Impact of Dynamic Capabilities on Performance in Dairy Sheep Farms in Spain

This study links the dynamic capabilities (DCs) theory with performance in dairy sheep farms in Castilla La Mancha (central Spain). The approach is novel as it seeks to understand how best results can be achieved by deploying DCs in farms. The proposal is that dynamic capabilities are interrelated to each other and present a positive impact on the farm’s economic sustainability. A mixed methods approach (a combination of qualitative and quantitative methods of research) was utilized. First, 30 indicators of dynamic capabilities (8 of absorption, 11 of integration, 9 of innovation, and 2 of profits) were selected by applying Delphi’s methodology. Second, a structural equation model (SEM) was applied over a random sample of 157 dairy sheep farms to measure the relationship between DCs and the impact of each capability on farms’ final performance. The existence of positive relationships amongst absorption, integration, and innovation capabilities was evidenced. Absorption and integration capabilities exhibited positive influences on a farm’s final performance. The knowledge of the relationships amongst dynamic capabilities is a new orientation to increase farms’ viability. These findings reveal that the application of the dynamic capabilities theory can explain best farms’ economic sustainability.


Introduction
In Spain, the dairy sheep industry suffers from an intense crisis as a consequence of a group of factors: lack of profitability, increased bureaucracy in management, absence of intergenerational relay that assures the continuity of the activity [1][2][3], changes in the habits of consumers with a reduction in the consumption of sheep meat and animal protein [4], and the need to mitigate climate change and reduce greenhouse gas emissions [5]. In Castilla La Mancha (central Spain), the producers have responded to the crisis: 33.7% of farms have abandoned the milking activity and the rest show a trend towards the intensification and specialization of production [6]. In Spain and within the framework of the Common Agricultural Policy (CAP), compensatory payments have been offered to producers for "loss of profits" as a strategy to preserve native breeds (Rural Development Plans, 2015-2020). However, ways to achieve them; it implies knowing the existing resources in the company and making current decisions about the future of the company [48,49]. Ultimately, the application of integration and innovation concepts from the dynamic capabilities approach favors reconfiguring organizational activities aimed at the creation of value [50]. In addition, all this is framed within the existing problems in livestock farming, such as excessive bureaucracy, loss of profitability, and lack of generational change, all taking place in a society committed to the reduction of greenhouse gases and the development of systems of production that are environmentally friendly.
There is a lack of research to assess quantitatively the impact of dynamic capabilities on the viability and sustainability of farms. There is a need to go deeply into this topic to find out answers to questions such as: what is the impact of an increase of dynamic capabilities on a farm's economic sustainability? How does the process of technology adoption take place from the dynamic capabilities approach?
To go deeply into these questions for the case of Manchego dairy sheep farms would be a novel approach, and it constitutes needed research aimed to improve the viability and sustainability of farms. The ones that are generating increased added value could promptly develop a sustainable competitive advantage. If validated, its applicability could be extended to other livestock activities in a critical situation. So, the relevance of this article consists of finding out if relationships among elements of the dynamic capabilities theoretical framework [32] significantly impact financial results in sheep farming in Spain.
The main objective of this research consists of analyzing how farms' dynamic capabilities impact on their final performance. A dynamic capabilities (DC) model, with 49 indicators identified, was applied to this research [32,34]. Nevertheless, several questions that had to be solved before applying the model were raised: were the selected indicators appropriate? How was it decided that each indicator should be categorized in the specific latent variable? What effect did each construct show on the final results of the farm? Questions were evaluated by applying a mixed methods approach, combining qualitative and quantitative methodologies. For the selection and grouping of capabilities' indicators, Delphi's methodology was used [2,8]. Later on, a structural equation model (SEM) was built to evaluate interactions between the different dynamic capabilities and the impact of dynamic capabilities on farms' performance for Manchego dairy sheep farms in Castilla La Mancha, Spain [1,51,52].

