Understanding the adoption of grazing practices in German dairy farming

Agricultural Systems 165 (2018) 230–239

Contents lists available at ScienceDirect

Agricultural Systems journal homepage: www.elsevier.com/locate/agsy

Understanding the adoption of grazing practices in German dairy farming ⁎

T

Henning Schaak , Oliver Mußhoff University of Goettingen, Department of Agricultural Economics and Rural Development, Platz der Goettinger Sieben 5, 37073 Goettingen, Germany

A R T I C LE I N FO

A B S T R A C T

Keywords: Technology acceptance model Structural equation model Grazing practices Dairy production

Due to a simultaneous decline in agricultural practice and an increased favorability and demand by society, grazing based milk production has become a topic of heightened interest in European agricultural policy, as well as dairy product marketing. This paper studies the behavior of German farmers with respect to the adoption of grazing practices. To do so, a structural equation model based on the technology acceptance model (TAM) is developed. Generally, the TAM hypothesizes that the perceived usefulness and the perceived ease of use are key determinants of the intention to use and the actual usage behavior of a technology. The results indicate that the perceived usefulness and perceived ease of use statistically significantly influence the adoption of grazing practices. Other important aspects are the production limitations on the individual farm, and the farmers' subjective norm towards grazing. Furthermore, the analysis reveals differences between conventional and organic farmers, showing that the influence of farmers' beliefs on the usage behavior tends to be greater for conventional farmers. The results show that farmers' subjective norm influences multiple other constructs of the model, including the intention to use. Under the assumption that farmers' perceptions of societal expectations depend on the public discourse, this indicates the relevance of public information and communication for the farmer's decision-making processes.

1. Introduction In many European countries, grazing practices have gained increased attention in social and political discourse in recent times. In discussions concerning preferable milk production systems, many stakeholders favor grazing-based systems. From the consumer perspective, this preference is driven by perceived advantages towards animal welfare (Weinrich et al., 2014). It is also reflected by a higher willingness to pay for pasture-based milk of some consumer groups (Ellis et al., 2009; Hellberg-Bahr et al., 2012). These findings have been acknowledged by the dairy sector, as dairy processors in Europe have started to market pasture raised milk separately (Fahlbusch et al., 2009; Kühl et al., 2016). Pasture-based milk production is also discussed with respect to pasture conservation issues. Grazing as a form of pasture usage is seen as an important measure in order to preserve pastures (Plachter and Hampicke, 2010). Related, also low-input milk production has gained attention (Bijttebier et al., 2017). Grazing can have positive effects on the welfare of cows (von Keyserlingk et al., 2009), which confirms consumer perceptions. From a perspective restricted to a single farm, there is a consensus that the economic viability of grazing depends on the on-farm conditions, other input costs, and the chosen management style (cf. Knaus, 2016; Peyraud et al., 2010; Thomet et al., 2011).

⁎

Regardless of these findings, the share of dairy farms utilizing pasture for grazing practices has been decreasing in many European countries (Reijs et al., 2013). The decline is driven by structural changes (such as increasing stock numbers per farm and changes in the availability of labor) and by other changes in the production system (e.g. changing calving patterns) (Hennessy et al., 2015). This indicates an existing gap between the developments in agriculture and the expectations of society. Therefore, the future development of pasture usage and its extent has become a lively political topic. Most recently a “Grazing-Charter” (“Charta Weideland Norddeutschland”; Grünlandzentrum, 2015), an industry agreement supported by policy measures, was introduced in northern Germany. The existing gap in the extent of grazing-based milk production indicates the need for a better understanding of the corresponding decision-making processes, as decision-making in economic contexts may not be purely driven by economic reasoning. Instead, it can also be influenced by intentions, attitudes, and beliefs of the decision maker (Ondersteijn et al., 2003; Willock et al., 1999). The decisions a farmer makes are not only directly influencing the low order farm systems (cf. the farming system scheme described by McConnell and Dillon, 1997). By influencing factors beyond immediate production, the decisions also influence agricultural systems beyond the individual farm. They have an impact on a local scale (e.g. on landscape features and local biodiversity) as well as the

Corresponding author. E-mail addresses: [email protected] (H. Schaak), oliver.musshoff@agr.uni-goettingen.de (O. Mußhoff).

https://doi.org/10.1016/j.agsy.2018.06.015 Received 18 August 2017; Received in revised form 9 March 2018; Accepted 25 June 2018

Available online 06 July 2018 0308-521X/ © 2018 Elsevier Ltd. All rights reserved.

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

farmers, where the overall general situation is quite different. Since up to 100% of Irish cows graze (Reijs et al., 2013), previous studies focused on specific aspects in dairy farming (e.g. grass growth measurement or the usage of rotational grazing techniques). In contrast, milk production in Germany is rather heterogeneous (Lassen et al., 2014, 2015; Reijs et al., 2013) and grazing is not a ubiquitous management technology. Given the similar situation in other western European countries (Reijs et al., 2013), the results of the study are more suited to be transferred to other settings. Second, compared to the previous work, the study takes a broader perspective by analyzing the drivers of the adoption of grazing management practices in general. We use the general term grazing for management practices which allow the herd access to pastures and the opportunity to graze there. Grazing requires several complementary identifiable actions and measures on the farm2. Third, as discussed above, the study is the first to apply the structural form of the TAM in the grazing context. Besides incorporating all constructs and relationships stated in the original TAM, as well as common extensions, the model also allows the analysis of the adoption extent in a continuous way. In contrast to previously used binary approaches, it allows for a more nuanced understanding of the adoption decision. Further, a comparison between conventional and organic farmers is carried out. The objectives of the paper can be summarized by three research questions: (1) what are the influences on the intention to use, and ultimately the actual usage of grazing?; (2) can differences between conventional and organic farmers be identified?; (3) does the analysis have implications beyond the perspective of an individual farm?

supra-regional (e.g. through groundwater leeching plant of nutrients) and the global scale level (e.g. through greenhouse gas emissions). This is particularly relevant to dairy production systems, where pastures are an essential part of the production system, as elaborated before. The specific dairy production system also becomes increasingly important with respect to the wider food sector, as it the production system is an increasingly important factor for product differentiation and labeling. Varying demand patterns of different production systems also influence preceding parts of the agricultural industry. In the past, several models have been developed to explain the behavior of an individual. A prominent early approach is the “Theory of Reasoned Action” (Ajzen and Fishbein, 1980) which assumes that behavioral intentions are the main predictors of behavior. It was later extended to the “Theory of Planned Behavior” (Ajzen, 1985), which additionally accounted for the individuals perceived control of the behavior. Although widely used, these approaches failed to produce reliable measures in the context of technology usage (Marangunić and Granić, 2015). In order to overcome this issue, the technology acceptance model (TAM) was developed. The TAM is a model for the analysis of acceptance processes of information technologies and was introduced by Davis (1986), and later refined by Davis (1989) and Davis et al. (1989). While originating from the prior models, the TAM has a different structure and relies on therefor conceptualized beliefs. Although initially developed in information systems research, it has been widely used to study technology-adoption behavior in a broader sense and in various domains (Venkatesh et al., 2007). An overview of the initial developments and later extensions of the model is given by Marangunić and Granić (2015). This study analyzes the impact of individual beliefs1 of German farmers on both the intention to use, and the actual application, of grazing management practices. Furthermore, individual farm specific conditions (such as herd size and available pasture) are taken into account. The analysis allows for the identification of possible differences between conventional and organic farmers. The study relies on a structural equation model, which augments the TAM (Davis, 1989; Davis et al., 1989). With respect to agriculture, the TAM has been applied to precision farming technologies, and the meat and dairy sector. The adoption of precision farming technologies was studied by Adrian et al. (2005) and Rezaei-Moghaddam and Salehi (2010). Arens et al. (2012) studied the acceptance of information systems by pig farmers. For the dairy sector, the adoption of technologies such as mineral supplementation and soil quality testing, was studied by Flett et al. (2004). Focusing on particular grassland management practices, Kelly et al. (2015) found that intention to implement a practice is strongly determined by the beliefs of the individual farmer. The adoption of different grazing related production technologies by new entrant farmers was studied by McDonald et al. (2016), who found a substantial influence of farmers' beliefs regarding a technology in the decisionmaking process. These studies used solely the conceptual basis of the constructs of the TAM. They also either considered the intention towards the adoption, or the actual adoption of grazing practices. These analyses then relied on binary regression frameworks. Best to the authors' knowledge, the structural form of the TAM has not been applied in the grazing context before. The structural form incorporates all constructs and relationships stated in the original TAM, including the influence of the intention towards adoption on the actual adoption. Further, these applications did not consider possible extensions of the TAM, which have been proposed in the literature. The present paper closes this research gap. Compared to previous research considering the adoption of grazing, the present study includes several novelties. First, while those applications of the TAM focused on Ireland, the focus here is on German

