Cost of wild boar to farmers in Sweden

Abstract

1. Introduction

Wild boar (Sus scrofa) are among the most widespread mammals in the world (Massei et al., 2015). Their natural range extends from Western Europe to East Russia, Japan and South-East Asia. In Sweden and many other countries, the wild boar population has the capacity to grow very rapidly, with a mean annual growth rate of almost 50 per cent during 2000–2010 (Lemel and Truvé, 2008; Jansson and Månsson, 2011; Gren et al., 2016). Wild boar are hard to control because they are difficult to hunt, they are generalists and they can adapt to most weather conditions. Sows can produce 10–13 piglets per year and they have few predators (Timmons et al., 2012).

Costs to farmers from wild boar arise from crop field damage by natural habitat selection and rooting behaviour during feeding (e.g. Frederick, 1998; Rao et al., 2002). Other costs are associated with damage to agricultural machinery due to wild boar nesting in the fields and damage to the quality of silage and grain due to admixture with soil. A third type of cost is associated with protective measures, such as hunting and the fencing of fields. In several countries, but not (yet) in Sweden, costs also arise from lost livestock through the transmission and dispersal of African swine fever (AFS) (Podgórski and Smietanka, 2018). AFS affects wild and domestic pigs and causes acute fever and death within 20 days of infection. Several studies have shown that these categories of costs can be considerable. In the United States, they total to approximately US$ 3.8 billion annually (Tanger et al. 2015; Holderieath, 2016; Mengak, 2016; Poudyal et al., 2017). Costs can be mitigated by fencing fields, frightening or trapping animals, feeding to lure them away from the crop fields and hunting (e.g. Geisser and Reyer, 2004; Náhlik et al., 2017).

The purpose of this study was twofold: to estimate the costs to farmers of wild boars in Sweden and to assess the determinants of the magnitude of these costs. Costs in terms of damage to crops and agricultural machinery and protection measures were included. With respect to the first purpose, data on different types of costs were collected through a survey of farmers. A stratified random sampling technique was used to ensure that observations were obtained from regions where wild boar are found and for different farm categories. Regarding the second purpose, data from the survey were combined with official data on landscape diversity at farm level and on wild boar abundance at a local level. Landscape diversity was included since it has been shown that wild boar population growth is positively correlated with landscape features such as deciduous trees and grass pasture (Gren et al., 2016). The specific econometric challenges were the censored distribution of costs to farmers obtained from the survey and the endogeneity of some explanatory variables. The resulting cross-sectional data were therefore analysed using the IV-Tobit econometric method.

There is a large body of literature calculating the costs of wildlife to farmers mainly in terms of costs of carnivores in developing countries (for a review see Gren et al., 2018). A few studies have calculated farmers’ costs of ungulates, mainly deer and wild boar. The studies estimating costs of wild boar include Pimentel et al. (2005), Schley et al. (2008), Lindblom (2010), Tanger et al. (2015), Anderson et al. (2016), Mengak (2016), Statistics Sweden (2016a) and Poudyal et al. (2017). Except for Pimentel et al. (2005) and Schley et al. (2008), all the studies applied a survey method to elicit the costs, which is the most common method used in the literature to calculate ungulate costs to farmers (Gren et al., 2018). Pimentel et al. (2005) did not define a cost, but simply assigned a cost of US$ 200 per wild boar, while Schley et al. (2008) used compensation payments for damage in Luxembourg.

The majority of studies based on farmer surveys cover only crop losses (Lindblom, 2010; Swedish Board of Agriculture, 2010; Anderson et al., 2016; Holderieath, 2016; Statistics Sweden, 2016a). The studies of farmers in Sweden indicate that the cost of wild boar could range from US$ 8/ha to US$ 55/ha. Anderson et al. (2016) used a large-scale survey and calculated costs from crop loss for 11 states in the USA, while Holderieath (2016) used an agricultural sector model to evaluate net welfare gains from removing feral pigs in nine states in the USA. A few studies included costs from both crop losses and other costs, mostly for cases in the USA, and found that the non-crop loss (e.g. replanting and additional field cultivation, losses of stored commodities, losses to hunting lease income, losses of livestock and repair and replacement of damaged equipment and farm infrastructure) could account for approximately 30 per cent of the total cost (Tanger et al. 2015; Mengak, 2016; Poudyal et al., 2017).