Sample Collected and Indicators
The study area was the Spanish region of Castilla La Mancha (38 • -41 • N; 1 • -5 • W), with a surface area of roughly 800,000 ha and Mediterranean continental climate: dry winters and hot and dry summers. A random sample of 157 Manchego sheep dairy farms was selected, 17.2% of the whole population of 907 Manchego sheep farms from the "La Mancha" region. Data were collected by using in situ visits to the farms by the same person, with 5 h of duration each one, from 2012 to 2014, updated according to data from existing farms in 2018. References [2,30] describe the Manchego dairy mixed system in Castilla La Mancha. The questionnaire included 203 questions related to the use of land, facilities, flock, labor force, feeding, grazing, breeding, animal healthcare, milking and milk quality, and economic and social issues [29,30]. In 2018, the economic data (prices and costs) of each farm were updated, operative farms were identified (87% of the sample), and missing data on dynamic capabilities were collected. Based on the model proposed by reference [32] for measurement of dynamic capabilities, the indicators were identified. Through the qualitative methodology in Section 2.2, we attempted to answer two questions: what are the indicators and why they were chosen? How was each item in the specific latent variable categorized? Later, through quantitative methodology, the relationship among constructs was found using SEM analysis. The model is proposed in Section 2.3, and the statistical analyses are described in Section 2.4.

Selection and Assignment of Dynamic Capabilities (DC) Indicators
Considering Delphi's methodology, 49 indicators of dynamic capabilities and performance were assessed (Supplementary Materials Table S1). The selection and assignment process consisted of experts' judgments by means of successive iterations of a questionnaire oriented to show convergence of opinions, and to identify dissent or non-convergence [2,11,12]. To do the Delphi, a workshop with face-to-face participation of 30 experts from the Animal Science Department at the University of Cordoba was held in May 2017. The workshop began with an introduction, oriented to explain the main objectives of the research and the concepts of DCs. The experts assessed each DC by using a Likert scale from one to five values, where one was the least important and five the most important. In the first round, the indicators were assessed and assigned by experts to dynamic capabilities, and a first draft of the metric was agreed upon.
In a second round, the indicators that did not reach consensus in first round were discussed. Once the indicators were finally chosen and assigned to dynamic capabilities by consensus, the questionnaire was sent to each expert to re-examine and reconsider his judgment.
Once the indicators of each DC were identified and considering the information obtained from the questionnaires sent to 157 farms, the proposed indicators were tested. Uncorrelated variables and the ones showing low coefficient of variation by pairs with linear dependence were removed. Detection capability indicators were excluded because they showed linear correlation with absorption and integration capabilities. Therefore, the initial model taken into consideration [32] was modified; only absorption, integration, and innovation capabilities were considered in a structural equation model. Finally, 30 indicators were selected, and four latent variables or constructs were built [53].
Most of the indicators were taking a value of 0 to indicate the absence of an attribute and a value of 100 to indicate its degree of presence. Certain attributes were quantified by several excluding options. Categorical questions with four response options were assigned values of 0, 33, 66, and 100. In addition, categorical questions with five response options were assigned values of 0, 25, 50, 75, and 100. Finally, two financial ratios normalized, return of assets (ROA) and return on investments (RF), were selected as economic sustainability indicators. The metric applied for the measurement of each variable is shown in Table S1.

Hypotheses
The general hypothesis is: a farm's dynamic capabilities (DC) will positively influence its economic sustainability. Subsequently, relationships amongst the dynamic capabilities of absorption (absorp), integration (integ), and innovation (innov) were also analyzed. Eight variables of absorption, 11 of integration, 9 of innovation capabilities, and 2 of profits were included in the final model. Five hypotheses were formulated as displayed in Figure 1.

Hypothesis 3 (H3).
The absorption capability (absorp) will positively influence a farm's results (profit); the return on assets (ROA) and return on investments (RF) are profit indicators (profit).

Hypothesis 5 (H5).
The innovation capability will positively influence a farm's final results (profit).
Hypothesis3 (H3). The absorption capability (absorp) will positively influence a farm's results (profit); the return on assets (ROA) and return on investments (RF) are profit indicators (profit).

