The preferences of farmers and non-farming citizens for farm catastrophic risk management policy alternatives

Abstract

Economic risks for farmers have increased during recent years due to various factors such as more extreme climate conditions and the volatility of agricultural markets. We analysed the preferences of Finnish farmers and non-farming citizens concerning catastrophic risk management policies in agriculture based on a survey addressed to both groups. Respondents were asked to rank their preferences regarding who should bear the costs from various disasters occurring on farms. Farmers and non-farming citizens did not prefer a single cost bearer for all risks, but they generally preferred either society or farmers’ insurance to bear the costs. The results indicate that citizens generally accept public spending on agricultural risk management, either through some ex post disaster aid or through subsidised insurance. Farmers’ preferences were generally aligned with those of non-farmers.

1. Introduction

Economic risks and the volatility of agricultural markets have increased during the past few years due to various factors. Climate change has increased the frequency of exceptional weather events, thus causing higher production risks and requiring adaptation (George et al. 2019; Wheeler and Lobley 2021). Besides the supply and stocks of agricultural commodities, trade disruptions and other disorders may cause market volatility in farm gate prices (Headey and Fan 2008; Headey 2011; Wright 2011). Market liberalisation has increased trade and international competition, as was seen in the case of the abolition of the European Union (EU) milk quota system (Cele et al. 2022; Frick and Sauer 2021). COVID-19 and Russian's invasion of Ukraine have had strong implications for food markets. These sudden changes in markets emphasise the need to assess and improve the resilience of farming systems with risk management (Meuwissen et al. 2019). The economic risks are becoming so severe that risk management is no longer only an issue for experts in the agriculture and food sector. Instead, it has become a challenge that affects the wellbeing of citizens, highlighting the importance of integrating citizens’ preferences into decision-making.

In Western countries, the average farm size has generally grown due to structural change. More hectares are being cultivated and more animals housed per farm (Niskanen et al. 2020). A larger scale might also imply higher risks at the farm level. In the USA, for example, larger farms were more likely to be categorised as under financial stress due to a lower off-farm income and higher farm debt, and a decline in farm income was predicted to increase the share of large farms under financial stress compared to midsize or small farms (Key et al. 2019). Smaller farms also tend to balance risks more through off-farm strategies than large farms (de Mey et al. 2016), implying that growing farms face higher risks through the scale of production and financial burden.

Institutional support has traditionally had a considerable role in the formation of agricultural income in Europe and to some extent also in risk management. EU member states differ in terms of the disaster aid and risk management tools they provide, but risk management tools generally have a low profile in European agriculture at present. Since the early days of the Common Agricultural Policy (CAP), the support has shifted from market support to coupled payments and, finally, to decoupled payments. From a risk management policy perspective, coupled and decoupled payments stabilise farm income to some extent but do not protect farmers from market instability (Baldock et al. 2010; Bardají et al. 2016). Protection from more severe income losses requires targeted risk management tools.

In this study, we focused on alternative management policy options for different types of catastrophic risks and how these options are preferred among farmers and non-farming citizens. We also analysed characteristics explaining differences in the preferences. Farmers, as the main beneficiaries of agricultural support, tend to favour the existing policy and object to major policy changes (Kay and Friesen 2011; Jost 2015; Vainio et al. 2021). For this reason, farmers and non-farming citizens might differently perceive a regime shift from income support towards a more risk-management-oriented policy. The differences between farmers and non-farming citizens in their views on agricultural policy require further research to improve the legitimacy of policies (Bernués et al. 2016). Wallner (2008) noted that failure to address policy legitimacy may compromise long-term policy goals and the interests of authoritative policymakers, as legitimacy ideally requires wide acceptance among both the affected stakeholders and the public. For this reason, the opinions of non-farming citizens regarding agricultural policies should also be considered.

Previous studies have demonstrated that citizens in the USA generally support government interventions in agriculture, although these would lead to higher taxes (Moon and Pino 2018). Moon and Pino (2018) found that basic sociodemographic factors, such as age, education, and gender, predict citizens’ preferences. According to the results of Ellison et al. (2010), US taxpayers support agricultural subsidies, as they believe that subsidisation secures food supply. Furthermore, Biedny et al. (2020) also found evidence for an increased preference for government interventions. These results are relevant from the risk management policy perspective, since insurance forms a large share of the total agricultural support in the USA.

