Direct experience and attitude change towards bears and wolves

Understanding how changes in the sizes of large carnivore populations affect the attitudes of the public is vital in order to mitigate social conflicts over large carnivore management issues. Using data from two Swedish postal surveys in 2004 and 2009, we examined the probable social effects of a continued increase in the Swedish populations of bear and wolf by comparing levels of direct experience of bears and wolves with public attitudes towards these animals. We report an increase in direct experience of bears and wolves, lower levels of acceptance of the existence of these animals, and a lower degree of support for the policy goals of both species in 2009 compared to 2004. We also find that these changes are more prominent in areas with local carnivore populations than in other areas of Sweden. Our results imply that attitudes towards bears and wolves are likely to become more negative as populations continue to grow. The uneven distributions of the carnivore populations are likely to generate more frequent social conflicts in the future as they could cause an increase in the attitudinal divide between those members of the Swedish public who have had direct experiences of carnivores and those who have not.


Direct experience and attitude change towards bears and wolves
Max Eriksson , Camilla Sandstr ö m and G ö ran Ericsson M. Eriksson (max.eriksson@umu.se) and C. Sandstr ö m, Dept of Political Science, Ume å University, , Dept of Wildlife,Fish,and Environmental Studies,Swedish Univ. of Agricultural Sciences,Sweden Understanding how changes in the sizes of large carnivore populations aff ect the attitudes of the public is vital in order to mitigate social confl icts over large carnivore management issues. Using data from two Swedish postal surveys in 2004 and 2009, we examined the probable social eff ects of a continued increase in the Swedish populations of bear and wolf by comparing levels of direct experience of bears and wolves with public attitudes towards these animals. We report an increase in direct experience of bears and wolves, lower levels of acceptance of the existence of these animals, and a lower degree of support for the policy goals of both species in 2009 compared to 2004. We also fi nd that these changes are more prominent in areas with local carnivore populations than in other areas of Sweden. Our results imply that attitudes towards bears and wolves are likely to become more negative as populations continue to grow. Th e uneven distributions of the carnivore populations are likely to generate more frequent social confl icts in the future as they could cause an increase in the attitudinal divide between those members of the Swedish public who have had direct experiences of carnivores and those who have not.
Th e Swedish parliament introduced population goals for the ' big fi ve ' carnivore species in the Swedish fauna -bear Ursus arctos , wolverine Gulo gulo , lynx Lynx lynx , wolf Canis lupus and golden eagle Aquila chrysaetos, in 2000. Th i what done to ensure the long-term survival in accordance with the criteria specifi ed in the EU Habitat Directive (Council Directive 92/43/EEC of 21 May 1992). Nine years later the defi ned minimum levels and interim targets had been achieved (Liberg 2010) and the government started to re-evaluate the status of the animals to update the policy (Dir.2010(Dir. :65 2010. Th e viability of large carnivore populations are aff ected both by a favorable ecological status and local acceptance, as acceptance is considered to be key to achieving sustainable long-term conservation of large carnivores (Guidelines for Population Level Management Plans for Large Carnivores 2008). A majority of Swedes support the current population goals set by parliament (Ericsson et al. 2006), but attitudes towards these policy goals and the animals as such are likely to change in the future as they are based on indirect information and poorly informed beliefs (Heberlein and Ericsson 2008). Th ese attitudes vary across Sweden geographically and inhabitants in rural areas with large carnivores are more negative to the population goals than other Swedes (Ericsson et al. 2006). Since 2000, the increasing numbers of bears and wolves, and their impact on farming, reindeer husbandry and hunting have given rise to intense debate, increased political polarization and societal confl icts, and as a consequence the revised large carnivore policy from 2013 marks a shift in policy; away from ecological concerns towards more socioeconomic aspects (Prop. 2012/13:191).
