1 Digital Technology Adoption Among Italian Farmers: An Extended Technology 1 Acceptance Model Approach in the Horticultural Sector 2 Elena Cozzi1, e-mail: elena.cozzi@unipr.it 3 Davide Menozzi1*, ORCID: 0000-0002-5241-1587, e-mail: davide.menozzi@unipr.it 4 Giulia Maesano2, ORCID: 0000-0001-9622-1067, email: giulia.maesano2@unibo.it 5 Maurizio Canavari2, ORCID: 0000-0003-0573-7880, email: maurizio.canavari@unibo.it 6 Cristina Mora1, ORCID: 0000-0003-3180-5496, e-mail: cristina.mora@unipr.it 7 1 Department of Food and Drug, University of Parma, Parco Area delle Scienze 47/A, 43124 8 Parma, Italy 9 2 Department of Agricultural and Food Sciences, Alma Mater Studiorum, Università di 10 Bologna, Via Zamboni, 33, 40126 Bologna, Italy 11 * Corresponding author 12 13 This article has been accepted for publication and undergone full peer review but has not been 14 through the copyediting, typesetting, pagination and proofreading process, which may lead to 15 differences between this version and the Version of Record. 16 Please cite this article as: 17 Cozzi E, Menozzi D, Maesano G, Canavari M, Mora C (2025). Digital Technology Adoption 18 Among Italian Farmers: An Extended Technology Acceptance Model Approach in the 19 Horticultural Sector, Bio-Based and Applied Economics, Just Accepted. DOI: 10.36253/bae-20 17364 21 Highlights 22 • A structured survey conducted with 251 Italian horticultural farmers 23 • The extended TAM3 explains 18% of the variance in the behaviour (the adoption of 24 water-smart technologies), and 65% of the variance in intention 25 • Behavioural intention is a significant predictor of the behaviour 26 • Perceived usefulness and social norms have a significant effect on adoption intention 27 • Perceived ease of use has no influence on adoption intentions 28 Abstract 29 The adoption of digital technologies in agriculture is essential for enhancing sustainability, 30 productivity, and resource efficiency. This study investigates the factors influencing Italian 31 mailto:elena.cozzi@unipr.it mailto:davide.menozzi@unipr.it mailto:giulia.maesano2@unibo.it mailto:maurizio.canavari@unibo.it mailto:cristina.mora@unipr.it 2 horticultural farmers' adoption of innovative water-smart agricultural technologies using an 1 extended Technology Acceptance Model (TAM3). The research employs a structured survey 2 conducted with 251 Italian farmers, analysing their perceptions of technology usefulness, ease 3 of use, social norms, and sustainability outcomes. Structural equation modelling (SEM) 4 confirms that perceived usefulness significantly influences adoption intentions, while perceived 5 ease of use plays a limited role. Social norms and sustainability-related benefits also emerge as 6 critical determinants. Results also indicate the impact of farm size and workforce on adoption 7 behaviour. These findings highlight the need for targeted policies, training programs, and 8 financial incentives to overcome adoption barriers. The study provides insights for 9 policymakers, technology developers, and agricultural stakeholders to foster digital innovation 10 in the horticultural sector, contributing to sustainable water management practices. 11 Keywords: Digital Agriculture; Farmer adoption; Technology Acceptance Model (TAM); 12 Horticultural Sector; Water-Smart Sustainable Farming. 13 14 3 1. Introduction 1 The agricultural sector is facing many unprecedented challenges. Given the increasing pressure 2 on agricultural systems, effective measures are needed to reduce agriculture's impact on natural 3 resources, in line with the European Green Deal and the United Nations 2030 Agenda 4 (Montanarella and Panagos, 2021). In this context, digital technologies and smart solutions are 5 key to improving efficiency, productivity, and sustainability in agriculture (Yigezu et al., 2018), 6 with irrigation practices playing a particularly important role (Asadi et al., 2020). Water scarcity 7 and drought are now recognised as a global problem of the utmost importance (Ermolieva et 8 al., 2022; Ungureanu et al., 2020), exacerbated by climate change (Kapsdorferová, 2024). Smart 9 water technologies offer a double benefit: help to save water and use it more efficiently; and 10 reduce operating costs and improve productivity by maximising output per unit of water 11 consumed (Gemtou et al., 2024). Innovations such as soil moisture sensors, automatic irrigation 12 systems and predictive modelling can effectively address the challenges of water scarcity and 13 climate variability (Adeyemi et al., 2017), as well as energy savings (Patle et al., 2019). A major 14 challenge in smallholder agriculture is the limited adoption of innovative technologies, 15 resulting in low overall technology penetration in the sector (Senyolo et al., 2018). 