This paper contains supplementary material that can be found online at http://journals.cambridge.org
The current rate of the extinction of species is estimated to be 1,000 times higher than in the fossil record (Pimm et al., Reference Pimm, Russell, Gittleman and Brooks1995). The main causes of this biodiversity crisis are human population growth and range expansion, and increased consumption of natural resources (Baillie et al., Reference Baillie, Griffiths, Turvey, Loh and Collen2010). Impacts on terrestrial mammals include widespread reduction in range, fragmentation and loss of habitat, loss of prey, population decline and local, regional and, in some cases, global extirpation of species (Baillie et al., Reference Baillie, Griffiths, Turvey, Loh and Collen2010). Species of the order Carnivora are especially susceptible to these threats because their high trophic position restricts them to low-density populations that are inherently vulnerable to demographic and environmental stochasticity (Woodroffe & Ginsberg, Reference Woodroffe, Ginsberg, Gosling and Sutherland2000). Greater proximity and reduced resource availability have intensified competition between people and carnivores, leading to increased human–carnivore conflict globally (Graham et al., Reference Graham, Beckerman and Thirgood2005; Inskip & Zimmermann, Reference Inskip and Zimmermann2009). Twenty-seven percent of carnivore species are considered extinct or threatened with extinction in the wild (IUCN, 2012).
In Africa many carnivore populations reside outside protected areas (Mackinnon & Mackinnon, Reference Mackinnon and Mackinnon1986), where they can pose a threat to human lives (Graham et al., Reference Graham, Beckerman and Thirgood2005). In agricultural areas carnivores may prey upon animals that have nutritional, financial or recreational value to people (Graham et al., Reference Graham, Beckerman and Thirgood2005). People frequently kill carnivores in an effort to prevent attacks on humans or livestock (Graham et al., Reference Graham, Beckerman and Thirgood2005; Inskip & Zimmermann, Reference Inskip and Zimmermann2009) and this has potentially severe implications for the conservation of threatened species. For example, a Namibian study attributed 47.6% of cheetah Acinonyx jubatus mortality to persecution by humans on farmland. Reducing indiscriminate persecution of carnivores in agricultural areas of Africa may therefore play an important role in conserving threatened species. Furthermore, indiscriminate persecution of meso-predators (e.g. black-backed jackals Canis mesomelas and caracals Caracal caracal) may trigger compensatory demographic responses such as increased natality and immigration (Prugh et al., Reference Prugh, Stoner, Epps, Bean, Ripple, Laliberte and Brashares2009), potentially increasing rates of predation (Marker et al., Reference Marker, Mills and Macdonald2003; Ogada et al., Reference Ogada, Woodroffe, Oguge and Frank2003; Thorn et al., Reference Thorn, Green, Dalerum, Bateman and Scott2012). Thus, in areas where meso-predators remain relatively abundant, lowering the rate of indiscriminate persecution may benefit both human and carnivore populations.
Reducing indiscriminate killing of carnivores requires a sound understanding of the factors motivating this behaviour, which include social and psychological factors that affect the decision-making process (St John et al., Reference St John, Edwards-Jones and Jones2010; White & Ward, Reference White and Ward2010). One such factor is attitude, defined here as a positive or negative response to carnivores. Attitudes affect community support for conservation and management initiatives (Naughton-Treves et al., Reference Naughton-Treves, Grossberg and Treves2003; Davenport et al., Reference Davenport, Nielsen and Mangun2010) and influence individual behaviour towards carnivores (Marker et al., Reference Marker, Mills and Macdonald2003; Lindsey et al., Reference Lindsey, du Toit and Mills2005b; Zimmermann et al., Reference Zimmermann, Walpole and Leader-Williams2005). Thus, conservation and conflict-mitigation initiatives often seek to change damaging human behaviour by altering underlying attitudes (St John et al., Reference St John, Edwards-Jones and Jones2010).
