## wpiea2025052-print-pdf

## Source details

**Canonical URL:** [wpiea2025052-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025052-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025052-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025052-print-pdf.pdf.json)

---

### Overview
- Farming is subject to risk; the increasing frequency of extreme weather events affects growth, productivity, and wellbeing.
- The paper concentrates on six South American countries: Argentina, Brazil, Colombia, Peru, Paraguay, and Uruguay.
- On average, agriculture constitutes about 8 percent of GDP in the sample and about 40 of the total value of its exports.
- Soy constitutes almost 60 percent of all agricultural production in Paraguay, and above 40 percent in Argentina, Brazil, and Uruguay.
- Countries in the sample are prepared for climate change to varying degrees, with Chile and Uruguay leading efforts (Figure 1).

### Data and Methodology
- Key data sources and measures:
  - Normalized Difference Vegetation Index (NDVI) used as a greenness index proxy for soy output.
  - Standard Precipitation and Evapotranspiration Index (SPEI) used to define droughts; drought defined as a month where SPEI < -1.5.
  - Land use data from MapBiomas to identify soy-producing cells (detailed crop categories available for Brazil).
  - Brazil municipal soy yield data from IBGE (2010–2020) to map NDVI to actual yields.
  - Crop calendar from USDA to align planting and harvesting and to select peak NDVI during harvest.
- Spatial harmonization:
  - Data adjusted to uniform 0.05° resolution; SPEI disaggregated via bilinear interpolation; higher-resolution land use data aggregated via nearest neighbor.
- Empirical identification:
  - Regression (1): max(NDVI)_{i,t} = β Droughts_{i,t} + η_i + δ_t + ε_{i,t}; droughts = number of months with SPEI < -1.5 before peak NDVI in an agricultural year. Standard errors clustered at cell level; estimates run pooled and by country.
  - Mapping NDVI to yield via regression (2) on Brazilian municipal data; assumed linear relationship Yield = κ + λ max(NDVI) + X + time and municipality fixed effects.
  - Mapping used: ΔYield = λ Δmax(NDVI) = 130 Δmax(NDVI) (equation (3)) based on λ estimate.

### Key findings: Effect of droughts on greenness and yield
- Descriptive statistics (cell-by-year, 2010–2020):
  - Average number of droughts per cell per year (whole sample): 1.37; maximum observed in a cell: 9 in one year.
  - Median of maximum NDVI across cells ranges from 0.66 to 0.77.
- Regression estimates (Table 2):
  - Pooled coefficient β = -0.008658*** (0.000183).
  - Country β coefficients (standard errors in parentheses):
    - ARG: -0.006534*** (0.000352)
    - BRA: -0.002829*** (0.000251)
    - COL: -0.025557*** (0.000556)
    - PER: -0.005598* (0.002567)
    - PRY: -0.05332** (0.01897)
    - URY: -0.021627*** (0.00099)
  - Coefficients are negative and statistically significant; effect sizes vary substantially across countries (around -0.002 in Brazil to around -0.05 in Paraguay).
- Mapping from NDVI to yield (Table 3):
  - λ estimates: 145.2** (62.66) and 130.5*** (46.45) with year and municipality fixed effects.
  - Regression controls: Temperature coefficient -12.09*** (column 1), -9.7*** (column 2); Precipitation coefficient 23.87*** (column 1), 15.87*** (column 2).
- Country-level yield and drought statistics (Table 4):
  - ARG: Yield 43.09; Average Number of Droughts per year 1.00
  - BRA: Yield 52.01; Average Number of Droughts per year 1.96
  - COL: Yield 41.61; Average Number of Droughts per year 1.23
  - PER: Yield 22.29; Average Number of Droughts per year 1.05
  - PRY: Yield 47.55; Average Number of Droughts per year 1.27
  - URY: Yield 32.70; Average Number of Droughts per year 0.93
- Estimated average physical impacts (ΔYield = λ β_i Droughts_i):
  - A farmer in Brazil or Argentina loses between 0 and 1 bushels per acre, or around 1 or 2 percent of agricultural output.
  - A farmer in Colombia, Paraguay, or Uruguay loses between 4 and 8 bushels per acre, or between 8 and 19 percent of agricultural output respectively.
  - In Paraguay, these losses would mean roughly 2 percent of total GDP.

### Heterogeneity and potential drivers
- Drought intensity, timing, and duration:
  - Panel A (SPEI distribution): droughts seem to have similar intensity across countries; Peru experienced more severe droughts during the sample yet had a smaller coefficient.
  - Panel B (timing): timing differs — Paraguay and Uruguay experience droughts mostly during planting; Brazil and Argentina experience them mostly during harvest or outside cultivation cycle.
  - Panel C (duration): duration distributions are similar across countries.
- Interpretation:
  - Soybeans are most vulnerable to drought stress during the early stage of growth; droughts during planting (Paraguay, Uruguay) likely cause larger yield impacts.
  - Results do not control for time-varying, cell-specific adaptation measures (e.g., irrigation, direct seeding, precision agriculture, soybean varieties).
  - Farmers in Paraguay and Uruguay may have less adaptation capacity, contributing to larger observed effects.