Europeans also consider public spending on agricultural policy acceptable. The Eurobarometer survey (European Commission 2020) asked EU citizens for their views about CAP spending. In total, 47 per cent of the survey respondents considered the current support level of the CAP to be appropriate and 39 per cent considered the support too low. In the case of Finland, the respective figures were 59 and 28 per cent. Against this background, we expect a considerable share of Finnish citizens to prefer society to compensate many agricultural risks. However, this does not necessarily imply that society should compensate all losses incurred by farmers. Farmers can manage several risks, and direct support from society may negatively affect their risk management strategies (OECD 2011). On the other hand, other parts of the food chain also affect the producer prices of agricultural products and thus the income farmers earn. It can be asked whether consumers, for example, should bear the costs through higher food prices.

This study contributes to the agricultural policy literature by focusing on catastrophic risk management policies. Because food is essential, agricultural disasters affect everyone through limited food availability or higher prices. Agricultural disasters imply high costs, but individual farmers have a very limited ability to bear the costs from disasters. In this study, we considered five types of crises that a farm can face and that threaten the continuation of the farm. These include major crop losses, extreme adverse changes in prices and the disruption of production due to severe illness of the farmer, for example. Our aim in this study was to determine what type of risk management policy farmers and non-farming citizens prefer. Previous studies surveying citizen views on agricultural policy have generally focused on environmental issues or the legitimacy of agricultural subsidies in general. Unlike previous studies, this study examined more directly how the management of catastrophic risks in agriculture should be financed, making it highly relevant from the perspective of legitimacy.

The rest of the article is structured as follows. Section 2 reviews the literature concerning risk types and policies related to risk management, as well the Finnish institutional background concerning risk management policies. Section 3 presents the method and data applied in the analysis, and Section 4 presents the results. The final sections provide discussion and conclusions concerning the alignment of empirical findings and the agricultural risk management policy. Policy recommendations inherently arise from the results.

2. Agricultural risk management and policy

An OECD report on agricultural risk management policies (OECD 2011) classified risks into three categories based on their need for a policy response. These include normal, marketable, and catastrophic risks. Normal risks occur often but cause relatively low damage, whereas catastrophic risks occur rarely but cause high damage. Marketable risks lie between these two. This study focused on catastrophic risks. According to the OECD report, normal risks do not require a policy response under a good governance, whereas catastrophic risks do need some policy response. Marketable risks, on the other hand, mostly require the establishment of markets to share them. The same policy response to all types of risks is not expected, according to the report.

As Finland belongs to the EU, the national agricultural policy follows the general guidelines set in the CAP. The European Commission (2017) mapped different European risk management instruments according to the OECD (2011) classification. In this mapping, cooperatives, forward contracting, and non-subsidised mutual funds and insurances were placed to solve only marketable risks, whereas market measures and all subsidised instruments, including mutual funds, insurances, and income stabilisation tools, mainly covered marketable and catastrophic risks. Ad hoc disaster aid was considered as a response to catastrophic risks, and decoupled payments were considered as an instrument to manage all types of risks. EU countries differ in terms of the adoption of subsidised risk management instruments and the ad hoc disaster aid they provide. Finland does not belong to the largest providers of ad hoc disaster aid and does not currently offer subsidised risk management instruments allowed in the CAP.

Farmers have several ways to manage normal risks. Commonly mentioned management options include non-farm income, savings, and the diversification of production, among others (Smith and Glauber 2012; Hardaker et al. 2015). Marketable risks can also be managed if suitable instruments exist. Finland has a long tradition of strong producer cooperatives, especially in milk and meat production. However, farmers cannot manage truly catastrophic risks as defined in OECD (2011), and these risks generally require additional cost bearers. On the other hand, perceptions of catastrophic risks differ. New Zealand, for example, has excluded market risks from disaster aid (OECD 2011), and individual farmers have subjective perceptions and attitudes towards risks (Ogurtsov et al. 2008; van Winsen et al. 2016). It should also be noted that not all experts consider catastrophic risks as totally unmanageable for individual farmers and private markets (Ogurtsov et al. 2008; Goodwin and Smith 2013).

If public compensation is not considered an option, then society can provide farmers other means to mitigate catastrophic risks. These include, among others, the creation of a legal and institutional environment for different financial instruments, facilitating the flow of information, and improving access to credit (Cafiero et al. 2007). Mutual funds could also be classified into this category if the role of society is limited to only establishing the system. The role of society must be further restricted, because expected or stated disaster aid from society affects the willingness to participate in a mutual fund (van Asseldonk et al. 2002).