We focused on attitudes to carnivores at privately owned commercial game and livestock farms because those land uses cover the majority of agricultural land in our study areas (Tladi et al., Reference Tladi, Baloyi, Marfo, Walmsley, Walmsley, Mangold and Kalule-Sabiti2002; de Klerk, Reference de Klerk2003). People inhabiting these areas co-exist with medium and large-sized carnivore species, including brown hyaena Hyaena brunnea, leopard Panthera pardus, caracal, black-backed jackal, cheetah and African wild dog Lycaon pictus. Current levels of persecution threaten brown hyaenas, cheetahs and African wild dogs (Friedmann & Daly, Reference Friedmann and Daly2004; Thorn et al., Reference Thorn, Green, Dalerum, Bateman and Scott2012). Indiscriminate persecution of jackals and caracals appears to be a widespread social norm (Thorn et al., Reference Thorn, Green, Dalerum, Bateman and Scott2012). Previous studies have assessed local determinants of persecutory behaviour towards carnivores (Thorn et al., Reference Thorn, Green, Dalerum, Bateman and Scott2012) but not the perceptions and attitudes that motivate such behaviour. This is an important omission because attitudes and behavioural intentions are often uncorrelated with actual behaviour (Romanach et al., Reference Romanach, Lindsey and Woodroffe2007; St John et al., Reference St John, Edwards-Jones and Jones2010; Heberlein, Reference Heberlein2012). If behaviour reflects attitudes to carnivores in our study area, investigating attitude determinants will inform the design and targeting of relevant conflict-mitigation activities. If not, conflict-mitigation strategies seeking to reduce persecution by altering attitudes may be a waste of precious resources.
This study seeks to evaluate the effect of social, economic and land-use variables that may influence attitudes to carnivores in South Africa. Predictions regarding these determinants are that attitudes are more positive among game farmers than those with other land uses (Lindsey et al., Reference Lindsey, du Toit and Mills2005b), more positive among English-speaking farmers than among Afrikaans-speaking farmers (Lindsey et al., Reference Lindsey, du Toit and Mills2005b), and negatively related to age (Lindsey et al., Reference Lindsey, du Toit and Mills2005b; Zimmermann et al., Reference Zimmermann, Walpole and Leader-Williams2005) and stock losses (Zimmermann et al., Reference Zimmermann, Walpole and Leader-Williams2005). We would expect stock holdings and farm size to be positively related to attitude, reflecting the relative affluence of land owners and the concomitantly reduced impact of financial losses as a result of predation (Romanach et al., Reference Romanach, Lindsey and Woodroffe2007). Our analysis seeks to test these predictions, evaluate which determinants exert the greatest influence on attitudes to carnivores and confirm whether attitudes are correlated with persecutory behaviour.
The study area comprised 116,320 km2 of commercial farmland in the North West Province of South Africa and the 11,090 km2 Waterberg area of the adjacent Limpopo Province, which includes the UNESCO Waterberg Biosphere Reserve and surrounding areas of the Waterberg District Municipality (Fig. 1). The Waterberg area has the highest proportion of grazing land in Limpopo (de Klerk, Reference de Klerk2003) and 54% of the North West Province is used for grazing (Tladi et al., Reference Tladi, Baloyi, Marfo, Walmsley, Walmsley, Mangold and Kalule-Sabiti2002). Farming of cattle and small-stock (pigs Sus scrofa, sheep Ovis aries and goats Capra aegagrus hircus) is common (Department of Agriculture, Forestry and Fisheries, 2010). Many commercial farmers combine game and livestock farming (henceforth, mixed farmers).
