Study area

This study focuses on the largest and longest-standing LGD programme in South Africa. Anatolian shepherds are the predominant breed of LGD used, and placements of the dogs occurs primarily along the northern border with Botswana and Zimbabwe. This area is characterised by semiarid savannah agricultural habitat Land use is predominantly agricultural production of sheep and goats, with some mixed livestock, game ranching and nature reserves. Both large commercial farms and small subsistence farms comprise the agricultural sector. Free-ranging carnivores in the region include leopard (Panthera pardus), brown hyena (Hyaena brunnea), spotted hyena (Crocuta crocuta), cheetah, caracal (Caracal caracal), African wild dog (Lycaon pictus) and black-backed jackal (Canis mesomelas) (Findlay 2016, Spencer et al. 2020). Farmers experience livestock losses to carnivores, ranging from minor to severe, and employ a range of protection systems, including non-lethal and lethal control methods (Rust et al. 2013, Lucas et al. 2020).

All dogs are placed and monitored through on-farm visits and telephone calls with a project manager (PM) employed by a non-government organisation (NGO), Cheetah Outreach Trust, herein referred to as the NGO. The logistics of the LGD programme operations have been described elsewhere (Rust et al. 2013, Whitehouse-Tedd et al. 2020), but, in brief, includes coverage of the first year's costs, regular and on-going monitoring and support. Farmer participants are asked to cease the use of lethal control methods whilst an LGD is operating and are responsible for the dog's health and welfare (with the exception of health and feeding costs in the first year), as well as regular completion of monitoring surveys and feedback interviews.

Primary data collection

Primary data were collected via a semi-structured interview involving both open and closed questions about the farmers' experiences prior to, and during, participation in the LGD programme. Semi-structured interviews provided the participants with some control over the interview process, allowing them to express what they believed to be important (Bryman 2012).

The interview question guide (Table S1) was arranged around three main constructs: knowledge of LGDs and the LGD programme, recruitment onto the LGD programme, and satisfaction with the LGD and LGD programme. The open questions provided responses for the qualitative element of the study. Closed questions were focused on sources of information about the LGD programme (selection of options to choose from), LGD training, future use of LGDs, and a 5-point rating scale of satisfaction with changes in livestock losses since LGD introduction (where 1 was “not satisfied” and 5 was “highly satisfied”).

The interview schedule was piloted with five programme participants to assess understanding of the questions, determine the average time required to complete an interview, and to ensure the questions generated data suitable for analyses. As per Arnott et al. (2014), advice was also sought from the PMs with regards to wording of the questions. Only minor changes were made to one question to improve understanding, therefore data collected from the pilot interviews (excluding those from the question that was modified) were included in the final analyses to maximise stakeholder group representation.

The NGO database of programme participants was used to identify potential interviewees and consisted of 136 farms; of these, 106 farms (each represented by one person) agreed to participate. Reasons for non-participation included invalid contact details (n = 7), no reply (n = 6), death (n = 4), lack of time (n = 2), uninterested (n = 1), or inclusion in a concurrent study and our desire to avoid response fatigue in these farmers (n = 10). For ethical reasons, no further details were requested from non-responders or those declining the interview. In the case of two farms, two interviews were conducted with different individuals representing the farm; both people were involved in the decision to join the LGD programme and assumed equal responsibility for the LGD during participation. As we wished to be as inclusive as possible and maximise stakeholder group representation, interview data from both were included in analyses (providing a total n = 108 interviews), with the exception of recruitment analyses, which were restricted to those provided by the person named in the NGO database (n = 106). The interviews were conducted via telephone (n = 79) for participants in remote areas, or face-to-face (n = 29).

The method of recruitment into the programme (described in detail under ‘data processing and analysis’) for each farm was determined from interviews with either the participant, or the PM responsible for the LGD’s placement. In the latter case, PMs were asked to recall data surrounding recruitment for every placement between 2005 (when the programme started) until the end of 2018, regardless of whether the farmer participated in an interview or not. This way, an additional 63 farmers were identified as having received an LGD, but had not been interviewed (e.g. through unavailability or received their LGDs after the interview data were collected). Two types of recruitment data were subsequently analysed; farmer-derived data from interview responses about discovery of the NGO programme (n = 106), and PM-derived data generated entirely by PM-recall (n = 169). Duplicate records (i.e. those in both datasets) were cross-checked through the process of triangulation, and instances of discrepancies (20% of cases) resolved according to PM-recall as this was typically supported by written records or other similar evidence and therefore considered the most reliable. Subsequently, the farmer-derived data for recruitment were analysed in a cross-sectional manner and facilitated assessment of recruitment type on a finer scale than PM-derived data, which were analysed on a longitudinal basis using only broad categories of recruitment type (see below).