Only two studies have examined the impacts of different explanatory variables on wild boar damage (Geisser and Reyer, 2004; Lindblom, 2010). Geisser and Reyer (2004) estimated the impact of three mitigation strategies (hunting, fencing and supplementary feeding) on damage intensity in the canton of Thurgau in Switzerland, using data on damage obtained in a survey of farmers. They did not estimate the costs, but only damage frequency, measured as the number of damage events per unit area of cropland. Using this as the dependent variable, they found that only hunting has a significant negative effect. In a survey of farmers in Sweden, Lindblom (2010) found that the probability of damage decreases with increasing distance to forest and potential bait stations.

The present study is believed to make two novel contributions: it offers an assessment of three types of costs of wild boar (crop loss, machinery and protection), which has only previously been undertaken in a few studies, and it also offers an evaluation of the explanatory power of management, landscape characteristics and wild boar abundance for the magnitude of costs. Costs of ASF in terms of livestock losses were not included since the disease has not (yet) been detected in Sweden. The possible effects on forestry of wild boar rooting behaviour, which may occur with young forest plants in particular, were also excluded. According to a survey of forest managers by Jansson and Månsson (2009), damage to forest from wild boar is not regarded as a serious problem by forest owners in Sweden.

The manuscript is organised as follows. Section 2 presents a simple bio-economic model of wild boar impacts on farmers, which provides the basis for the design of the questionnaire and the econometric analyses. This is followed by the description of the survey design and presentation of the results in Section 3. The econometric analysis of the determinants of the costs and associated results are presented in Section 4. The study ends with a brief summary and conclusions.

2. A simple bio-economic model of the costs of wild boar to farmers

The cost of wild boar is defined as the impact on optimal profits over an infinite period of time. Cultivation of several crops was assumed, where $πi$ is the unit profit of a crop, which is assumed to be constant over time, with i = 1,…,n crops without any wild boar damage. The yield of each crop was assumed to depend on the area of land allocated to it, $Qi(Lti)$⁠, where $Lti$ is the area of land allocated to crop i in time t. If wild boar do appear, a share of the crop yield, $0≤dti≤1$⁠, is damaged. Costs other than from crop losses may arise due to damage to machinery and other infrastructure on the farm, $ct$⁠. The costs of crop losses and damage to infrastructure can be mitigated by protection measures, $Atr$⁠, where r = 1,…,m are the measures such as repairs to fields damaged by pigs, creation of open areas, construction of fences, and frightening activities by humans or dogs. The damage also depends on the given landscape characteristics, $K$⁠, where a more diversified landscape with field elements attracts wild boar because of hiding opportunities. Hence, the damage to crops and infrastructure depends on the wild boar population, W_t, protection measures and landscape characteristics: $dti=di(Wt,At1,…,Atm;K)$ and $ct=c(Wt,At1,…,Atm;K)$ where $dWi≥0,dAi≤0$⁠, $dKi≥0$⁠, $cW≥0,cAr≤0$ and $cK≥0$⁠.

The denominator on the right-hand side of equation (3) is the increase in the number of animals from a marginal feeding, and the condition thus shows that the unit hunting value of an animal equals the cost per animal of a marginal increase in feeding.

Equation (4) shows the familiar condition that the benefit of a marginal increase in the population in a period on the left-hand side of equation (4) equals the marginal cost on the right-hand side (e.g. Conrad, 2010). The latter includes the foregone return on the net value of the harvest, and marginal damage to crops and infrastructure. Without any marginal stock effects, i.e. when $∂di∂Wt=∂c∂Wt=0$⁠, the population level is determined solely by the discount rate and growth rate. Consequently, the farmer is indifferent between a marginal increase in wild boar growth and eliminating another wild boar and investing the income to obtain the return $ρ$⁠. Note that the condition is the same for $πS<0$⁠, but the interpretation of marginal benefits and costs is slightly changed. The left-hand side of equation (4) then expresses the cost of a marginal increase in population, and the right-hand side includes marginal benefits in terms of the avoided cost of hunting (the first term) minus the marginal cost of crop losses and damage to infrastructure (second and third terms).