Since 2000, Sweden has experienced an increase in the numbers of bears and wolves (Kindberg et al. 2011, Svensson et al. 2012). Elsewhere, a fast growth in bear and wolf populations have correlated with a change towards more negative attitudes towards these animals (Kellert 1987, Bath and Buchanan 1989, Mech 1995, Rodriguez et al. 2003, Bisi et al. 2010 and towards other species of large carnivores . Livestock depredation, competition for huntable game and clashes between diff erent views of nature arises with growing populations of bear and wolves, and as these issues involves diff erent groups in society in diff erent ways social confl icts are likely to increase (Rodriguez et al. 2003, Skogen and Krange 2003). Central to this discussion is the role of direct experience on these societal reactions when large carnivore species e.g. wolves reappear in the fauna (Ajzen 1989, Ericsson and Heberlein 2003, Heberlein 2012. We argue that previous research on attitude change may have been too limited in its scope (Ericsson and Heberlein 2003) or contained unrepresentative sampling (Treves et al. 2013) to test the hypothesis of direct experience on attitudes towards large carnivores conclusively. Th e fi eld also suff er from a general lack of reliable time series data (Williams et al. 2002) making comparisons over time problematic. In an attempt to improve on these shortcomings, we analyzed data from two large representative samples of the Swedish public taken fi ve years apart.
Attitudes are mental structures made up of cognitive and aff ective components ( Fig. 1) which aff ect our evaluation of attitude objects (Heberlein 2012). An attitude object can be an actual physical object or something more abstract such as an idea or a situation (Olson and Zanna 1993). Over time, repeated interaction with an attitude object forms the basis of an attitude which acts as a roadmap for a response when faced with the same, or a similar, attitude object in the future (Olson and Zanna 1993). Th us, attitudes serve as a mental shortcut for the individual when evaluating an attitude object, cutting down on the costs of decision-making and possibly infl uencing behavior (Alwin and Krosnick 1991, Eagly and Chaiken 1993, Olson and Zanna 1993. Attitude patterns are assumed to be socialized early and then generally strengthened over time as a result of confi rmation bias (Eagly and Chaiken 1993, McFarlane and Boxall 2003, Heberlein and Ericsson 2005, making them stable mental structures that govern the creation of our identity, our world view and our actions (Olson and Zanna 1993).
Th ough normally resistant to change, attitudes may also change rapidly (Zaller 1992, Olson and Zanna 1993, Eagly and Chaiken 1998, Heberlein 2012. Th e conviction with which the pre-existing attitude is held determines the cost associated with attitude change, while the relevance of the new input for the attitude in question determines the benefi t gained from attitude change; if the benefi t exceeds the cost, an attitude change takes place (as suggested by Whittaker et al. 2006, Fazio et al. 1983, Fazio 1995. Th us, attitude change is most likely to occur in situations where an attitude with a low cost of change (henceforth referred to as a loosely held attitude) is challenged by new information or a new experience (Ajzen 1989). Direct experience of an attitude object has been shown to be more likely to change attitudes than more indirect experiences and information (Fazio et al. 1982). Th e attitudes of the Swedish public towards bears and wolves are likely to be loosely held, as they are based largely on second-hand information and a majority of the population have no direct experience of bear and wolf. Th is should make them susceptible to change (Williams et al. 2002, Heberlein andEricsson 2008) and makes both new information and direct experience possible drivers of attitude change (Heberlein andEricsson 2008, Houston et al. 2010).
