Drought is one of the main environmental risks in Mediterranean Europe. Climate change predictions indicate that droughts will intensify in the 21^st century in the region (IPCC, 2012). One of the main adaptation tools to climate change in arid zones is irrigation. Drought does not only affect rainfed agriculture; it also affects the availability of water for irrigation. Meteorological droughts predate hydrological droughts, reducing the availability of surface and groundwater sources, and giving rise to water shortages both for consumption uses and for environmental flows. Water scarcity in irrigated farming systems results in potentially greater economic losses than in rainfed systems, due to the higher investments being made on the farm.

To manage water scarcity, deficit irrigation has been studied and implemented, particularly to ligneous crops, showing satisfactory results (Ruiz-Sanchez et al., 2010). Water markets are commonly used in Australia, Chile and the United States under widely different regulatory frameworks, facilitating the reallocation of water resources among users, therefore allocating water to high-value uses and improving water use efficiency. At the European Union (EU) level, only Spain has developed a regulatory framework to permit water trading (Easter and Huang, 2014; Rey et al., 2014). Insurance is also a means to manage water supply instability (Pérez & Gómez, 2013; Rey et al., 2015).

The Spanish Crop Insurance System (SCIS) is one of the most advanced and widespread systems in the EU, serving as a benchmark for international comparisons of insurance systems (OECD, 2011).Yet, irrigated agriculture is unprotected against losses due to the lack of water for irrigation. In Spain, irrigated land covers 13% of the useful agricultural surface, contributing to 50% of the total agricultural production value (INE, 2012). In spite of the recurrent drought episodes that affect Spain (Estrela & Vargas, 2012), irrigated surface is still increasing. In the period 2004-2014, irrigated surface increased on average 1% per year (MARM, 2014a). In consequence, there is an evident need for an insurance policy to cover the risk of water scarcity, as has already been demanded by the agricultural sector.

As far as we know, the only experience on drought insurance for irrigated agriculture is implemented in the USA. The Multiple Peril Crop Insurance program in the USA covers the unexpected failure of irrigation water supply. Irrigated yield losses are covered when water shortages are due to a naturally occurring event and cannot be foreseen at the time of the insurance take-up (prevented planting insurance is a possible alternative when a water shortage is expected). The insurance scheme has detailed provisions in order to avoid moral hazard. Besides, decreased water allocations due to a diversion of water for environmental reasons, selling water to municipalities, or other causes are not covered (Risk Management Agency - Topeka, 2015).

Several researchers have proposed other insurance schemes covering water shortage risk. Traditional schemes, using crop production functions or crop simulation models to assess the economic impact of drought have been proposed in some irrigated regions of Spain (Quiroga et al., 2011; Pérez & Gómez, 2013; Ruiz et al., 2013, 2015). Index insurance mechanisms have been proposed in the literature seeking to cover financial risks for water managers (Brown & Carriquiry, 2007; Zeff & Charaklis, 2013) and offering coverage for agricultural production (Zeuli & Skees, 2005; xLeiva & Skees, 2008).

The aim of this article is to propose an insurance tool to protect irrigators against water scarcity in the framework of the SCIS. The tool would mitigate the economic cost of water shortages. The novelty of our approach, with respect to other insurance schemes proposed, is the use of an expert panel formed by 8 individuals representing all stakeholders to select which design is more adapted to the needs of the Spanish irrigators. The Panel included experts and stakeholders related to both water management in the Spain and the SCIS. As a result of the expert panel (the next section offers a detailed description), the best option to ensure irrigated crops in Spain would be a Hydrological Drought Index Insurance (HDII) contracted by the Irrigation District (ID). Index insurance to protect irrigators against water scarcity had already been proposed in the literature by Zeuli & Skees (2005) and Leiva & Skees (2008). However, we propose a simpler design, detail the conditions that the ID should fulfill in order to be insured under this policy, and analyze the hedging effectiveness of the instrument comparing the gross margin of the ID with and without the insurance contract. Similarly to Brown & Carriquiry (2007), we propose a unitary indemnity, but based on crop production and adapted to the ID. HDII is applied to the Bardenas Irrigation District V (ID-V). Besides, to respond to the interest shown by some expert panel participants, we discuss the possibility to subsidize the proposed insurance scheme within the EU legislative framework.

