Data

The data comprised comprehensive questionnaires completed by farmers, who rear one of the two local cattle breeds, namely either the GA or the RDP breed. Both breeds are local in Schleswig-Holstein, which is the most Northern state of Germany. The initial list of possible respondents consisted of 200 questionnaires (100 per breed) and was selected by the breeding organisation RSH (Rinderzucht Schleswig-Holstein e.G., Germany) due to farmers’ individual breed activity and engagement. The design and sending of the questionnaires as well as the collection of completed surveys were also done by the breeding organisation. In total, 158 farmers (79%) filled out the questionnaires, thereof 78 participants for GA (78%) and 80 participants for RDP (80%). Each questionnaire consisted of a total of 12 queries, open-ended and multiple-choice questions including information on sub-items, such as farm, herd, expectations, reproduction, traits and difficulties as well as handwritten farmers’ personal opinions on the breeds’ SWOT (Supplementary Material S1).

Quantified strategy decision tool

To identify strategies using an organised approach, the FAO (2010) suggested employing the SWOT analysis from Weihrich (Reference Weihrich1982). A SWOT analysis is used to identify SWOT in order to enhance the profitability of an individual production system. This qualitative method has the disadvantage of high subjectivity during its application in decision-making (Hill and Westbrook, Reference Hill and Westbrook1997; Pesonen et al., Reference Pesonen, Kurttila, Kangas, Kajanus and Heinonen2000; Martín-Collado et al., Reference Martín-Collado, Díaz, Mäki-Tanila, Colinet, Duclos, Hiemstra and Gandini2013). Therefore, Kurttila et al. Reference Kurttila, Pesonen, Kangas and Kajanus(2000) and Saaty and Vargas (Reference Saaty and Vargas2001) developed a combination of the Analytic Hierarchy Process (AHP) method, which is a quantitative tool from Saaty (Reference Saaty1980), and the SWOT analysis to obtain more complex and reliable decisions. This hybrid method improves the quantitative information basis and forces the decision-maker to think over and analyse the situation more precisely and in more depth (Kurttila et al., Reference Kurttila, Pesonen, Kangas and Kajanus2000). SWOT-AHP methodology has been applied in the fields of forestry (Kangas et al., Reference Kangas, Pesonen, Kurttila and Kajanus2001), silvopasture adoption (Shreshta et al., Reference Shreshta, Alavalapati and Kalmbacher2004) and livestock (Wasike et al., Reference Wasike, Magothe, Kahi and Peters2011).

In the current study, the SWOT-AHP was used to quantify identified SWOT. Statistical analyses and computations for eigenvalues, priority vectors and quality control parameters within the SWOT-AHP methodology were performed by using the R-software (R Core Team, 2018). The SWOT-AHP approach was divided into five steps (Figure 1) and is described below:

1. (I) Assignment of items

   The farmers’ responses were allocated as items by an expert team into four single SWOT groups of SWOT. Expert team members were previously identified through the application of certain qualification criteria. These criteria comprised: (1) long-time experiences with both local cattle breeds in practice, (2) deep knowledge of respective production systems and the local environment and (3) profound background in population genetics and management of small populations. In total, four scientists from two German animal breeding departments passed the qualification criteria, and thus, formed the team of experts. Experts discussed their attitudes and valuations as soon as they had reached consensus. The number of items and SWOT factors allocated for each SWOT group by the team of experts exhibited substantial flexibility and was allowed to differ between both breeds.

2. (II) Defining SWOT factors

   The farmers’ items were defined as SWOT factors by the expert team (Tables 1 and 2). Generally, the number of SWOT factors within each SWOT group can vary depending on the number of allocated items.