Interview methods and questionnaire design
We surveyed commercial game and livestock farms throughout the North West Province between October 2006 and September 2008, and in the Waterberg area of the adjacent Limpopo Province during March–August 2011. Respondents were recruited initially by opportunistically approaching as many attendees as possible at local farming forums, and further contacts were supplied by each participant. Interviews were administered in person and in private (usually at the respondent's home) by MT and MG. Each lasted c. 40 minutes and was based on a semi-structured questionnaire (Supplementary Material 1). The questionnaire was piloted prior to use to ensure clarity, resulting in minor amendments. It contained four sections, based on (1) characteristics of respondents and their farms, (2) carnivore species present and predator–prey interactions, (3) use of anti-predation measures, including lethal control of carnivores, and (4) attitudes towards carnivores. All respondents were assured of anonymity and confidentiality.
Attitudes to carnivores were explored using 10 statements (Table 1), chosen following a literature review and preliminary discussions with local researchers, conservation workers and land owners. This process also informed the selection of the predictor variables used in generalized linear models. Responses to each statement were measured on a five-point Likert scale (strongly agree, agree, unsure, disagree, strongly disagree).
Cultural group was inferred from the respondent's first language. Respondents estimated game and livestock abundance as well as the number of animals killed by carnivores. The accuracy of abundance and predation estimates for extensively managed animals (usually plains game but sometimes cattle) may vary because those animals can be difficult to detect and are counted infrequently (usually annually or bi-annually). However, despite potential bias, such data were not available from any other source. Many respondents cited lack of population growth in game species as indirect evidence of predation but such losses were not recorded because they were unquantifiable and could have been caused by factors other than carnivore predation.
We coded responses to the attitude statements on a scale of one (strongly negative) to five (strongly positive), the sum of which gave a total attitude score per respondent. The reliability of the total scores was tested using Cronbach's α. Questions that were not internally consistent were removed, ensuring that the total of the remaining questions (henceforth, the composite attitude score) was truly additive and reflected overall attitude towards carnivores (Cortina, Reference Cortina1993).
We then evaluated the effect of possible determinants, using a generalized linear regression model with a Poisson error distribution and a log link function. We first created a Spearman's rank correlation coefficient matrix and excluded predictor variables that were inter-correlated by Spearman's ρ (r s ) > 0.7 (Kolowski & Holekamp, Reference Kolowski and Holekamp2006). There was a strong positive correlation between farm size and total stock holdings (r s = 0.74, P < 0.001) and therefore farm size was excluded. The full model included the composite attitude score as the response variable, with respondents' estimated total stock holdings, estimated proportion of stock predated, age, culture, and land use as predictor variables. Some respondents omitted one or more of the attitude statements and therefore maximum composite score was modelled as an offset variable.
Our set of candidate models contained all additive combinations of predictors as well as all two-way interactions. Missing observations were excluded case-wise. We ranked candidate models in order of parsimony, using Akaike's Information Criterion adjusted for small sample size (AICc; Burnham & Anderson, Reference Burnham and Anderson2002). Models within 7 ΔAICc units of the model with the lowest AICc were considered to have sufficient relative support to be included in the final set of explanatory models (Burnham & Anderson, Reference Burnham and Anderson2002; Burnham et al., Reference Burnham, Anderson and Huyvaert2011). We excluded any model that was simply a more complicated version (i.e. containing additional parameters) of a nested model with lower AICc, to reduce selection of overly complex models (Burnham & Anderson, Reference Burnham and Anderson2002). We then used Akaike model weight to determine the relative probability of each model being the most parsimonious in the candidate model set, and summed weights to evaluate the relative importance of determinants (Burnham & Anderson, Reference Burnham and Anderson2002). We predicted effect sizes from regression coefficients (β). Analysis was conducted using SPSS v. 17 (SPSS, Chicago, USA). Financial values are given in South African rand (ZAR 1 = GBP 0.08 = USD 0.13).
We interviewed 78 people in the North West Province and 92 in the Waterberg area. With a refusal rate of 4%, non-response bias was assumed to be negligible. All but five respondents were white, reflecting typical land-tenure patterns on South African commercial farms (Tladi et al., Reference Tladi, Baloyi, Marfo, Walmsley, Walmsley, Mangold and Kalule-Sabiti2002). As in Marker et al. (Reference Marker, Mills and Macdonald2003) we considered commercial farmers to be a homogeneous group and therefore assumed our respondents to be approximately representative of the wider population of commercial farmers in the study area. Although not inherently unrealistic, this assumption could not be verified because there were no reference statistics available for land ownership by ethnic group in the study area.