Direct income support has traditionally been at the core of the CAP, and the current policy model with decoupled payments determines to a much lesser extent what and how much should be produced (Rude 2008). However, a possibility to launch supported agricultural insurance policies has also more recently been adopted in Europe. The second pillar of the CAP includes the most relevant insurance tools (European Court of Auditors 2019). They are based on defining a threshold level to trigger compensation. In this system, compensation is allowed if the income is less than the threshold level compared to the average of the preceding 3 years or optionally the preceding 5 years with the highest and lowest entries excluded. The World Trade Organization has determined levels that do not distort trade, where a 30 per cent threshold level of losses in production or income is commonly accepted. However, the current EU legislation also allows 20 per cent threshold levels in supported crop insurances under tool M17.1 and in a sectoral income stability tool.

Unlike in most European countries, subsidised insurance systems in the USA and Canada provide a major instrument to manage risks. The subsidisation of insurances is often considered necessary due to the absence of a private market (Miranda and Vedenov 2001; Hardaker et al. 2015; Liesivaara and Myyrä 2017). Insurance can provide an efficient tool to steer disaster aid (Meuwissen et al. 2001; Antón et al. 2013; Hardaker et al. 2015), but Hardaker et al. (2015) noted that the actual implementations of subsidised insurance systems have rarely proven successful. Goodwin and Smith (2013) and Smith and Glauber (2012) questioned the need for subsidised insurance programmes, as private insurance markets can insure more severe non-agricultural risks than systemic agricultural losses. They further argue that subsidised insurances affect production decisions and generally increase moral hazard, thus leading to negative welfare impacts at the taxpayers’ expense.

Besides subsidising insurances, society can bear the costs from catastrophic events more directly through ex post disaster aid, which is often given on an ad hoc basis. Disaster aid has been provided both in the USA and in the EU (Belasco and Smith 2022; European Commission 2017). Although the provision of disaster aid is well justified in the case of severe catastrophic events, the practical implementation of the aid has potential for shortcomings, including political pressure to support farmers too frequently (Hardaker et al. 2015; OECD 2011). Belasco and Smith (2022) further demonstrated that ad hoc disaster aid in the USA has concentrated on large farms. It should also be noted that disaster aid may cause conflicts if society already subsidises insurances. Studies have demonstrated that the expectation of aid reduces farmers’ demand for insurances (van Asseldonk et al. 2002; Liesivaara and Myyrä 2017). It has been noted for this reason that the aid should only be provided to those farmers having insurance (Bardají et al. 2016; Belasco and Smith 2022).

Since farmers’ likelihood of surviving and recovering from financial losses depends on the price they receive for their products, retailers and consumers could also bear the costs from catastrophic risks through higher food prices. Although the position of farmers is not easily improved through policy measures, consumers appear to want farmers to obtain a higher share of the consumer price for their products. In the German context, Busch and Spiller (2016) demonstrated that consumers perceive the position of farmers in the food chain as unfair and would increase the revenue of farmers by lowering the shares received by others, especially the share of retailers. Consumers may in some cases state willingness to pay a premium in return for a fairer price for farmers. This has especially been noted in the case of organic products (Zander and Hamm 2010; Briggeman and Lusk 2011). A consistent application of fairness perceptions could imply paying a higher price to farmers when they face catastrophic losses.

Although consumers and retailers could, in theory, bear the costs from agricultural disasters, very few policy measures exist to fully implement this. Farmers are in a relatively weak position in relation to retailers in the food chain, and for this reason they may not be able to share or transfer much of the costs to retailers and consumers. The EU has recently taken actions to improve the position of farmers, for example, by allowing the establishment of producer organisations to increase the market power of producers. In general, compensation from retailers and consumers would require voluntarily actions, and these could be facilitated by society. Based on the reviewed literature, farmers, society, consumers, farmers’ insurance, and retailers were considered as alternative cost bearers in this study. These alternatives imply different policy measures and the distribution of costs in society, and citizens were asked to state their preferences concerning them.

3. Data and methods

3.1 Data collection

An Internet survey was implemented to collect representative data on the risk management preferences of Finnish citizens. We applied two different samples. The first sample comprised both farmers and non-farmers and represented the sociodemographic profiles of Finnish citizens. It was drawn from the Internet panel of the private survey company Taloustutkimus. This panel comprises a large number (approximately 30,000) of respondents representing the population and recruited using random sampling. As the share of farmers in the labour force is low (1.8 per cent), the second sample strengthened the representation of farmers. To obtain as representative a sample of farmers as possible, the second sample of farmers was drawn from the register of the Finnish Food Authority. The data from both samples was collected as an Internet survey by Taloustutkimus.