2. Theoretical framework As mentioned before, the decision of whether to adopt a management technology may not only depend on economic reasoning. The intentions, attitudes, and beliefs of the decision maker can also influence the decision (Austin et al., 1998; Willock et al., 1999). By adapting the TAM to dairy farming, we follow previous research (cf. Flett et al., 2004; Kelly et al., 2015; McDonald et al., 2016). Based on the TAM, a theoretical framework to analyze behavioral drivers towards the application of grazing is developed. The graphical representation of the model can be found in Fig. 1. The core of the model is the initial TAM model. Suitable extensions and their corresponding linkages are derived from the literature. Extensions are considered to be suitable, when they are a) widely applied in the literature and b) applicable to other contexts than information systems research. Further, an additional construct, which accounts for possible limitations of the actual farm of the individual, is introduced. In the following, the research hypotheses regarding the adoption of grazing are described and reasoned. The technology acceptance model (TAM) introduces the concepts of Perceived Usefulness (PU) and Perceived Ease of Use (PEOU). The PU is defined as “the degree to which a person believes that using a particular system would enhance his or her job performance” (Davis, 1989). The PEOU is defined as “the degree to which a person believes that using a particular system would be free of effort” (Davis, 1989). Both beliefs are used to explain the Intention to Use (IU) a technology (also referred to as the “Behavioral Intention”). Further, the PU is influenced by the PEOU. The target construct of the TAM is the actual Usage Behavior (UB). It is influenced by the IU a technology. These relationships have been found to be robust in various technologies (King and He, 2006; Marangunić and Granić, 2015). The first four hypotheses of our model represent these relationships (cf. Fig. 1): H1:. The intention to use (IU) positively influences the usage behavior (UB) of grazing. H2:. The perceived ease of use (PEOU) positively influences the

1 Generally, beliefs “refer to a person's subjective probability judgments concerning some discriminable aspect of his world” (Fishbein and Ajzen, 1975, p. 131).

2 Overviews over different approaches and their implementations for grazing practices are e.g. given by Blanchet et al. (2000) and Hodgson (1990).

231

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

Shen et al. (2006) found the SN also influences the PEOU. We follow the authors and hypothesize that: H9:. The subjective norm (SN) towards grazing positively influences perceived ease of use (PEOU) of grazing. In the context of information systems, the POQ of the technology is understood as the extent to which the technology's performance is believed to be adequate for the required tasks (Marangunić and Granić, 2015), or more generally, the individuals perception of how well the system performs (Venkatesh and Davis, 2000). In context of this study, POQ is understood as the farmers' belief about the product quality (e.g. milk or meat quality) achieved by grazing. Following the literature, we hypothesize that: H10:. The perceived output quality (POQ) positively influences perceived usefulness (PU) of grazing. Fig. 1. Theoretical Framework. Source: Own elaboration based on Venkatesh and Davis (2000)

Last, we include the farmers' age (AGE) in the model. There are three reasons to do so. First, in the context of the TAM, the influence of an individual's age on his or her beliefs and intentions has frequently been shown (Marangunić and Granić, 2015). Second, the individuals' age partially represents his or her experience in context of agriculture (Ondersteijn et al., 2003). For example, Tauer (1995) showed that farm productivity varies between farmers of different age classes. Third, agricultural research has shown an influence of age on attitudinal components in other contexts. While Arbuckle et al. (2013) has shown that increasing age positively influences farmers' views towards adaptation and mitigation measures for climate change, Vanslembrouck et al. (2002) found that younger farmers have a more positive environmental attitude. Related, Wheeler (2008) found that older agricultural professionals (e.g. researchers and advisors) have a less positive attitude towards organic agriculture than younger ones. Therefore, we hypothesize that:

intention to use (IU) grazing. H3:. The perceived usefulness (PU) positively influences intention to use (IU) grazing. H4:. The perceived ease of use (PEOU) positively influences the perceived usefulness (PU) of grazing. The possibility to adopt grazing on a farm depends on the individual setup of the farm. Even given that a farmer has a high IU a grazingbased production system on his farm, various factors can limit or even prevent the UB. Thus, the UB cannot only depend on the IU, but must also depend on the conditions a farmer faces on his or her farm. Therefore, we include a construct called On-Farm Condition (OFC) in the model. As a low OFC may limit grazing on an individual farm, H5 is as follows:

H11:. The farmers' age (AGE) negatively influences his or her beliefs towards grazing.

H5:. The on-farm condition (OFC) positively influences the usage behavior (UB).

In economic studies related to agriculture and food production, researchers frequently differentiate between conventional and organic farms and farmers. Researchers not only studied production-related differences, but also differences in behavioral aspects. Egri (1999) found different preference patterns between conventional and organic farmers. In context of dairy farming, Power et al. (2013) found differences in multiple attitudinal constructs like environmental attitudes. Sullivan et al. (1996) even report differences in the attitude towards farming itself. Also, Mzoughi (2014) found that organic farmers state higher levels of life satisfaction than conventional farmers. Finally, Serra et al. (2008) found differences in risk attitudes of conventional and organic farmers. Researchers also examined potential drivers for the adoption of organic practices (Fairweather, 1999; Lunneryd and Öhlmér, 2009; Mzoughi, 2011). With respect to grazing practices, there is another reason to consider potential differences between conventional and organic farmers. Grazing, or at least pasture access is typically required by organic farming certifications. Therefore, the impact of IU on the UB may be lower for organic farmers, as their UB is partially determined by the production style. For the other relationships hypothesized in the model, similar observations may be present. Given this consideration and the results found in the literature, the following overall hypothesis is formulated (not shown in Fig. 1):

The farmers PEOU should (at least partially) reflect the conditions which he or she experiences on his or her farm. This implies that the OFC on a farm positively influences the PEOU for the farmer and leads to the following hypothesis: H6:. The on-farm condition (OFC) positively influences the perceived ease of use (PEOU) of grazing. The TAM has been extended and modified in numerous studies (Marangunić and Granić, 2015; Venkatesh et al., 2007). For example, researchers regularly included other beliefs and attitudes. These extensions are generally more specific to the initial domain of the TAM, and not all of them are transferable to other domains (like agriculture). In this paper only widely used extensions, which the authors consider to be justifiable, are included in the model. One influential extension by Venkatesh and Davis (2000) introduced the concepts of Subjective Norm (SN) and Perceived Output Quality (POQ). The SN towards the technology reflects “the influence of others on the user's decision to use or not to use the technology” (Marangunić and Granić, 2015). It represents whether the individual perceives positive or negative societal norms and/or expectations towards the usage of the technology. The common extensions of the TAM show that the SN positively influences the PU and the IU (Marangunić and Granić, 2015):

H12:. There are differences in the adoption drivers of grazing between conventional and organic farmers.