Statistical Analyses
The values of each variable were calculated for each one of the 157 farms visited. Thirty indicators of dynamic capabilities and economic sustainability were analyzed.
The estimation of the structural equation model (SEM) evaluated the relationships amongst the different constructs (absorp, integ, innov, and profit), through path coefficients, significance level, and cross-validated redundancy. The focus was on graphs without feedback loops between nodes. The model was estimated by means of partial least squares (PLS) path modeling, which allows for estimating complex cause-effect relationships among and latent variables. To ensure the stability of the results, the number of bootstrap subsamples was set to 5000. The estimated parameters from the subsample's models were used to derive standard errors for the estimates and, consequently, the bootstrap 95% confidence intervals [51,56,57].

Dynamic Capabilities Model
Quality criteria considered for the constructs in the final model are presented in Table 1. The square root of the average variance extracted (AVE) of each construct should be higher than its highest correlation with any other construct [58]. The variability of a latent variable attributable to indicators compared to variance derived from measurement errors gives a construct's AVE. Values

Statistical Analyses
The values of each variable were calculated for each one of the 157 farms visited. Thirty indicators of dynamic capabilities and economic sustainability were analyzed.
The estimation of the structural equation model (SEM) evaluated the relationships amongst the different constructs (absorp, integ, innov, and profit), through path coefficients, significance level, and cross-validated redundancy. The focus was on graphs without feedback loops between nodes. The model was estimated by means of partial least squares (PLS) path modeling, which allows for estimating complex cause-effect relationships among and latent variables. To ensure the stability of the results, the number of bootstrap subsamples was set to 5000. The estimated parameters from the subsample's models were used to derive standard errors for the estimates and, consequently, the bootstrap 95% confidence intervals [51,56,57].

Dynamic Capabilities Model
Quality criteria considered for the constructs in the final model are presented in Table 1. The square root of the average variance extracted (AVE) of each construct should be higher than its highest correlation with any other construct [58]. The variability of a latent variable attributable to indicators compared to variance derived from measurement errors gives a construct's AVE. Values for the AVE above 0.5 are acceptable [59]. This criterion is satisfied in Table 1, confirming convergent validity. Measures used to evaluate internal consistency reliability were Cronbach's alpha, Dillon-Goldstein's rho and eigenvalue analysis [60]. In the literature, there is no minimum acceptable value of Cronbach's alpha, sometimes levels above 0.7 are acceptable [61], and other values above 0.5 are consistent with models enjoying internal reliability [62]. Two other metrics used to ascertain the correct outer model specification were also considered. Dillon-Goldstein's rho values were acceptable as they were higher than 0.7. First, eigenvalues were greater than 1, while second, eigenvalues were less than 1. This is as expected of reasonable constructs [60]. R 2 for innovation was high (0.6394) and low (below 0.3) for integration capability and profit. Average redundancy showed the ability of exogenous constructs to explain the average variation in the indicators of dependent latent variables. As the use of redundant items has adverse consequences for the measures' content validity, researchers are advised to minimize the number of redundant indicators. In this research, absorption, integration, and innovation predicted 21.7% of the variability of profit indicators. Communality shows how much reflective indicator variability is reproducible by the latent variable. A communality of 50% may be expected from a reflective block [61]. The R 2 and redundancy for absorption were both zero, as the construct was an exogenous latent variable [51,63].

