Opinion: The real obstacle to Brazilian rural insurance

The phrase that "Brazilian rural insurance is bad" has become commonplace in official speeches and hasty analyses. It sounds simple, but it hides a deeper problem, shifting the focus from institutional design and data infrastructure to the easiest scapegoat: insurance itself. When viewed from a distance, the picture is different. What is in crisis is not the idea of ​​rural insurance, nor the premium subsidy mechanism, but the informational base upon which a market is being built that currently covers something between 10% and 15% of the cultivated area, compared to about 60% in the United States, despite formally similar subsidy models.

In 2024, the insured area shrank from approximately 14 million to 7 million hectares, a collapse of around 50%. This was not a sudden loss of interest from producers, nor a rejection of the instrument itself; it was the direct result of budget cuts to the Rural Insurance Premium Subsidy Program (PSR) and an environment of increasing uncertainty in pricing, largely fueled by a scarcity of reliable productivity and climate data at the farm level. As the budget fluctuates and information remains scarce, insurance is criticized for what it is materially unable to deliver.

Mature markets constantly remind us that insurance is, above all, about information management. In the United States, producers wishing to access the federal program must submit, year after year, their production history by plot, the so-called Actual Production History (APH). These records feed into a national database managed by the Risk Management Agency (RMA/USDA), in which each farm carries its own "statistical memory." Rates, coverage, and product design are a function of this information.

In Brazil, the scenario is almost the opposite. The expected productivity of insured crops is frequently derived from municipal statistics from IBGE (Brazilian Institute of Geography and Statistics) or estimates from Conab (National Supply Company). These data are useful for supply and macroeconomic planning, but too weak to discriminate risk at the farm level. A municipality might have an average of 60 sc/ha of soybeans, constructed from farms that harvest 40 sc/ha and others that harvest 80 sc/ha; in the absence of individualized data, the actuary is forced to treat all as if they were averages. The efficient producer finds themselves paying for an asymmetry they did not create; the producer in a marginal area benefits from a rate that does not reflect their real risk. Over time, adverse selection does the rest: those with above-average performance tend to leave, those with below-average performance tend to stay.

The result is reflected in the claims ratio. In the last decade, the sector collected around R$ 14 billion in premiums and paid approximately R$ 40 billion in indemnities. This balance only works with strong public support and a reduction in supply in subsequent harvests, precisely when the producer most needs protection. It's not that insurance "doesn't work"; it's that it's trying to manage risk in a sophisticated way with a data infrastructure that doesn't keep pace with either the technological stage of the field or the size of Brazilian agribusiness.

Within the farm gate, the country operates connected machines, sensors, satellite imagery, and management software that allow for near real-time monitoring of what is happening in each plot. Estimates indicate that more than 70% of medium and large producers currently use some form of digital management tool, from productivity maps to precision meteorology. Outside the farm gate, however, agricultural policy and insurance continue to operate with aggregated data, incomplete records, and outdated historical series.

This dichotomy creates an uncomfortable paradox. The information exists, it is produced daily by the production system itself, but it lacks clear, secure, and well-regulated channels to feed into public and private risk modeling. Producers, cooperatives, and farm management technology companies possess a wealth of productivity, management, and climate data that could significantly improve pricing, but in practice, it remains confined to private silos due to a lack of incentives, interoperability standards, and a data governance policy that provides legal security to all involved.

Meanwhile, the PSR fluctuates. In years with higher funding, such as 2020 and 2021, the insured area increases and approaches 17% of the grain area; in years of cuts, such as 2023 and 2024, it falls again to levels close to 8–10%. The program reacts to the budget, but fails to move forward in terms of informational design. Spending is done on premiums, but the type of data that would make this spending more intelligent in the next cycle is not required in return.

A hasty reading might lead one to conclude that the problem is the subsidy itself. But international comparison suggests otherwise. Countries that today exhibit high penetration of agricultural insurance, such as the United States, Canada, and Spain, also operate with subsidized models. The difference lies less in the principle of public support and more in the predictability and sophistication of the informational counterpart.

Despite legitimate criticisms, the PSR (Program for the Support of Insurance) has proven capable of boosting the market when it has had a stable budget. In 2021, with approximately R$ 1,2 billion in subsidies, more than R$ 60 billion in insured value was protected, and the number of subsidized policies exceeded 200. The basic design—premium support, offerings by private insurers, and state supervision—is in line with international best practices. What is lacking is treating it as a state policy, linked to a deliberate effort to build a database, and not as an annual item adjustable to each fiscal constraint.

To say that rural insurance is "bad" because it depends on subsidies is to ignore that this dependence is the rule, not the exception, in complex climate risk markets. What truly differentiates Brazil from successful cases is that, abroad, each dollar of subsidy brings in return a flow of data: on production, losses, management practices. Here, the subsidy often finances a policy that is born and dies without leaving a minimally usable statistical trace for future calibration.