H7:. The subjective norm (SN) towards grazing positively influences perceived usefulness (PU) of grazing.

3. Material and methodology

H8:. The subjective norm (SN) towards grazing positively influences the intention to use (IU) grazing.

In the following segment, Subsection 3.1 describes the sample which was examined for the analysis. Subsection 3.2 briefly discussed the applied methodology and then describes and reasons the used

Still, an influence of the SN on other constructs can be imagined. 232

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

measurement model.

Table 1 Descriptive Statistics Characteristics

3.1. Sample In order to collect the data, an online survey was conducted at the beginning of the year 2016. The farmers were contacted via newsletters of consulting collectives, professional associations, and a magazine specialized in milk production. The only requirement for participation was that the farmer is housing dairy cattle. Participants from the new federal states of Germany (referring to the area of the former German Democratic Republic) were excluded from the sample. This was motivated by the observation that only few farmers were from that area, while the respective farms were the largest outliers with respect to the farm size (in terms of herd size and ha of arable land)3. Thus, the sample only consists of farmers from the old German federal states. After the removal of incomplete surveys, the final sample consists of data from 334 farmers. The descriptive statistics of the farmers and their dairy production are presented in Table 1. If available, it also shows the German average of the respective variables. The share of female farmers is above the German average of 8% (Gurrath, 2011). The farmers are between 22 and 69 years old. On average, the farmers manage twice as much land as the average German farm (Destatis, 2017). For the majority, farming represents the main source of income. On average, around 50% of the farms' area is pasture land. Of this pasture, the farmers consider 62% feasible for grazing. Roughly one fifth of the farms are operated organically, a high share compared to the German average. In the German context, a farm is considered to be run organically, when it at least fulfils the standards of the European Union regulation No. 834/2007 (“EU-Eco-regulation”). Most farmers received a formal agricultural education, ranging from an apprenticeship to a university degree. The average herd size is one third larger than the German average (BLE, 2016). The average yearly milk yield per cow and year of the sample farms is close to the German average of 7628 kg per cow and year (BLE, 2016). 43% of the farms are specialized in milk production. The most frequent secondary operation is crop production. None of the farms have more than six different operational branches (out of 13 predefined branches). The extent of grazing is shown in Fig. 2. The left side shows the distribution of average grazing days per year, and the right side shows the distribution of average grazing hours per day. As seen in both graphs, 84 farmers (25.15%) in the sample do not conduct any grazing activities. For the farmers who apply grazing, the herd has pasture access for on average 171.67 days per year (SD: 51.87) and 11.41 h per day (SD: 5.95).

Gender Age Farming as main income source Farming system: Organic Specialized dairy farm Arable land Pasture land Thereof: grazing land Education Agricultural education No. of cows Milk yield

Mean

SD

Female Years Yes

10.18% 45.81 88.62%

10.69

Yes Yes ha ha ha No. of school years Yes

19.46% 43.11% 46.08 47.44 29.41 11.21

54.25 34.86 31.90 2.76

6.00%a n.a. 58.09b 21.06b n.a. n.a.

91.60% 80.17 7884.30

75.38 1549.08

n.a. 58.00c 7628.00c

kg per cow and year

German mean 8.00%a n.a. 42.92%b

Source: Own calculations; N = 334. a Gurrath (2011). b Destatis (2017). c BLE (2016).

identification of the appropriate measurement model for the constructs depends on their conceptualization (Hair et al., 2017). For example, the causal priority between the indicator and the constructs (Diamantopoulos and Winklhofer, 2001), meaning whether the combination of indicators forms the construct (Rossiter, 2002), or if the indicators are interchangeable (Jarvis et al., 2003), needs to be taken into account. The consequences of model misspecification have been widely discussed (Diamantopoulos and Riefler, 2008; Diamantopoulos and Winklhofer, 2001; Roy et al., 2012). Commonly, the constructs of the TAM and its extensions are defined reflectively (Roy et al., 2012). Therefore, IU, PEOU, PU, SN and POQ are applied as reflective constructs. The indicators applied in the model can be found in Table 2. AGE is reflectively applied, as single item measures are always defined as reflective in the SEM-PLS algorithm (Hair et al., 2017). Both UB and OFC are combined representations of multiple indicators that are not interchangeable, and represent different aspects of the overall UB and OFC respectively. They are therefore applied as formative constructs. The wording of the reflective indicators is based upon norms applied in the literature (Davis, 1989; Davis et al., 1989; Venkatesh et al., 2003; Venkatesh and Davis, 2000). The wording was adapted to fit the dairy and farming context. Answers were given on a 5-Point-Likert scale.4 For formative measures, the selected indictors need to capture the overall construct domain (Diamantopoulos and Winklhofer, 2001). For OFC, the indicators were selected because they are critical aspects required to a large extent for grazing. The first indicator is the size of the farm's herd (No. of cows). An increasing number of animal increases the number of underlying requirements which need to be fulfilled (e.g. management of grass growth, additional requirements on herd lanes or time requirements for herd driving). This also increases the required managerial skills of the farmer. The second indicator is the average milk yield per cow and year. If the farmer aims to achieve high milk yields per cow, the extent to which he or she can apply grazing practices is limited. In extreme cases, grazing practices also may not be possible at all. Related to this, the third indicator is pasture area (which is suitable for grazing) per cow. High grazing volumes are restricted by pasture area per cow, in order to ensure sufficient feed intake during the day. UB is measured by three indicators, which form the overall usage behavior of grazing. In the context of this study, grazing is primarily defined as the extent of access to pastures. UB1 is the average days of

3.2. Methodology and measurement model In order to estimate the structural model described in Section 2, we apply structural equation modeling, based on the partial least squaresapproach (SEM-PLS). In contrast to covariance based approaches such as LISREL, SEM-PLS is a variance-based approach to structural equation models. Considering the given data, there are two aspects which support the application of SEM-PLS. First, SEM-PLS allows smaller sample sizes than covariance-based approaches, especially for complex models (Hair et al., 2011). Second, SEM-PLS does not make assumptions regarding the distribution of the indicators. As we observe non-normally distributed data, this also requires a PLS-based estimation (Hair et al., 2011). In the applied model, only AGE and OFC are exogenous, as seen in Fig. 1. The remaining constructs are endogenous (Hair et al., 2011). The

4 Other potential indicators where considered for the measurement model. In order to achieve a valid measurement model (cf. Hair et al., 2017), these indicators were not included in the final analysis.

3 This can be explained by historical differences between the agricultural systems of the former German Democratic Republic and the Federal Republic of Germany (between 1949 and 1990).