Hypotheses Assessment
The inner-plot is shown in Figure 2, and Table 2 presents the structural model, testing the hypotheses presented above. Statistically significant relationships have p-values lower than 0.05. Absorption capability positively impacted integration at a 1% level of significance. Integration had a positive influence on innovation at a 1% significance level. Absorption and integration capabilities positively influenced economic sustainability at 1% and 5%, respectively. Innovation had no statistically significant influence on profit. Although innovation appeared to negatively influence profitability, the relationship was not statistically significant. Hypothesis 5 may, therefore, be rejected.
The DC models' goodness-of-fit was 0.5380. This index was calculated as the geometric mean of the average communality and the average R 2 value. A severe shortcoming with this index is that there was no threshold allowing to determine its statistical significance [56,57]. Although this was below the threshold level of 0.7 to be considered "acceptable", the prediction power was fairly substantial. The higher this measure is, the more appropriate the model is [61]. Bootstrapping results are shown in Table 2. Original path coefficients are compared with the results after bootstrapping. Confidence intervals assist in ascertaining the significance of the relationships examined. At a 95% confidence level, innovation did not impact profitability in this industry, given the available data. The DC models' goodness-of-fit was 0.5380. This index was calculated as the geometric mean of the average communality and the average R 2 value. A severe shortcoming with this index is that there was no threshold allowing to determine its statistical significance [56,57]. Although this was below the threshold level of 0.7 to be considered "acceptable", the prediction power was fairly substantial. The higher this measure is, the more appropriate the model is [61]. Bootstrapping results are shown in Table 2. Original path coefficients are compared with the results after bootstrapping. Confidence intervals assist in ascertaining the significance of the relationships examined. At a 95% confidence level, innovation did not impact profitability in this industry, given the available data. The impact that each construct had on the rest of the latent variables is shown in Table 3. The total effects are the sum of direct and indirect effects. Additionally, the impact of absorption (absorp) over innovation capability (innov) is shown in Table 3; this last relationship was not considered in the initial model (Figure 1).  The impact that each construct had on the rest of the latent variables is shown in Table 3. The total effects are the sum of direct and indirect effects. Additionally, the impact of absorption (absorp) over innovation capability (innov) is shown in Table 3; this last relationship was not considered in the initial model ( Figure 1).
Thirty indicators of dynamic capabilities were selected by a panel of experts. A random sample of 157 dairy sheep farms in Spain was selected. Results showed the association amongst the capabilities of absorption, integration, and innovation (Hypotheses 1 and 2 were accepted) and the impact of dynamic capabilities in farms' profit was evaluated (Hypotheses 3 and 4 were accepted and hypothesis 5 was rejected). Absorption capability showed a positive influence on integration (Hypotheses 1) and on the economic result (Hypotheses 3) in a direct and indirect way. Farms recognize the value of external information and new technologies, assimilate and apply them to commercial ends [2,9]. Farmers in central Spain mainly responded to the economic crisis by increasing the sizes of their flocks, increasing the intensification and specialization in a productive activity, and reducing diversity and resilience of the farming system [3,6,30]. This situation entails a high level of technology adoption by establishing priorities to those technologies that impact on final results in the short term. In this sense, farmers that have implemented local breeding programs achieve improvements in their farms' viability [2].
The technologies implemented must be assembled with other areas of the farm, such as feed and animal health. This need to articulate knowledge is reflected in the existence of a direct effect (0.2927) between the absorption capability and economic results, and one indirect effect (0.1352) that generates a total effect absorp → profit (0.4280) ( Table 3). These results agree with previous models [32]. On the contrary, these farms maintain a strategy opposed to the dual-purpose systems described in Latin American and Caribbean countries, with very low technological levels [44,76,77]. For smallholders of dual-purpose cattle farms in Mexico, low technological levels are described [9]. In this case, their strategy is oriented to small-scale production, low levels of technology, scarce external inputs, and low risk.
In this research, integration capability showed a high positive influence over innovation (hypothesis 2 accepted); although, at the same time, integration did not show any indirect effect on innovation (Table 3). That is, DCs' impacts appeared to happen sequentially and cumulatively [17]. The interaction of activities demonstrated the use and interchange of information amongst functional units. In the dairy sheep industry, those producers adopting quality improvement programs need an integrated technology adoption in different fields of the farm [31]. In this same sense, producers included within the breeding planning need to improve other technological areas (health, feeding, etc.) to achieve the success of the program [2]. DCs consider the organization as a whole, as well as its relationship with the environment, facilitating the coordination of employees and the development of behavior patterns [78]. In this regard, authors of [8,9] explain the need to assemble the different technological areas in the company (feeding, reproduction, health, management, land use, quality of the milk) to achieve higher levels of efficiency. The results showed a direct effect of the integration capability on a farm's performance (hypothesis 4 accepted).
Integration capability is the ability to integrate different patterns of interaction through contribution, representation, and interrelation [32,41,73]. This way, integration capability refers to the ability to properly combining human, technological, and organizational resources [19]. Firms may have similar resources, but the way they combine them can be different and, therefore, turn in a variety of degrees of integration capability [68,71].
We suggest improving the 18 organizational routines and practices in absorption and integration capabilities (Table S1); e.g., migrating from a batch control to individual control, by productive objectives and applying IT in decision-making in dairy sheep farms.
Contrary to expectations, innovation capability did not exhibit a significant effect on results and hypothesis 5 was rejected. Bravo and Herrera [27] found a positive effect between innovation and economic performance. In the same vein, a positive relationship between innovation capability and growth in SMEs in Spain was verified [43]. To a contrary extent, references [1,11] indicate that the technical innovations allowed an increase in the gross margin in the short term, although it was not sustainable in the long term. Further, in adverse political and economic conditions, innovations are not enough to mitigate the negative effects of crises. Apart from this, advances in animal health innovation and milk quality are bound by the country's regulations, which causes them to have a negative or null impact on the results [9].
These findings suggested that dairy sheep farmers have made a great effort for implementing technology adoption in Spain (absorption capability) [2,29]. Currently, they are adapting it to the production process and to the other areas of the farm (integration capability) [7,27]. However, they have not yet reached the innovation phase that favors the creation of additional value and the deployment of sustainable competitive advantages. Once the integration in processes is optimized, the innovation phase, where the productive models must be redefined, arises [44]. The adoption sequence of capabilities could be an explanation, as no positive relationship between innovation and profit in the dairy sheep sector was found.
The increase of a farm's performance is proposed from the deployment of dynamic capabilities in dairy sheep farms in Spain. An integrative and complementary vision to previous ones aimed to explain the reasons why some farms have developed certain dynamic capabilities that have allowed them to reach the best results would be of great interest. Dynamic capabilities (intangible factors) have been related to farms' final performance through an SEM model. The deployment of DCs allows improving economic results in the short run and thinking in a strategic change of the productive model in the long run.
The research presented significant limitations. First, because the conceptualization of dynamic capabilities is relatively new, much remains to be learned about the nature of the synergies and trade-offs between absorption, integration, and innovation components [79][80][81][82]. There is little empirical research upon which to base the selection of one modeling approach, however, there are other approaches theorizing in the dynamic capabilities context that should be evaluated [83]. Likewise, it would be necessary to confront the proposed model with other species, production systems, and socio-economic contexts that would allow it to be completed and validated [9]. This research used a cross-sectional research method. In the future, longitudinal studies in the farms should be considered for understanding how the growth of these capabilities influences economic sustainability [2].

