1. Introduction
Low-income economies are severely affected by climate change, limiting sustainable development [
1]. These effects are strengthened in regions with underdeveloped financial services [
2,
3]. Unusual weather events lead to disruptions in worldwide food security [
4]. According to recent studies, the effects of climate change, such as droughts and extreme heat conditions, have reduced the yield of staple crops worldwide by approximately 10% [
5,
6].
Agriculture in Colombia is one of the most significant socioeconomic sectors, contributing to 7.4% of the GDP [
7]. Some of Colombia’s main agricultural products are coffee, corn, potatoes, and palm oil, among others. According to the National Agricultural Survey, ENA 2019, the total production recorded was estimated at more than 63 million tons, 67% of which corresponds to the agro-industrial product group [
8]. The export of high-value crops such as coffee accounts for a sizeable portion of the GDP total. In 2020, coffee was the third most exported product in Colombia, with a value of
$2.54 billion (about
$8 per person in the US), and was ranked as the fourth largest coffee-exporting country in the world [
9]. Nevertheless, according to the World Bank [
10], Colombia is highly exposed to hydrometeorological events, known as El Niño and La Niña. These weather events result in drought or excess rainfall which directly impact the agricultural sector and affect the country’s GDP. In 2010/11, La Niña’s damages were estimated at 2% of GDP and about 3.5 million people were affected [
10]. It is estimated that these changes in the amount of precipitation in the next 30 years will harm coffee crops [
11].
Several studies have demonstrated a direct correlation between time series analysis and climate change [
12,
13]. These studies allow for the identification of ways to mitigate the consequences of climate change on agriculture. Trends in agrometeorological and hydrological data sets are frequently evaluated using the nonparametric Mann–Kendall test [
14,
15,
16]. This test allows researchers to determine the significance of a time series trend of weather variables [
17].
“Agricultural index insurance is a financial tool for risk management that has demonstrated promise for encouraging economic development and resilience” [
18]. The main difference between this insurance and standard insurance is that payouts are based on the measure of climate variables such as rainfall and temperature. The values across which indemnity payments are made in this type of insurance are established by set thresholds and restrictions. Thresholds provide the starting point for indemnity payments. Once the threshold is met, the payout grows steadily as the index’s value approaches its limit or maximum indemnity [
19]. Weather index insurance offers advantages and shortcomings (
Table 1).
To highlight some points, index insurance is advantageous as it overcomes problems of traditional insurance such as moral hazard, adverse selection, and high costs; in addition, it can be measured remotely [
26]. Despite these advantages, in index insurance, basic risk arises when the measurement does not match with actual losses because of poorly designed tools [
26] or if the distance between the index measurement location and the crop’s location is far [
20]. One way to handle this risk is by using data or images based on satellites. For instance, payouts can be based on a Normalized Difference Vegetation Index (NDVI), which allows for estimating vegetation conditions from a color map showing relative biomass [
27]. Nevertheless, NDVI sometimes does not include site calibration, and so should not be widely used in the design of index-based insurance products [
28]. The benefits it offers to individual farmers are limited, so there may be more advantages to using a reinsurance tool used by corporations that have a relationship with many agricultural producers, agricultural banks, or cooperatives [
21,
22].
Although the insurance index started to be mentioned in the land regime law in 1944, it was not until the crisis caused by the La Niña phenomenon in 1992 that interest was renewed. As a result, Law 69 was approved in 1993, establishing the regulations for agricultural insurance in the country. This law regulates the entities that can be insurers, which can be public, private, or mixed, but must be under the supervision of the financial superintendence [
29].
The agricultural insurance market has existed in Colombia since 1998. In 1998, the first program was designed for banana crops, which based its indemnity on verifiable losses. Seven years later, two companies launched the first index insurance program for corn and cotton crops, but the results were poor and fueled producer distrust [
10,
29]. The main problem with this program was that farmers could choose between different thresholds that triggered indemnity and the premium price varied depending on the trigger level selected [
18]. Farmers tended to choose the cheapest premium, which resulted in lower coverage. In practice, when crops suffered damage due to rainfall, the insurance did not activate, so no payment was generated, resulting in distrust among banana producers [
30].
By 2017 there was already a presence of public and private traditional insurers in Colombia, which launched programs to underwrite crops, forestry, and livestock. Among the participants are Sura, Mapfre, Seguros Bolivar, Allianz, and Liberty, and also one state insurer, La Previsora. There is also Proagro, which is a Mexican insurance company that was registered in Colombia. This company promotes a range of agricultural insurance products and services to the market [
31].
In 2018, an emerging company formed by a consortium of insurers launched its first pilot program of weather-indexed insurance for coffee producers in Caldas, a Colombian department. A year later, it received premium subsidies from the government, which allowed it to cover more production costs at risk and offer its insurance to other coffee producers [
32]. In 2020, this program was expanded, reaching approximately 9000 smallholders and 2200 hectares. This can be interpreted as the sustainable success of the program [
33]. The insurance offered in the program covers farmers against heavy rains and floods and shortages based on the analysis of available information from weather stations and satellite data. For index-based insurance programs, satellite weather data has been increasingly employed in developed economies [
34]. However, in low-income economies, satellite information is difficult to access.