233

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

Fig. 2. Distribution of Pasture Access. Source: Own illustration; N = 334 Table 2 Constructs and Indicators. Construct

Indicator

Unit/statementa

Ageb (AGE) Intention to usec (IU)

Age IU1 IU2 IU3 IU4 OFC1 OFC2 OFC3 PEOU1 PEOU2 PEOU3 POQ1 POQ2 POQ3 PU1 PU2 PU3 PU4 SN1 SN2 SN3 UB1 UB2 UB3

in years I try to implement as much grazing as possible. If it were possible, I would extend grazing. Even if grazing is possible on my farm, I do not implement it.d I can imagine obliging myself to implement grazing. Number of cowsd Milk yield per cow and year Grazing area per cow (in ha) Every cow is suitable for unlimited grazing. Grazing simplifies herd management. Grazing is easy to implement. Grazing increases the meat quality of slaughtered cows. Grazing increases the milk quality and composition. Milk from grazing cows yields a higher quality than milk from pure in-housed cows. Grazing increases the productivity of my milk production. Grazing increases the costs of my milk production.d Grazing increases the profit of my farm. Grazing is not worth the additional effort.c Grazing is demanded by society. My private environment demands grazing. Grazing is demanded by my colleagues. Days/Year Hours/Day Access for the offspring

On-farm conditionb (OFC)

Perceived ease of usec (PEOU)

Perceived output qualityc (POQ)

Perceived usefulnessc (PU)

Subjective normc (SN)

Usage behaviorc (UB)

Source: Own elaboration. a related to a 5-point-Likert scale. b exogenous construct. c endogenous construct. d reversed coded.

between conventional and organic farmers, the sample was divided into two corresponding subsamples.

access to grazing pasture per year. UB2 are the average hours of access to grazing pasture per day. UB3 indicates the access to pasture for the offspring (measured in the levels “no”, “yes, partially”, and “yes”, assuming an underlying continuous distribution). A differentiation between different grazing systems such as set stocking, rotational grazing or strip grazing (Hodgson, 1990), is not feasible, as the use of binary indicators in endogenous constructs violates basic assumptions of the estimation algorithm (Hair et al., 2012).

4.1. Evaluation of the measurement model In order to evaluate the quality of the measurement model, several steps are required. Established evaluation routines can, e.g., be found in Hair et al. (2011, 2017). The following elaborations closely follow (Hair et al., 2017). Note that the required confirmation steps differ between reflective and formative constructs. Each confirmation step is carried out three times: first for the full sample (FULL, N = 334), second for the subsample of conventional farmers (CONV, N = 269), and third for subsample of organic farmers (ORGAN, N = 65). The reflective measures were evaluated in terms of the composite reliability, indicator reliability, convergent validity and discriminant validity. The respective statistics can be found in Tables 3 and 4.

4. Results and discussion The following subsection 4.1 presents the evaluation of the measurement model test the model specification. It is followed by an assessment of the results of the structural model (Subsection 4.2). The calculations were carried out using the software “SmartPLS 3”, Version 3.2.4 (Ringle et al., 2015). In order to test for possible differences

234

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

Table 3 Evaluation Statistics of the Measurement Model. λ/γ

IU IU1 IU2 IU3 IU4 POQ POQ1 POQ2 POQ3 PEOU PEOU1 PEOU2 PEOU3 PU PU1 PU2 PU3 PU4 SN SN1 SN2 SN3 UB UB1 UB1 UB3 OFC OFC1 OFC2 OFC3

AVE

FULL

CONV

ORGAN

0.904*** 0.696*** 0.785*** 0.832***

0.902*** 0.716*** 0.769*** 0.828***

0.832*** 0.003 0.695*** 0.836***

0.833*** 0.922*** 0.922***

0.826*** 0.917*** 0.911***

0.668*** 0.894*** 0.921***

0.789*** 0.869*** 0.866***

0.821*** 0.871*** 0.872***

0.498*** 0.855*** 0.779***

0.833*** 0.754*** 0.879*** 0.857***

0.831*** 0.765*** 0.880*** 0.859***

0.802*** 0.526*** 0.814*** 0.718***

0.794*** 0.834*** 0.554***

0.795*** 0.824*** 0.618***

0.474* 0.888*** 0.495*

0.546*** 0.282*** 0.364***

0.537*** 0.320*** 0.332***

0.454* 0.410* 0.567***

0.347*** 0.351*** 0.626***

0.403*** 0.122 0.738***

0.296 0.566*** 0.628***

CR

FULL

CONV

ORGAN

FULL

CONV

ORGAN

0.652

0.651

0.468

0.881

0.881

0.725

0.799

0.784

0.698

0.922

0.916

0.872

0.709

0.731

0.528

0.880

0.891

0.762

0.692

0.697

0.524

0.900

0.902

0.811

0.545

0.565

0.419

0.777

0.793

0.664

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

Source: Own elaboration; N(FULL) = 334, N(CONV) = 269, N(ORGAN) = 65; λ: loadings of reflective indicators, γ: weights of formative indicators, CR: composite reliability, AVE: average variance extracted; significance test based on bootstrapping with 5000 runs, *: 10%; **: 5%; ***: 1% significance level; the single itemconstruct AGE is not displayed.

model due to their impact on further model results. This also applies for the two subsamples.5 Additionally, a bootstrapping procedure was used to assess the significance of the indicator loadings. Each indicator is highly statistically significant in the full sample, and at least one time in the subsamples. The AVE is higher than 0.5 for the full sample and the conventional subsample in all cases, indicating that for each construct at least half of the variance is explained by the model (Hair et al., 2017). In ORGAN, the value is below the threshold for the constructs IU and SN. In order to evaluate the internal consistency reliability, it is suggested to calculate the CR. Here, all values are above the suggested threshold of 0.7 (Hair et al., 2011). In order to test whether the model establishes discriminant validity, the Heterotrait-Monotrait ratio of correlations (HTMT) (Henseler et al., 2015) was calculated. The results for all three groups can be found in Table 4. All values are below the suggested threshold of 0.9 (Henseler et al., 2015). Assessing the measurement model for formative constructs requires other steps. First the selected indicators need be reasoned by theory (see Section 3.2). Next, the selected indicators need to be checked for collinearity issues. For formative SEM-PLS models, variance inflation factors (VIF) above 5 are considered problematic (Hair et al., 2011). The highest observed VIF in all three samples was 2.151, which implies that there are no collinearity issues present. Finally, consideration of the significance and relevance of the indicators is required. The outer weights of all formative indicators are significant in the full sample (Table 3). In the two subsamples, one indicator weight is non-significant in each case. As the loadings of the respective indicators are

Table 4 Heterotrait-Monotrait Ratios. IU

POQ

PEOU

PU

SN

FULL IU POQ PEOU PU SN AGE

0.787 0.835 0.857 0.786 0.125

0.787 0.737 0.707 0.062

0.838 0.647 0.112

0.666 0.101

0.095

CONV IU POQ PEOU PU SN AGE

0.769 0.826 0.844 0.759 0.164

0.798 0.700 0.681 0.061

0.843 0.601 0.094

0.608 0.120

0.172

ORGAN IU POQ PEOU PU SN AGE

0.538 0.831 0.750 0.517 0.160

0.540 0.644 0.478 0.060

0.721 0.641 0.214

0.633 0.155

0.261

AGE

Source: Own elaboration; N(FULL) = 334, N(CONV) = 269, N(ORGAN) = 65.

Table 3 displays the indicator loadings (λ) for the reflective constructs, respectively the indicator weights (γ) for the formative constructs, the composite reliability (CR), and the average variance extracted (AVE). The indicator loadings are, with two exceptions, above the common threshold of 0.7 for the full sample (Hair et al., 2011). Following Hair et al. (2017), the indicators below the threshold are retained in the

5 In the organic subsample, one indicator (IU2) has a loading that would indicate an immediate removal from the model (for the organic subsample). As this only applies for one subsample, it is still retained in the model.