### Risk management and agricultural insurance: market features and constraints
- Risk management instruments:
  - Crop diversification, intercropping, flexible use of inputs, vaccines and quarantines for livestock, commodity futures, self-insurance (savings, contingent credit), microfinance, remittances, cooperative/mutual insurance.
- Financial and market context:
  - Agricultural lending penetration: on average about 7 percent of total credit lent by the financial system in 2023 was to the agricultural sector.
  - Latin America agricultural insurance market size estimates:
    - USD 2 billion in 2024 (5 percent of total global market)
    - USD 1.6 billion in 2015
    - USD 780 million in 2009
- Index-based insurance (Box 1): features, benefits, and challenges:
  - Settles claims based on local weather conditions rather than individual damages; reduces underwriting and processing expenses.
  - Primary challenge is basis risk — index may not trigger despite actual losses.
  - Simplicity, transparency, objectivity, and quicker disbursement increase appeal; modelled data enable introduction in markets lacking historical claims.
  - Regulatory transparency and client education are vital for adoption.
- Market penetration and costs:
  - Agricultural insurance accounts for only about 2 percent of regional non-life premiums.
  - Regional penetration: around 0.6 percent of agricultural output (compared to 1 percent in Europe and 5 percent in North America).
  - Paraguay: penetration estimated at 0.03 percent of GDP and 0.6 percent of agriculture output.
  - Brazil: about 20 percent of farming activity is covered by insurance on average (SUSEP).
  - Premium distribution concentration: Brazil, Argentina (and México) account for 90 percent of gross written premiums (GWP).
  - Cost structure: a decade ago total expenses for provision of agricultural insurance in Latin America and Caribbean were estimated to be 11 percent higher than average expenses in other regions, and they were rising.
- Supply-side constraints:
  - High fragmentation of rural clients; complexity of value chains; lack of granular public data; technical demands; high administrative and transaction costs; reinsurance access challenges; reinsurers averse to underwriting MPCI for individual growers due to systemic risk accumulation.
- Demand-side constraints:
  - Limited incomes of small- and medium-sized farms make premiums relatively expensive; limited farmer understanding of insurance benefits.
- Role of reinsurers and fiscal support:
  - 15 reinsurance companies present in the region.
  - Approximately 65 percent of the total direct written premiums for agricultural insurance in the region are conceded to this market.
  - Common public support: premium subsidies, R&D for products, direct purchase/provision of insurance and reinsurance, programs targeted to small and marginal farmers.
  - Country examples:
    - Peru: transition from ad hoc disaster compensation to formal crop and livestock insurance with government subsidies.
    - Brazil: public sector disaster relief alongside subsidized crop insurance.
    - Paraguay: no direct government involvement in insurance provision or subsidization; post-disaster outlays financed by post-disaster loans; administrative forbearance during 2021-22 and 2024 droughts allowed banks to relax loan service obligations while liquidity was available from the central bank.

### Dynamic general equilibrium model: calibration and productivity gains
- Framework:
  - Dual-sector model (agriculture and manufacturing) where agricultural risk affects use of intermediate goods and productivity.
  - Production functions:
    - Manufacturing: Y^m_t = A N^m_t (sector-neutral TFP A; manufacturing is numeraire).
    - Farming: y^a_t = z_t A x_t^γ n^ζ_a_t, where γ + ζ < 1, z_t i.i.d.
  - Households: E_0 Σ β^t u(c_a_t, c_m_t) with u(c_a_t, c_m_t) = α log(c_a_t − a) + (1 − α) log(c_m_t), a > 0 subsistence.
  - Timing: choose x_t prior to shock z_t realization; savings in agricultural goods subject to depreciation δ.
- Exogenous parameter values:
  - β = 0.96
  - α = 0.005
  - ζ = 0.42
  - γ = 0.40
  - p_x = {1.17, 2.28, 1.50}
  - τ (manufacturing labor tax proxy) ψ̂ = {0.36, 0.44, 0.34} used as τ
  - σ_z = {0.419, 0.499, 0.438}
- Calibration targets and model matches: selected values include a = {0.023, 0.026, 0.025}, z bounds, A = {0.140, 0.226, 0.150} with corresponding target/model comparisons reported in the calibration table.
- Model experiments:
  - Comparative experiment: economies without insurance versus with complete insurance (interpreted as an upper bound).
- Quantitative model results (Table 7: percentage changes when introducing insurance):
  - Brazil:
    - Agricultural Productivity: 2.67 %-change
    - GDP per capita: 0.99 %-change
    - Intermediate Input Share: 2.66 %-change
    - Agricultural Labor Share: -1.30 %-change
  - Paraguay:
    - Agricultural Productivity: 7.45 %-change
    - GDP per capita: 2.85 %-change
    - Intermediate Input Share: 9.34 %-change
    - Agricultural Labor Share: -5.89 %-change
  - Uruguay:
    - Agricultural Productivity: 3.60 %-change
    - GDP per capita: 1.51 %-change
    - Intermediate Input Share: 4.50 %-change
    - Agricultural Labor Share: -2.18 %-change
- Mechanisms:
  - Insurance raises the mean real intermediate share and reduces misallocation induced by risk, increasing measured agricultural TFP and allowing higher input use.