Although every attempt was made to interview those individuals primarily responsible for joining the programme according to the NGO, in some instances (n = 11) a spouse was interviewed instead, ten of which were female. In four cases, the task was actively delegated to a spouse, and in other cases the spouse was the only person available for an interview. In all cases, the spouse had detailed knowledge of the LGD and its placement. Interview length did not vary significantly between telephone and face-to-face interview techniques (median 14 minutes and 15 minutes, respectively: U(1) = 979.5, z = –1.153, n = 108, P = 0.249) but interview length varied considerably (3-42 minutes). All responses were manually recorded.

Nine participants required the use of a translator (a native Afrikaans-speaking educator) to complete the interview in Afrikaans. Responses were translated into English whilst the participant was answering, but these interviews were both initiated and concluded by the first author, in English. When necessary, the translator was asked to rephrase questions, except for the rating scale measuring satisfaction, which was scripted to ensure that meaning and rating criteria were consistent between English and Afrikaans.

Secondary analysis of existing data

Qualitative and quantitative data on the livestock farmed, the status of all LGD placements, and participant status within the programme for the interviewed participants were extracted from monitoring reports generated by the PMs. Participant status referred to their current form of engagement with the programme. Status was based on whether the oldest surviving LGD on their farm was working, or not, at the time of the study, as determined from qualitative information in progress reports. Interview participants were categorised as ‘active’ when their LGD was working, whilst other LGD conditions (retired, dead, removed, sold) were used to categorise a participant as ‘inactive’ within the programme. Extraction and analysis of existing data occurred concurrently with the interviews but was not used to inform primary data collection or sample selection; the intention was to interview as many farmers as possible and to supplement interview data with existing data where necessary. Reasons for LGD retirement, death, removal, or sale have been explored previously (Whitehouse-Tedd 2020).

Data processing and analysis: qualitative data

The data obtained via the semi-structured interviews were analysed using principles of thematic analysis (Braun & Clarke 2012). Responses were grouped by common themes and NVivo (QSR International Pty Ltd. Version 12, 2018) was used to code statements relevant to satisfaction, motivation and recruitment. The themes identified were also verified by co-authors (KWT and JS). Responses were initially categorised at the semantic level, identifying key phrases and words that link to a specific theme. This was used to generate the over-arching thematic structure, followed by analysis at the latent level, picking up on the underlying conceptualisations, opinions and ideologies in relation to the questions/statements.

Data processing and analysis: quantitative data

The participants' responses to closed-ended questions (including demographic and scale items) were coded numerically, and quantitatively described using descriptive statistics. Length of time in the programme was calculated as the difference between placement date of the first LGD and end date (LGD removal, death, or interview date).

Recruitment type was categorised from farmer-derived data as either direct or indirect (Oldenburg & Glanz 2008) followed by further classification (Fig. S1). A participant was classed as a direct recruitment if they initially learned about the LGD programme from a representative of the programme (the NGO or project partner) and subsequently decided to join. This category includes participants that may already have known about the programme but did not join until contact was made by the NGO. Indirect recruitment refers to cases where the participant initially found out about the LGD programme via means unrelated to direct contact with an NGO representative and the participant subsequently contacted an NGO representative to request an LGD.

Recruitment data from PM sources which were not concurrently available in the farmer-derived dataset (n = 63) could only be classified as either ‘direct’ or ‘indirect’. As such, analysis of recruitment type over time using PM-derived data (n = 169) was restricted to these two categories alone.