A recent study suggested that knowledge about black bears infl uences attitudes toward recovery strategies (Morzillo et al. 2010). However, the fi ndings of quantitative studies vary. While some studies have discovered a positive relationship between knowledge and support for wolves, other studies point to a negative or nonexistent relationship (Kellert 1985, Biggs 1988, Lohr et al. 1996, Bjerke et al. 1998, Williams et al. 2002. Consequently it is still unclear to which extent knowledge makes people more or less supportive of wolves. A comparison of two samples both taken in 2004, revealed that municipalities with a carnivore presence tended to have more negative attitudes towards the wolf policy goal (Ericsson et al. 2006). Th is warrants oversampling of small rural municipalities, as these tend to also have a large carnivore presence. Th ese less positive attitudes in the carnivore area could potentially be explained by experience (Karlsson and Sj ö str ö m 2007), or by the predominantly rural locations of the bear and wolf populations (Ericsson et al. 2006, Treves et al. 2013. Sweden has an urbanization rate of approximately 85% ( Ͻ www. scb.se Ͼ ). Th is means that our random national sample will, mostly refl ect the attitudes of urban respondents, while the proportional multiple municipal sample design used in the carnivore areas will refl ect the attitudes of rural respondents (Ericsson et al. 2006, Heberlein andEricsson 2008).
Th e predominantly rural nature of the carnivore area gives carnivore issues strong overtones of urban-rural confl ict, often linked to personal experiences (Heberlein and Ericsson 2008). Typically, rural inhabitants see the protection of bears and wolves as something that is ultimately controlled by external forces, i.e. the urban majority Krange 2003, Heberlein andEricsson 2008), while the rural population are the ones aff ected by the cost of living with these animals in terms of predation damage to livestock, reindeer, and hunting dogs (Swedish Government Offi cial Report, SOU 2012:22). Th e eff ect of direct experience of large carnivores on attitudes towards them may be hard to determine due to the subjective nature of such experiences (Williams et al. 2002, Heberlein andEricsson 2005, Karlsson and Sj ö str ö m 2007, Treves et al. 2013). Some studies have suggested a negative correlation between experience and the wolf policy goal (Ericsson et al. 2006), indicating that people become less positive with experience. However, others have maintained that hearing a wolf howl or seeing a wild wolf can be valued as a positive experience (Heberlein and Ericsson 2005), while other hypotheses posit that the negative eff ects caused by a population increase will vanish over time (Zimmermann et al. 2001, Majic andBath 2010).
Studies worldwide suggest a link between change in large carnivore population size, experience of large carnivores, and a change in attitudes of the public towards them and their policy goals (sensu Williams et al. 2002). We treat knowledge as one possible mediator in this process, where direct experience on a personal level reinforces already known facts (Fazio et al. 1982), and is more likely to alter beliefs that a person holds towards the bears or wolves. Th is may then in turn result in an attitude change (Heberlein andEricsson 2008, Heberlein 2012).
Regardless of the specifi c paths and direction of eff ects, we predict that the spatial expansion and increase in the populations of bear and wolf will have an eff ect on attitudes towards these animals and their related policy goals.
Th is process should be driven by both indirect and direct experience (Ajzen 1989), with direct experience having the strongest eff ect. Direct experience is considered to be fi rst-hand experience, such as seeing an animal or an animal track, while indirect experience is information gathered through other sources such as the press. Attitudes towards bears and wolves should change over time, and this change should be greater in areas where these animals are present than in other places, as these areas are subjected to an increase in both direct and indirect experience. Based on this literature review, we make the following predictions: Levels of experience of carnivores should be higher in -2009 than in 2004. Levels of experience of carnivores should be higher in the carnivore area.
Attitudes towards the existence of bears and wolves and their policy goals should be diff erent in 2009 than in 2004. Attitude diff erences between 2009 and 2004 should be greater in the carnivore area than in the non-carnivore area.