Most respondents (75%) were Afrikaans-speaking, 19% were English-speaking and 6% were from other cultural groups (3% German-speaking and 3% Bantu language-speaking). Median age was 44 years (Q1 = 36, Q3 = 54, range 23–82).
The total area of the farms surveyed was 5,454 km2 (median = 13.1 km2, range 0.06–960 km2), which represents 7.5% of farmland within the study area. The sample contained an equal proportion of game and livestock farmers (37% each) and the remaining 26% were mixed farmers. Total stock holdings were estimated at 183,049 animals (median = 478, Q1 = 200, Q3 = 900). The estimated total number of animals predated in the year prior to the interview was 2,268 (median = 5, Q1 = 0, Q3 = 30). The estimated median proportion of stock holdings predated by carnivores was 1% (Q1 = 0, Q3 = 4%).
Agreement rates (the proportion of respondents who agreed or strongly agreed) for the 10 attitude statements are summarized in Table 1. Attitude statements four and seven were excluded from the composite score, giving Cronbach's α = 0.8, indicating good internal consistency. In keeping with our predictions, game farmers had more positive attitudes (higher median composite attitude scores) than mixed or livestock farmers (Fig. 2). English-speaking farmers had more positive attitudes than Afrikaans-speaking farmers and other cultural groups (Fig. 2). Age and stock losses were negatively related to composite attitude scores, whereas total stock holdings were positively related to attitudes (Fig. 2). The same is true of farm size, which was strongly positively correlated with stock holdings. Respondents who killed carnivores (58%) had significantly lower median composite attitude scores than those who did not (U = 2,730.5, Z = −3.658, P < 0.001, n = 170), indicating a clear link between attitude and persecutory behaviour.
The global additive model showed no evidence of a lack of fit (χ 156 2=132.12, $\hat c = 0.85$ , n = 164). Three of 42 candidate models were within 7 Akaike units of the top-ranked model (Table 2). Land use (summed Akaike weight 99.9) and culture (summed Akaike weight 95.8) were by far the most influential variables. Age (summed Akaike weight 4.3), proportion of stock predated (summed Akaike weight 4.2) and total stock holdings (summed Akaike weight 0.1) had comparatively little effect. The top-ranked model was an interaction between land use and culture (Table 3), indicating that the effect of those predictors differed among their sub-categories. Regression coefficients were taken from the top-ranked model, which received considerably greater support than any other model (0.92 Akaike weight). The coefficients indicate that, after controlling for the effect of other variables, predicted composite attitude scores among English-speaking game farmers were 14% higher than expected scores among the reference category (cultural group = other, land use = mixed). Predicted scores for Afrikaans-speaking mixed farmers and livestock farmers in the cultural group ‘other’ were 8% and 41% lower, respectively, than the reference category.
* Reference category: cultural group = other × land use = mixed purpose
Quantifying the attitudes of landowners towards carnivores and understanding the factors that determine them is instrumental to conservation planning and increasing tolerance of carnivores (Naughton-Treves et al., Reference Naughton-Treves, Grossberg and Treves2003; Zimmermann et al., Reference Zimmermann, Walpole and Leader-Williams2005; Romanach et al., Reference Romanach, Lindsey and Woodroffe2007; Anthony et al., Reference Anthony, Scott and Antypas2010). Although negative attitudes and the intention to kill carnivores are not always correlated with actual carnivore removals (Romanach et al., Reference Romanach, Lindsey and Woodroffe2007), strong positive correlation in our study area indicates that influencing attitudes to carnivores can reasonably be expected to produce concomitant alterations in persecutory behaviour.