The majority of participants provided more than one motivating factor, and some participants provided more than one point of concern regarding joining the LGD programme. Therefore, each statement relating to motivation (n = 162) or concern (n = 36), rather than each participant, was treated as the unit of analysis. Statements that referred to the present situation only (hence deemed unrelated to motivation prior to receiving an LGD) were excluded. Categories of motivations were labelled using terms as similar to the original statement as possible, with minimal inferences made by researchers. Occasionally statements relating to motivation were assigned to more than one category (n = 12) when they included distinct components. The satisfaction scale was divided into low (1-2), medium (2.5-3.5) and high (4-5) for analysis, accounting for participants (n = 4) that provided fractional ratings.

In some cases, more than one theme or category of response was provided per question. Likewise, for some interviews, some questions were unanswered because of a lack of participants' time (and unavailability for follow up), inability to formulate a response or understand the question, or preferred not to answer. As such, the number of units of analysis varied across sets of results.

Demographics

All 106 farms represented by interviewed participants were using private land for the purpose of commercial livestock production and had LGDs placed with sheep (n = 45), goats (n = 41), cattle (n = 11), mixed small stock (n = 8) or game (n = 1). Apart from two black, Tswana-speaking farmers, all participants were white, Afrikaans-speaking farm owners or managers. Whilst the majority of participants (n = 108) were male (n = 88), 20 were female. For participants responding to questions about their status in the programme (n = 103), the time they had been involved in the programme ranged from 1-100 months (mean 35 ± 24 months). Sixty participants were active in the programme at the time of the interview. The remainder's LGD had died (n = 19) or been retired, removed, or sold (n = 24).

Knowledge of the NGO programme and the use of dogs to protect livestock

Of those participants asked about the use of dogs for livestock protection (n = 100), 72 were already aware of this function of dogs prior to joining the NGO programme, while 32 had used LGDs previously, 28 of which used a breed other than Anatolian shepherds. When asked about livestock protection methods (n = 101), 26 volunteered that they had used lethal methods to control carnivores, and 17 stated the methods used (poison, shooting, pitfall traps and leg-hold (gin) traps).

Of 99 participants responding to questions about their awareness of the programme prior to joining, 85 had heard someone else (other than the NGO) speak about the programme before they joined. Word-of-mouth was the most frequently cited source of information (n = 69) about LGDs and the NGO programme among the 108 participants. Of these, 65 provided information about whom they had heard information from, but only 64 provided sufficient detail to permit classification of the information as positive or negative. Most referred to another person by name (n = 22; many of whom named the same person/people). Others had either spoken to neighbours (n = 17), farmers of unspecified relationship to the participant (n = 11), friends (n = 7), family (n = 5), or colleagues (n = 3); none of these data overlap. In regards the detailed nature of the information shared, it was more often wholly positive about the programme (n = 36), whereas the number of mixed responses (n = 13) was approximately equivalent to the number of wholly negative (n = 15) ones. Participants sometimes did not provide specifics about positive feedback they had heard, but used terms such as, ‘good’, ‘positive’ and ‘happy’. Negative feedback heard by participants most often included anecdotal reports of behavioural problems with the dogs, e.g.: “some people said they were prone to hunting and roaming” (T031); and “I have heard negative things: that the dogs are lazy when it’s hot” (T104). Eleven participants indicated that they were not affected by negative word-of-mouth, e.g.: “if they complain, I listen to how they’ve treated the dog” (T105) and, “there are people talking but they say a lot of nonsense. They are the people who don’t have the dogs” (T082).

Fig. 1.

Recruitment type, as recalled by Cheetah Outreach Trust project managers, for all Livestock-Guarding Dog programme participants (n = 169), from inception in 2005 until end of 2018. The numbers in each column represent newly recruited participants (either directly or indirectly) per year.

When asked for sources of information perceived to be best in regards the LGD programme (n = 96), 68 named the NGO and specific persons representing the organisation. Participants were not asked about PMs but a sense of trust was evident between them, as reflected in participants' statements, such as, “I trusted (PM), he has good knowledge and background” (NW018).

Recruitment type

Three quarters (n = 80) of participating farms involved in interviews (n = 106), first joined the programme through indirect means, whilst 26 had been recruited directly (Fig. S2). Of the interviewed participants, one could not remember how they first joined the programme such that NGO records and PM recall were used to classify their recruitment for inclusion in this analysis.