Study area
We report data from two postal surveys using up to four contacts (Dillman et al. 2009). In both surveys, we used a two-tiered sample design: one random, proportional sample on a national level with a sample size of 1001 ( Our study design allowed us to isolate the eff ects of direct experience, using the national sample as a control group, while the municipal sample acted as a treatment group. Th e national sample is representative of the majority of Swedes; a largely urban group, with less direct experience of bears and wolves, while the municipal sample provided data on attitudes towards bears and wolves from rural groups with a greater level of direct experience or being close to individuals with direct experience (Ericsson et al. 2006).

Sampling and response rates
Th e 2004 survey had a response rate of 55% for the national sample, and 62% for the carnivore area sample. In 2009, the corresponding rates were 46% for the national sample and 52% for the carnivore area sample.
Drop-out analysis has been carried out with respect to sex and age. Th e diff erences in the overall response rates between our samples and the respondents as a whole were to be expected, given previous wildlife surveys. Women ' s participation was lower than for men and older respondents had higher response rates. Th us, given the fi ndings of previous research, this sampling bias will likely result in somewhat more negative attitudes towards carnivores, and a slightly higher degree of direct experience.

Data analysis
All data analyses were carried out using STATA 13. Population weights were employed when analyzing the municipal samples to achieve proportional samples refl ecting population size. Two-way ANOVAs were conducted for each experience, existence and attitude item separately, with measurement point and sample as independent variables. Th is was followed by Tukey -Kramer post hoc mean comparison tests between sampling groups (Dunnett 1980, Hayter 1984. Scales were created for each set of items using a combination of Cronbach ' s alpha-based coherence tests and principal component analysis (Sokal and Rohlf 1981).

Measurements
We measured experience with two questions repeated for each species with the answer alternatives: never, once, and more than once. Translated from Swedish, the questions were: " Have you ever seen bear/wolf tracks? " and " Have you ever seen a wild bear/wolf? " Th e attitude items were measured with two questions repeated for each species. In both cases, the alternatives ranged from 1 to 5 where 1 or 2 represented a negative attitude/a wish to reduce the policy goal and 4 or 5 represented a positive attitude/a wish to increase the policy goal; 3 was the neutral/acceptable alternative. Translated from Swedish, the questions were roughly: " What is your opinion of the fact that bears, wolverines, lynx and wolves exist in Sweden? " and " In the spring of 2001, the Swedish parliament decided how many large carnivores we should have in Sweden. Th e fi rst goals were set for the number of reproducing females, corresponding to at least 1000 brown bears (1500 lynx, 400 wolverines and 200 wolves). What is your opinion of the goals set by the parliament for large carnivores in Sweden? " A high degree of correlation (r p ) among the attitude items within each group of items (0.74 between the existence items and 0.71 between the policy items) encouraged the use of principal component analysis to avoid multicollinearity (Morgan and Sonquist 1963).
Th e PCA produced one-component solutions for all three set of items (k Ͼ 1) with internal reliability values (Chronbach ' s alpha) of 0.70 for the experience items, 0.84 for the existence items, and 0.82 for the policy items. To facilitate interpretation, we substituted the component scores with the raw sums of the items as these correlated highly (r p Ͼ 0.99). Th is process resulted in three 10-step scales where a low value indicated a lesser degree of direct experience, a less positive attitude towards the existence of the species in Sweden, or a willingness to reduce the policy goals of the species.

Direct experience
In both samples more respondents reported having direct experience with bears and wolves in 2009 than in 2004, and direct experience with bear (Table 1) was more common than with wolves (Table 2). In the carnivore area 73% of the respondents had seen bear tracks in 2009 compared to 68% in 2004. In 2009 36% had seen a wild bear compared to 31% in 2004. In the national sample 27% and 21% of the respondents had seen bear tracks in 2009 and 2004 respectively, and 13% had seen a wild bear at both measurement points.
In the carnivore area wolf tracks had been seen by 38% of the respondents in 2009 and 30% in 2004. 18% had seen a wild wolf in 2009 compared to 13% in 2004. In the national  Respondents in the carnivore area had more direct experience with carnivores than the respondents in the national, more urban, sample in both surveys; the diff erence between the two time periods was greater in the carnivore area than in the national sample. Th e ANOVA produced signifi cant simple main eff ects (p Ͻ 0.05) with regards to both measurement point in time and sample across all experience items, as well one statistically signifi cant interaction eff ect (p Ͻ 0.05) between measurement point and sample for the seen bear tracks item, F 1,12593 ϭ 10, 92 Indicating that being a part of the carnivore area sample led to a more rapid increase in the direct experiences in relation to bear tracks over time, than in the national sample.

