One size policy does not fit all: latent farmer groups in crop insurance markets in Finland

Transcription

1 One size policy does not fit all: latent farmer groups in crop insurance markets in Finland Sami Myyrä and Petri Liesivaara Abstract: This paper assesses how farmers differ in their willingness to pay (WTP) for crop insurance. Data from a choice experiment were analysed using the latent class approach to reveal the number of latent groups and differences in farmers WTP for crop insurance in Finland. The analysis identified three homogeneous groups that differed significantly from each other. Farmers in these classes were characterized as catastrophic loss preventers, risk lovers and shallow loss preventers based on their implicit prices for insurance attributes. The outputs provide valuable information when these latent groups are connected to farm typology. The results indicate that the EU Common Agricultural Policy (CAP) is not sufficiently flexible to take into account the differing needs of farmers for agricultural risk management. If the EU is to implement efficient risk management policies for agriculture, flexibility will be needed in the legislation, and shallow loss insurances will also need to be introduced to cover all risk prevention needs equally. Keywords: crop insurance; choice experiment; latent class model; yield; CAP The authors are with the Natural Resources Institute Finland (Luke), Latokartanonkaari 9, Helsinki, Finland. (Corresponding author: Sami Myyrä, sami.myyra@luke.fi.) Farms and farmers are not a homogeneous group. They differ markedly in their objectives regarding farming, risk exposure and attitudes to risk. However, agricultural policies aimed at stabilizing farm income do not take into account these farm-level differences. Decoupled income support is preferred in the Common Agricultural Policy (CAP) of the EU (that is, policies support all farms in the same way, irrespective of differences in farmers preferences and susceptibility to income fluctuations). In contrast, in the USA agricultural policy has shifted away from decoupled support to risk management policies in which farmers receive payments only when they are willing to pay a proportion of the insurance premium themselves. The main tools of this policy include crop and income insurances aimed at reducing income volatility at the farm level. The EU has not had a common policy for yield risk management. However, the markets for crop insurance products are developing in the EU. One of the incentives that the EU is paying increasing attention to is agricultural risk management (Meuwissen et al, 2013). In the CAP, from 2015 onwards, premium subsidies could be as large as 65% of the premiums faced by farmers. At the same time, the EU has limited the crop insurance deductible to at least 30%. The deductible is the share of losses that the farmers must pay before an insurer will pay any compensation. Subsidies for crop insurances based on indices are also permitted. However, basis risk, a fundamental feature of index insurances, is not tolerated (European Commission, 2013). Basis risk determines the efficiency of index insurance. The higher the positive correlation between the yield of an individual farm and the index, the lower the basis risk. Crop insurance is promoted with premium subsidies throughout the world, and with subsidies governments are able to regulate crop insurance markets. Due to international commitments, for example, the pressure to limit the minimum threshold for the farmers deductible is high (WTO, 1994). However, farmers are heterogeneous in terms of their attitudes to risk (Koundouri, 2009). In developing crop insurance markets, one of the main problems to be solved is the setting of appropriate Outlook on AGRICULTURE Vol 44, No 4, 2015, pp doi: /oa

2 premiums and subsidy levels for crop insurance schemes in order to fulfil the policy makers objectives for a high participation rate among farmers. This is a challenging task due to the heterogeneity among farms regarding their risk positions. Thus, it is typical that high-risk farmers start to become overrepresented and low-risk farmers to be underrepresented in the risk pool, which leads to losses on the side of the insurers (Goodwin et al, 1994). In Finland, the Crop Damage Compensation (CDC) scheme has been the main tool to compensate for crop losses. The scheme is fully financed by the government and free of charge for farmers. However, in 2016, the CDC scheme is to be abolished. The scheme has been found to suffer from major moral hazard problems (Myyrä and Pietola, 2011); however, the main problem relates to its design. Compensation payments are based on a threshold level determined from the average yield measured at the regional level. Thus, farms operating with high yield levels are not covered due to the very large deductible. This