In Colombia, the average penetration rate of crop insurance between 2010 and 2015 was only 1.76% (hectares seeded vs. insured hectares) [
35]. The limited traditional compensation processes available for Colombian crops have typically relied on traditional insurance frameworks based on yield loss or damage [
11]. Some reasons for the low coverage rates in the country are related to local conditions, such as regional conflict and the resulting insecurity, which makes field inspections difficult [
36]. Although there are many studies on weather-indexed insurance in the literature, there is still a lack of research that includes product design in low-income economies [
19].
State-of-the-art studies show diverse types of index insurance models where indices are mostly built from mereological, temperature, meteorological drought, hydrological drought, climate, and vegetation indices. For the methods of yield–index relationship estimation, the most common methods used to build indices are regression, correlation, copulas, production functions, and, more recently, machine-learning methods [
20]. Regarding the risk evaluation methods, most studies focus on risk-reducing effectiveness evaluation methods, such as variance or semi-variance deviation, value at risk, or conditional value at risk [
37,
38,
39,
40,
41]. However, few of these methods address the relevance of measuring future financial risks in highly volatile regions. For example, Siebert [
42] used a Gerrit Skill Score (GSS) to assess the way indices forecast payouts or non-payouts during loss and non-loss years. This was seen as advantageous, as the GSS score weighs both sides of basis risk, assigning false payout and failure to pay. Recommendations from these state-of-the-art reviews include more in-depth studies emphasizing index derivation during critical crop growth periods and the employment of models to aid in index design that considers future climate conditions. This study addresses these two recommendations.
Referring to the costs for index insurance programs, Kerer [
43] states that one of the main shortcomings in these programs is the high cost for low-income farmers in developing countries, as most of their income is absorbed by their basic needs such as food and housing. This results in low uptake of this insurance tool. Di Marcantonio [
24] states that while high premium prices for weather index insurance programs are the main constraint when adopting weather index insurance, this is offset by public subsidies and support donors who help to promote insurance expansion. Some studies discuss whether public subsidies are the best way to promote the use of index insurance. Instead, these authors promote the use of credit subsidies and fertilizer subsidies in conjunction or separately. Ricome et al. [
44] showed first that the potential benefits of public subsidized weather index insurance are more evident in the driest areas. Second, in zones with extreme precipitation, insurance premiums that are publicly subsidized by 60% induced farmers to adopt insurance; however, this incentive was less efficient than subsidizing credits and technologies such as fertilizers, or just cash transfers. One Colombian index insurance example of expansion is the Cafe Seguro program. In this program, one of the key factors that promote the adoption of index insurance in coffee crops has been the close work by insurers with farmers. Farmers receive clear information about the rationale of the insurance tool, and questions are answered regarding their needs and the expectations of the insurance program. In addition, these insurance programs are successful because of recent public and private subsidies [
33].
This paper aims to propose a new methodology based on a data-driven framework that allows us to explore the use of different statistical models in the design of indexed insurance in agriculture. The methodology considers the effect of future changes in precipitation levels in low-income economies such as Colombia. We believe that the results from this research could be useful in the design of new policies that would benefit the actors involved in Colombian agriculture.
4. Conclusions
This study offers a different methodology to quantify the financial weather risk in agriculture. It is done by using a basic index insurance model to compute future payouts and a ratio of payments in exceedance. This can be applied to short time series precipitation data from weather stations near a crop’s productive zones. This new methodology allows us to identify highly volatile weather crop production areas in the current- and mid-term by applying a data-driven approach such as SVD techniques and a financial ratio that accounts for a risk metric. This metric comprises the POERs or payouts in exceedance which could be interpreted as future disbursals due to tendency components.
Studies highlight the need to promote insurance programs to acquire financial resilience, especially in low-income economies whose income relies on agricultural activities. This study suggests a specific methodology to assess financial risk involving future weather instances in crop seasons. This is because state-of-the-art index insurance reviews recommend deeper research to model and assist future index insurance scenarios, as well as an emphasis on index derivation during critical crop seasons. We found that in the mid-term, a hypothetical financial entity would have to disburse from 20 to 60% of payouts in addition to regular seasonal disbursals due to the tendency of weather components in dry coffee seasons.
As is the case for many other countries where a significant part of their productive agricultural activities occurs in risky-weather regions, it is necessary to account for possible extraordinary disbursements or overpayments before losses occur. In the case studied here, these regions are the villages located in (i) Caldas: Samana, Cocorna; (ii) Antioquia: Briceño, San Carlos, and San Francisco; and (iii) Santander: El Guacamayo. Using the POERs measures recently proposed in the literature, it is possible to identify risk regions with an increasing trend of payments that could impact the quality of coffee and the sustainability of ongoing insurance programs.
Public and private subsidies have been successfully implemented and have promoted the expansion of index insurance in the department of Cauca, Colombia. One example is the program Cafe Seguro. One of the key factors of this program’s success is the use of public and private subsidies along with fostering a close working relationship with farmers.
This analysis can be implemented and customized in other crops and regions to promote the wider use of index insurance programs. Therefore, programs should consider the implementation of insurance products by maintaining a close work relationship with farmers and by funding programs in order to ensure their adoption and the reduction in costs.
These results could be improved with the inclusion of more weather indices, as well as larger time series to capture climate changes. Crop yield data could be used to enhance the limits and thresholds of the index model. Nevertheless, we believe that this study serves as a point of reference for further debate and the elucidation of Colombia’s agricultural development efforts.