So far it has been assumed that the farmer owns the land and has the right to hunt. This is not the case for tenant farmers in Sweden, who constitute 38.5 per cent of all farmers (Swedish Board of Agriculture, 2018). The benefits of reducing the wild boar population are then reduced by the associated value of hunting. In this case, there are only costs associated with marginal increases in the wild boar population. These costs may rise due to the absence of hunting as a protection option (if this option is included in the optimal solution for a farmer owning the land).

3. Design and results of the survey

Farmers in Sweden were already experiencing the cost of wild boar back in the 18^th century, when the animal had been in the country for over a thousand years (Tham, 2004). Wild boar were consequently eradicated in the late 1770s, but minor populations were kept in fenced areas over the centuries. Some individuals escaped in the 1970s and the feral population had increased to approximately 200,000 by 2015 (Gren et al., 2016).

3.1. Design of the questionnaire

The survey of the costs to farmers today (in 2015) was based on the simple bio-economic model in Section 2, which identified three classes of questions: (i) costs (crop losses, machinery damage, protection actions), (ii) feeding and protection measures and (iii) farm characteristics (farm size, crop patterns, land ownership). The questionnaire was developed and tested on a focus group of 13 farmers in February–May 2014. These farmers were selected based on having a good awareness of the problems associated with the wild boar population and substantial knowledge of different wild boar mitigation strategies. The questions were mainly considered satisfactory, but the farmers’ focus group had strong opinions about questions related to feeding. Feeding of wild boar has been a sensitive issue for some time (Sveriges Riksdag, 2016). Fifty-eight per cent of Swedish farmers are over the age of 55 (Swedish Board of Agriculture, 2014). They have typically remained in the same area for a long period of time and are therefore perceived to be well acquainted with conditions facing agriculture in their local area. Given the advice of the focus group as well as these facts, it was decided to design the questionnaire in line with their comments. Hence, a direct question about feeding might result in an untruthful answer or non-response, despite the anonymity of the survey. Therefore, questions on feeding wild boar were reformulated to inquire about feeding in the neighbourhood, irrespective of who is responsible. A translation of the final questionnaire from Swedish to English can be found in the Supplementary data (at ERAE online).

Crop losses were calculated based on the impact on four major crop categories: grain and oilseeds, forage and fodder crops, potatoes and sugar beets, and miscellaneous crops such as field-grown vegetables, bio-energy crops etc. The respondents were asked to grade the damage incurred in terms of losses inflicted upon land, with at least 5 per cent of the crop affected by extensive damage due to rooting behaviour or wild boar nesting in the field. For the area affected, the respondents were provided with 11 alternatives for damage assessment, ranging from 0 to 5 per cent to more than 50 per cent of crop yield. In addition, a check question on ‘no crop damage’ was introduced to examine the consistency of the answers. This check was motivated by the fact that some farmers in the focus group test raised the issue of possible ‘over-reporting’ of damage. Costs due to crop losses were calculated based on the responses in the questionnaire and data per unit crop value for the region (Andersson et al., 2016).

The question about machinery repairs reflected only repairs/maintenance and not replacement. Normally, to the best of the authors’ knowledge, damage to machinery would not be sufficiently great to require replacement. Costs of protection measures were obtained from the reported number of hours of farmers’ labour needed to mitigate the damage caused by wild boar. Costs were estimated as the average labour cost per hour multiplied by the reported hours. The costs for equipment, such as fences, traps etc. (excluding labour costs), were specifically addressed. Wild boar traps were implicitly included in this question since it was related to a question about ‘additional costs due to wild boar’, which included traps, fences etc. However, it should be noted that specific traps for wild boar have played a very minor role in controlling the wild boar population in the period 2010–2012 (SOU, 2014). In 2013, there were only eight legally permitted traps, which were subject to fairly extensive regulations in terms of the period during the year and time of day when they can be used.