one-size-fits-all policy has also been extensively criticized by farmers. In this study, our aim was to identify separate farmer groups that differ according to their preferences regarding crop insurance attributes. We applied a choice experiment, a stated-preference technique, to crop insurance markets. Thus, we dealt with a hypothetical insurance product about which the farmers had no pre-existing knowledge. We allowed the choice experiment data the farmers choices to reveal underlying and latent groups. It is important from an agricultural policy perspective to recognize these groups, since one of the key mechanisms through which agricultural support policies, even decoupled ones, may influence production decisions in their effect on farmers risk aversion (Hennessy, 1998; USDA, 2004; Sckokai and Antón, 2005; Koundouri et al, 2009). Farmers preferences regarding risk management differ according to both farm and farmer characteristics. Multiple factors have been identified to explain the demand for crop insurance (Sherrick et al, 2003; Smith and Baquet, 1996; Goodwin, 1993; van Asseldonk et al, 2002). In addition, farmers preferences and perceptions of risk affect the adoption of crop insurance (Ogurtsov et al, 2009; Smith and Goodwin, 1996). A common practice in crop insurance studies is to include farm and farmer characteristics as explanatory factors in models explaining the demand for crop insurance. In our analysis, we first identified different groups among farmers based on their preferences for crop insurance attributes, after which we then compared farm characteristics between the respective groups. As the EU is paying increasing attention to crop insurance, and many EU countries are considering the use of crop insurance premium subsidies, it is important to develop techniques to evaluate the demand for crop insurance in order to ensure viable public policies. In addition, it is important to examine whether subsidies ultimately support the farmers who are the intended recipients targeted by the policies. In the following sections, the estimation method and choice experiment data are presented, followed by the results and conclusions. Estimation method and choice experiment data Choice experiments (CEs) are often used in analysing the relative importance consumers assign to product characteristics. A choice experiment consists of several choice sets with two or more alternative goods described by their attributes. In this study, farmers were asked to choose one of the alternative insurance products based on their differing insurance characteristics. According to Lancastrian consumer theory and random utility theory, these choices reveal consumer trade-offs between the attributes of the goods (Lancaster, 1966). Before the choice experiment data used in this study are described, the applied estimation method is outlined. Latent class model The most typical way to model CE data econometrically is with a conditional logit model. When such a model is used, it is assumed that the respondents have a similar preference structure. This implies that we assume that all farmers have the same preferences across all crop insurance attributes. If we had had a strong a priori assumption concerning farmers preferences for crop insurance attributes, we could have applied different models to the sub-populations. However, as crop insurances are entirely new products for Finnish farmers, we could not make such assumptions. Thus, a latent class model was used to investigate the different farmer segments. The utility, U, of farmer n from insurance i takes the form: U ni s = β s X ni + ε ni s, (1) where β is the vector of parameters, X is the vector of attributes and ε is the random component in the utility function. The heterogeneity is included in the model with a class s. The farmer classes were determined based purely on the choices made by the individuals in the choice experiment. The latent class model is: exp(β s X ni ) P n (i B) = Σ S s=1 (2) Σ J j=1 exp(β s X nj ) where P n is the probability that a farmer will select insurance i from set of alternatives J. The parameters β s for the attributes are estimated in an iterative manner using a maximum likelihood estimation, where the number of segments S is given and the estimation is repeated several times with different numbers of segments. The best model having the optimal number of consumer classes is selected by using model fit criteria such as the Bayesian information criterion (BIC) or Akaike information criterion (AIC). These are log likelihood scores with correction factors for the number of observations and the number of parameters. With discrete choice models, the utility of respondents is measured. Thus, the estimated model coefficients are not interpretable in economic terms. Therefore, to reveal the overall WTP for a crop insurance product, implicit price (IP) estimates of crop insurance attributes are calculated as: β k IP k = (3) β p 2 Outlook on AGRICULTURE Vol 44, No 4