The proposed insurance scheme has some issues related to its implementation that merit a few comments. Firstly, the implementation of the collective policy may confront some difficulties since all irrigators in the ID should agree on the decision to insure and on contract conditions. In case farmers from the ID do not find a consensus that satisfies all of them, one alternative would be to establish individual contracts (so contract conditions C1 & C4 would no longer be applicable). Individual contracts following the scheme proposed can only be established in the case water allocation between the farmers is fixed and not subject to the irrigators water allocation’s requests. However, this is unlikely, since usually irrigators make a formal request to the ID concerning the water allocation they need. Then, ID tries to meet all irrigators’ requests taking into account water availability and water rights. As a consequence, insurance is to be based on the average water allotment granted to the ID by the Basin Agency. The insurance design would then offer excessive coverage to irrigators cultivating low water demanding crops and would fall short offering coverage those cultivating high-water demanding crops. Another alternative would be to let the irrigators set freely the liability, as proposed by Zeuli & Skees (2005). The unitary indemnity received would change as a consequence and so would change the premium rate. The scheme would then work as a weather derivative, since the unitary indemnity is no longer estimated from past records of water allotments and GMs in the farm. Therefore, this could lead to some disagreements with crop insurance laws, given that it could provide a coverage beyond the actual farm loss.

Secondly, the insurance scheme requires that water management rules should be implemented in full over the validity period of the insurance policy and no significant changes in water infrastructure should be observed. In case it does not hold, the model and all the calculations will need to be updated. Since there would not be historic data on DI and water allotments for the new situation, we could only rely on water simulation models to generate water allotments and DI values under the new situation. In case of technical change in crop production, it would not affect the index insurance model that estimates water allotments from a DI. However, it might affect the water allotment guaranteed and also the unitary indemnity estimation, since technical changes would result in larger farmer’s income, yielding higher GM for the same water allotment.

Thirdly, the probability and severity of droughts can also be influenced by climate change. Climate change introduces an additional factor of uncertainty in drought risks (Bielza et al., 2008). Increasing drought risk resulting from climate change affects the price of insurance in two ways. First, ambiguity and catastrophe loads may increase because uncertainty associated with future climate change impacts leads insurers to plan for the worst likely scenario when establishing these loads. Historical return periods may not be valid since they might underestimate the likelihood of agricultural losses in the future. Second, increasing drought risk changes the pure risk (Collier et al., 2009). In consequence, insurance parameters have to be adjusted over time to effectively hedge future weather risk (Kapphan et al., 2012).

Finally, unitary indemnity estimation does not completely offset economic losses that might affect ligneous crops in case of drought, especially when drought affects production in subsequent years.

Some limitations specifically related to our findings for the Bardenas case study are associated to data availability. The unitary indemnity estimation and hedging effectiveness analysis rely on provincial and regional data, and not on local data. Besides, some of the crop patterns linked to water allotment scenarios, used for the unitary indemnity calculation, are designed based on the discussion with irrigators, due to the lack of longer data series. This could be a source of inaccuracies in the results that could be addressed with appropriate and longer data series. Collaboration among insurance companies, producers associations, and/or public entities would be required to have access to a more appropriate data. Besides, we have detected the presence of outliers in water allotments in the period analyzed, which had to be addressed including a Binary variable to the model.

Additional insight into the quality of our results is gathered by comparing our findings to previous literature results. Firstly, the liability of our insurance scheme (equal to 752 €/ha) is smaller than the liability reported by Ruiz et al. (2013) in a drought whole-farm insurance with on-field loss adjustment covering irrigated crops, also in Bardenas (equal to 1488 €/ha). This difference is due to the fact that our index insurance scheme takes into account that farmers not being able to irrigate may still have the alternative of producing rainfed crops (when cultivating annual crops), and thus earning some revenue (except for the case of rice, which is cultivated on marginal soils). Secondly, our unitary indemnity estimation can be compared to Lorite et al. (2012) works in a southern river basin in Spain. We obtained a unitary indemnity of 0.115 €/m³ that is in the range of Lorite et al. (2012) results: wheat (0.054 €/m³), sunflower (0.092 €/m³), and maize (0.15 €/m³).