Conclusions
The appropriate variables to assess each of the dynamic capabilities and profit indicators were agreed by 30 experts in dairy sheep farms. In the case of Castilla La Mancha (Spain), the dynamic capabilities of absorption, integration, and innovation were identified as significant intangible constructs. After the selection of the 30 indicators, data were collected from 157 farms. A structural equation model (SEM) measured the relationships of these variables and their effect on the improvement of production results. The results evidenced the existence of interrelations between absorption, integration, and innovation (hypotheses 1 and 2 were accepted). In addition, absorption and integration capabilities showed a positive impact on the results (hypotheses 3 and 4 were accepted), although it cannot be confirmed for the case of innovation (hypothesis 5 was rejected).
This study addressed a gap between the dynamic capabilities perspective and the economic sustainability of the farms, linking from an interdisciplinary way, strategic management theory with animal science. This research took a novel approach as it quantified the relationship amongst dynamic capabilities (absorption, integration, and innovation) and their impacts on farms' final performance.
The improvement of dynamic capabilities favors an increase in farms' viability. Technological absorption is a first step, although it is not sufficient to create competitive advantages. These results suggest that the farms should develop organizational routines and practices oriented to create absorption and integration capabilities aimed to reach the best managerial results.
Thus, dynamic capabilities theory provides an interesting theoretical foundation for studying the link between absorption, integration, and innovation capabilities and firms' performance. In addition, it allows us to deepen in the knowledge of the mechanisms by which organizations achieve superior performance levels and sustainability. Finally, the DCs perspective could allow farms to reach sustainable competitive advantages.