3 where β k is the parameter of the kth attribute, and β p is the price coefficient. Choice experiment design and data description Crop insurance does not exist in Finland. Thus crop insurance products are hypothetical products for Finnish farmers. In the questionnaire, respondents were presented with six crop insurance product cards. Different price structures were constructed in the CE design to examine the effect of starting point bias in crop insurance markets (Liesivaara and Myyrä, 2014). One of the three structures contained the whole price range (4, 8, 12, 16, 24 and 32 euros/hectare), which was used in this analysis to reduce the effect of the price range in the cluster formation process. We chose to have a sufficiently wide price range as we did not have a clear idea of the true value of crop insurance in Finland due to the lack of data and nonexistence of crop insurance products. Each choice card presented two different crop insurance products with varying attributes. Farmers were asked to select the most suitable crop insurance product for them. Respondents could also choose an out option that is, not to purchase crop insurance. Other attributes besides the price chosen for the insurance products were the insurance cover, type and expected indemnity (scale). Insurance cover and expected indemnity are essential features of crop insurance (Barnett and Skees, 1995). Coverage determines the share of the loss (deductible) that is covered by the farmer. The expected indemnity attribute defines the level of indemnity payment a farmer receives if the farm yield (or a particular index in the case of index insurances) falls below the trigger level. In many of the insurance products developed in the USA, farmers can change the scale and modify the insurance product to suit their farm. Thus, the scale is treated as an individual attribute in CE designs, and it describes the most likely indemnity payment of a farm, if it is eligible for compensation. An example of a choice card presented to farmers is given in Figure 1. In this study, the expected indemnity was determined for one hectare, and it could have three different levels, namely 100, 300 or 600/ha. The insurance coverage could be 90%, 80% or 70%. In the choice sets, the term deductible was used instead of cover to describe better the share of losses farmers must cover by themselves. The insurance type could be index or farmspecific insurance. In farm-specific insurance, inspection is needed if the farm experiences a crop loss. In index insurance, compensation is based on regional indices: for example, regional yield. If the value of the index falls below the deductible level, the insured farmer is eligible for compensation, even if the farm has not experienced crop damage or loss. Because Finnish farmers have no prior experience of market-based crop insurance, the surveyed farmers were provided with information on the insurance attributes on the choice cards before they were asked to complete their insurance choices. All four attributes were described in detail. The choice experiment survey was conducted in The survey was sent to 5,000 farmers in Finland. Altogether, 1,105 questionnaire forms were returned, 388 of which contained the whole price range. After missing responses were removed from the dataset, we obtained 2,019 completed insurance INSURANCE CARD 1 Insurance premium /hectare Insurance 1 Figure 1. Example of crop insurance choice card. Table 1. Descriptive statistics for the dataset and the farming population in Finland in Attribute Sample dataset Finland (2012) Number of farms ,042 Age (mean) Cultivated hectares (mean) Animal husbandry (% of farms) 19.3% 37.2% Crop farming (% of farms) 80.7% 62.8% South Finland (% of farms) 31.7% 32.3% West Finland (% of farms) 44.1% 38.2% East Finland (% of farms) 14.7% 24.6% North Finland (% of farms) 9.5% 5.0% choices from 349 farmers. The sample farms in the dataset are compared to Finnish farms in Table 1. The share of animal husbandry farms in the dataset is somewhat lower than in Finland on average. The share of responses from the eastern part of the country is also smaller than the share of farms in that region. Results and discussion Deductible 20% 20% Insurance type Expected compensation /hectare MY CHOICE Yield index insurance, farm inspection is not needed. Insurance 2 Farm yield insurance, inspection of loss at the farm is needed No buy I would not purchase insurance Latent class model results In the questionnaire, 44.5% of the offered insurance products were chosen. Moreover, 25.6% of respondents did not select any of the insurance choices offered on the six choice cards they were given. All observations are equally important for the analysis. By choosing no buy, farmers