A pilot survey (N = 202) was conducted in August 2020. After the pilot, two on-line focus groups were implemented to discuss the topics of future agriculture and to test the measures of the survey. The focus groups examined, for example, what kinds of ideas the questions evoked in the participants, in general, and how understandable the survey instructions and information sections were. A modified questionnaire was further tested in a second pilot study (N = 205) in November 2020.

The questionnaire is available as supplementary material in Pouta et al. (2022)¹. The final survey data were collected during January and February of 2021. For the first sample including farmers and non-farming citizens, a random sample of 10,362 respondents was selected and 2014 respondents completed the survey (response rate 19.4 per cent). This data set represented the population rather well, as presented in Table 1. Regarding farmers, an invitation e-mail was sent to 4,827 farmers and 518 responses were received (response rate 10.7 per cent). However, one of the farmer respondents filled in the survey incompletely regarding the risk management question and was consequently removed from the data analysed in this study. The total number on farmer respondents was 517. Farmer respondents from the first sample were combined with the second sample to obtain data on farmers. The first sample, after the removal of farmer respondents, comprised the data set on non-farming citizens.

3.2 Dependent variable: preferences for risk management

For the survey, we selected risk types that differed regarding farmers’ opportunities to take them into account in farm management. Natural hazards were included, as well as market and social risks. We tested seven different risk types in the pilot survey², but five risk types possibly faced by Finnish farms were eventually selected to simplify the question for the final survey. The risk types refer to severe, catastrophic risk, which cannot be considered as part of normal farm business volatility. The questions (Q1–Q5) focused on the following events:

• A pest or disease epidemic that stops production (Q1 EPIDEMIC).

• Crop losses due to weather conditions or wild animals (e.g. migratory birds) (Q2 NATURE).

• Serious illness of the farmer or employees (Q3 ILLNESS).

• Considerable changes in prices: a major increase in fuel and fertilizer prices or decrease in product prices (Q4 PRICES).

• An environmental disaster, such as a chemical spill or nuclear fallout, originating outside the farm (Q5 ENV.DISASTER).

Respondents were asked to rank their preferences for different risk management options. This enabled us to obtain more information than a question asking only the most preferred alternative. The alternative risk management options were defined with the following cost bearers: (1) the farmer, (2) society, (3) consumers, (4) farmers’ insurance, and (5) retailers.

The question asking respondent to provide ranks had the following formulation: ‘In your opinion, who should be responsible for securing farmer income under the following crises? Please select from one to three responsible parties in order of importance (most important (1), second most important (2) and third most important (3)).’ The ranking task produced a panel data set with twenty-five rows per respondent, as there were five risk types and five management options with different cost bearers. The cost bearers were coded as dummy variables and the risk types as categorical variables, with society as a reference class. For each combination of risk type and cost bearer, a rank was coded in ascending order, implying that the option ranked first was 1, the second was 2, and so on. Non-ranked options were all ranked last. This variable was the dependent variable in the following statistical model.

3.3 Statistical model

This study applied rank-ordered logistic regression (ROL), which was first introduced by Beggs et al. (1981). Instead of asking respondents to give the most preferred option, the data of this study consists of ranks in which respondents were asked to rank the three most preferred out of five options. This produced more information compared to eliciting the preferred option only and so increased the accuracy of the model (Hausman and Ruud 1987). ROL has its foundation in the random utility model (Allison and Christakis 1994). Although respondent i’s utility |${U}_{ij}$| for option j is unobserved, the respondent is expected to prefer alternative j over k if |${U}_{ij} > {U}_{ik}$|⁠. It should be noted that, in the present application, j differs between respondents because respondents were allowed to rank less than three options. The utility of respondent i for option j can be approximated as a linear function of respondent and option-specific characteristics. Only the characteristics of respondents were considered in this study, so that utility consists of respondent characteristics |${z}_i$| and a random error |${\varepsilon }_{ij}$| (Equation 1).

It is assumed that the error term |${\varepsilon }_{ij}$| follows the extreme value distribution, and that the individual errors are independently and identically distributed. The estimation of the likelihood of a certain ranking, i.e., |$pr[ {{U}_{i1} > {U}_{i2} > \ldots > {U}_{ij}} ]$|⁠, requires a series of choices. The first choice coincides with multinomial logit model in which only the preferred option is selected. The second choice is made from the remaining set of alternatives and so on. With the sample of n respondents, the log-likelihood function shown in Equation 2 is maximized as presented in Allison and Christakis (1994).