### Limitations and possible extensions
- Limitations:
  - Relationship between yield and reflectance (NDVI) may be localized and not easily extendable across areas.
  - Mapping between yield and NDVI is based on Brazilian data and extended to other countries; assumes greenness-to-yield translation identical across countries.
  - Results do not control for time-varying, cell-specific adaptation measures.
- Possible extensions:
  - Availability of granular yield data for all countries to permit country-specific yield functions.
  - Incorporating IPCC RCP scenario projections to assess future drought effects on yield.
  - Further refinement of NDVI measures by overlaying cropland masks or narrowing growing-season windows.

### Policy recommendations and conclusions
- Key empirical conclusions:
  - Droughts produce substantial heterogeneity in soy yield losses across countries: on average, droughts produce close to a 2 percent reduction in soy output in Argentina and Brazil; reductions reach 8 percent in Uruguay and 19 percent in Paraguay.
- Public-private partnership objectives and measures:
  - Spread the risk through pooling (farmer groups, insurers, governments), potentially at regional level.
  - Orient insurance industry focus toward the agribusiness value chain, not only individual farmers.
  - Governments could participate in risk financing on top of catastrophic layers and redefine premium subsidies and public support to avoid moral hazard and indirect support.
- Improve data availability:
  - Governments should invest in public infrastructure to establish agricultural and weather databases that are regular, up-to-date, granular, and historic to enable enhanced risk modeling and index insurance solutions.
- Strengthen regulations:
  - Strengthen regulatory regimes, contract enforcement, and supervisory frameworks; establish clear regulations for indices, triggers and payout structures.
- Reduce costs and expand access:
  - Establish efficient distribution networks and leverage mobile technology, automated weather stations, and remote sensing to reduce administrative and claims settlement costs.
  - Expand coverage to remote regions; leverage mobile money for premium collection and settlement where telecommunication infrastructure permits.
  - Awareness campaigns and financial literacy training to build trust among small farmers.
- Market outlook and cautions:
  - Global agricultural insurance market projected to grow from the current USD 40 to 60 billion by 2030.
  - Agricultural insurance is not one-size-fits-all; calibrated and targeted public policies and proactive social assistance to the most vulnerable farmers should replace reactive policies.

*Source: IMF Working Paper — Introduction, Box 1, and Conclusions (wpiea2025052-print-pdf).*

### Introduction _____________________________________________________________________________________ 3

### Introduction

### Major sections in this content unit
- Extreme Weather Events and Agricultural Output (page 5)
- Risk Management: Agricultural Insurance (page 15)
- Agricultural Insurance and Productivity (page 19)
- Conclusions (page 23)

### Figures listed within the content unit
- Figure 1. Vulnerability and Preparedness for Climate Change (page 3)
- Figure 2. Selected Agricultural Statistics (page 4)
- Figure 3. Approach and Mapping (page 5)
- Figure 4. Crop Calendar for Soy (page 6)
- Figure 5. Yield and Maximum NDVI in Brazil (page 9)
- Figure 6. Agricultural Output Loss from Droughts (page 10)
- Figure 7. Comparing Droughts Across Countries (page 12)
- Figure 8. Benefits of Agricultural Insurance (page 15)
- Figure 9. Agricultural Insurance: Availability and Penetration (page 17)
- Figure 10. Premia and Claims in Brazil and Paraguay (page 18)
- Figure 11. Total Fiscal Support to Agriculture (page 18)
- Figure 12. Models of Support for Agricultural Insurance (page 19)

### Tables listed within the content unit
- Table 1. Descriptive Statistics (page 7)
- Table 2. Regression Results: Effect of Droughts on Greenness (page 8)
- Table 3. Mapping Between Yield and Greenness: Results (page 9)
- Table 4. Parameters for Loss Estimation (page 10)
- Table 5. Risk Management Strategies and Mechanisms (page 14)
- Table 6. Calibration and Model Fit (page 22)
- Table 7. Model Results (page 23)

### Annexes and reference materials included
- Annex I. Literature Review (page 26)
- Annex II. Agricultural Insurance Products by Type (page 29)
- References (page 29)

*IMF WORKING PAPERS Extreme Weather Events, Productivity and Agricultural Insurance: Evidence from Latin America — Introduction (wpiea2025052-print-pdf)*

### Introduction

### Introduction

### Overview
- Farming is subject to risk; the increasing frequency of extreme weather events affects growth, productivity, and wellbeing.
- Agricultural insurance is part of farmers’ risk management tools in both developing and industrialized economies, but existing data are scattered across regions, risks, and instruments.
- The paper concentrates on selected South American countries with large agricultural sectors exposed to many extreme events and limited risk management strategies.
- Countries in the sample are prepared for climate change to varying degrees, with Chile and Uruguay leading efforts (Figure 1).