indicated that the insurances presented did not increase their utility. A model with three consumer classes was selected based on Bayesian information criterion (BIC) statistics and reasonable class sizes. The Wald p- values in Table 2 indicate that the attributes were jointly significant, while the Wald* p-values show that only the price attributes were class-dependent. However, this is very important in the economic interpretation of the results. In the latent class model, the price coefficient was negative, as expected, and significant in all farmer groups. This implies lower utility and a lower probability of choice of insurance with an increasing price. The deductible attribute was also negative, as expected. As farmers own share of the risk increased (increasing deductible: that is, decreasing cover), they were less willing to purchase insurance. The scale coefficient was positive and significant. A larger expected indemnity payment increased the utility of farmers from insurance. The Outlook on AGRICULTURE Vol 44, No 4 3

4 Table 2. Latent class model for insurance choices. Model for choices (class size) Class 1 (0.42) Class 2 (0.30) Class 3 (0.28) Wald p-value Wald *p-value Coefficient Coefficient Coefficient Constant (1) Constant (2) Constant (3) Price > Deductible > Scale > Overall R² R²(0) Note: The reference level for all attributes is set to 0. Constant 3 refers to the no-buy option. All estimated parameters are significant at the 99% level based on z-statistics. insurance type attribute was excluded from the analysis and all three intercepts were included in the model. The first and second intercepts were negative (insurance choice) and the third was positive, which suggests that no insurance presented to the farmers was preferred over the others. Implicit prices (IPs) are the marginal rates of substitution between price and product attributes. These reveal how willing farmers are to trade one attribute for another, given that the bundle of these attributes gives constant utility. Because farm and farmer characteristics were not included in the model, implicit prices provide a guideline for the naming of latent farmer groups revealed by the estimation. Farmers in Class 1 (representing 42% of farmers) were named as catastrophic loss preventers. Their IP for the deductible is higher compared to the rest of the farmers. With increasing deductible levels, which they are willing to accept, they need to cover a larger part of the yield risk by themselves. At the farm level, this can be observed in the way that farmers believe in their possibilities to prepare against inter-annual yield variation. However, severe losses in one year are considered as a threat to the continuation of the business, and these farmers thus seek insurance products that are more of the catastrophe prevention type. Their IP for the scale is in the same range as for Class 3 farmers (see Table 3). We named farmers in Class 2 (30% of farmers) risk lovers. They have the lowest IP for the scale. By examining the scale separately from other attributes, they are willing to pay for the example scale of 300/ha a premium of only 4.19/ha. Risk lovers do not accept high deductible levels for their yield risk insurances. When the IPs from the deductible and scale are summed, their WTP for insurance products on average turns negative. Thus, their estimated WTP is below any considerable price of yield insurance. However, a negative WTP for yield insurance does not necessarily indicate that their overall production behaviour is risk-loving. They might apply farming practices that are already selected based on risk aversion. We did not consider farmers risk preferences in this study, and producers actual risk preferences are not therefore observable. Thus, the term risk lover only refers to derived WTP measures. The label risk lovers indicates in this case that they prefer to handle typical inter-annual Table 3. Implicit prices (IPs) for insurance attributes and WTP estimates for an insurance product example. IP ( /ha) Class 1 Class 2 Class 3 Deductible (10%) Scale ( 1/ha) WTP a Note: a Deductible 30% and scale 300/ha. yield variation as agronomists, and do not need insurance in this portfolio. The third farmer class, comprising 28% of farmers, was named shallow loss preventers. They were willing to pay premiums of 6.5/ha for each scale of 100/ ha they received. While having a higher and negative IP for the deductible than Class 1 farmers, they would not opt for higher deductible insurances. This indicates that they prefer insurance against yield variation closer to the long-term yield average. Compared to catastrophic loss preventers, shallow loss preventers have a completely opposite approach to yield risks. Farm characteristics of latent farmer groups Since adverse selection has been a prominent problem in yield insurances, we described latent