16 Indeed, despite efforts to promote a fair transition to digital agriculture, the adoption of smart 17 technologies remains uneven, shaped by a complex mix of individual, technological, and 18 contextual factors (Shang et al., 2021). Previous studies have shown that there are significant 19 differences in adoption rates among farmers (Paustian and Theuvsen, 2017). Farmers' decision-20 making processes is shaped by perceived benefits, ease of use and external pressures, making 21 4 it a key factor in understanding the adoption landscape (Cimino et al., 2024; Schulze Schwering 1 et al., 2022). Given the limited technological penetration of the agricultural sector and the 2 potential benefits offered by digital technologies, it is crucial to investigate the farmers' 3 behaviour, their willingness to adopt smart solutions, the determinants of smart technology 4 adoption and identifying strategies to promote broader adoption of water technologies (Gemtou 5 et al., 2024). 6 Although prior research has examined farmers' adoption patterns, it has often focused on large-7 scale farms or specific regions with advanced technological infrastructures (Paustian and 8 Theuvsen, 2017). Additionally, studies have highlighted barriers such as limited digital literacy, 9 financial constraints, and a lack of institutional support for small and medium-sized farms 10 (Senyolo et al., 2018; Shang et al., 2021). Despite the increasing research on technology 11 adoption in agriculture, there are still notable gaps. Firstly, little attention has been paid to the 12 adoption of water-saving technologies in the horticultural sector, which is crucial for 13 sustainable agriculture, in contrast to the more commonly studied large-scale cereal production 14 (Adeyemi et al., 2017). Secondly, while factors like such as farm size and socio-demographic 15 characteristics have been investigated, the influence of sustainability motivations and social 16 norms on technology adoption has not yet been sufficiently researched (Gemtou et al., 2024). 17 Finally, the adoption of digital technologies in Italian agriculture, which is characterised by 18 fragmented land ownership, regional diversity and varying levels of technological readiness, 19 has been insufficiently studied (Baldoni et al., 2018). 20 5 This study aims to fill these gaps by analysing the factors that influence Italian farmers in 1 adopting digital technologies for better water management and the barriers they face, with a 2 focus on horticultural crops. Horticulture has been considered for some reasons: first, because 3 of the importance of this sector in the Italian agricultural system; secondly, for the relevance of 4 the irrigation in this cropping system (Patle et al., 2019); third because of the relevance of smart 5 precision in horticulture (Adeyemi et al., 2017). The technologies studied relate to smart water 6 management through a three-stage technology complexity: the first (basic) stage is represented 7 by the introduction of soil moisture sensors, which proceeds to a system that combines sensors 8 with an automatic irrigation system, and in the last stage the sensors are connected to an 9 automated system, which in turn is connected to and dialogs with predictive models1. 10 Understanding farmers' perceptions and adoption of water-efficient innovations is essential for 11 shaping targeted policies and effective incentives, offering valuable insights for policymakers 12 and stakeholders aiming to promote sustainable agriculture. 13 14 2. Literature review and theoretical background 15 As the existing literature shows, the process of adopting new technologies is inherently complex 16 and dynamic (Montes de Oca Munguia et al., 2021). In particular, the decision-making process 17 is influenced by various factors that affect the rate of technology adoption by farmers (Gemtou 18 1 Specifically, automatic irrigation systems are connected to sensors that monitor soil moisture and activate valves wirelessly; instead, predictive modelling integrates the first two solutions (soil moisture sensors and automatic irrigation systems) into predictive models that merge real-time data with historical data, analyse it, and make autonomous irrigation decisions thanks to water delivery schedules that optimize dosing based on specific crop requirements and environmental conditions. 