In Equation 2, |${\delta }_{ijk}$| equals 1 if the rank of option k is greater than or equal to option j and zero otherwise.

Earlier studies (Fok et al. 2012; Palma 2017) have noted that individuals tend to rank the most preferred alternatives more accurately and consider the rest of the alternatives indifferently. In this case, rankings do not reflect the true utilities of individuals in the case of the least preferred alternatives. This problem was circumvented in this study by allowing respondents to state only the number of preferred alternatives they considered important. As there were five risk types, respondents gave five similar rankings, which enabled us to treat the data as a panel. Consequently, a random effects model was estimated. Equation 3 shows the general model estimated in the study.

In this equation, an index t refers to risk types. The main difference compared to the basic single-equation model in Equation 1 is the individual-specific constant term |${\alpha }_{ij}$|⁠. A random effects specification further allowed us to relax the independence of irrelevant alternatives assumption on which ROL in its basic form leans (Revelt and Train 1998).

The ROL model (Equation 3) was estimated separately for non-farmer and farmer citizens to allow comparison between the two groups. Both models had the same set of independent variables, which consisted of different sociodemographic factors. The results of farmer and non-farmer models were compared based on parameter confidence intervals. Furthermore, an additional model for farmer respondents was estimated in which the preferences were explained by farm characteristics.

Due to fundamental differences between the different crises presented to respondents, question specific constant terms were added to the model. Random effects were further allowed to correlate between each other. This seems plausible, as systematic connections between the alternatives are likely to exist. The estimation of random parameters requires simulation, and 1,000 Halton draws were applied in this study. We used NLOGIT 6 for the analysis.

3.4 Independent variables in the model

To estimate the ROL model, several sociodemographic factors were considered as the independent variables to explain the ranks of risk management alternatives. Studies examining public opinions have generally applied basic sociodemographic variables as explanatory variables. Different sociopsychological theories can be applied to explain why individual characteristics and contexts affect attitudes and preferences. Blekesaune and Quadagno (2003), for example, discuss self-interest and ideological arguments, which provide a means to understand why individual characteristics would affect policy opinions. Sociodemographic factors, such as gender, age, and socio-political ideology, also explain risk perceptions (Byrnes et al. 1999; Rolison et al. 2012; Choma et al. 2013), which is a relevant aspect in the current context.

Table 2 presents the list of final variables with descriptions. The variables were basic sociodemographic factors. Gender, age, education, and income were included. The income and age of the respondents were categorised into three groups based on the distributions in the data. Respondents aged under 35 formed the group of younger and those aged over 65 the group of older respondents. With this classification, the group of younger respondents approximately corresponded to the lowest quartile and the group of older respondents to the highest quartile. In the case of income, family income determined the income group if the respondent had a spouse, and personal income was used if the respondent had no spouse.

The impact of political ideology measured with voting behaviour in the last election, was estimated using three categories. Finland has four larger parties and several smaller ones. Three out of the four largest parties represent a clear ideological view relevant for this study, and the voters for other parties and the respondents with no data on political ideology formed the reference group. The variable KESK refers to the Centre Party (Keskusta), which has traditionally advocated farmers and rural areas. The variable SDP refers to the Social Democratic Party, the party that has traditionally supported a stronger role of government but has stated more recently that the role of subsidies in agriculture should be decreased. The last variable, KOK, refers to the National Coalition Party (Kokoomus), which belongs to the political right and supports liberal markets. The place of residence was included in the model, and it was based on the European NUTS 2 classification. Southern Finland was a reference class, and the other three regions were Helsinki—Uusimaa, which is the capital region, Western Finland, and Northern and Eastern Finland.

Farmer respondents were further analysed using additional data. Specialisation in animal production, specialisation in plant production, and organic production were included in the model to analyse the differences in the preferences of farmers from different farm categories. Specialised production was compared with mixed production so that a farm was considered specialised if the respondent reported only one production type. Although the farmers in these farm categories are highly heterogeneous and the categories are somewhat artificial, some differences can be expected in the preferences. Farmers in organic production are often less risk averse than conventional farmers (Flaten et al. 2005; Gardebroek 2006), so they may support a smaller role of government. Specialised production is inherently riskier than diversified production (Katchova 2005), and the preferences of specialised farms may therefore differ from those of diversified farms. Furthermore, there are some differences in the exposure of crop and livestock farms to different types of risks, which may lead to differing preferences between them.