235

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

of a bootstrapping procedure. The significance tests of the differences between the conventional and the organic subsample are the result of the multi-group analysis procedure (MGA) introduced by Sarstedt et al. (2011).6 As assumed by the TAM, the usage behavior of grazing is positively influenced by the intention to use (H1). While this holds for all three samples, the MGA shows that the effect size in the conventional subsample is significantly larger than in the organic sample. Both the PEOU and the PU show a statistically significant influence on the IU. Again, the result holds for all three samples, supporting H2 and H3. The MGA indicates a significant difference in the effect size for H3. The fourth hypothesis derived from the TAM (H4) is also supported by the results. Here the effect size in the conventional sample is significantly larger than in the organic sample. The results for these hypotheses confirm the mains assumptions of the TAM (cf. Marangunić and Granić, 2015) in context of grazing usage. Given structural analysis and the broader setting, they further extend and underline previous results found in the literature (Kelly et al., 2015; McDonald et al., 2016). The on-farm conditions represented by the OFC shows a positive effect, both on the PEOU and UB (H5 and H6)7. Thus, the UB indeed depends on the conditions the farmer face on his or her farm. This effect is statistically significantly larger for the organic subsample than the conventional subsample. This is only case observed in the analysis, as the differences are either not significantly different, or larger for the conventional subsample. The SN norm influences the PU, the IU as well as the PEOU (H7, H8 and H9) in the expected way. The only exception is observed in the organic subsample. Here, the influence of the SN on the IU is not statistically significant. Additionally, the MGA indicates a significant difference between the conventional and the organic subsample for the effect on the IU. Regarding the practical implications, the role of SN should be noted. The results show that the farmer's perception of societal expectations significantly influences the farmer's PU, PEOU and IU. To some extent, this contradicts previous findings by Arens et al. (2012), who found that the SN does not have a significant effect in context of information systems in pig production. Still, the different results may be attributed to the different technologies, and their presence in public discourse. It is reasonable to assume that the farmer's SN itself is influenced by media coverage of grazing practices and related public communications. Thus, targeted information may influence the farmer's beliefs and intentions through this channel. H10 is supported for the full sample, as the perceived output quality achieved by grazing practices positively influences the PU. The construct AGE shows that an increasing age has a negative effect on the IU and the PU for the full sample. For the conventional subsample, an additional negative effect on the SN is present. This indicates that older conventional farmers perceive lower levels of societal norm and expectations towards grazing then younger ones. Still, as AGE does not show statistically significant effects on the beliefs and intention of the organic subsample, H11 is only supported partially. Nevertheless, the farmers' age has an influence on his or her beliefs and intentions towards grazing. If significant, the observed effects show a negative influence of the farmers age. The negative effect on the perceived usefulness of grazing leads to two interpretations. On the one hand, it could be argued that older (and therefore more experienced farmers) predominantly perceive the negative aspects of grazing. On the other hand,

Table 5 Quality Criteria for Endogenous Constructs. FULL R IU POQ PEOU PU SN UB

2

0.648 0.001 0.398 0.599 0.005 0.599

CONV Q

2

0.414 −0.003 0.273 0.402 −0.002 0.409

R

2

0.644 0.000 0.373 0.586 0.018 0.585

ORGAN Q

2

0.408 −0.007 0.264 0.391 0.008 0.397

R2

Q2

0.365 0.004 0.355 0.478 0.023 0.429

0.126 −0.006 0.142 0.184 −0.013 0.140

Source: Own elaboration; N(FULL) = 334, N(CONV) = 269, N(ORGAN) = 65.

larger than 0.5, they are retained in the model (Hair et al., 2017). Overall, the required results for the measurement model are met for the full sample. For the subsamples, the requirements are also generally met, with a few exceptions for the subsample of organic farms. This shows that the selected measures are suited for this analysis. Additionally, as the items of the formative measures represent the different dimensions of the overall construct domain, the indicator weights reported in Table 3 allow for further interpretation. For the UB, all indicator weights are significant, thus all indicators contribute substantially to the construct. For the full and the conventional subsample, the most important indicator is the number of grazing days per year (UB1). For organic farms, the access of the offspring (UB3) contributes the most to the UB. This implies a difference in the impact of the different dimensions on the UB between conventional and organic farms. While the differences in the construct for conventional farms are most importantly influenced by differences in the days of pasture access per year, the access of the offspring is most important for organic farms. The grazing area per cow (OFC3) contributes most to the OFC in all subsamples. The non-significance of the indicator weights of OFC1 and OFC2 in the organic and the conventional subsample respectively can be interpreted in the sense that the indicators are less important in comparison to the other (significant) indicators (Hair et al., 2017). Thus, the milk yield per cow does determine the OFC, but is relatively unimportant in the case of conventional farms. Similarly, the number of cows is relatively unimportant for organic farms.

4.2. Evaluation of the structural model Following the evaluation of the measurement model, this subsection presents the estimation results of the structural model. Table 5 shows the coefficients of determination (R2) (Hair et al., 2017) and the predictive relevance of the model (Stone-Geisser-Q2) (Geisser, 1974; Stone, 1974) for the endogenous constructs. For the endogenous constructs taken from the TAM (IU, PEOU, PU and UB) the R2 shows that the model explains between 35 and 65% of the observed variance for the full sample. For the remaining endogenous constructs (POQ, SN), depending only on the construct AGE and OFC, the R2 are close to zero. This is reflected in the estimated Q2-values. For the constructs taken from the TAM, they reveal a predictive relevance of the model for the constructs. For POQ and SN, the Q2-values are negative, indicating that the model lacks such relevance. The results for the conventional and the organic subsample are comparable with those for the full sample. Overall, the observed R2- and Q2-values are slightly smaller (for the organic more distinctly than for the conventional subsample). The only exception is observed for the Q2 of SN in the conventional subsample. Here the value is positive, showing that the model has a predictive relevance for the SN. Table 6 presents the estimated path coefficients for the relationships hypothesized by the model (compare Fig. 1). Table 6 contains the estimates for the full sample, as well as the conventional and the organic subsample. It further shows the differences of the estimations for the subsamples. The significance tests for the path coefficients are the result

6 Note that this does not allows a comparison of the absolute extent of the effect, e.g. on actual hours per day, as the indicators are standardized for the PLS algorithm, and the observed ranges differ between the conventional and organic subgroup (e.g. as the organic farmers are obliged to allow for pasture access). Thus it only allows for a comparison of the relative effect. 7 A reviewer suggested that the OFC may also function as moderator on the effect of IU on UB. When including the respective relationship in the model, we find the respect to be close to zero and statistically insignificant. We therefore do not include the relationship in to model examined in the paper. The remaining results are robust to the introduction of the moderating effect.

236

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

Table 6 Path Coefficients of the Structural Model. Hypothesis

H1 H2 H3 H4 H5 H6 H7 H8 H9 H10 H11

Path

FULL

IU → UB PEOU → IU PU → IU PEOU → PU OFC → UB OFC → PEOU SN → PU SN → IU SN → PEOU POQ → PU AGE → IU AGE → POQ AGE → PEOU AGE → PU AGE → SN

CONV

|Δ|a

ORGAN

β

HSb

β

HSb

β

HSb

0.655*** 0.318*** 0.380*** 0.486*** 0.203*** 0.438*** 0.149*** 0.239*** 0.332*** 0.254*** −0.081** 0.025 0.082* −0.132*** −0.071

Yes Yes Yes Yes Yes Yes Yes Yes

0.670*** 0.324*** 0.365*** 0.552*** 0.174*** 0.419*** 0.130** 0.248*** 0.337*** 0.179*** −0.091** 0.000 0.085 −0.140*** −0.133**

Yes Yes Yes Yes Yes Yes Yes Yes

0.411** 0.355*** 0.378*** 0.323*** 0.381*** 0.462*** 0.210* −0.094 0.317*** 0.364*** 0.030 0.064 0.095 −0.169 0.152

Yes Yes Yes Yes Yes Yes Yes No

Yes Partially

Yes Partially

Yes No

0.259** 0.031 0.013 0.230** 0.207** 0.043 0.080 0.342** 0.020 0.185* 0.121 0.064 0.010 0.028 0.285**

Source: Own elaboration; N(FULL) = 334, N(CONV) = 269, N(ORGAN) = 65; a: absolute difference between CONV and ORGAN; b HS: Hypothesis supported; *:10%, **: 5%, ***:1% significance level (based on bootstrapping results (5000 runs) for β, on PLS-MGA results for |Δ|).