### Sample and Importance of Agriculture
- Study sample: six South American countries — Argentina, Brazil, Colombia, Peru, Paraguay, and Uruguay.
- On average, agriculture constitutes about 8 percent of GDP in the sample and about 40 of the total value of its exports (Figure 2).
- Paraguay, Argentina, and Uruguay have the largest agricultural sectors as a percentage of GDP and exports.
- Rural population shares vary: about 37 percent in Paraguay to 4 percent in Uruguay; Peru employs the highest number of people, notably women, in agriculture.
- Soy constitutes almost 60 percent of all agricultural production in Paraguay, and above 40 percent in Argentina, Brazil, and Uruguay.
- Production figures exclude the informal sector and ancillary industries such as food processing and distribution; agriculture is a major source of foreign exchange and fiscal revenues.

### Extreme Weather Risks and Regional Exposure
- Extreme weather events (floods, droughts, frosts, heavy rainfall) are increasing in frequency, intensity, and coverage.
- Drought affects agricultural production in almost all countries; El Niño/a events have divergent effects across countries.
- Hailstorm risk is important in Argentina, Uruguay, and southeastern Brazil; northeastern Argentina, eastern Paraguay, Uruguay, and southern Brazil are heavily exposed to tornadoes.
- Assessing economic impacts of extreme weather on output and effects of risk management strategies is crucial for policy.

### Data and Methodology
- Focus of empirical analysis: effects of droughts on agricultural output.
- Built a unique dataset of high-frequency satellite and weather station data:
  - Normalized Difference Vegetation Index (NDVI) — a greenness index used as a proxy for agricultural output in soy.
  - Standard Precipitation and Evapotranspiration Index (SPEI) — used to define droughts; drought defined as a month where SPEI < -1.5.
  - Land use data from MapBiomas to identify soy-producing cells (detailed crop categories available for Brazil).
  - Brazil municipal soy yield data from IBGE (2010–2020) used to map NDVI to actual yields.
  - Crop calendar from USDA to align planting and harvesting periods and select peak NDVI during harvest.
- Spatial harmonization:
  - Data adjusted to uniform 0.05° resolution; SPEI disaggregated via bilinear interpolation; higher-resolution land use data aggregated via nearest neighbor.
  - Fuzzy matching strategy: median temperature, monthly precipitation, and median NDVI for Brazilian soy cells used to infer potential soy cells in other countries.
- Empirical identification:
  - Regression (1): max(NDVI)_{i,t} = β Droughts_{i,t} + η_i + δ_t + ε_{i,t}; droughts = number of months with SPEI < -1.5 before peak NDVI in an agricultural year.
  - Standard errors clustered at cell level; estimates run pooled and by country.
  - Mapping NDVI to yield via regression (2) on Brazilian municipal data; assumed linear relationship Yield = κ + λ max(NDVI) + X + time and municipality fixed effects.
  - Mapping used: ΔYield = λ Δmax(NDVI) = 130 Δmax(NDVI) (equation (3)) based on λ estimate.

### Key Findings: Effect of Droughts on Greenness and Yield
- Descriptive statistics (cell-by-year, 2010–2020):
  - On average, cells experience 1.37 droughts per year; maximum observed in a cell is 9 droughts in one year.
  - Median of maximum NDVI across cells ranges from 0.66 to 0.77.
- Regression results (Table 2):
  - Pooled coefficient β = -0.008658*** (standard error 0.000183).
  - Country coefficients (β) (standard errors in parentheses):
    - ARG: -0.006534*** (0.000352)
    - BRA: -0.002829*** (0.000251)
    - COL: -0.025557*** (0.000556)
    - PER: -0.005598* (0.002567)
    - PRY: -0.05332** (0.01897)
    - URY: -0.021627*** (0.00099)
  - Coefficients are negative and statistically significant; effect sizes vary substantially across countries (around -0.002 in Brazil to around -0.05 in Paraguay).
- Mapping to yield (Table 3):
  - λ estimates: 145.2** (column 1), 130.5*** (column 2 with year and municipality fixed effects).
  - Regression controls: Temperature coefficient -12.09*** (column 1), -9.7*** (column 2); Precipitation coefficient 23.87*** (column 1), 15.87*** (column 2).
- Estimated average yield losses (ΔYield = λ β_i Droughts_i), using parameters in Table 4:
  - Country-level average soy yields (bushels/acre) and average number of droughts per year:
    - ARG: Yield 43.09; Average Number of Droughts per year 1.00
    - BRA: Yield 52.01; Average Number of Droughts per year 1.96
    - COL: Yield 41.61; Average Number of Droughts per year 1.23
    - PER: Yield 22.29; Average Number of Droughts per year 1.05
    - PRY: Yield 47.55; Average Number of Droughts per year 1.27
    - URY: Yield 32.70; Average Number of Droughts per year 0.93
- Aggregate interpretation of losses (Figure 5):
  - A farmer in Brazil or Argentina loses between 0 and 1 bushels per acre, or around 1 or 2 percent of agricultural output.
  - A farmer in Colombia, Paraguay, or Uruguay loses between 4 and 8 bushels per acre, or between 8 and 19 percent of agricultural output respectively.
  - In Paraguay, these losses would mean roughly 2 percent of total GDP.