In recent years, parametric insurance has gained traction in the national debate as an alternative capable of reducing operational costs and making the system more agile. In theory, it makes sense: by linking compensation to a measurable index—rainfall, temperature, vegetation index—much of the subjectivity in claims settlement is eliminated. In countries that have invested decades in dense meteorological networks and continuous historical series, with homogeneous production systems, this model fulfills a role.

In Brazil, however, the reality is harsher. The density of INMET (National Institute of Meteorology) weather stations in rural areas falls short of international recommendations, especially in agricultural frontier areas like Matopiba, precisely where the risk to productivity is most volatile: comparative studies indicate a standard deviation in soybean productivity of over 20% in some of these regions, compared to around 7% to 10% in North American national averages. When the index is measured dozens of kilometers from the field, the underlying risk becomes inescapable: it may rain during the growing season and there may be a lack of rain on the farm, or vice versa, undermining the credibility of the product. And make no mistake, remote sensing data, agroclimatic models, and others depend on the same data to be calibrated to the local context; therefore, the differences still remain, although attenuated.

The technology is available, the financial instruments too, but they all run into the same fundamental obstacle: the lack of consistent historical series, an adequate observational network, and sufficient field data to reliably calibrate indices. Without addressing this deficit, parametric insurance risks being sold as a "silver bullet" and, in practice, only further exposing the disillusionment with insurance.

From an economic standpoint, informational fragility translates into concentration and high costs. The Brazilian insured area is heavily focused on a few crops; soybeans, corn, and wheat account for more than 80% of the insured value, and in some regions: the South and parts of the Southeast and Midwest. High-value-added crops, such as fruits, vegetables, and regional segments, remain on the margins less due to a lack of demand and more due to the absence of minimum data series that would allow pricing these risks without prohibitive premiums. In parallel, the portfolio as a whole suffers from a high average loss ratio, fueled by severe weather events and the inability to distinguish "good risk" from "bad risk" based on evidence.

From the perspective of insurers and reinsurers, the Brazilian picture, viewed from an office in Zurich or London, is that of a country without a single statistical "truth": two official crop data sources with distinct methodologies, flaws in meteorological series, gaps in loss data, and a production base whose intra-municipal variability is largely unknown. The rational response to this type of uncertainty is to charge higher prices, restrict areas, and limit crops. From the producer's side, the interpretation is the opposite: the premium is high, the understanding of the calculation is low, and the perception of fairness is limited. Distrust feeds on itself.

International examples suggest that there is a possible path, but it is neither short nor glamorous. In the United States, APH (Agricultural Preservation Program) has been consolidated over decades. In Canada, provincial public insurers store, plot by plot, the productivity of producers participating in the programs; this data is returned to the producer, who can compare their performance with the regional average, and informs agricultural policy. In Spain, the centralization of policies in the Agroseguro consortium created, by design, a single national database of risks and claims, fed by highly standardized appraisal norms.

Brazil has, on a smaller scale, components that could make up something similar: the Proagro database, with decades of loss records (but in need of a thorough review); the Rural Insurance Atlas, with open data on subsidized policies; the Rural Environmental Registry (CAR), with property geometries; the network of public and private meteorological stations; and the data held by producers, cooperatives, input industries, and technology companies. What is lacking is the political will to organize this architecture, with clear governance rules, robust incentives for sharing, and a long-term horizon in which data is not seen as a byproduct, but as a central asset of public policy.

To say that Brazilian rural insurance is "bad" is, ultimately, to confuse symptom with cause. What exists today is a partially developed, concentrated market, dependent on subsidies and subject to fiscal and climatic shocks that any fragile system would have difficulty absorbing. The difference is that, here, insurance is expected to perform within a system that has not yet built the necessary informational foundations.

Low penetration in the field, the unbalanced relationship between premiums and indemnities, the recent contraction in coverage, and the comparison with markets that demand informational reciprocation in exchange for each dollar of subsidy, all point in the same direction. The central problem is not the choice between model X or Y, nor the abstract discussion about the "appropriate size" of the State. It is the practical decision to treat agricultural data as strategic infrastructure, on par with roads, warehouses, and credit.

Until this decision is made, we will continue in a predictable cycle: good years alleviate the pressure, bad years reignite criticism, budget cuts interrupt expansion trajectories, and the phrase "Brazilian rural insurance is bad" will continue to circulate as an easy diagnosis. The way out of this cycle involves a quieter and more demanding change: patiently and systematically building the information base that will then allow us to demand more from insurance, and rightfully so.

About the author:

Daniel Miquelluti is the co-founder and Head of New Markets at Picsel, an insurtech company specializing in agricultural insurance. An Agricultural Engineer with a master's degree in Experimental Statistics and a doctorate in Applied Economics from Esalq/USP, he has worked for over a decade developing innovative solutions in rural insurance.

*The text above is the responsibility of the author and does not necessarily reflect the opinion of Insper Agro Global.




Are you already following our official channels? Click and follow our whatsapp channel, follow us on LinkedIn And sign up to receive our bi-weekly newsletter. Stay up-to-date with research and knowledge on global agribusiness issues.