6 et al., 2024; Osrof et al., 2023). Although the existing literature has explored the mechanisms 1 of innovation diffusion, there does not seem to be a unified set of theories or models that could 2 explain the phenomenon. Some authors have highlighted the specificity of theories in modelling 3 different aspects of the technology adoption process (Dissanayake et al., 2022; Osrof et al., 4 2023), while others have expressed doubts about the generalist ability of theories to represent 5 different technologies and practices (Montes de Oca Munguia et al., 2021). Indeed, there is still 6 confusion about the methods of analysis and the choice of explanatory variables that should be 7 used to model the adoption process (de Oca Munguia and Llewellyn, 2020). To illustrate, Shang 8 et al. (2021) argue that the mechanisms of adoption and diffusion of digital agricultural 9 technologies need to be understood at both the farm level and the system level. They also 10 suggest that the focus in determining technology diffusion should be on system interactions in 11 combination with individual characteristics. Given the evidence presented in the literature, it 12 can be assumed that the categories of individual, technological, social and economic factors 13 influencing technology adoption can describe the entire decision-making process (Dissanayake 14 et al., 2022). There is a clear lack of convergence and consistency in the results regarding the 15 impact and statistical significance of the individual factors assessed in the adoption models (de 16 Oca Munguia and Llewellyn, 2020). This discrepancy can be attributed to the fact that most 17 adoption studies do not include variables on technologies or practices. It is recognized that the 18 use of multiple paradigms in modelling technology adoption and diffusion can increase the 19 explanatory power of the models. However, it is important to consider the factors and their 20 7 interactions in a way that is consistent with the objectives and context of the study within a 1 specific food system (Dentoni et al., 2023). 2 Literature offers a large range of theoretical frames exploring the intention to adopt; among 3 others the Theory of Planned Behaviour (TPB) (Ajzen, 1991; Fishbein and Ajzen, 2010), the 4 Social Cognitive Theory (Bandura, 1986), and the Unified Theory of Acceptance and Use of 5 Technology (UTAUT) and its further extensions (Venkatesh et al., 2003). In the present work 6 we applied the Technology Acceptance Model (TAM) (Davis, 1989) for measuring the 7 intention of Italian farmers to adopt innovative smart technologies. The TAM model has been 8 chosen for designing the model since this theory was precisely conceived focusing on 9 technological solutions, hereby offering the ideal setting to place the research. The other 10 theories might be adapted to a technology frame, whereas the TAM, being already tailored on 11 the observed object of the analysis, better suits the purpose of the current investigation. 12 According to the TAM paradigm, two dispositions towards a new technology (perceived 13 usefulness and ease of use) determine a person's attitude towards using that technology and 14 influence their desire to use it. Perceived usefulness is the extent to which a person believes 15 that job performance can be enhanced by using the new technology, whereas perceived ease of 16 use is the extent to which a person believes that using the new technology is effortless. Some 17 extensions of the original TAM conceptualization have been proposed, such as the TAM3 18 version (Venkatesh and Bala, 2008). The TAM3 extension introduces new constructs and 19 determinants that affect the core variable perceived ease of use and proposes new relationships 20 between the constructs. The factors influencing perceived ease of use in the TAM3 version are 21 8 computer self-efficacy, perception of external control, computer anxiety, computer playfulness, 1 perceived enjoyment, and objective usability, whereas perceived usefulness is affected by 2 subjective norm, image, relevance to work, output quality and demonstrability of results. Other 3 innovations introduced by this extension include: (i) the correlation between perceived ease of 4 use and perceived usefulness, (ii) the correlation between perceived ease of use and intention, 5 and (iii) the concept of anxiety. The latter factor, which expresses the degree of emotional fear, 6 apprehension, nervousness, or stress experienced when interacting with a new technology, is 7 supposed to negatively affect the perceived ease of use. The more anxiety a person feels, the 8 less likely they are to perceive the technology as easy to use. 9 The TAM3 conceptualization was adopted since flexible adjustments to context specific needs 10 compared to the original TAM version are possible. Indeed, Venkatesh and Bala (2008) added 11 some constructs (i.e. experience, subjective norm, and image) which resulted to be key factors 12 in the exploratory phase: farmers decision rely on peer recommendation as well as on previous 13 experience on digital tools. Therefore, only an extended version of the very first 14 conceptualization could confirm or not these clues. Furthermore, in order to comprise some 15 small adjustments