### Heterogeneity and Potential Drivers
- Investigated drought intensity, timing, and duration:
  - Panel A (SPEI distribution): droughts seem to have similar intensity across countries; Peru experienced more severe droughts during the sample yet had a smaller coefficient.
  - Panel B (timing): timing differs — Paraguay and Uruguay experience droughts mostly during planting; Brazil and Argentina experience them mostly during harvest or outside cultivation cycle.
  - Panel C (duration): duration distributions are similar across countries.
- Interpretation:
  - Soybeans are most vulnerable to drought stress during the early stage of growth; droughts during planting (Paraguay, Uruguay) likely cause larger yield impacts.
  - Regression controls for time fixed effects and cell fixed effects; results do not control for time-varying, cell-specific adaptation measures (e.g., irrigation, direct seeding, precision agriculture, soybean varieties).
  - Farmers in Paraguay and Uruguay may have less adaptation capacity, contributing to larger observed effects.

### Risk Management and Agricultural Insurance
- Risk management methods include: crop diversification, intercropping, flexible use of inputs, vaccines and quarantines for livestock, commodity futures, self-insurance (savings, contingent credit), microfinance, remittances, cooperative/mutual insurance.
- Commercial credit and financing:
  - Input suppliers and traders finance commercial soybean farmers in Brazil, Argentina, and Paraguay.
  - Agricultural lending penetration is low: on average about 7 percent of total credit lent by the financial system in 2023 was to the agricultural sector (variation across countries).
  - Loan access can motivate farmers to acquire agricultural insurance (mandatory by lenders or to obtain better credit conditions).
- Role and benefits of agricultural insurance (Figure 8):
  - Strengthens resilience by compensating crop losses and supporting food security and economic stability.
  - Supports productivity and growth by enabling farmers to invest (purchase quality inputs, adopt better practices).
  - Enhances budget stability for governments by transferring some risks to the private sector and enabling proactive social protection.
  - Unlocks job creation potential and reduces corruption by promoting transparency and fair, timely payouts.
- Market characteristics and products:
  - Latin America agricultural insurance market size estimated at more than USD 2 billion in 2024 (5 percent of total global market), up from USD 1.6 billion in 2015 and USD 780 million in 2009.
  - Main policies: Multi-Peril Crop Insurance (MPCI), crop-named perils, and index-based insurance; MPCI is most common.
  - Two thirds of premiums are written for crop, named-peril and individual-farmer multiperil crop.
  - NDVI-based crop insurance introduced in 2006 in Mexico; index-based and crop revenue insurance are leading innovations.
- Market gaps and constraints:
  - Overview reveals low coverage and a market characterized by many constraints on both demand and supply sides.
  - Cooperatives and mutual insurer schemes accounted for significant shares of premiums in 2015: 61 percent in Paraguay, 45 percent in Argentina, 37 percent in Uruguay.

### Dynamic General Equilibrium Modeling and Productivity Gains
- Using a dynamic general equilibrium model calibrated to sample countries, the paper quantifies gains from broadening insurance coverage:
  - Agricultural productivity enhancements estimated as:
    - Paraguay: 7.5 percent
    - Brazil: 2.7 percent
    - Uruguay: 3.6 percent

### Limitations and Possible Extensions
- Limitations:
  - Relationship between yield and reflectance (NDVI) may be localized and not easily extendable across areas.
  - Mapping between yield and NDVI is based on Brazilian data and extended to other countries; assumes the greenness-to-yield translation is identical across countries.
  - Results do not control for time-varying, cell-specific adaptation measures.
- Possible extensions:
  - Availability of granular yield data for all countries would permit country-specific yield functions.
  - Incorporating IPCC RCP scenario projections could shed light on future effects of droughts on agricultural yield.
  - Further refinement of NDVI measures by overlaying cropland masks or narrowing growing-season windows to reduce nonagricultural vegetation capture.

### Key Statistics and Exact Figures (as reported)
- Sample countries: Argentina, Brazil, Colombia, Peru, Paraguay, Uruguay.
- Agriculture share: about 8 percent of GDP on average in the sample.
- Agriculture share in exports: about 40 (value presented in source as "about 40 of the total value of its exports").
- NDVI data resolution: 0.05° (approximately a 5km x 5km grid).
- SPEI data resolution: 0.5°; drought defined as SPEI < -1.5.
- Average number of droughts per cell per year (whole sample): 1.37; maximum observed in a cell: 9 in one year.
- Median of maximum NDVI across cells ranges from 0.66 to 0.77.
- Regression pooled β: -0.008658*** (0.000183).
- Country β coefficients:
  - ARG: -0.006534*** (0.000352)
  - BRA: -0.002829*** (0.000251)
  - COL: -0.025557*** (0.000556)
  - PER: -0.005598* (0.002567)
  - PRY: -0.05332** (0.01897)
  - URY: -0.021627*** (0.00099)
- Mapping coefficient λ (with fixed effects): 130.5*** (46.45); alternate estimate 145.2** (62.66).
- Yield (bushels/acre) and average droughts per year by country (Table 4):
  - ARG: 43.09; 1.00
  - BRA: 52.01; 1.96
  - COL: 41.61; 1.23
  - PER: 22.29; 1.05
  - PRY: 47.55; 1.27
  - URY: 32.70; 0.93
- Estimated productivity enhancements from broader insurance coverage:
  - Paraguay: 7.5 percent
  - Brazil: 2.7 percent
  - Uruguay: 3.6 percent
- Latin America agricultural insurance market size estimates:
  - USD 2 billion in 2024 (5 percent of total global market)
  - USD 1.6 billion in 2015
  - USD 780 million in 2009