farm groups according to their farm characteristics (Table 4). Our analysis did not provide a clear distinction between the groups. However, small farms were found to be less likely to buy insurance, while farms with between 50 and 100 cultivated hectares were very likely to buy crop insurance. Larger farms appeared to rely on their capacity for yield risk management and were more interested in catastrophic risk management. Surprisingly, this was also revealed for farms in the size range ha, indicating non-linearity between farm size and interest in yield insurance. In Finland, climate conditions become harsher and the average yield decreases from south to north and west to east. This is also reflected in farmers insurance choices. Farmers in the western and southern parts of Finland did not feel such a high level of insecurity concerning yield variation around the average yields. Thus, they accepted higher deductible levels. Insurances 4 Outlook on AGRICULTURE Vol 44, No 4

5 Table 4. Farm distribution among the latent risk management groups. Catastrophic loss Shallow loss Risk lovers preventers preventers 30% 42% 28% Production line Animal husbandry Crop farming Farm size (ha) Region West East North South Note: West: North Ostrobothnia, Ostrobothnia, South Ostrobothnia and Satakunta. East: Southeast Finland, North Savo, South Savo and North Karelia. North: Kainuu and Lapland. South: Uusimaa, Southwest Finland, Häme, Pirkanmaa and Central Finland. targeted at yield catastrophes would suit these regions. However, the opposite situation was apparent in the eastern and northern regions. Concluding comments Finnish farmers have no prior experience of commercial crop insurance. This study examined the demand for crop insurance in Finland. We conducted a latent class analysis using choice experiment data to reveal latent groups of Finnish farmers in yield insurance markets. The analysis revealed three homogeneous groups that differed significantly from each other in their WTP for crop insurance attributes. The main division used earlier in the literature for risk-averse and risk-loving behaviour was also recognized in this study. Our analysis, applying possible insurance products in Finland, revealed that 30% of farmers were risk lovers and would not buy insurance products of any kind. The remaining farmers showed riskaverse behaviour and had a positive willingness to pay for insurance products. A new finding in this study was that stated preferences methods could be used efficiently in dividing risk-averse farmers into more distinct groups. By applying well recognized econometric methods from other industrial areas, we defined farmers as being catastrophic loss and shallow loss preventers. These groups have earlier been recognized in quality terms that is, based on their preferences for risk taking but have not been measured in a quantitative manner. In the data representing Finnish farmers, the number of catastrophic loss preventers exceeded the number of shallow loss preventers. Currently, the Finnish government is forming its policy regarding catastrophic assistance. As it will not introduce crop insurance premium subsidies after 2015, the government focus has shifted to the future role of catastrophic assistance. However, uniform catastrophic assistance rules would be challenging to implement because of heterogeneity among Finnish farmers in risk-aversion values. Based on our results, we conclude that a single type of catastrophic assistance would not suit Finnish yield insurance markets. Farmers preferences should be taken into consideration and EU Commission guidelines for yield insurances could easily be applied as they are. The only adjustment needed is that premium subsidies for yield insurances having a deductible larger than 30% should be considered as catastrophic assistance. In this manner, farmers needs for catastrophic assistance would be equally considered, since those farmers calling for catastrophic assistance would be covered. In addition, introducing catastrophic assistance in this manner would prevent the Finnish government from facing a major obstacle otherwise caused in crop insurance markets (Liesivaara and Myyrä, 2016; van Asseldonk et al, 2002, 2013). The shallow loss type of yield insurance has attracted considerable interest in the USA. Our results support the introduction of this type of insurance into EU markets. In 2014, the Finnish government decided not to promote crop insurance with premium subsidies. Our results reveal that farmers are not a homogeneous group regarding their WTP for crop insurance. Thus, only some of the farmers would benefit from such a policy in Finland, and in this respect the decision made by the Finnish government appears sound. The CAP for limits the scope of subsidized crop insurance policies, with the deductible limited to 30%. Moreover, the use of