3. (III) Computation of priority scores for SWOT factors

   The computations of the factor priority vector (FPV), the group priority vector (GPV) and overall priority vector (OPV) were performed with AHP rankings by the named expert team above based on the frequency of SWOT factors among farmers. The fundamental scale from Saaty (Reference Saaty1980) was used for such AHP rankings (Table 3). This scale represented a transformation from verbal judgements into numerical judgements to determine the relative importance of each element. There were five verbal judgements, which were ranked by their importance from equal to extreme (equal, moderate, strong, very strong and extreme). These judgements were translated into the numeric values of: 1, 3, 5, 7 and 9 with four intermediate values (2, 4, 6, 8) for compromises in importance. In other words, the scale indicated how important or dominant one element was over another element (Saaty, Reference Saaty2008). The elements for FPV were individual SWOT factors within each SWOT group and for GPV the four single SWOT groups. The elements were ranked within AHP matrices of different levels: Level 1 for the SWOT factors within each SWOT group and Level 2 for the SWOT groups (Figure 1). Kahraman et al. Reference Kahraman, Demirel and Demirel(2007), Borajee and Yakchalie (Reference Boraje and Yakchali2011) and Görener et al. Reference Görener, Toker and Uluçay(2012) defined the set of elements as E={E_j | j=1, 2, …, n}. The results of the n-ranked elements were resumed in an evaluation matrix A (n * n). Each element a_ij (i, j= 1, 2, …, n) was the quotient of the weights of the elements w_ij (i, j= 1, 2, …, n). In this matrix, the element a_ij=1, when the weights of elements w_i=w_j. The ranked elements were also expressed via a square and reciprocal matrix:

   $$\[A=({{a}_{ij}})=\left[ \begin{matrix} 1 \and {{{w}_{1}}}/{{{w}_{2}}}\; \and \cdots \and {{{w}_{1}}}/{{{w}_{n}}}\; \\ {{{w}_{2}}}/{{{w}_{1}}}\; \and 1 \and \cdots \and {{{w}_{2}}}/{{{w}_{n}}}\; \\ \vdots \and \vdots \and \ddots \and \vdots \\ {{{w}_{n}}}/{{{w}_{1}}}\; \and {{{w}_{n}}}/{{{w}_{2}}}\; \and \cdots \and 1 \\ \end{matrix} \right];{{a}_{ij}}=\frac{{{w}_{i}}}{{{w}_{j}}},{{a}_{ij}}\ne 0\]$$
   Each matrix was normalised and their relative weights (A_w) were determined. The relative weights were given by the correct eigenvector (w) corresponding to the largest eigenvalue (λ _max), as
   $${A_w}{} = \lambda _{{\rm{max}}}}{\rm{*}}w$$
   The largest eigenvalue (λ _max) was computed by forming the sum of the single normalised eigenvector (w) per row multiplied by the computed priority vector per corresponding column:
   $$\lambda _{{\rm{max}}}} = \sum \left( {\mathop \sum \limits_{ij} {w_{\left( {{\rm{row}}} \right)}}*{{\mathop \sum \nolimits_{ij} {w_{\left( {{\rm{column}}} \right)}}} \over {\sum \left( {\mathop \sum \nolimits_{ij} {w_{\left( {{\rm{column}}} \right)}}} \right)}}} \right){\rm{}}$$
   If the pairwise comparisons were completely consistent, the matrix A had rank 1 and λ _max = n. In this case, weights could be obtained by normalising any of the rows or columns of A. The priority vectors of FPV and GPV were computed by dividing the single normalised eigenvectors (w) per column by the sum of all single normalised eigenvectors per column for the single matrices:
   $${\rm{FPV}}/{\rm{GPV}} = {{\mathop \sum \nolimits_{ij} {w_{\left( {{\rm{column}}} \right)}}} \over {\mathop \sum \left({{\mathop \sum \nolimits_{ij} {w_{\left( {{\rm{column}}} \right)}}} }}\right)$$
   It should be noted that the quality of the results was related to the consistency of the comparison judgements. As a quality control, the inconsistency ratio (iCR) of the ranked elements was computed as a ratio between the consistency index (CI) and random index (RI):
   $${\rm{iCR}} = {{{\rm{CI}}} \over {{\rm{RI}}}}$$
   The iCR provided information on the inconsistency of the ranked elements. An iCR ≤ 0.1 was acceptable as a criterion (Kahraman et al., Reference Kahraman, Demirel and Demirel2007; Borajee and Yakchali, Reference Boraje and Yakchali2011). Thus, the ranking of the elements had to be repeated by the team of experts when the iCR exceeded this threshold. The CI was computed by using the subsequent formula:
   $${\rm{CI}} = {\lambda _{\rm{max}}-{\rm n} \over {\rm n - 1}}$$
   The RI was defined as an expected value of the CI depending on an individual number of ranked elements n. These fixed values were obtained from the study of Aguarón and Moreno-Jiménez Reference Aguarón and Moreno-Jiménez(2003) for further computations of the CI parameter (Table 4). The authors simulated 100,000 matrices for several sizes of n and calculated standard values for RI.