Using PM-recalled data (n = 169), direct recruitment decreased relative to indirect recruitment, from programme initiation in 2005 until 2018 (Fig. 1).

Motivation for using an LGD

The importance of having an LGD was predominantly related to prevention of livestock-carnivore interactions (Table 1). A further ten participants also made reference to livestock theft by humans. These types of motivation accounted for 102 of the 162 recorded motivation statements.

Twenty-one participants stated they disliked using lethal methods and expressed a desire to ‘farm with nature’.

Table 1.

Reasons cited as motivation to join a Livestock Guarding Dog programme by South African farmers (n = 104) and relating to respondent satisfaction of reduction in losses (n = 100).

continued

For some farmers, adopting an LGD was a last resort. These participants demonstrated concern or despondency through their responses.

Of the 98 participants, 28 had prior concerns with LGDs while 68 did not. Despite all participants subsequently deciding to obtain an LGD, 18 of 98 participants providing a response stated they initially had fears about LGD exhibiting poor behaviours. Thirty four reasons for these concerns were provided; of these, the most frequently cited (n = 13) concerns were behaviour harming (biting) the livestock, roaming away from the herd, or being difficult to train behaviour. Other responses (n = 9) showed concern for wildlife (predominantly game) being affected by LGDs. In contrast, one participant indicated concern for the carnivores, e.g. “if the predators move away from my farm, I'm worried they will be killed by a neighbour” (T060). Additional responses included scepticism about whether the LGD programme would be successful (n = 6), and perceived negative attributes of Anatolians (n = 3) as being “too big” (T094) or “very aggressive” (L014). Two responses detailed concern for how an LGD would interact with existing farm dogs, and one response revealed a fear that the LGD could “get killed by the neighbour” (T106).

Satisfaction with LGDs and the programme

Of the 105 participants asked about their satisfaction with the LGD and programme, 90 rated their satisfaction with changes in livestock losses as ‘high’, whilst two each rated it as ‘medium’ or ‘low’ (Table 1). The mean rating was 4.7 out of 5. A few (n = 11) felt unable to give a satisfaction rating, but provided qualitative feedback; some held mixed views (n = 2), e.g. “I had lots of problems: he killed a couple of sheep and impala. Sometimes it worked well – whilst there he was effective” (L081). Some participants with positive comments (n = 4), e.g. “she was definitely doing a good job” (T077), could not assign a numerical rating. Likewise, some with negative feedback (n = 3), e.g. “I was a bit disappointed with the dog, I still lost lambs to caracal” (T047), did not provide a numerical rating.

Participants who provided a numerical rating also often supplemented this with qualitative detail indicating that satisfaction was related to there being a cessation in livestock losses with an LGD and also the ability of the LGD to deter predators. Seventeen commented on negative behaviours of their LGDs. These negative comments appeared unrelated to the rating score provided for changes in livestock losses.

As an indicator of satisfaction, when asked if they had ever recommended an LGD to anyone else, 93 (of 99) said ‘yes’, and four participants expressed a willingness to recommend LGDs, although they had not done so yet. The remaining two participants would not recommend an LGD.

Ninety five (of 107) participants said ‘yes’ they would use an LGD in the future, although 15 of these indicated that it would depend on certain conditions, either related to their farm (n = 12) or factors relating to the dogs (n = 3) (Table 1). Despite saying that they would use an LGD again, six raised concerns; two were about having more than one LGD at the same time, two were concerned about hunting, one had experienced roaming behaviour, and one had experienced livestock damage.

Of the nine participants who stated they would not use an LGD in future, two alluded to the intense training required but only one indicated this was due to LGD behavioural problems. Four reported circumstances around farming and one participant explained that they were “too sad to get another” (T061), since both LGDs were lost to predators early in the training period. One participant did not elaborate as to why they would not use an LGD in future, and a minority (n = 3) remained undecided.

Forty-nine (of 108) participants referred to negative LGD behaviours (without specific questioning) at least once during their interview; behaviours included harming livestock, chasing wildlife, and not staying with the herd. Of these participants, 32 had experienced such behaviours with their own LGD, while the remaining 17 had either heard of others experiencing these concerns or were worried that they may experience these issues in future. Two participants had nothing positive to say about the dogs. One had suffered depredation despite experiencing two LGD placements. The other stated that, “it caused us lots of problems. It killed lambs” (T087) and demonstrated animosity, through their answers, towards Anatolian dogs and their neighbouring farmer, who used LGDs.