With respect to the second category of questions, feeding practices may be applied to lure the animals away from agricultural land, but also to concentrate them in an area where they are relatively easy to hunt. Wild boar are difficult to hunt because they are active at night and have excellent hearing. The questionnaire included two questions on feeding. One pertained to the frequency, with six choices ranging from feeding less than every month to feeding when baits are empty. The other question concerned the spatial frequency, divided between the number of feeding plots/area unit or number of farmers with feeding plots in the vicinity. With regard to protection measures, the questionnaire contained questions with binary responses on five alternatives: hunting piglets in fields, creating open strips around tillable land, frightening pigs, constructing fences and others. A hunting licence is needed to hunt wild boar or any other game, which is issued by the Swedish Environmental Protection Agency. In addition, hunters without their own land with hunting rights need to access such land by renting it individually or as a member of a hunting team. If these conditions are met, hunting of wild boar is only restricted with respect to the prohibition of hunting sows with piglets. Otherwise, the animal is not subject to any additional regulations and can be hunted at most times of the year.

With respect to the third class of questions, the questionnaire included questions on farm size, hunting rights and land ownership. Tenant farmers may have some limited rights to conduct hunting for protection purposes. However, according to a government investigation (SOU 2014: 54), only 12 per cent of tenant farmers have the full right to hunt on the land they lease and just 4 per cent have a special agreement that enables them to hunt for protection purposes. The questionnaire contained questions on the area of rented agricultural land.

3.2. Survey implementation

The final questionnaire was distributed to a sample of famers. In total, there are approximately 67,000 operating farms in Sweden (Swedish Board of Agriculture, 2016b). In order to obtain observations from regions where wild boar are found and from different farm categories, a stratified sampling technique was applied (Scott and Smith, 1975; Krosnick, 1990). The distribution of wild boar is limited by the cold Nordic climate since piglets are sensitive to cold winters. Wild boar are therefore rare in the five northern counties of Sweden. They are also not found on Gotland, a Swedish island located in the Baltic Sea. Wild boar are present in the other 15 counties of Sweden (Supplementary data at ERAE online). The farms were stratified based solely on tillable acreage in order to ensure that a sufficient number of farms in each size category of tillable land were included in the sample. It should be noted that 58.5 per cent of all farms in Sweden operate less than 20 hectares of tillable land (Swedish Board of Agriculture, 2016a), hence a random sample would be less likely to accurately reflect the conditions facing larger farms.

The Swedish Board of Agriculture (SBA) administered the questionnaire, which was initially sent to a sample of 3,200 farmers, stratified by county, in the 15 counties with wild boar populations. Two hundred farmers with at least 5 ha were selected in each county except in substantially larger counties geographically such as Skåne and Västra Götaland, where 300 farmers were selected. Farmers with smaller areas could not be included since they are not registered in the SBA data set that is needed for obtaining data on landscape characteristics discussed in Section 4. The questionnaire was sent out in April and the initial response rate was around 35 per cent. A reminder was sent to 2,109 farmers in May 2015. In a second reminder in early June, the questionnaire was sent to 1,527 farmers and a supplementary questionnaire sent to 825 farmers. Altogether, a total of 4,025 farmers received the questionnaire and the final response rate was 61.9 per cent.

A telephone interview was conducted with 140 randomly selected farmers (stratified by county) of the population that had received but not answered the survey during the regular period. SBA was able to contact 72 of these farmers, i.e. 51.4 per cent, for a telephone interview. This survey of the non-respondents showed that the most common reasons were forgetfulness (32 per cent), questionnaires completed but not reaching SBA (35 per cent), and only having minor problems with wild boar (18 per cent). Another crucial variable was the tenancy ratio that measures the share of tillable land that is rented. The share of rented land of non-respondents amounted to 37.6 per cent, which was slightly lower than the weighted average of the survey. This ratio amounted to 42.9 per cent. This may partly be explained by the fact that the farms in the survey were slightly larger, with the tenancy ratio typically increasing with farm size (Andersson, 1995).