4 Results

Figure 1 describes the preferences stated in both non-farmer and farmer samples regarding who should bear the costs from various disasters occurring on farms. Each respondent gave the most preferred option, but it should be noted that not all gave their second and third most preferred option, as it was not required. Preferences differed between questions, but citizens clearly preferred society and insurance over other alternatives. It is interesting to note that farmers were relatively often considered as the third most preferred alternative and generally preferred over consumers and retailers. Consumers were the least preferred alternative, except in the situation where production suffers from unfavourable changes in input and output prices.

Farmers generally preferred society over insurance to bear the costs from various disasters occurring on farms, unlike non-farming citizens. It should be noted that relative differences between different options remained surprisingly similar between farmer and non-farmer respondents. The respondent preferences differed between risk types. Insurance was considered as the main tool to manage human risks, whereas respondents preferred society to bear the costs from an environmental disaster. While these outcomes were expected, some others were not. A surprising result was that citizens clearly preferred society to bear the costs from price risks.

The rank ordered logit model showed which variables explained the respondents’ ranks of the management alternatives. Three different models were estimated: (1) a non-farmer model, (2) a basic farmer model, and (3) a developed farmer model. Models 1 and 2 enabled the analysis of the differences between farmers and non-farming citizens. Model 3 for farmers was developed further to test whether farm characteristics associate with risk management preferences.

Parameter identification requires the selection of reference categories. Society represented the reference class among the risk management alternatives, and for this reason parameter estimates for other alternatives describe the difference relative to society. As shown in the descriptive analysis, society was often the first or second most preferred alternative. The random coefficient estimates were negative for this reason. Since the questions represented fundamentally different risks, fixed coefficients for each cost bearer alternative had risk-specific dummies to capture the systematic differences between risks. Q5 ENV.DISASTER served as the reference class in this case, because it represented a disaster completely beyond an individual farmer's or company's capacity to manage. For this reason, the question-specific coefficients were expected to be positive in general.

Farmers’ preferences were expected to differ from those of non-farmers. Specifications in models 1 and 2 were exactly the same. The complete results are presented in the  Appendix 1. Figure 2 illustrates general results for the preferences of both farmers and non-farmers. Society and insurance were highly preferred over other alternatives to bear the costs from various disasters occurring on farms, which confirmed the results from the descriptive analysis. Society was the most preferred alternative in the case of PRICES and ENV.DISASTER, whereas insurance was the most preferred in the rest of the cases. The descriptive analysis revealed differences between questions, and the model further confirmed this finding. Preferences in the case of PRICES differed considerably from other risk types.

Differences between farmer and non-farmer respondents were of special interest. Unsurprisingly, farmers less often preferred to bear the costs themselves either directly or by paying for insurance. The two groups particularly differed in their preferences concerning the relative emphasis between society and insurance. However, the relative difference of society and insurance between the groups was fairly small in the case of ILLNESS and PRICES. Preferences concerning farmers, consumers, and retailers were closely aligned.

The results of the non-farmer model ( Appendix 1) demonstrate that the general preference for society and insurance was not greatly affected by sociodemographic factors. Risk-type dummies were almost always larger in absolute values than the fixed coefficients associated with sociodemographic factors. However, this does not imply that sociodemographic factors had no impact. Factors were highly significant, and impacts were mostly as initially hypothesised. To ease the interpretation of the results, we present elasticities in Table 3.

The elasticities indicate the change in probability compared to the reference group, i.e., how much the probability would change if the respondent had, for example, a higher income instead of a middle income. FEMALE had a positive elasticity value in the case of insurance but was otherwise negative, so women tended to prefer insurance over society but society over other alternatives more often than men. In fact, differences between men and women were absolutely the highest compared to any other sociodemographic group. In addition, in the case of YOUNGER, HIEDUC, HIGHERINC, SDP, KOK, and UUSIMAA, the probability of choosing insurance was higher, indicating that younger people, those with a higher income and voters for the National Coalition Party, in particular, preferred insurance. The elasticities for KESK and NORTHEAST were negative in the case of insurance and farmers. Thus, voters for the Centre Party and people living in northern and eastern parts of the country preferred a lower responsibility for farmers. These outcomes were expected. It is also interesting to note that the elasticities of YOUNGER were relatively high in the case of farmers and consumers, implying that younger people do not consider the two large institutions, society and insurance, to be as appealing as collective cost bearing through higher prices and farmers' own risk management.