Inter- and intra-stakeholder knowledge transfer and trust

Project managers played a major role in adoption of LGDs because they were trusted by farmers as sources of information. Inter-stakeholder trust is beneficial for efforts toward human-wildlife coexistence (Anthony et al. 2010) and considered essential for conservation initiative sustainability (Madden & McQuinn 2014) and participation (Bogezi et al. 2021).

Three quarters of farmers were recruited via indirect means, generally involving word of mouth; importantly, specific individuals were referred to on multiple occasions. Individuals with a high social status and the drive to implement conservation can act as “local champions” (Young et al. 2018, Niemiec et al. 2019); these “model landowners” are generally trusted amongst their peers because of their shared experience and values (Niemiec et al. 2019). In contrast, conflicts with neighbouring farmers were also reported in our study, emphasizing the significance of both negative and positive experiences or communications (Bhatia et al. 2020, Volski et al. 2021).

LGD programme recruitment and satisfaction

Given the importance of word of mouth in recruitment, high programme satisfaction likely explains the increasing proportion of indirect recruitment observed over time for this programme. The increasing role of indirect recruitment reduces NGO effort (temporal and financial) in marketing and advocacy (not measured), allowing for greater focus on delivery and refinement. Overall, this is likely to yield a more sustainable and optimised programme.

Satisfaction is likely related to the ability of a technique to meet user expectations (Pannell et al. 2006, Greiner et al. 2009, Bogezi et al. 2021). Here, farmer expectations appear to have been largely met and high satisfaction was reported. Similar findings have previously been reported for other LGD programmes (e.g. Marker et al. 2020, van der Weyde et al. 2020). However, our approach elucidated a complexity in determining overall satisfaction, whereby both costs and benefits may be equally weighted, or difficult to untangle for some participants. Additionally, despite being unable to assign a numeric value to their level of satisfaction, some participants clearly wished to express their personal views and have their voice heard. The extraction of qualitative data overcame the barrier posed by score-based measures.

Practical challenges to participation and retention

Dog behavioural problems were a recurring theme throughout our study. Although negative hearsay affected only a small proportion of farmers, this may bely its social significance in regards LGD use given the importance of word of mouth in recruitment. The need for appropriate training, along with suitable care of the dogs, has been highlighted previously in relation to LGD effectiveness and behavioural management (Potgieter et al. 2013, Whitehouse-Tedd et al. 2020). Whilst considered a cost or concern by some farmers, this also reflects an acknowledgement of responsibility and the important role that farmers play in optimizing programme success. Behaviour emphasising the importance of farmer involvement in dog training would ensure stakeholders were aware of their respective responsibilities prior to joining the programme, and likely reduce the occurrence of problem behaviours in the dogs during the placement.

Tangible motivators

Despite concerns for wildlife occasionally being apparent, most farmers were motivated to obtain an LGD for livestock protection reasons. Nonetheless, participation represents a success for both conservation and farmer stakeholders, regardless of motivation. Conservation goals achieved as indirect or secondary outcomes to primarily economic or other goals are still achievements and such differences in stakeholder objectives are not necessarily mutually exclusive.

Limitations

Reliance on participant recall may have limited the accuracy of our findings, but the high degree of agreement between farmer recall and PM recall indicates this to be negligible. Additionally, despite changes made following the pilot study, some questions were not answered by all participants, in part because of poor understanding, thereby contributing to missing data.

Since all interviewed farmers were participants of the programme, it is possible that a noteworthy number of farmers were sufficiently deterred by negative hearsay that they consequently did not participate in the LGD programme, and were thereby unavoidably excluded from our study. This limitation is common across intervention studies (in this region and internationally), whereby sampling is inherently biased towards recruited farmers. However, a number of participants provided frank and detailed comments in our survey, reflecting negative perceptions of LGDs and animosity towards the programme overall, suggesting an absence of social desirability bias, at least in these cases. Additionally, participants for whom an LGD placement was not successful were included in our sample, confirming our sample representativeness.