R2-value of an endogenous construct if the specific exogenous construct is omitted from the construct (Hair et al., 2017). The results presented in the appendix show that in those cases where the MGA indicates significant differences between conventional and organic farms, differences (in terms of no, small, medium or large effect sizes) in the f2values are also present. Additionally, the quality criteria for the endogenous constructs show that the model has the highest predictive power if all farmers are jointly analyzed, followed by the independent analysis of the conventional subsample. The predictive relevance of the model is lowest for the organic subsample.

Table 7 f2-values for the structural model. Path

IU → UB PEOU → IU PU → IU PEOU → PU OFC → UB OFC → PEOU SN → PU SN → IU SN → PEOU POQ → PU AGE → IU AGE → POQ AGE → PEOU AGE → PU AGE → SN

FULL

CONV

ORGAN

f2

f2

f2

0.824 0.131 0.177 0.316 0.079 0.286 0.038 0.116 0.164 0.038 0.018 0.001 0.011 0.042 0.005

0.856 0.130 0.160 0.389 0.058 0.255 0.029 0.127 0.163 0.202 0.021 0.000 0.011 0.046 0.018

0.256 0.134 0.141 0.152 0.220 0.325 0.068 0.011 0.150 0.079 0.001 0.004 0.013 0.053 0.024

5. Conclusions Due to increasing societal interest in grazing as a form of pasture use, this paper analyzed possible behavioral drivers towards the adoption of grazing practices for German dairy farmers. Taking the TAM as a basis, an augmented structural equation model was developed. The model additionally takes the influence of the POQ achieved by grazing and the farmer's SN into account. It also controls for the effects of a farmer's age, as well as on-farm condition (OFC). The results show that the conceptualization of the adoption decision process, based on the TAM, leads to relevant insights regarding the adoption of grazing by German dairy farmers. With respect to the first research question, concerning potential influences on IU and UB, the results support previous findings regarding the importance of PU and PEOU with respect to the adoption of grazing practices specifically (Kelly et al., 2015; McDonald et al., 2016), and more generally, to technologies in agriculture (e.g. Adrian et al., 2005; Arens et al., 2012; Flett et al., 2004). The extensions applied in the model give further insights into the drivers of adoption decision. Considering the second research question (whether differences between conventional and organic farmers exist), the analysis revealed different effects of the OCF and the IU. Age was also found to only have significant negative effects on the behavioral constructs of conventional farmers. Since age can be seen as a proxy for experience, this implies that more experienced conventional farmers perceive grazing practices more negatively. Turning towards the implications beyond the perspective of an individual farm (research question three), the results for the subjective norm indicate that the farmer's beliefs and intentions are influenced by his or her perception of societal expectations. If these perceptions are shaped by the expectations communicated in the broader public and in political discourse, the farmer's beliefs and intentions may be influenced through these channels. If the preservation and extension of pastures and its grazing usage continue to be a policy goal, future policy

Source: Own elaboration; N(FULL) = 334, N(CONV) = 269, N(ORGAN) = 65; values of 0.02, 0.15 and 0.35 indicate small, medium and large effects of the exogenous on the endogenous construct, respectively (Hair et al., 2017).

the result could indicate that younger farmers are more open to this more traditional, but often little considered practice. The negative effect on the intention to use could potentially be explained by practical reasons, such as that successful grazing requires intensive management practices (Hodgson, 1990). With respect to the MGA results, the only statistically significant difference between the effects of AGE is present for the SN. Here the estimated coefficient for the conventional farmers is significantly larger than for the organic ones. As already pointed out, the MGA reveals statistically significant differences in the effects between the conventional and the organic subsamples, which supports H12. The results of the MGA are confirmed when applying the permutation test of the analysis of multiple groups (Sarstedt et al., 2011). Considering the two subsamples, the model overall performs better for the conventional subsample. The results of the MGA show that in those cases where a significant difference between the subsamples exists, the organic effect size is, with one exception, smaller. This indicates that the overall (relative) extent of grazing applied on conventional farms depends more on farmers' beliefs and intention and less on the on-farm conditions. This interpretation is also supported by a comparison of the calculated f2-values for the different subsamples (Table 7). The f2-value measures the change of the 237

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

measures should be thoughtfully designed with respect to information and communication strategies. This especially applies to conventional farmers who may be the more responsive, thus the more important targets of such measures. As the general results of the present study are in line with the existing literature on the TAM, they are likely to be transferable to other countries. Still, readers should be cautious when transferring the more specific results of the study. They may only be transferred to countries with similar conditions of production (e.g. other western European countries). Also, unaccounted cultural differences may limit the transferability of the results. The apparent limitation of the study is the way the construct OFC is developed. As already discussed, the creation of formative constructs requires cautious reasoning. Without a doubt, a construct that is applied in the way of the present study cannot fully capture the complex realities of a farm. While the creation of higher order constructs is also certainly possible, a high differentiation of construct domains would imply a high number of binary indicators, in order to map different management tools. Such an approach is not feasible, as the introduction of categorical indicators requires particular caution, and leads to problems (Hair et al., 2012, 2017). In order to avoid endogeneity problems, it also has to be evaluated whether a possible indicator is a long-term on-farm condition, and not an action which follows from the adoption of grazing. Still, construction of the OFC (and similarly the UB) allows for an improvement in future research. It should be carefully examined, which continuous indicators could be used to improve the coverage of the overall construct domains and to better account for potential usage barriers. A possible further extension of the model could be differentiation by grazing system (e.g rotational and permanent stocking systems), and an estimation of suitable separate models. Further research could also explicitly consider non-continuous usage barriers, like the absence of pastures suitable for grazing, or climate conditions. In such research, complementary statistical approaches like multivariate regression techniques may be used, also to control for the potential endogeneity issues. Considering the influence of SN, future research could also assess the causal effect of potential public actions. Such a causal-effect-framework would require panel data. While this would raise additional demands for data collection and analysis, it could also be fruitful for aspects such as the improvement of agricultural policy communication.