The third model specification (Appendix 3) revealed the differences in preferences between different types of farms. Table 4 presents the results in elasticities for the variables indicating farm characteristics. All the elasticities were statistically significant. Society and insurance were generally preferred, irrespective of the farm characteristics. However, some differences between farm types were found. All the elasticities were negative in the case of CROPPROD and ANIMALPROD, implying that farms specialised in animal or crop production were more likely to prefer society over other alternatives compared to mixed farms. All the p-values of the estimated model coefficients were < 0.01, except in one case, which demonstrates the statistically significant difference between specialised and mixed farms (Appendix 3). The preferences of organic farms differed significantly in the cases of insurance, farmers, and retailers (p < 0.01) from those of conventionally producing farms. ORGANIC predicted a slight increase in the probability of selecting insurance. A noteworthy result was that organic farms also were more willing to let farmers themselves bear the costs. This may reflect the less risk averse attitudes of organic farmers, as hypothesized. However, estimated elasticities related to ORGANIC were fairly small.

5 Discussion

Finnish farmers and non-farming citizens did not prefer a single cost bearer for all the risk types under study, but their preferences were divided between either society or farmers’ insurance as cost bearers. Alternatives for cost bearers were not restricted, making it possible for a respondent to select any cost bearer for a specific crisis despite potential feasibility or fairness issues. The exclusion of certain combinations would have been highly subjective and potentially restricted the amount of information obtained from the respondents. The results indicate that the majority of citizens accept public spending on agricultural risk management either through some ex post disaster aid or subsidised insurance. The result can be considered robust. If respondents had been allowed to name more than one preferred alternative, then it is unlikely that the main conclusion would have changed. However, the relative importance of the rest of the alternatives might have been affected by the way the respondents were asked to state their preferences. Some respondents might have given slightly different rankings if equal preferences and the ranking of all alternatives would have been allowed.

An interesting feature of the results was the preference differences between the price risk and other risk types considered. Disaster aid from society was highly preferred in the case of price risk, but the differences between other alternatives were relatively small. Although society could compensate the most severe losses, efficient and timely compensation from society would be difficult to provide. On the other hand, insurance should be an income insurance or a similar product in practise, because tailoring insurances for different prices and situations would be infeasible. Perhaps the policy should, in this case, focus on the most efficient distribution of the costs instead of selecting a single cost bearer or policy measure.

Farmers’ preferences were generally aligned with those of non-farmers. The main difference between the farmers and non-farmers was the role of insurance. Farmer respondents clearly preferred society over insurance as the cost bearer, which was an expected outcome. However, the preferences revealed in this study do not indicate any major conflicts between farmers and non-farming citizens. Whether the future agricultural risk management policy is insurance-led or disaster aid-led should not cause major conflicts of interest, according to the results. However, the main instrument should vary between risk types. Insurance, for example, was preferred in the case of human risk but not in the case of price risk.

Although disaster aid from society generally has strong legitimacy, the aid has a risk of becoming frequent and inefficiently targeted (Hardaker et al. 2015; Belasco and Smith 2022). Insurance has the advantage of being formally defined and quick to pay out as the crisis hits (Antón et al. 2013). Subsidised insurances have played a major role in North American agricultural policy, and the CAP has taken steps towards the wider adoption of such systems. Although some European countries have already established national insurance systems, Finland has no such system and has not shown any serious intention of establishing one soon. The results of this study, however, indicate that both farmers and non-farming citizens preferred insurance as a major risk management instrument.

The respondents were not asked whether the insurance system should be fully private or a public–private partnership. The latter implies public spending, and the sums have been considerable, as for example in the USA (Goodwin and Smith 2013). In the case of catastrophic risks, some public involvement is likely to be required due to market failure (Hardaker et al. 2015). This does not seem to cause conflicts of interest, given that respondents expressed the acceptance of public spending on agricultural risk management. It should be noted that the coexistence of disaster aid and subsidised insurances may not be a viable policy option if the former is not conditional on the latter. The results of van Asseldonk et al. (2002) and Liesivaara and Myyrä (2017) demonstrated that willingness to pay for insurance decreases if society is expected to provide disaster aid. Therefore, policymakers must carefully coordinate the provision of disaster aid if an insurance system exists. An efficient policy response to agricultural crises generally requires clearly defined boundaries for catastrophic risks and some predefined framework for the implementation (OECD 2011; Hardaker et al. 2015).