Existence
Respondents in the carnivore area supported the existence of bears and wolves (Table 3)  Th e ANOVA produced signifi cant simple main eff ects (p Ͻ 0.05) with regards to both measurement point in time and sample across all existence items, but showed no statistically signifi cant interaction eff ects between measurement point and sample.

Policy
In the carnivore area sample, acceptance for the current bear policy was lower in 2009 (47%) than in 2004 (53%). Th e wolf-related items showed little or no change in terms of acceptance (Table 4). However the support for increasing the wolf policy goal for wolves was lower in 2009 than in 2004, and the support for decreasing the policy goal was higher in 2009 than in 2004.
Th e ANOVA produced signifi cant simple main eff ects (p Ͻ 0.05) with regards to both measurement point in time and sample across all policy items, and a signifi cant (p Ͻ 0.05) interaction eff ect between measurement point and sample F 1,12470 ϭ 7.44 with regards to the bear policy item. Th is indicates that the bear policy acceptance have decreased faster in the carnivore area than in the national sample.
Th e national sample showed a lower " should be increased " rating for bears and for wolves, and a lower support for increasing and for accepting the wolf policy when comparing 2009 to 2004. Policy support in the carnivore sample was lower than in the national sample with regards to all items except the 2004 measurement of acceptance of the wolf policy goal. Th e carnivore area sample exhibited a statistically significant (p Ͻ 0.05) correlation between the experience scale and the other two scales within each time period, with correlations (Pr) of -0.10 within each time period, while no such correlation was found in the national sample. Eff ect sizes ranged from 0.02 to 0.08, which is to be expected given the slow rate of attitude change.
Th e carnivore area exhibited higher levels of direct experience in 2009 than in 2004, and less positive attitudes to both the existence scale and the policy scale in 2009 compared with 2004. In the non-carnivore area, the policy scale exhibited a signifi cant diff erence (p Ͻ 0.05) with less positive attitudes in 2009 compared to 2004. In the non-carnivore area, there was a statistically signifi cant diff erence in the policy scale (p Ͻ 0.05), with the 2009 score being lower than the 2004 score. Th e carnivore sample also exhibited a higher proportion of respondents reporting direct experience of bears and wolves in 2009 compared to 2004. Th e degree of acceptance towards the existence of bears and wolves was higher in 2004 than in 2009, while no such diff erences could be established in the national sample. Th ese fi ndings are all in line with our theoretical predictions, supporting a negative direction of eff ects between direct experience and attitudes towards bears and wolves, and thus making a causal link between direct experience with bears and wolves and attitude change plausible.