The size of farm, the share of rented land and the share of forage crops in the sample among the population was also compared (Andersson et al., 2016). The average farm size in Sweden was 42.5 hectares of tillable land in 2014 (Swedish Board of Agriculture, 2015). The unweighted average farm size in the sample was 109.4 hectares, but given the stratified sampling approach with sample weights attached to each observation, the weighted average of the sample revealed a farm size of 51.4 hectares, which is close to the farm size in the entire population of farms during 2014. Similarly, the weighted average share of rented land and forage crop in the sample was 0.39 and 0.36, respectively (Table 1). The corresponding shares in the population are 0.38 and 0.42.

Based on the follow-up study of non-respondents and the comparison of the weighted size and share of rented land and forage crops between the sample and the population, no systematic bias was expected in the responses. However, some responses were deleted due to inconsistencies, such as the area reported to be affected by wild boar being larger than the total area on the farm. In total, the dataset contained 2,484 usable observations.

3.3. Survey results

Results from the survey revealed that 39 per cent of respondents had experienced at least one of the three types of costs caused by wild boar. Since the questions and responses on costs were measured in SEK (8.44 SEK = 1 US$, and 9.36 SEK = 1 euro on average in 2015), the results below are presented in SEK. The average total cost of wild boar amounted to 28,842 SEK/farm, the majority (63 per cent) due to crop losses, 9 per cent to machinery damage and 28 per cent to protection costs (Table 1).

The variation in costs per farm was large, ranging from 0 to 3,448 thousand SEK. The largest cost is explained by crop losses on large farms in mid-Sweden, which has large populations of wild boar (Jansson and Månsson, 2011; Gren et al., 2016). When the average cost was calculated for farmers who reported wild boar damage, the cost increased to 71,480 SEK/farm. When relating the cost to the area instead of per farm, the total cost was SEK 305/ha and the cost of crop loss SEK 175/ha. The cost of crop loss may be compared with similar Swedish studies where the cost ranges between SEK 66/ha and 479/ha, depending on the type of crop (Lindblom, 2010). The average cost to farmers with reported losses amounted to SEK 743/ha, of which 57 per cent came from crop losses and the remaining 42 per cent from repair and protection costs.

With regard to the responses to the second type of questions, the results in Table 1 revealed that 42.1 per cent of farmers reported the existence of feeding, most of which occurred far away from their fields (15.6 per cent). Approximately 7 per cent of the farmers reported feeding close to or quite close to their own fields. The frequency was evenly distributed between the three options, ranging from 5.1 per cent to 6.1 per cent (low frequency). The share of farmers reporting any type of protection measure was similar to that of feeding, and amounted to 43.8 per cent. Hunting was most common and practised by almost 30 per cent of all the farmers.

With respect to farm characteristics, grain and oil seeds were cultivated on 52.9 per cent of the arable land area, and forage on 35.6 per cent of the area. It was also noted that the average share of rented land was 0.385. Approximately 30 per cent of farmers reported no rented land, but less than 10 per cent had no land of their own. Not all farmers had a hunting license and almost 40 per cent did not hunt.

4. Econometric analysis of the determinants of costs

The survey generated data on the necessary dependent variable, i.e. costs of wild boar, and several, but not all, of the explanatory variables. As shown in the bio-economic model in Section 2, there was a need for data on landscape characteristics and wild boar population. Data on landscape diversity were obtained from SAM, a database containing extensive information on each Swedish farm unit that serves as the basis for direct income payments as part of the EU CAP payment programme (Swedish Board of Agriculture, 2016b). The data include information on the type and area of crops grown, livestock operations, and additional data on the configuration of agricultural land. One measure of landscape diversity is the number of ‘blocks’ on a farm. A ‘block’ is a tract of land defined by natural borders such as roads, forest land, pasture land, rivers, ditches, trenches or other landscape elements. Consequently, the number of blocks and area per block provide indirect information concerning the nature and diversity of the agricultural landscape. The results showed that the average number of blocks per farm was 27.746 (Supplementary data at ERAE online).