*Source: IMF Working Paper — Introduction (wpiea2025052-print-pdf).*

### Box 1. Index-Based Agricultural Insurance

### Box 1. Index-Based Agricultural Insurance

### Index-Based Insurance: Features, benefits, and challenges
- Settles claims based on local weather conditions rather than on the specific damages suffered by an individual.
- Covers losses often excluded by traditional insurance while significantly reducing expenses associated with claim underwriting and processing.
- Use of modelled data enables introduction in markets lacking historical claims data for accurate actuarial evaluations.
- Simplicity, transparency, objectivity, and quicker disbursement increase appeal to low-income individuals who may lack experience, awareness of risks, or trust.
- Primary challenge: basis risk — possibility that a claim is not activated by the index despite an actual loss occurring, creating pressure to align triggers with actual losses.
- Continuous innovation in index development and advances in satellite technology are crucial for expanding reach.
- Increasing education and awareness among clients is essential to promote adoption.
- Regulatory transparency and support are vital for successful implementation of parametric insurance.
- Reference: FSI Insights (2024): Uncertain waters: can parametric insurance help bridge NatCat protection gaps?

### Agricultural insurance market size and penetration in the region
- Agricultural insurance accounts for only about 2 percent of regional non-life premiums.
- Regional penetration: around 0.6 percent of agricultural output (compared to 1 percent in Europe and 5 percent in North America).
- Country examples:
  - Paraguay: penetration estimated at 0.03 percent of GDP and 0.6 percent of agriculture output (World Bank (WB), 2023).
  - Brazil: about 20 percent of farming activity is covered by insurance on average (Superintendência de Seguros Privados (SUSEP)).
  - Colombia: schemes being implemented but current penetration and geographical coverage are limited; some schemes not scaled beyond pilot stage (Country Climate and Development Report (CCRD), WB, 2023).
  - Peru: non-life insurance penetration is one of the lowest; agriculture’s share is even lower; of the 18 active insurance companies in Peru, only five offer catastrophic insurance (CCRD, WB, 2023).
- The SAC catastrophic instrument covers only 8 of 25 regions and about 8.9 percent of small and medium-sized farmers on average.

### Premium distribution, costs, and recent loss experience
- Premium distribution is uneven across sub-lines and countries: Brazil, Argentina (and México) account for 90 percent of gross written premiums (GWP).
- Cost of insurance is higher compared to other regions; a decade ago total expenses for provision of agricultural insurance in Latin America and Caribbean were estimated to be 11 percent higher than average expenses in other regions, and they were rising.
- Severe 2021-22 drought effects:
  - Paraguay: losses exceeded premiums by 42 percent.
  - Brazil: significant increase in premiums following the drought.
- Brazil: evolution chart indicates premia and claims series (SUSEP); average: 116.3 (unit context from figure).

### Supply- and demand-side constraints
- Supply-side challenges:
  - High fragmentation of rural clients.
  - Complexity of agricultural value chains.
  - Lack of granular public data on agricultural and weather trends.
  - Significant technical demands requiring insurers to develop specialized technical units or outsource underwriting.
  - High administrative and transaction costs.
  - Reinsurance access challenges in developing areas due to small business volumes, lack of data, and regulatory impediments.
  - Reinsurers averse to underwriting MPCI for individual growers due to systemic risk accumulation (drought, flood).
  - Increasing frequency of extreme weather has led reinsurers to adjust risk strategies, increasing premiums.
- Demand-side challenges:
  - Limited incomes of small- and medium-sized farms make premiums relatively expensive.
  - Farmers’ limited grasp of insurance benefits leading to confusion or skepticism.
- Legal and regulatory weaknesses at domestic level, including in application of international regulatory standards.
- Industry note: main challenge for crop revenue insurance is lack of developed local commodity futures markets with enough open interest for forward positions.