to the basic model (like the sustainability dimension), an extended version 16 with more variables seemed to be more appropriate than the earliest edition. Minor adjustments 17 were made to the original TAM3 version by Venkatesh and Bala (2008) to better suit the 18 purpose and context of the analysis. First, all constructs were considered in the context-specific 19 environment, i.e. the adoption of new water-smart agricultural technologies by Italian 20 horticultural farms. Moreover, some aspects were evaluated as very important and emerged 21 9 explicitly from the exploratory phase with the participants, such as technology self-efficacy and 1 quality of outcomes. Other TAM3 characteristics, such as (computer) playfulness or perceived 2 enjoyment, do not apply to the context of the current research and were therefore excluded from 3 the model design. Then, the need to include sustainability dimensions when dealing with digital 4 technologies in the agri-food sector is pressing (Alexandrova-Stefanova et al., 2024; Gemtou 5 et al., 2024; Khanna, 2021; McFadden, 2022). Consequently, a category based on with the 6 Sustainability Assessment of Food and Agriculture Systems (SAFA) aspects (FAO, 2014) was 7 included in the model since this method represents a comprehensive and international reference 8 for assessing sustainability along agri-food value chains. More specifically, the themes inspired 9 by the SAFA indicators, were (i) the reduced water-used thanks to the optimization of the 10 irrigation system, (ii) the improved skills the employees and the holder/farmer need to reach to 11 use the technology, and (iii) new employees recruited thanks to their technological skills. 12 Therefore, we tested the following main hypotheses on the factors influencing the adoption of 13 new water-smart agricultural technologies by Italian horticultural farms (Figure 1): 14 H1: perceived usefulness is positively affected by output quality (H1a), by sustainability 15 outcomes measured by SAFA indicators (H1b), and by subjective norms (H1c); 16 H2: perceived ease of use is positively affected by technology self-efficacy (H2a), and is 17 negatively affected by anxiety (H2b); 18 H3: perceived ease of use has a positive impact on farmers' intention to adopt new technologies 19 (H3a), and is positively affecting the perceived usefulness of new technologies (H3b); 20 H4: perceived usefulness has a positive impact on farmers' intention to adopt new technologies; 21 10 H5: subjective norms have a positive impact on farmers' intention to adopt new technologies; 1 H6: the farmers' intention to adopt new technologies is positively affecting the behaviour, i.e. 2 the new technology adoption. 3 4 Figure 1. Model testing the factors affecting the adoption of water-smart agricultural new 5 technologies by Italian horticulture farms. 6 7 8 Moreover, individual factors, such as socio-demographic and organizational characteristics, 9 which determine the natural and structural conditions of the farm, have been found to correlate 10 with farmers’ decisions. In particular, farmers’ education level, gender, age, technology 11 literacy, were among the individual drivers more frequently included in studies investigating 12 the smart farming technologies adoption (Osrof et al., 2023). Farmers differentiate for their 13 level of human capital (i.e. experience, education, and other relevant training) (Huffman, 2001; 14 Okoye et al., 2025). These individual-level differences shape their technology adoption: the 15 11 higher the human capital, the more likely the technologies are adopted (McFadden, 2022). Farm 1 size, mostly expressed in total acreage farmland, is a prominent factor among the organizational 2 ones, since larger farm size is consistently seen as pivotal for achieving economies of scale 3 when adopting smart farming technologies (Okoye et al., 2025). Farm income is another key 4 element, as farmers with a higher income are more willing to invest in new technologies (Osrof 5 et al., 2023). Farm location is also a notable barrier within this theme, showing mixed effects 6 in past studies. Some research indicates that it might negatively affects farmers’ motivation to 7 adopt smart technology, particularly if farms face unfavourable climate conditions or soil 8 quality (Paxton et al., 2011). In Italy, farms in the northern regions are generally more 9 competitive, due to larger farm sizes, advanced mechanisation, and stronger market integration. 