Land‑intensive perennial export crops drive most of the response

Elasticities are highly heterogeneous across commodities:

  • Large, positive elasticities for:
    • Palm oil
    • Cocoa
    • Rubber
    • Stimulants/spices/aromatics
    • Nuts
    • Coffee
    • Sugar crops
    • Soybeans
    • Cattle/pasture
  • These crops share three features:
    1. They are land‑intensive perennial systems with high fixed establishment costs and relatively low marginal costs of expanding onto nearby forest land.
    2. They are high‑value, export‑oriented commodities, making frontier conversion profitable even where transport and clearing costs are high.
    3. Their agro‑climatic suitability closely overlaps with humid tropical forests, so the main margin of expansion is forest rather than already cleared land.
  • Small or statistically insignificant elasticities for:
    • Staples: cassava, rice, maize, pulses/legumes, other cereals (in many regions),
    • Horticultural crops: fruits and vegetables in most regions,
    • Some fibre and oilseed crops, depending on the continent.

In some cases, staples such as cassava show negative or near‑zero elasticities, suggesting that expansion may occur on non‑forest land, or that these crops are more easily intensified on existing agricultural areas.

3. Effects are concentrated in tropical forest frontiers

Geographically, foreign‑demand and export elasticities are largest in:

  • South America (notably Brazil and its neighbors),
  • Southeast Asia, and
  • Parts of Africa and the rest of Asia.

In these regions:

  • Land‑intensive export crops (palm, cocoa, rubber, coffee, nuts, soybeans, sugar, cattle) show large positive elasticities, often well above the pooled average.
  • Agro‑climatic suitability for these crops overlaps strongly with remaining tropical forests, so forest land is the primary expansion margin.

By contrast:

  • Europe and North Asia exhibit weaker or even negative elasticities for many commodities.
  • In North & Central America and Oceania, positive elasticities are present for some high‑value export crops (e.g., coffee, cocoa, nuts, palm, stimulants), but the overall deforestation response is smaller relative to South America and Southeast Asia.

At the commodity–continent level, I find that:

  • The largest elasticities cluster in classic frontier combinations, such as:
    • Palm, rubber, cocoa, coffee, nuts, stimulants, soybeans, and cattle in South America, Southeast Asia, and parts of Africa and the rest of Asia.
  • Staples and many horticultural crops often show weak or no link between export demand and deforestation, even in frontier regions.

Policy Relevance

  • International demand matters, but effects are highly concentrated.
    • A relatively small set of high‑risk commodity–region combinations—palm, soy, cattle, cocoa, coffee, rubber, nuts, stimulants in South America, Southeast Asia, and parts of Africa—accounts for a disproportionate share of demand‑driven deforestation.
    • Targeting these frontiers is likely to yield the largest environmental returns per unit of regulatory effort.
  • Implications for global deforestation accounting and modeling.
    • The estimated elasticities are suitable as behavioral parameters for:
      • Quantitative trade models that incorporate land use, and
      • Climate and land‑use models interested in simulating the impact of trade‑linked policies (e.g., EUDR‑type regulations, zero‑deforestation commitments).
    • Because the empirical framework accounts for zero deforestation and is estimated at the country–commodity level, it can be integrated with spatial models to explore leakage, indirect land‑use change, and interactions across commodities and regions.

JEL Codes and Keywords

  • JEL Codes: Q17, Q56, F18
  • Keywords: Deforestation; foreign demand; export sales; international trade; PPML; shift–share; land use

Full Paper