4. (IV) Calculation of the GPV

   The GPVs were computed for each corresponding SWOT group by the expert team based on the frequency of farmers’ items included in each SWOT group. Computation details for the GPV are described in step III computation of priority scores for SWOT factors.

5. (V) Calculation of the OPV

   However, the OPVs were calculated by multiplying the single FPVs of the single SWOT factors with the corresponding GPVs of each SWOT group:

   $${\rm{OPV}} = {\rm{GPV*FPV}}$$
   The number of computed OPVs should correspond with the number of computed FPVs.

Figure 1

Methodical implementation of AHP based on farmer surveys for local cattle breeds. SWOT=strengths weaknesses opportunities threats; AHP=Analytic Hierarchy Process; GPV=group priority vector; FPV=factor priority vector; OPV=overall priority vector.

Table 1

Priority scores of ranked SWOT factors for GA

SWOT=strengths weaknesses opportunities threats; GA= German Angler; GPV=group priority vector; FPV=factor priority vector; OPV=overall priority vector.

Table 2

Priority scores of ranked SWOT factors for RDP

SWOT=strengths weaknesses opportunities threats; RDP=Red Dual-Purpose cattle; GPV=group priority vector; FPV=factor priority vector; OPV=overall priority vector.

Table 3

Fundamental scale to determine the relative importance of each element for the German Angler and Red Dual-Purpose cattle by Saaty Reference Saaty(1980) and Yüksel and Dağdeviren Reference Yüksel and Dağdeviren(2007)

Table 4

Fixed values of RI dependent on the size of matrices to obtain strategy consistency for the German Angler and Red Dual-Purpose cattle by Aguarón and Moreno-Jiménez Reference Aguarón and Moreno-Jiménez(2003)

n=ranked elements; RI=random index.

Conservation and development strategies at breed level

In general, SWOT strategies embracing single breeds should be identified at breed level (Martín-Collado et al., Reference Martín-Collado, Díaz, Mäki-Tanila, Colinet, Duclos, Hiemstra and Gandini2013) and not just from a selected team of experts. Especially, in small endangered populations the number of experts or stakeholders is limited and a multi-stakeholder approach for single local breeds may be not achievable. Thus, identified strategies can be misleading. Furthermore, the identification of SWOT factors is often unbalanced (Karppi et al., Reference Karppi, Kokkonen and Lähteenmäki-Smith2001) and a multi-stakeholder approach was preferred in several studies to avoid subjectivity (Impoinvil et al., Reference Impoinvil, Ahmad, Troyo, Keating, Githeko, Mbogo, Kibe, Githure, Gad, Hassan, Orshan, Warburg, Calderón-Arguedas, Sánchez-Loria, Velit-Suarez, Chadee, Novak and Beier2007; Martín-Collado et al., Reference Martín-Collado, Díaz, Mäki-Tanila, Colinet, Duclos, Hiemstra and Gandini2013). Quantified SWOT have been identified by applying the quantified strategy decision tool of SWOT-AHP in the step before. To improve the limited amount of stakeholder groups, quantified SWOT factors with the three highest OPVs per SWOT group were comprehensively discussed face-to-face with unbiased representatives of further stakeholder groups, that is, farmers, breeders and staff of the breeding organisation (breeding board). Face-to-face discussions were performed in one group meeting with the following respective representatives: 3 (farmers), 3 (breeders) and 3 (breeding board) per breed. All representatives were equally weighted and discussed their attitudes as soon as they had reached consensus. Then promising conservation and development strategies at breed level were collectively transformed together. This formation process was refereed and protocolled by the expert team.