10 In contrast, farms located in central and southern regions often face structural constraints, 11 including smaller farms and lower productivity (Baldoni et al., 2018; Okoye et al., 2025). Other 12 studies emphasize the importance of social factors and access to information for the adopting 13 of innovative smart technologies (Blasch et al., 2022). In this context, being a member of a 14 farmers’ associations or a producer organizations (POs), where knowledge transfer is one of 15 the main objectives, might facilitate adoption (Okoye et al., 2025). Therefore, we controlled the 16 main endogenous variables of the model, i.e. perceived usefulness, adoption intention and 17 behaviour, with individual factors, namely farmers’ age, education level and years of 18 experience in the agricultural sector, and organizational ones, including farm size, farm location 19 (expressed by the latitude of the province where the farm is located), number of employees, and 20 membership in a cooperative or a producer organization (Figure 1). 21 12 1 3. Data and Method 2 3.1. Data collection 3 The data collection consisted in two phases: first we conducted a preliminary exploratory 4 phase with qualitative, unstructured interviews. We interviewed 10 stakeholders, including 5 agronomists, technicians from Producer Organizations (POs), representatives of the Tomato 6 Interbranch Organization (OI Pomodoro Nord Italia), farmers, and vegetable/tomato 7 producers. The aim of the exploratory interviews was to identify relevant aspects to be 8 included in the final model and to highlight those that could be omitted. In this way, relevant 9 points such as the output quality and self-efficacy were included in the final surveys. The 10 questions focused on previous experience with smart technologies, skills in using them, public 11 financial support for the adoption of technical solutions and the farm structure, as well as 12 farmers’ previous personal background. In the second phase, we conducted a survey among a 13 sample of Italian horticultural farms. An initial pilot phase (n=21 interviews) was carried out 14 to test the questionnaire. In this pilot phase the model was tested on 21 farms equally 15 distributed in the country (10 in Centre and Southern Italy, 11 in the Northern regions). Most 16 of them cultivate tomatoes, while a smaller group (n=4) grows only other fresh vegetables, 17 such as pumpkin, celery, fell, broccoli, etc. The majority of the farms interviewed do not use 18 any kind of technology; only a few rely on autonomous driving systems (n=3 farms), drones 19 (n=1 farm), and sensors (n= 4 farms). Word-of-mouth, peer recommendations, and previous 20 experience are the preferential ways the technologies are adopted. 21 13 Once the pilot phase was completed, the main study was conducted in the period from October 1 to November 2024 by an international market research company using the CATI (Computer 2 Assisted Telephone Interview) method. The survey lasted approximately 30 minutes. The final 3 sample consisted of 251 Italian farmers. It included both tomato growers (50% from northern 4 Italy and 50% from the south) and farmers cultivating other fresh vegetables, such as carrots, 5 peppers, eggplants, lettuce, etc., distributed across northern, central and southern Italy (30%, 6 17%, and 53%, respectively). The geographical distribution was designed to be representative 7 of horticultural farms, based on data from the Italian National Institute of Statistics (ISTAT). 8 The inclusion of multiple administrative regions across the country ensures a comprehensive 9 understanding of cultivation practices and highlights the heterogeneity of technological 10 adoption. 11 12 3.2. Measures 13 Together with the socio-demographic information and the descriptive indicators, the 14 questionnaire was designed to test the model hypotheses. Overall, it included 14 constructs, 15 with a total of 45 items. The constructs included in the final model (Figure 1) were aimed to 16 understand the drivers for the adoption of innovative water-smart agricultural technologies by 17 Italian horticultural farms. All TAM3 items were measured on a 7-point scale (from ‘strongly 18 disagree’ to ‘strongly agree’) (see Annex Table A1). 19 Subjective norm was assessed by three items (e.g., “Many producers I know have already 20 adopted this innovation”) and was adapted by Fishbein and Ajzen (2010). This construct 21 14 refers to the perceived social pressure to perform or not perform a particular behavior. These 1 social influences shape intentions and actions, and, in this study, the construct was explored 2 considering peer adoption, which could lead to direct recommendations or serve as a concrete 3 example (Kutter et al., 2011). The same construct considered also the customers’ perception, 4 which is more and more important, especially if linked to sustainability issues (such as water 5 saving practices). 