index insurances has been limited due to basis risk (European Commission, 2013; Liesivaara and Myyrä, 2015). If the EU pushes for efficient risk management policies for agriculture, flexibility will be needed in the legislation, and shallow loss insurances will also need to be introduced to cover all risk-prevention needs equally. The third group of farmers referred to in this study are risk lovers, who will always opt out of premium subsidy schemes, which is natural based on their values and needs. No further action from the government side is needed for such farmers related to premium subsidies for yield insurance. Finally, from the perspective of insurance companies, valuable information can be obtained when latent groups are connected to farm typology. However, no clear linear connection between latent groups and farm characteristics could be found in this study. This suggests that farmers choices to insure against crop losses are driven by farmer characteristics such as attitude to risk, which are difficult for the insurer to measure. References Barnett, B., and Skees, J. (1995), Region and crop specific models of the demand for federal multiple peril crop insurance, Journal of Insurance Issues, Vol 19, No 2, pp European Commission (2013), Regulation (EU) No 1305/2013 of the European Parliament and of the Council of 17 December 2013 on support for rural development by the European Agricultural Fund for Rural Development (EAFRD) and repealing Council Regulation (EC) No 1698/2005, European Commission, Brussels. Goodwin, B.K. (1993), An empirical analysis of the demand for multiple peril crop insurance, American Journal of Agricultural Economics, Vol 75, No 2, pp Outlook on AGRICULTURE Vol 44, No 4 5

6 Goodwin, B., Vandeveer, M., and Deal, J. (1994), An empirical analysis of acreage effects of participation in the federal crop insurance programme, American Journal of Agricultural Economics, Vol 86, No 4, pp Hennessy, D.A. (1998), The production effects of coupled and decoupled agricultural income support policies, American Journal of Agricultural Economics, Vol 80, No 1, pp Koundouri, P., Laukkanen, M., Myyra, S., and Nauges, C. (2009), The effects of EU agricultural policy changes on farmers risk attitudes, European Review of Agricultural Economics, Vol 36, No 1, pp Lancaster. K. (1966), A new approach to consumer theory, Journal of Political Economics, Vol 74, No 2, pp Liesivaara, P., and Myyrä, S. (2014), Willingness to pay for agricultural crop insurance in the northern EU, Agricultural Finance Review, Vol 74, No 4, pp Liesivaara, P., and Myyrä, S. (2015), Feasibility of area-yield insurance scheme in the EU: evidence from Finland, EuroChoices (forthcoming), doi: / X Liesivaara, P. and Myyrä, S. (2016), The demand for public private crop insurance and government disaster relief, Journal of Policy Modeling (forthcoming). Meuwissen. M.P.M., Assefa. T.T., and van Asseldonk, M.A.P.M. (2013), Supporting insurance in European agriculture: experience of mutuals in the Netherlands, Eurochoices, Vol 12, No 3, pp Myyrä, S., and Pietola, K. (2011), Testing for moral hazard and ranking farms by their inclination to collect crop damage compensations, European Association of Agricultural Economists, 2011 International Congress, Zurich. Ogurtsov, V.A., van Asseldonk, M.A.P.M., and Huirne, R.B.M. (2009), Purchase of catastrophe insurance by Dutch dairy and arable farmers, Review of Agricultural Economics, Vol 31, No 1, pp Sckokai, P., and Antón, J. (2005), The degree of decoupling of area payments for arable crops in the European Union, American Journal of Agricultural Economics, Vol 87, No 5, pp Sherrick, B.J., Barry, P.J., Schnitkey, G.D., Ellinger, P.N., and Wansink, B. (2003), Farmers preferences for crop insurance attributes, Review of Agricultural Economics, Vol 25, No 2, pp Smith, V.H., and Baquet, A.E. (1996), The demand for multiple peril crop insurance: evidence from Montana wheat farms, American Journal of Agricultural Economics, Vol 78, No 1, pp Smith, V.H., and Goodwin, B.K. (1996), Crop insurance, moral hazard, and agricultural chemical use, American Journal of Agricultural Economics, Vol 78, No 2, pp USDA (2004), Decoupled Payments in a Changing Policy Setting, Agricultural Economic Report No 838, Economic Research Service, United States Department of Agriculture, website: van Asseldonk, M.A.P.M., Meuwissen, M.P.M., and Huirne, R.B.M. (2002), Belief in disaster relief and the demand for a public private insurance program, Review of Agricultural Economics, Vol 24, No 1, pp van Asseldonk, M.A.P.M., Pietola, K., and Niemi, J.K. (2013), Trade-offs between catastrophic assistance and subsidized insurance in European agriculture, Outlook on Agriculture, Vol 42, No 4, pp WTO (1994), Agreement on agriculture, GATT, Marrakesh Agreement 1994, World Trade Organization, Geneva. 6 Outlook on AGRICULTURE Vol 44, No 4