Discussion
Attitude change on an aggregate level is normally considered to be a slow process, driven by broad societal trends (Williams et al. 2002) or generational value shifts (Inglehart 1995). However more rapid attitude change have been observed in connection with growth of large carnivore populations (Duda et al. 1998, Treves et al. 2013. Our results show an attitude change in the general public with regards to bears and wolves between 2004 and 2009. Th is change is likely the result of people being subjected to direct experience with an attitude object forcing them to reevaluate their loosely held attitudes towards that attitude object (Fazio et al. 1982, 1983, Olson and Zanna 1993, Eagly and Chaiken 1998, Williams et al. 2002, Ericsson and Heberlein 2003, Heberlein and Ericsson 2008; as previous research has indicated that the Swedish context is characterized by low levels of direct experience (Williams et al. 2002, Heberlein andEricsson 2008), and attitudes based mostly on indirect information (Heberlein and Ericsson 2008). Our empirical fi ndings is consistent with a causal connection between direct experience of, and attitude change towards, bears and wolves and their respective policies.
Such a causal connection, primarily linked to change in population size of, in this case, brown bears and grey wolves has a number of interesting implications. It means that differences in attitudes between urban and rural areas need to be understood in terms of overlap of carnivore areas and rural areas in addition to the cultural factors observed in previous research (Skogen and Krange 2003, Heberlein and Ericsson 2005, Krange and Skogen 2007, thus making diff erences in attitudes an issue of proximity (sensu Karlsson and Sj ö str ö m 2007) and direct experience.
Increases in the populations of bears and wolves leads to a more rapid change of attitudes in the areas housing those populations compared to areas with few or bears and wolves. Th e role of the wolf as a driver of attitudes and attitude change ) off ers an important explanation of the social confl icts arising in connection to growing numbers of wolves and reintroduction of wolves in diff erent contexts (Kellert 1987, Bath and Buchanan 1989, Ericsson and Heberlein 2003, Dickman 2010. Th e wolf, is not only a symbol of, for example, urban power over rural areas (Skogen and Krange 2003) but also have a more direct eff ect on attitudes via direct experience (Fazio et al. 1982, to be increasing with the carnivore area having less positive, and more rapidly changing attitudes than the national sample. Our empirical fi ndings thus support the theoretical assumption that direct experience drives attitude change. Th e carnivore area sample saw a larger attitude change than the national sample, and displayed an increase in direct experience which was not found in the national sample. Th is makes direct experience of bears or wolves a plausible link between changes in the size of the animal populations and attitude change. Th is would make the uneven geographical distribution of the populations of bears or wolves in Sweden a potential social problem, as rural discontent is likely to increase in tandem with growing carnivore populations and increased direct experience with bear and wolf. Williams et al. 2002, Heberlein andEricsson 2008). Furthermore, the uneven distribution of the animals geographically naturally creates diff erent rates of attitude change in rural and urban areas, which generates friction between urban and rural groups in society.
From a theoretical perspective, it is also important to diff erentiate between the attitudes towards the species per se and the policy objectives linked to the species. While there are changes in attitudes to both the species and the policy objectives related to these species in areas where people have direct experience of bear or wolf, another pattern emerges in areas with limited direct experience of bears or wolves. Although attitudes towards bears and wolves are more positive in urban areas, the attitudes to the policy become less accepting or even negative over time. Our fi ndings thus show that people are able to distinguish between diff erent attitude items and evaluate policy issues and their eff ects without direct experience, indicating that attitudes to different policy alternatives are probably easier to change and infl uence through indirect experience than attitudes to the existence of bears and wolves in the fauna.
Th ere are several lessons to be learnt from our fi ndings in terms of conservation policy. Since the diff erence in acceptance between rural and urban is, in part, related to the distribution of the animals and direct experience rather than solely created by slow changing cultural diff erences, this suggests that social confl icts could be ameliorated more quickly than previously assumed. Th is, however, calls for a more diff erentiated policy that would take into consideration regional and local aspects including local and adaptive management (Sandstr ö m 2009).
More direct experience seems to lead to lower policy acceptance. Since Sweden can sustain considerably larger populations of bears and wolves from a biological perspective, and populations are currently growing and expanding their geographical range (Swedish Government Offi cial Report SOU 2012:22) the issue becomes political in nature rather than biological as the ultimate limits to growth is given by the policy goals, which is connected to public acceptance. Th e revision of Sweden ' s national large carnivore policy (Prop. 2012/13:191) essentially meant a lowering of the populations targets, and could refl ect the lower levels of acceptance of the policy from 2000 observed here.
Future research needs to repeat, solidify and conceptually improve the understanding of the underlying mechanisms behind our fi ndings with respect to experience. Our work stops at establishing zero order correlations at an aggregate level. It would be rewarding to expand the model to account of other variables in addition to direct experience, lengthen the time series in order to say something about development over time, and complement these results with a panel study design in order to follow the development of respondents over time on an individual level in order to track attitude change over time on the individual level.

Conclusions
Acceptance of bears and wolves in Sweden is generally high, but the previously observed pattern seen in 2004 (Ericsson et al. 2006) of lower acceptance of these animals in the carnivore areas was repeated in 2009. Th ese variations seems