A major challenge was to obtain data on wild boar populations at the local level as there are no official data. Lack of data on population size is a common problem, not only for wild boar, but also for most other types of wildlife. In ecology, different methods and constructs have been developed to approximate population abundance (e.g. Maunder et al., 2006). A commonly used measure is the relationship between killed animals and a pressure variable. Data on the number of wild boar killed by hunters is available from Viltdata (2016). An ideal measurement of pressure in the present study would have been the number of hunters and their hunting efforts, but these data were not available. Instead, the number of hunting licenses granted was used as an approximation of hunting effort. The average number of wild boar killed, so-called bags, per hunting license is approximately 0.70, but it may reach 4.3 in some municipalities (Supplementary data at ERAE online). Owing to the simplified approximation of hunting effort, an alternative measure of population size was constructed and tested: number of bags/ha. On average, the number of wild boar bags/ha amounted to 8.038 per 1,000 hectares, but it varied considerably and in extreme cases it was 52 per 1,000 hectares (Supplementary data at ERAE online). Irrespective of the construct, the variable may show a positive relationship with farmers’ total costs, but may show a negative relationship when hunting is used as a protective measure.

Since it was difficult to disentangle the impacts on damage of the different types of feeding categories, a composite feeding index was constructed, Feeding_COMP (Supplementary data at ERAE online), which included all feeding choices. Principal component analysis was used in which the feeding categories were weighted according to their relative contribution to the explanation of the total variance (see e.g. OECD (2008) for details). The weights were used when at least 70 per cent of the total variance was explained in principal component analysis (Andersson et al., 2016). A similar composite index was also constructed for the protection measures, denoted Protection_COMP (Supplementary data at ERAE online).

4.1. Econometric specification

Econometric estimation of the determinants of damage cost(s) associated with wild boar in the present dataset was hampered by two main econometric challenges. First, approximately 60 per cent of farmers in the dataset do not incur any costs from wild boar, which resulted in a censored distribution of the cost of damage (Supplementary data at ERAE online). Second, wild boar abundance and game protection measures by farmers are potentially endogenous. For instance, high boar bag density increases agricultural damage, but areas with high food abundance, and hence high damage, may also attract hunting and increase bag density. Furthermore, protection measures can reduce the cost of damage, but adoption of protection measures could be in response to a high level of damage experienced by farmers. Therefore, the possibility of reverse causality cannot be ignored. These factors mean that ordinary least squares (OLS) estimation of the determinants of costs is likely to induce biases.

As in all IV models, the instrument(s) must fulfil the exclusion restriction criteria of being correlated to the endogenous covariate(s) and only correlated with the outcome variable through the endogenous covariates. Snow depth, change in temperature and total hunting licenses in a municipality were used as instruments for wild boar abundance and wildlife protection measures (Swedish Meteorological and Hydrological Institute, 2018a, 2018b). Snow depth is a major factor influencing wildlife abundance. Greater snow depth reduces food availability for wildlife since potential sources of food are covered by snow. As a result, wild boar have to expend a considerable amount of energy digging to clear the snow for a small amount of food energy, resulting in starvation and low reproduction rates (e.g. Lindström et al., 1994). In view of the above, changing climate or weather patterns will have implications for the distribution of wild boar populations. As a result, the deviation in temperature during the survey year from the long-run average temperature in the municipality was included as an additional instrument.

It should be noted that while temperature and snow cover may to a certain extent be correlated, the variable here was a measure of climate change rather than an absolute temperature in a given year. This variable was expected to be largely uncorrelated with other determinants of costs of crop damage. Even though climatic factors such as temperature and snow are generally determinants of crop choice, the present study area was largely homogenous in terms of the crops cultivated. An analysis of cropping patterns in 2015 (Swedish Board of Agriculture, 2016b) revealed that grain typically accounts for 30–50 per cent of the tillable acreage in the counties, with the exception of some of the more forested areas. Sugar beet is the only crop concentrated in one region (Skåne), but accounts for just 4.2 per cent of the tillable acreage in the region. The concentration in Skåne is mainly due to structural changes in the location of the sugar processing industry and not due to weather patterns.

Finally, the total number of hunting licenses in a municipality could indirectly determine hunting pressure and hence wild boar abundance. Arguably, a hunting license combined with hunting rights allow the shooting of all game on the land in question, with moose and deer regarded as the most popular game (Engelman et al., 2018). Thus the abundance of wild boar would not be expected to have an impact on the number of licenses, so this variable satisfied the criteria for instruments. A correlation matrix between these instruments and the variables used in the regression analysis shows no evidence of a significant correlation between the instruments and the other explanatory variables (Supplementary data at ERAE online).