O., Grieve, R., Sutherland, A., 1998. Empirical models of farmer behaviour using psychological, social and economic variables. Part I: linear modelling. Agric. Syst. 58, 203–224. Bijttebier, J., Hamerlinck, J., Moakes, S., Scollan, N., Van Meensel, J., Lauwers, L., 2017. Low-input dairy farming in Europe: exploring a context-specific notion. Agric. Syst. 156, 43–51. http://dx.doi.org/10.1016/j.agsy.2017.05.016. Blanchet, K., Moechnig, H., Dejong-Hughes, J., 2000. Grazing Systems Planning Guide. University of Minnesota, Extension Service. Bundesanstalt für Landwirtschaft und Ernährung (BLE), 2016. Milchwirtschaft auf einen Blick in Deutschland nach Kalenderjahren. Bonn. Davis, F.D., 1986. A Technology Acceptance Model for Empirically Testing New End-User Information Systems: Theory and Results. Massachusetts Institute of Technology. Davis, F.D., 1989. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Q. 13, 319–340. Davis, F.D., Bagozzi, R.P., Warshaw, P.R., 1989. User acceptance of computer technology: a comparison of two theoretical models. Manag. Sci. 35, 982–1003. Destatis, 2017. Bodennutzung der Betriebe (Landwirtschaftlich genutzte Flächen) (No. Fachserie 3 Reihe 3.1.2), Land- und Forstwirtschaft, Fischerei. Wiesbaden. Diamantopoulos, A., Riefler, P., 2008. Formative Indikatoren: Einige Anmerkungen zu ihrer Art, Validität und Multikollinearität. Z. Betriebswirt. 78, 1183–1196. http://dx. doi.org/10.1007/s11573-008-0099-7. Diamantopoulos, A., Winklhofer, H.M., 2001. Index construction with formative indicators: an alternative to scale development. J. Mark. Res. 38, 269–277. http://dx. doi.org/10.1509/jmkr.38.2.269.18845. Egri, C.P., 1999. Attitudes, backgrounds and information preferences of Canadian organic and conventional farmers: implications for organic farming advocacy and extension. J. Sustain. Agric. 13, 45–72. http://dx.doi.org/10.1300/J064v13n03_05. Ellis, K., Billington, K., McNeil, B., McKeegan, D., 2009. Public opinion on UK milk marketing and dairy cow welfare. Anim. Welf. 18, 267–282. Fahlbusch, M., Bahr, A., Brummer, B., Spiller, A., 2009. Der Markt für Milch und Milcherzeugnisse. Agrarwirtschaft 58, 52–71. Fairweather, J.R., 1999. Understanding how farmers choose between organic and conventional production: results from New Zealand and policy implications. Agric. Hum. Values 16, 51–63. http://dx.doi.org/10.1023/A:1007522819471. Fishbein, M., Ajzen, I., 1975. Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research. Addison-Wesley, Reading, MA. Flett, R., Alpass, F., Humphries, S., Massey, C., Morriss, S., Long, N., 2004. The technology acceptance model and use of technology in New Zealand dairy farming. Agric. Syst. 80, 199–211. http://dx.doi.org/10.1016/j.agsy.2003.08.002. Geisser, S., 1974. A predictive approach to the random effect model. Biometrika 61, 101–107. http://dx.doi.org/10.1093/biomet/61.1.101. Grünlandzentrum, 2011. Charta Weideland Norddeutschland. Ovelgönne. Gurrath, P., 2011. Landwirtschaft auf einem Blick. Statistisches Bundesamt, Wiesbaden. Hair, J.F., Ringle, C.M., Sarstedt, M., 2011. PLS-SEM: indeed a silver bullet. J. Mark. Theory Pract. 19, 139–152. http://dx.doi.org/10.2753/MTP1069-6679190202. Hair, J., Sarstedt, M., Ringle, C., Mena, J., 2012. An assessment of the use of partial least squares structural equation modeling in marketing research. J. Acad. Mark. Sci. 40, 414–433. http://dx.doi.org/10.1007/s11747-011-0261-6. Hair, J.F., Hult, G.T.M., Ringle, C., Sarstedt, M., 2017. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM), 2nd ed. Sage Publications. Hellberg-Bahr, A., Steffen, N., Spiller, A., 2012. Marketingpotentiale für Weidemilch. Jahrbuch der Österreichischen Gesellschaft für Agrarökonomie 21, 3–12. Hennessy, D., Delaby, L., Van den Pol-Van Dasselaar, A., Shalloo, L., 2015. Possibilities and constraints for grazing in high output dairy systems. Grassl. Sci. Eur. 20, 151–162. Henseler, J., Ringle, C.M., Sarstedt, M., 2015. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J. Acad. Mark. Sci. 43, 115–135. http://dx.doi.org/10.1007/s11747-014-0403-8. Hodgson, J.G., 1990. Grazing Management: Science into Practice, Longman Handbooks in Agriculture. Longman Scientific & Technical, Harlow. Jarvis, C.B., MacKenzie, S.B., Podsakoff, P.M., 2003. A critical review of construct indicators and measurement model misspecification in marketing and consumer research. J. Consum. Res. 30, 199–218. http://dx.doi.org/10.1086/376806. Kelly, E., Heanue, K., Buckley, C., O'Gorman, C., 2015. Proven Science Versus Farmer Perception (No. 229067), 2015 Conference of the International Association of Agricultural Economists, August 9–14. Milan, Italy, pp. 2015. King, W.R., He, J., 2006. A meta-analysis of the technology acceptance model. Inf. Manag. 43, 740–755. http://dx.doi.org/10.1016/j.im.2006.05.003. Knaus, W., 2016. Perspectives on pasture versus indoor feeding of dairy cows. J. Sci. Food Agric. 96, 9–17. http://dx.doi.org/10.1002/jsfa.7273. Kühl, S., Fahlbusch, M., Ermann, M., Bruemmer, B., Schulze-Ehlers, B., 2016. Der Markt für Milch und Milcherzeugnisse im Jahr 2015. German J. Agricul. Econ. 65, 18–39. Lassen, B., Nieburg, H., Kuhnert, H., Sanders, J., 2014. Status-quo-Analyse Ausgewählter Nachhaltigkeitsaspekte der Milcherzeugung in Niedersachsen (No. 28), Thünen Working Paper. Thünen-Institut für Betriebswirtschaft, Braunschweig, Germany. Lassen, B., Nieberg, H., Kuhnert, H., 2015. Status quo-Analyse ausgewählter Nachhaltigkeitsaspekte der Milcherzeugung in Schleswig-Holstein (No. 48), Thünen Working Paper. Johann Heinrich von Thünen-Institut, Braunschweig, Germany. Lunneryd, D., Öhlmér, B., 2009. The influence of values on strategic choices: the choice of organic milk production by Swedish farmers. Acta Agric. Scand. Sec. C — Food Econ. 6, 1–20. http://dx.doi.org/10.1080/16507540903178415. Marangunić, N., Granić, A., 2015. Technology acceptance model: a literature review from 1986 to 2013. Univ. Access Inf. Soc. 14, 81–95. http://dx.doi.org/10.1007/s10209014-0348-1. McConnell, D.J., Dillon, J.L., 1997. Farm Management for Asia: A Systems Approach. In: FAO Farm Systems Management Series. FAO, Rome.