### Role of reinsurers and fiscal support
- Reinsurers: 15 reinsurance companies present in the region.
- Approximately 65 percent of the total direct written premiums for agricultural insurance in the region are conceded to this market.
- Crop hail and named-peril crop insurance programs have adequate reinsurance capacity because this business is not subject to catastrophic losses.
- Government participation motives:
  - Systemic risk and constrained reinsurer capacity.
  - Absence of insurance infrastructure and private sector services in rural areas.
  - Prohibitively high start-up costs for developing agricultural insurance products.
  - High administrative costs of underwriting.
  - Farmers’ affordability issues due to high premiums.
- Common forms of public support:
  - Funding of premium subsidies.
  - Research and development of insurance products.
  - Direct purchase or provision of insurance and reinsurance.
  - Setup of specific insurance programs targeted to small and marginal farmers.
- Country examples:
  - Peru: transition from ad hoc disaster compensation to formal crop and livestock insurance programs executed by private sector with government subsidies for premiums or reinsurance protection.
  - Brazil: public sector disaster relief alongside subsidized crop insurance.
  - Paraguay: no direct government involvement in insurance provision or subsidization; post-disaster or emergency outlays financed by post-disaster loans; administrative forbearance during 2021-22 and 2024 droughts allowed banks to relax loan service obligations while liquidity was available from the central bank.
- Regulatory recommendation: need to establish clear regulations for indices, triggers and payout structures, and provide enabling supervisory framework with clear expectations and guidance (BIS, 2024).

### Models of public-private cooperation
- Various models observed in practice with varying cost-benefit outcomes (Figure 12; source Iturrioz (2009)).
- Historical note: agricultural insurance was provided by public sector insurers from the 1950s up to the end of the 1980s but were terminated by 1990 due to poor results.

### Dynamic general equilibrium model: linking insurance to productivity
- Framework: dual-sector model (agriculture and manufacturing) emphasizing how agricultural risk affects use of intermediate goods and productivity.
- Key mechanisms:
  - Intermediate inputs chosen prior to realization of random productivity shocks (z_t) cause risk to influence optimal input use.
  - Subsistence requirement implies decreasing relative risk aversion and makes equilibrium price of agricultural output influence risk aversion through subsistence cost.
  - Risk induces farmers to use fewer intermediate inputs (self-insurance), leading to misallocation and disproportionate impacts on poorer households.
- Production technology:
  - Manufacturing: Y^m_t = A N^m_t (sector-neutral TFP A; manufacturing is numeraire).
  - Farming: y^a_t = z_t A x_t^γ n^ζ_a_t, where γ + ζ < 1, z_t i.i.d. across households and time, intermediate input price p_x ≥ 1.
- Households maximize expected utility E_0 Σ β^t u(c_a_t, c_m_t) with u(c_a_t, c_m_t) = α log(c_a_t − a) + (1 − α) log(c_m_t), a > 0 is minimum subsistence agricultural consumption.
- Timing: ordering phase (choose x_t), shock z_t realized, production phase (labor allocation, production, consumption, savings decisions). Savings are in agricultural goods subject to depreciation δ.

### Calibration strategy and parameter choices
- Baseline calibration for Brazil, Paraguay, and Uruguay integrating aggregate statistics and micro-level NDVI variation.
- Parameters set exogenously:
  - β = 0.96
  - α = 0.005
  - ζ = 0.42
  - γ = 0.40
  - p_x = {1.17, 2.28, 1.50}
  - τ (manufacturing labor tax) estimated via regressions; ψ̂ = {0.36, 0.44, 0.34} used as τ.
  - σ_z = {0.419, 0.499, 0.438} computed from NDVI residuals after cell and year fixed effects.
- Parameters calibrated jointly:
  - Shock distribution: mean-zero truncated log-normal with bounds {z, z̅} chosen to match top and bottom 1% of NDVI distribution.
  - Subsistence requirement a calibrated to match agricultural employment share.
  - Sector-neutral TFP A set to reflect real GDP per worker ratio relative to the U.S.
- Table of calibration and model fit (selected parameter values and model/target matches):
  - α = 0.005 (target and model equal for BRA, PRY, URY)
  - β = 0.960
  - a: set values {0.023, 0.026, 0.025} with target values {0.087, 0.170, 0.080} and model values {0.047, 0.154, 0.082} for BRA, PRY, URY respectively.
  - γ = 0.400; ζ = 0.420
  - σ_z = 0.419, 0.499, 0.438
  - z lower bounds: 0.760, 0.725, 0.780 (BRA, PRY, URY)
  - z̅ upper bounds: 0.900, 0.875, 0.930 (target) and model values 1.148, 1.111, 1.152 (BRA, PRY, URY) and 1.091, 1.117, 1.185 in another model column.
  - τ = 0.360, 0.440, 0.340
  - p_x = 1.170, 2.280, 1.500
  - A = 0.140, 0.226, 0.150 (parameter) with target values 3.750, 4.550, 2.320 and model values 4.073, 4.893, 2.773 (BRA, PRY, URY).