Conditional on the instrument validity, the IV-Tobit estimation allowed an examination of the influence of wild boar abundance and farm feeding and protection measures on the cost of wild boar to farmers in the present dataset. A key limitation of Tobit class models is the strong assumption of homoscedastic residuals. To this end, standard errors were clustered at county level to reduce the extent of bias and improve efficiency of the standard errors.

4.2. Econometric results

The results of the IV-Tobit confirmed the a priori expectations that wild boar presence and protection measures are endogenous. In all regressions, the p-value Wald test of exogeneity was highly significant, thus rejecting the null hypothesis of no endogeneity. The first-stage results also showed a strong correlation between the instruments and the set of endogenous variables. Four different models were estimated; costs measured as total cost and only crop losses, and wild boar abundance as hunting bags per license and per 1,000 ha (Table 2).

The results showed a significant positive impact of wild boar abundance as measured by both constructs. Other robust results were significant with negative effects of share of land with grain, and significant positive effects of number of blocks. However, area of arable land and the index constructs of feeding and protection practices showed no significant impact in any of the models.

The negative effect of land cropped with grain may partly be due to grain crops typically being grown on relatively large fields with homogeneous landscape conditions. Another reason may be the lack of feed during winter with these crops, which instead puts grassland at risk from the presence of wild boar. The significant positive effect of number of blocks indicated that heterogeneity in landscape elements obviously contributes to the magnitude of damage cost by wild boar. Blocks on farmland provide shelter for wild boar, particularly during the day, since the animals scavenge mainly at night.

One difference between the two models of wild boar constructs was the estimate of rented land. The results showed a positive impact of the proportion of all farmland rented on the cost of damage, but it was significant only for wild boar per hunting license, and only at a relatively low significance level. Hunting rights are associated with land rights, so permission to hunt belongs to the landowner. Thus, farmers who rent land have limited options to hunt and manage land to avoid damage, so the estimated positive sign for rented land was expected.

Although area of land was not significant in any model, the sign was positive in all models, as expected. The low coefficients show a relatively small impact of area of land on damage. The coefficient for feeding index was also positive for all models, indicating that feeding contributes to damage costs, possibly through positive impacts on the population size. However, the index for protection measures showed different signs, depending on the choice of the wild boar abundance measure.

To assess the effects on costs of marginal changes in the significant explanatory variables, the impact of the variables on the total cost and costs of crop damage were calculated using wild boar per licence as the construct of population abundance. In the authors’ view, this construct is more appropriate than wild boar per unit area, since it reflects the impact of pressure. Animals killed per unit area could be a reflection of extensive efforts, and not necessarily of a large population abundance. A marginal increase in killed wild boar per license would involve an increase in killed animals by approximately 142,000, from approximately 98,000 in 2015 (Viltdata, 2016), which does not seem realistic. Instead the cost impacts of a 10 per cent increase in killed wild boar per licence were calculated. This was also done for the significant explanatory variables measured as shares, which included the share of grain cropped and rented land relative to total arable land per farm. Blocks are measured in numbers, and the effects were calculated of a marginal increase in this variable on cost. In order to compare the magnitude of impact between the variables, the elasticities were calculated. All the values were calculated by taking the derivative of the cost function in terms of the respective variable, and then evaluating the mean value of the cost and variable in question. The results are presented in Table 3.

For both total costs and costs of crop damage, there was a higher estimated change in magnitude and elasticity (in absolute values) of changes in wild boar abundance and share of land cropped with grain. A decrease in the share of grain acreage reduced total cost and damage from crop losses by SEK 3,683 and SEK 3,860, respectively. This indicates that policies promoting grassland with an associated decrease in the cultivation of grain, for example, would add a considerable cost to farmers. Policies increasing the area of land with permanent crops to reduce leakage of nutrients have been widely implemented in Sweden (EC, 2018). A land use change generating a 1 per cent decrease in the share of grain acreage would increase the total cost and cost of crop damage by 1.28 and 2.12 per cent, respectively. The elasticities of the two cost types with respect to changes in wild boar abundance were also relatively high, while they were below unity for the share of rented land and number of blocks.