Acknowledgements We thank anonymous reviewers and the editor for helpful advice and comments on an earlier version of this paper. The study was supported by the Ministry for Science and Culture of Lower Saxony (MWK) within the collaborative research project SAM, Analysis of Dairy Production: Grazing versus Indoor Housing of Dairy Cows, Support Code: ZN 2864. References Adrian, A.M., Norwood, S.H., Mask, P.L., 2005. Producers' perceptions and attitudes toward precision agriculture technologies. Comput. Electron. Agric. 48, 256–271. http://dx.doi.org/10.1016/j.compag.2005.04.004. Ajzen, I., 1985. From intentions to actions: A theory of planned behavior. In: Kuhl, P.D.J., Beckmann, D.J. (Eds.), Action Control, SSSP Springer Series in Social Psychology. Springer Berlin Heidelberg, pp. 11–39. http://dx.doi.org/10.1007/978-3-642-697463_2. Ajzen, I., Fishbein, M., 1980. Understanding Attitudes and Predicting Social Behavior. Prentice-Hall, Englewood Cliffs, NJ. Arbuckle, J.G., Morton, L.W., Hobbs, J., 2013. Farmer beliefs and concerns about climate change and attitudes toward adaptation and mitigation: evidence from Iowa. Clim. Chang. 118, 551–563. http://dx.doi.org/10.1007/s10584-013-0700-0. Arens, L., Plumeyer, C.-H., Theuvsen, L., 2012. Akzeptanz von Informationssystemen durch Schweinemäster: Eine Kausalanalyse. In: Balmann, A. (Ed.), Unternehmerische Landwirtschaft Zwischen Marktanforderungen Und Gesellschaftlichen Erwartungen, Schriften Der Gesellschaft Für Wirtschafts- Und Sozialwissenschaften Des Landbaus e.V. Landwirtschaftsverlag Münster, Münster, pp. 289–300. Austin, E.J., Willock, J., Deary, I.J., Gibson, G.J., Dent, J.B., Edwards-Jones, G., Morgan,

238

Agricultural Systems 165 (2018) 230–239

H. Schaak, O. Mußhoff

Serra, T., Zilberman, D., Gil, J.M., 2008. Differential uncertainties and risk attitudes between conventional and organic producers: the case of Spanish arable crop farmers. Agric. Econ. 39, 219–229. http://dx.doi.org/10.1111/j.1574-0862.2008.00329.x. Shen, D., Laffey, J., Lin, Y., Huang, X., 2006. Social influence for perceived usefulness and ease-of-use of course delivery systems. J. Interact. Online Learn. 5, 270–282. Stone, M., 1974. Cross-validatory choice and assessment of statistical predictions. J. R. Stat. Soc. Ser. B Methodol. 36, 111–147. Sullivan, S., McCann, E., Young, R.D., Erickson, D., 1996. Farmers' attitudes about farming and the environment: a survey of conventional and organic farmers. J. Agric. Environ. Ethics 9, 123–143. http://dx.doi.org/10.1007/BF03055298. Tauer, L., 1995. Age and farmer productivity. Rev. Agric. Econ. 17, 63–69. http://dx.doi. org/10.2307/1349655. Thomet, P., Cutullic, E., Bisig, W., Wuest, C., Elsaesser, M., Steinberger, S., Steinwidder, A., 2011. Merits of full grazing systems as a sustainable and efficient milk production strategy. Grassl. Sci. Eur. 16, 273–285. Vanslembrouck, I., Huylenbroeck, G., Verbeke, W., 2002. Determinants of the willingness of Belgian farmers to participate in Agri-environmental measures. J. Agric. Econ. 53, 489–511. Venkatesh, V., Davis, F.D., 2000. A theoretical extension of the technology acceptance model: four longitudinal field studies. Manag. Sci. 46, 186–204. http://dx.doi.org/ 10.1287/mnsc.46.2.186.11926. Venkatesh, V., Morris, M.G., Davis, G.B., Davis, F.D., 2003. User acceptance of information technology: toward a unified view. MIS Q. 27, 425–478. Venkatesh, V., Davis, F., Morris, M.G., 2007. Dead or alive? The development, trajectory and future of technology adoption research. J. Assoc. Inf. Syst. 8, 1. von Keyserlingk, M.A.G., Rushen, J., de Passillé, A.M., Weary, D.M., 2009. Invited review: the welfare of dairy cattle—key concepts and the role of science. J. Dairy Sci. 92, 4101–4111. http://dx.doi.org/10.3168/jds.2009-2326. Weinrich, R., Kühl, S., Zühlsdorf, A., Spiller, A., 2014. Consumer attitudes in Germany towards different dairy housing systems and their implications for the marketing of pasture raised milk. Int. Food Agribusiness Manag. Rev. 17, 205. Wheeler, S.A., 2008. What influences agricultural professionals' views towards organic agriculture? Ecol. Econ. 65, 145–154. http://dx.doi.org/10.1016/j.ecolecon.2007. 05.014. Willock, J., Deary, I.J., Edwards-Jones, G., Gibson, G.J., McGregor, M.J., Sutherland, A., Dent, J.B., Morgan, O., Grieve, R., 1999. The role of attitudes and objectives in farmer decision making: business and environmentally-oriented behaviour in Scotland. J. Agric. Econ. 50, 286–303. http://dx.doi.org/10.1111/j.1477-9552.1999.tb00814.x.

McDonald, R., Heanue, K., Pierce, K., Horan, B., 2016. Factors influencing new entrant dairy Farmer's decision-making process around technology adoption. J. Agric. Educ. Ext. 22, 163–177. http://dx.doi.org/10.1080/1389224X.2015.1026364. Mzoughi, N., 2011. Farmers adoption of integrated crop protection and organic farming: do moral and social concerns matter? Ecol. Econ. 70, 1536–1545. http://dx.doi.org/ 10.1016/j.ecolecon.2011.03.016. Mzoughi, N., 2014. Do organic farmers feel happier than conventional ones? An exploratory analysis. Ecol. Econ. 103, 38–43. http://dx.doi.org/10.1016/j.ecolecon. 2014.04.015. Ondersteijn, C.J.M., Giesen, G.W.J., Huirne, R.B.M., 2003. Identification of farmer characteristics and farm strategies explaining changes in environmental management and environmental and economic performance of dairy farms. Agric. Syst. 78, 31–55. http://dx.doi.org/10.1016/S0308-521X(03)00031-3. Peyraud, J., Van den Pol-Van Dasselaar, A., Dillon, P., Delaby, L., 2010. Producing milk from grazing to reconcile economic and environmental performances. Grassl. Sci. Eur. 15, 865–879. Plachter, H., Hampicke, U. (Eds.), 2010. Large-scale Livestock Grazing. Springer Berlin Heidelberg, Berlin, Heidelberg. Power, E.F., Kelly, D.L., Stout, J.C., 2013. Impacts of organic and conventional dairy farmer attitude, behaviour and knowledge on farm biodiversity in Ireland. J. Nat. Conserv. 21, 272–278. http://dx.doi.org/10.1016/j.jnc.2013.02.002. Reijs, J.W., Daatselaar, C.H.G., Helming, J.F.M., Jager, J., Beldman, A.C.G., 2013. Grazing dairy cows in north-West Europe. LEI Rep. 2013, 1. Rezaei-Moghaddam, K., Salehi, S., 2010. Agricultural specialists intention toward precision agriculture technologies: integrating innovation characteristics to technology acceptance model. Afr. J. Agric. Res. 5, 1191–1199. http://dx.doi.org/10.5897/ AJAR09.506. Ringle, C.M., Wende, S., Becker, J.-M., 2015. SmartPLS 3. SmartPLS GmbH. Rossiter, J.R., 2002. The C-OAR-SE procedure for scale development in marketing. Int. J. Res. Mark. 19, 305–335. http://dx.doi.org/10.1016/S0167-8116(02)00097-6. Roy, S., Tarafdar, M., Ragu-Nathan, T.S., Marsillac, E., 2012. The effect of misspecification of reflective and formative constructs in operations and manufacturing management research. Electron. J. Bus. Res. Methods 10, 34–52. Sarstedt, M., Henseler, J., Ringle, C.M., 2011. Multigroup analysis in partial least squares (PLS) path modeling: Alternative methods and empirical results. In: Sarstedt, M., Schwaiger, M., Taylor, C.R. (Eds.), Measurement and Research Methods in International Marketing, Advances in International Marketing. Emerald Group Publishing Limited, pp. 195–218.

239