### Model experiments and quantitative results
- Comparative experiment: economies without insurance versus with complete insurance (interpretation as an upper bound given data and modeling constraints).
- Key model results (Table 7: percentage changes when introducing insurance):
  - Brazil:
    - Agricultural Productivity: 2.67 %-change
    - GDP per capita: 0.99 %-change
    - Intermediate Input Share: 2.66 %-change
    - Agricultural Labor Share: -1.30 %-change
  - Paraguay:
    - Agricultural Productivity: 7.45 %-change
    - GDP per capita: 2.85 %-change
    - Intermediate Input Share: 9.34 %-change
    - Agricultural Labor Share: -5.89 %-change
  - Uruguay:
    - Agricultural Productivity: 3.60 %-change
    - GDP per capita: 1.51 %-change
    - Intermediate Input Share: 4.50 %-change
    - Agricultural Labor Share: -2.18 %-change
- Mechanisms driving productivity gains:
  - Insurance raises the mean real intermediate share, directly increasing agricultural labor productivity.
  - Elimination of risk improves allocative efficiency among farmers (reduces misallocation) even at a constant mean realization, enhancing measured agricultural TFP.
- Cross-country heterogeneity:
  - Larger effects in Paraguay tied to being poorer and less productive, facing less favorable input costs, and encountering higher, more dispersed agricultural risks leading to more misallocation.

*Source: IMF Working Paper — Box 1. Index-Based Agricultural Insurance (content unit: wpiea2025052-print-pdf).*

### Conclusions

### wpiea2025052-print-pdf - Conclusions

### Key findings on weather shocks and agricultural output
- Using satellite data, the effect of a weather shock (drought) on agricultural output can be mapped.
- Regional influences significantly impact agricultural output, displaying considerable variation:
  - On average, droughts result in a close to 2 percent reduction in soy output in Argentina and Brazil.
  - The decrease is substantially larger in other countries, reaching 8 percent in Uruguay and 19 percent in Paraguay.
- The disparity in drought impacts could be attributed to:
  - Timing of droughts occurring at different stages of the crop cycle across countries.
  - Differences in the adoption of drought-resistant seeds and technologies.
- Potential benefits of other adaptation measures on agricultural productivity:
  - Paraguay: 7.5 percent enhancement.
  - Brazil: 2.7 percent enhancement.
  - Uruguay: 3.6 percent enhancement.

### Insurance market outlook and barriers
- As compounding shocks become more prevalent, innovative insurance solutions are increasingly critical.
- Global market agricultural insurance is projected to grow from the current USD 40 to 60 billion by 2030.
- Multiple impediments to broader development and penetration of insurance in the region:
  - Lack of analytical capacity at a local level, including limited knowledge and information systems to monitor and evaluate agricultural risks appropriately.
  - Expensive reinsurance at the international level, dampening insurers’ capacity to offer crop insurance products at the national level.
- Agricultural insurance is not a one-size-fits-all solution; there is no single strategy for overcoming development challenges.
- Calibrating and targeting public policies is crucial (Obolensky, 2024).
- Proactive social assistance to the most vulnerable farmers with a long-term view should replace reactive policies.

### Public-private partnership model: objectives and measures
- Spread the risk:
  - Farmer groups, insurance companies, and governments could examine pooling agricultural risks, including on a regional level, guided by international examples such as African Drought Insurance and Caribbean Catastrophe Risk Insurance Facility.
  - Consider orienting the insurance industry’s approach from a focus on farmers to a broader focus on the agribusiness value chain.
  - Governments could participate in risk financing on top of catastrophic risk layers to complement reinsurance markets while redefining the role of agricultural insurance premium subsidies and other public support mechanisms to avoid moral hazard or indirect support (e.g., regulatory forbearance), constituting a proactive rather than reactive approach.
- Improve data availability:
  - Governments have a vital role in enhancing data availability for insurers and other key stakeholders (such as modeling and forecasting companies) by investing in public infrastructure.
  - Establish agricultural and weather databases, including at a regional level, that provide regular, up-to-date, granular, and historic data on average yields and rainfall.
  - Such databases would enable enhanced risk modeling and provide the basis for potential index insurance solutions.
- Strengthen regulations:
  - Governments should strengthen regulatory regimes, including contract enforcement, and facilitate access to international good practices on underwriting, policy terms and conditions, and loss adjustment procedures.
  - Regulators should ensure stability, fairness, and transparency of the insurance sector to increase the chances that local insurers can access international reinsurance services.
  - Legal and regulatory frameworks will need to keep up with ongoing changes in product design and delivery mechanisms to maintain a healthy retail and reinsurance market.
- Reduce costs:
  - Establish efficient distribution networks and leverage technology to reduce administrative and claims settlement costs.
  - Mobile technology and automated weather stations have shown effectiveness in Latin America.
  - Experimentation with remote sensing technology for index-based insurance shows promise for reducing operational costs by streamlining risk assessment.
- Improve access and build trust:
  - Expand insurance coverage to include remote regions, particularly in the Andean countries.
  - Technologies, including mobile and internet-based solutions, are increasingly important for delivering insurance.
  - Further growth of mobile money technology in premium collection and settlement will require well-functioning telecommunication infrastructure, necessitating public-private cooperation.
  - Awareness campaigns and financial literacy training can help build trust in and awareness of insurance as a financial product among small farmers.

*Source: Conclusions section, wpiea2025052-print-pdf - Conclusions*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025052-print-pdf.pdf_