Attitudes are functions of affective (emotional) and cognitive components (Vlek & Steg, Reference Vlek and Steg2007; Heberlein, Reference Heberlein2012). Changing attitudes therefore involves communicating relevant and credible messages whose emotional and rational content appeals to target recipients (Vlek & Steg, Reference Vlek and Steg2007), and responses to the individual attitude statements provide a good starting point for the development of such material. Many respondents felt that carnivores were over-abundant (35%) and inflicted financial losses (62%) and that lethal control of suspected predators was the most cost-effective method of limiting predation losses (31%). These beliefs constitute compelling motives for killing carnivores but there is a lack of empirical data with which to address them. Robust, contemporary estimates of carnivore population size and distribution are lacking for most South African carnivore species (Friedmann & Daly, Reference Friedmann and Daly2004), as is information on carnivore ecology outside protected areas. Likewise, there have been few studies of human–carnivore conflict in South Africa (Thorn et al., Reference Thorn, Green, Dalerum, Bateman and Scott2012).
Recommended themes for future research therefore include the effect of predators on farming profitability, carnivore diet and predation rates on farmland, carnivore population status, the efficacy and cost of non-lethal anti-predation measures, and the effect of persecution on carnivore populations. Such studies would provide information about the effects of carnivores on people and the implications of different management and conservation options. Species-specific information would be useful, given that attitudes and behaviour towards carnivores often vary according to species (Romanach et al., Reference Romanach, Lindsey and Woodroffe2007). Local and regional measurements would help to place the data in an appropriately broad context. Communicating research findings to farming communities that can benefit from them should also be an integral element of future studies.
In the meantime it may be helpful to highlight the utilitarian benefits of carnivores (Davenport et al., Reference Davenport, Nielsen and Mangun2010) in terms of the financial value of sustainable hunting (Marker et al., Reference Marker, Mills and Macdonald2003; but see Treves, Reference Treves2009) and carnivore-based eco-tourism (Marker et al., Reference Marker, Mills and Macdonald2003; Lindsey et al., Reference Lindsey, Alexander, du Toit and Mills2005a), pest and intra-guild population control (Prugh et al., Reference Prugh, Stoner, Epps, Bean, Ripple, Laliberte and Brashares2009) and improving the health of prey populations by removing weak or diseased animals (Davenport et al., Reference Davenport, Nielsen and Mangun2010). It may also be possible to influence the customers of hunting, tourism and animal production businesses and increase demand for ecologically sustainable products as consumers are often willing to pay more for products that are perceived as ecologically responsible (Brécard et al., Reference Brécard, Hlaimi, Lucas, Perraudeau and Salladarré2009). Our modelling results (Tables 2 & 3) elucidate characteristics that can be used to determine target recipients for conflict-mitigation activities. The results indicate that in northern South Africa the attitudes of commercial farmers towards carnivores are influenced by cultural and land-use attributes more than by economic factors such as stock holdings or rates of loss. Thus, activities aimed at improving attitudes should focus on areas where livestock, and particularly mixed-purpose, farming is prevalent. Demographically, areas with high proportions of Afrikaans-speaking farmers are also a priority. Interventions aimed at fortifying positive attitudes should focus on game-farming areas, especially those with large English-speaking populations. Cultural influences on attitudes can be difficult to address because culture incorporates social and psychological factors that are central to self-perception (Vlek & Steg, Reference Vlek and Steg2007). Nevertheless, it is possible to influence deeply ingrained attitudes if the source of communication is perceived as fair, legitimate and trustworthy (Anthony et al., Reference Anthony, Scott and Antypas2010) and if the information appropriately incorporates the norms, values and beliefs of the target cultural group (Vlek & Steg, Reference Vlek and Steg2007).