5. Summary and conclusions

The cost of wild boar to farmers in Sweden were calculated based on data collected in a survey of 4,025 respondents, representing 5 per cent of the total number of farmers in Sweden. The response rate was 61 per cent, and the results revealed that a majority (approximately 60 per cent) of the respondents did not incur any costs from wild boar. The average damage cost for all farms amounted to 28,842 SEK per farm, which corresponds to approximately 17 per cent of average net farm income in 2015 (Statistics Sweden, 2016b). Similarly, the average cost per farm amounted to 7 per cent of the total value of the crops produced, which is substantially higher than that reported by Tanger et al. (2015). Most of the cost (63 per cent) was attributable to crop losses, 9 per cent to machinery damage, and 28 per cent to other losses, such as various forms of protection costs. The cost due to crop losses was within the range reported in the few previous studies available on the costs of wild boar to farmers in Sweden (Statistics Sweden, 2016a).

With respect to the determinants of the damage costs of wild boar, a method (IV-Tobit) was used that accounted for the large number of zero damage respondents and the risk of endogenous explanatory variables. Owing to a lack of data on wild boar abundance, an approximation was constructed relating killed wild boar to the number of hunting licenses. Another measure, killed animals/1,000 ha, was also tested. A specific feature of the study was the inclusion of landscape diversity as an explanatory variable. Robust results were obtained using the IV-Tobit model, which indicated that wild boar abundance and landscape diversity contribute to high damage costs, while the share of land cropped with grain reduces damage costs. A 10 per cent increase in wild boar abundance increased total costs by 3,680 SEK or approximately 15 per cent. Similarly, landscape diversity, measured as the number of fields separated by different landscape elements, increased damage. This highlights the cost of providing diversity in the agricultural landscape, which to the authors’ knowledge has not previously been identified.

Furthermore, these results showed that reducing the share of land cropped with grain by 10 per cent reduces the costs of wild boar per farm by SEK 3,682. This may reflect a less well-known negative side effect of several policy provisions with specific subsidies targeted at dairy and beef production, which require forage/grass land with relatively low shares of grain production. Hence, these results once again reveal that some policies designed to promote biodiversity and other environmental goals indirectly affect the costs attributable to the wild boar population. However, it could be argued that damage by wild boar and other ungulates is captured by the market price of land, which should then be reduced in regions with abundant wildlife. If so, wildlife damage associated with some biodiversity measures, such as set-aside of land, is captured by the reduction in the opportunity cost of land.

Wild boar are excellent survivors and it is unlikely that populations will decrease unless hunting pressure increases. A full-fledged analysis of the optimal level of hunting must include the costs and values of wild boar and numerical information on population dynamics (e.g. Zivin et al., 2000). Lack of data on population dynamics is a problem for optimal analysis of wildlife in general, but studies on costs and benefits at given population sizes abound (see Gren et al. (2018) for a review). Two studies have estimated the hunting value of wild boar in Sweden. Engelman et al. (2018) considered different game animals as attributes of hunting and used a choice experiment to estimate willingness to pay for hunting bags of wild boar, which amounted to approximately SEK 390/bag. Mensah and Elofsson (2017) arrived at a much higher value when using a hedonic method, where the rental rate of land for hunting was determined by the number of bags of different game. They showed that the hunting value of a wild boar amounts to 2,400 SEK/bag on average. However, even this higher estimate is well below the present estimated damage cost of SEK 3,680 from a 10 per cent increase in wild boar bags per licence. In order to reduce the wild boar population, it may be necessary to compensate hunters. These results indicate that farmers’ losses due to wild boar are sufficiently large to make compensation payments acceptable. The design of such compensation schemes and the impact on the price of arable land are interesting topics for future research.

Acknowledgements

We are much indebted to the Wildlife Fund at the Swedish Environmental Protection Agency for financial support for the project ‘Economic analysis of wild boar’ (grant no. 12/133) and to three anonymous reviewers for useful comments.

Review coordinated by Ada Wossink

References