Although society and farmers’ insurance were highly preferred, farmers were generally the third most preferred option. Perhaps surprisingly, farmers were clearly preferred over consumers and retailers in most cases, despite the fact that farmers are in a much weaker position in the food chain compared to retailers. Therefore, farmers must continue to develop their own risk management strategies, and these efforts could be supported with some policy measures, for example, facilitating the establishment of mutual funds or financial instruments in the private market. It should be noted that subsidies from society very likely reduce farmers’ willingness to adopt fully private risk management instruments. If farmers are wanted to adopt fully private risk management instruments and to develop their own risk management strategies, then disaster aid and heavily subsidised insurance systems should be forgone.

This study focused on the preferred cost bearers of disasters occurring on farms, but mixes of policy instruments may be relevant in the design of actual policies. What can be stated about the policy mix based on the previous literature is that trade-offs inherently exist. This suggests the selection of a single policy rather than a mix of policies. A subsidised insurance system appears to be the best candidate based on the results and the previous literature. On the other hand, a mix could be the most viable option in the case of price risks, as they most directly affect the whole chain. Our results demonstrated that the preferences between different instruments were most dispersed in the case of price risks, indicating the legitimacy of a policy mix. This could be a mix of subsidised income insurance and policies to increase the market power of farmers. The feasibility and implementation of these policy instruments were not further analysed. This could, perhaps, be considered as a limitation but certainly a topic for further research. In practice, this would imply combining citizens’ preferences and the economic feasibility of different policy measures. Combining these two aims in policy design may require balancing between legitimacy and efficiency.

Previous studies surveying citizen views on agricultural policy have generally focused on the legitimacy of agricultural subsidies in a broad sense. American and European studies have found that citizens support agricultural subsidies, and the results of the present study confirmed this. On the other hand, the results did not provide evidence for consumers’ perceptions of fairness noted in some consumer studies. Consumers in these studies stated a higher willingness to pay for a product if farmers received a higher share of the price. The retailers’ share has also been considered excessive (Busch and Spiller 2016). However, in the present study, consumers and retailers were not generally considered as potential alternatives to bear the costs from agricultural disasters. It should be noted that the survey was carried out before Russian's invasion of Ukraine. This event led to the realisation of catastrophic risks to realise, and European farms suffered from soaring input prices and consumers from soaring food prices. In Finland, the role and responsibility of retailers, in particular, has gained considerable public attention. Thus, it would be interesting to repeat the survey to examine whether a concrete and very recent crisis changes the attitudes of citizens.

The results from this survey provide a direct signal to Finnish policymakers, but they cannot be directly applied in other EU member states. However, the results revealed a discrepancy between the current policy paradigm and the preferences of the public regarding the role of insurance. The respondents recognised insurance as a central risk management tool, but the actual role of insurance is negligible in Finland. This result is something that could also be considered in other member states, as the preferences of farmer and non-farmer populations may differ from the current policy. A similar survey conducted in other member states may reveal the discrepancy in the preferences between population groups and from the current policy.

6 Conclusions

Agricultural risks have increased due to climate change and market liberalisation, among other factors. Since farms have a highly limited capacity to manage catastrophic risks and recover from major economic losses, the question about the cost bearer has become more relevant and current. This study revealed that Finnish citizens prefer society and farmers’ insurance system to bear the costs from major losses occurring on farms. According to the results, the two were, in general, equally preferred among non-farming citizens. Farmers, on the other hand, had lower preferences for insurance compared to non-farming citizens. From a policy perspective, a subsidised insurance system could provide the best compromise, which would fulfil the preference for both society and insurance and considers the preferences of farmers and non-farming citizens. It appears that disaster aid also has legitimacy from the public perspective, especially in the case of catastrophic price risk.

In Finland, agricultural risk management does not currently involve considerable public spending compared to other agricultural subsidies. Given that Finland currently lacks a well-established policy framework for agricultural risk management, this study provides valuable information for policymakers in constructing the policy. According to the results, the legitimacy of the policy does not appear to depend so much on a social reference group, but more on the risk type and instrument. Disaster aid and subsidised insurances have an existing legal framework, and it appears that they also have legitimacy. The future challenge for the policy is how to design and implement consistent and efficient policy measures that neither weaken private insurance markets nor reduce farmers’ incentives to manage risks.

Funding statement

We thank the Academy of Finland [310205] for the financial support.

Data availability statement

The data and software code underlying this article are available in online supplementary material.

Footnotes

Appendix

References