Game farming is increasingly common in South Africa (Bothma, Reference Bothma2005) and therefore it is promising that game farmers hold more positive attitudes to carnivores. It is somewhat surprising, however, because few of the anti-predation measures available for livestock farmers (e.g. herding, guarding animals, kraaling, deterrent devices) are feasible for extensively managed game (Shivik, Reference Shivik2006). It may be that game farmers are less aware of (cognitive component) and so less distressed by (affective component) predation losses because their stock is dispersed over a larger area and is counted less frequently than intensively managed stock. The socio-economic status of game-farm owners may also differ from other farming groups. For example, game farming is considered to require lower input of capital and offer greater profitability than cattle or small-stock farming (Absa, 2003), and therefore game farmers may be comparatively affluent and less adversely affected by predation losses. Further investigation of such factors may produce useful insights into the attitude gap between farmers with different land uses.
Nevertheless, reducing the frequency of persecutory behaviour may require more than just a change in attitudes to carnivores. Experience is also an important driver of behaviour (Heberlein, Reference Heberlein2012) and therefore communicating persuasive educational information may not necessarily change behaviour if farmers have negative experiences of carnivores (Heberlein, Reference Heberlein2012). It may be necessary to engage in outreach and extension activities that accurately quantify and/or reduce predation to acceptable levels, encourage good husbandry practices and provide carnivore-based economic benefits (e.g. carnivore-based tourism). Marker et al. (Reference Marker, Mills and Macdonald2003) reported positive results from extension activities among Namibian farming communities that are culturally similar to those in our study area. Greater awareness of conservation issues and the ability to accurately diagnose the cause of stock losses meant that persecution of cheetahs was reduced and was more closely linked with actual rather than perceived predation problems.
Human–wildlife conflicts are a product of complex interactions between ecological, socio-economic and political systems (Graham et al., Reference Graham, Beckerman and Thirgood2005; Anthony et al., Reference Anthony, Scott and Antypas2010; Davenport et al., Reference Davenport, Nielsen and Mangun2010). Unravelling this complexity and reducing conflict requires adaptive, interdisciplinary approaches that facilitate informed dialogue between key stakeholders (Davenport et al., Reference Davenport, Nielsen and Mangun2010; St John et al., Reference St John, Edwards-Jones and Jones2010), including the rural communities who typically experience adverse effects of wildlife. Engaging these people in management planning and conflict resolution is particularly important in South Africa because the institutions responsible for managing human–wildlife conflict and enforcing biodiversity legislation are hampered by a lack of financial and human resource capacity (Anthony et al., Reference Anthony, Scott and Antypas2010). Comprehensive changes in persecutory behaviour cannot be compelled and sustainable conflict-mitigation strategies often rely on persuasion rather than enforcement. Our findings therefore provide insight into previously under-studied human dimensions of conflict between people and carnivores on South African farmland and a starting point for designing acceptable and practical conflict-mitigation activities. Such activities may also influence wider regional metapopulation function as a result of spatial and genetic linkages between carnivore populations in the study area and populations in adjacent countries (Lindsey et al., Reference Lindsey, du Toit and Mills2004, Reference Lindsey, Marnewick, Davies-Mostert, Rehse, Mills and Brummer2009; Thorn et al., Reference Thorn, Green, Dalerum, Bateman and Scott2012).
We are grateful to the interview respondents and to Lapalala Wilderness and Welgevonden Private Game Reserve for logistical assistance. We thank S. Uzzell and the Earthwatch Institute for their support, and two anonymous reviewers for helpful comments. We also thank Land Rover Centurion for sponsoring the project, and the Endangered Wildlife Trust, Knowsley Safari Park, the Rufford Small Grants Foundation, the University of Brighton, and the Leverhulme Trust for providing funding.
Michelle Thorn and Matt Green's interests include mammal distribution, practical field protocols for conservation monitoring and the management of human–animal conflicts. Kelly Marnewick is the manager of the Endangered Wildlife Trust's Carnivore Conservation Programme. She is particularly interested in cheetahs outside protected areas and in human–wildlife conflict. Dawn Scott's research focuses on mammalian conservation biology and landscape ecology and she is the principal investigator of the Brown Hyaena Research Project in South Africa.