There are some issues about the use of the Aux I, which is required in case the selected DI might be subject to manipulation. One disadvantage of its use is the increase in basis risk. In our case study, the use of an Aux I increases basis loss (0.11% for NFD and 0.03% for FD) and basis gain (0.08% for NFD and 0.8% for FD). Another important fact to take into account is that the Aux I would not reflect some impediments that might prevent water from being delivered to irrigators, such as a breakdown in the canals or conveyance system, or a pollution problem. Under these circumstances, irrigators would not receive an indemnity, but would suffer economic losses. This would need to be clearly stated in the policy wording. If water management is performed according to predefined rules, there would be no need of such an Aux I, which may be warranted when the selected DI (e.g. water stocks in the reservoir) might be subject of manipulation. This would need further discussion with all stakeholders involved.

It is interesting to observe how basis risk is distributed among basis loss and basis gain. Ideally basis loss and basis gain should be similar, so the insurance scheme is favoring neither the farmer nor the insurance company. In Bardenas, basis loss in the analysis period is larger than basis gain in all cases, meaning that the insurance scheme is overestimating water allotments. This is due to the fact that in several years the DI was above the trigger, and irrigators received a water allotment below the guaranteed water allotment. These cases would require further investigation. Despite this, the hedging effectiveness analysis shows that NFD and FD insurance schemes are effectively reducing risk exposure, even considering a 40% premium load to cover administrative and capital costs of the insurance company. Note that premium loading for the FD scheme might be smaller than for the NFD scheme, because the more frequent the payoffs take place, the higher the administrative costs that the insurance company is going to charge.

Results suggest that the insurance scheme would be useful to provide economic stability to IDs. It would constitute a means for irrigators to adapt to climate variability and it could encourage investment in irrigation technologies as a means of adaptation for dryland farmers. It would also promote a more efficient use of the irrigation water. This insurance can be generalized to any ID fulfilling the conditions mentioned in Section “Index insurance design, hedging effectiveness, and contract conditions”. IDs not fulfilling condition C3 might be subject to intertemporal adverse selection. Intertemporal adverse selection comes from the fact that preseason weather information can influence crop insurance decisions (Carriquiry & Osgood 2012), as farmers could use this information to purchase insurance only in years with enhanced drought risk and probability of payout (Luo et al., 1994). Further research is needed to propose other insurance schemes dealing with intertemporal adverse selection.

The use of an expert panel to determine the type of insurance scheme that better fits with the Spanish framework is expected to bridge the gap between theory and practice, contributing to a more attractive and adapted insurance scheme to local conditions. Further discussions with the expert panel exploring the implementation difficulties of the insurance scheme selected and developed would be of interest for the research.

Public support to index insurance is currently accepted in EU legislation, under certain conditions. Specifically, EC (2014) and EU (2013) refer to insurance premiums6 of schemes (i) using “weather indexes (including quantity of rainfall and temperature) established at local, regional or national level” to quantify the economic losses. For that, (ii) “the occurrence of an adverse climatic event has to be formally recognized as such by the competent authority of the Member State concerned” and (iii) financial contribution to premiums for crop insurance shall only be granted for insurance contracts which “cover for loss caused by an adverse climatic event, which destroys more than 30% of the average annual production of the preceding three-year period or the 5-year Olympic average”. Concerning our HDII, although EU legislation and guidelines allows the use of weather indexes (including quantity of rainfall and temperature), streamflows or water stocks (in which our DI is based) are not explicitly mentioned as weather indexes and could be manipulated. In the case streamflows and specially water stocks would not be accepted as weather variables, Aux Is could be used, with the consequent increase in basis risk. The compliance with condition (ii) would not represent a problem as it would easily be reached if the DI triggering the indemnity was at the same time an official indicator that indicates a natural disaster, which is the case in our insurance scheme. Particularly, the FD scheme is triggering a payment whenever the DI indicates a status of Emergency. Lastly, the use of a γ might help comply with condition (iii). In our case study, the FD scheme has established a γ at 8.6% of the water allocation. Although water allocation and annual production might not follow a linear relationship, a γ of less than 10% of water deficit would probably be far from being equivalent to the 30% required (moreover, in the period analyzed, there are not historic records showing a 30% loss in terms of water allotment). Although the insurance scheme in our case study will not be considered to be compatible with EU legislation7, FD scheme in other regions would be, provided irrigators would be exposed to water shortages destroying more than 30% of the average annual production and the DI would be accepted as a weather variable.

The main strength of this insurance scheme is the compatibility with water markets, water banking, transfer of water rights, and groundwater use. Another important strength is that farmers might receive the economic compensation as soon as the DI is measured, far before the end of the crop season, allowing them to have ready cash for any eventuality arising in the crop season, such as the possibility of participating in water market mechanisms.