6 We measured the perceived usefulness with four items (e.g., “This innovation could improve 7 my productivity”) and perceived ease of use with two items (e.g., “This technology should be 8 easy to use”) (Davis, 1989). Output quality, i.e. the perception of the quality of the technology 9 in performing the task, was measured by four items (e.g., “Using this technology will improve 10 the quality of my products”), each adapted from Venkatesh and Bala (2008). SAFA-based 11 aspects (i.e. the sustainability-related outcomes of the new technology adoption) were assessed 12 by four items (e.g., “By using this innovation, I could help reduce water consumption”). 13 Technology self-efficacy, i.e. the belief in how well someone can perform actions to achieve 14 performance outcomes, was measured by three items (e.g., “I would use this innovation easily 15 if I had technical support”), whereas anxiety was assessed by three items (e.g., “New 16 technologies make me feel uncomfortable”). All these latter items were adapted from 17 Venkatesh and Bala (2008). Finally, we used three items to assess behavioural intention (e.g., 18 “I intend to use this technology in the near future”) (Fishbein and Ajzen, 2010). 19 The study focused on the three levels of water-smart technologies described above: Level 1) – 20 soil moisture sensors, Level 2) – a system combining sensors with an automatic irrigation 21 15 system, and Level 3) – sensors connected to an automated system, which in turn is connected 1 to and interacts with predictive models. If a farmer indicated they have adopted a certain level 2 of the technology, the items were framed for the next level. For instance, if a respondent have 3 already adopted soil moisture sensors, we asked about the intention to adopt the sensors 4 connected to an automated system. If no adoption was reported, we asked about their intention 5 to adopt soil moisture sensors (Level 1), whereas when they reported the highest level of 6 adoption, we asked about the intention to adopt more advanced predictive models. Thus, the 7 behaviour was assessed with a single item, ranging from 1 to 4, considering the different 8 adoption levels (1=no technology; 2=Level 1; 3=Level 2; 4=Level 3). We acknowledge that in 9 TAM3 formulations intention and behaviour typically refer to the same “future-oriented” action 10 (Venkatesh and Bala, 2008). Given the cross-sectional nature of our study, we adopted a 11 graduated behavioural scale to capture the level of adoption by farmers of increasingly 12 sophisticated water-smart technologies. This decision was necessary given the multi-stage 13 nature of the technology adoption in precision agriculture (Finger, 2023), including smart-water 14 management technologies. Indeed, in many cases farmers adopt these solutions incrementally, 15 in a progression from basic systems (in our case, soil moisture sensors) to more complex 16 systems (such as predictive models). Accordingly, behavioural intention was framed toward the 17 next logical step in technology adoption, while behaviour captured the current level of adoption. 18 Therefore, the intention-behaviour link in this study, tested in H6 (Figure 1), should be 19 interpreted as a sequential progression in the adoption pathway, as measured by the behavioural 20 scale, and not as a direct action-intention relation. 21 16 1 3.3. Data analysis 2 We performed the statistical analysis using SPSS v.29.0 and AMOS v.29.0 statistical software 3 (IBM Corporation, Armonk, NY, USA). Means, standard deviations, median and interquartile 4 range (IQR) were calculated for each questionnaire item and its related construct. Structural 5 equation modelling (SEM) was used to test hypotheses H1–H6 and the theoretical framework 6 in Figure 1. SEM allows models to be specified with both latent (e.g., perceived usefulness) 7 and observed variables (e.g., farm size) (Kline, 2016). 8 To improve our understanding of farm-level attitude and structural variables, we have 9 considered two models using the same dataset: in Model 1 we included only the variables of 10 the extended-TAM3 model, i.e. behaviour, behavioural intention, subjective norm, perceived 11 usefulness, perceived ease of use, output quality, sustainability-related constructs (SAFA), 12 technology self-efficacy and anxiety. Then, we controlled for the effects of individual factors 13 (i.e., farmers’ age, educational level, and years of experience in the agricultural sector) and 14 organizational factors (i.e., farm size, farm location, number of employees, membership in a 15 cooperative or a producer organization) on the endogenous variables (i.e., perceived usefulness, 16 behavioural intention, and behaviour), building a second model, which is an extension of Model 17 1. The full model, containing all factors and their interactions, was fitted to the data; then, the 18 least significant parameters were removed through a backwards deletion method, using 19 Akaike’s Information Criterion (AIC) for assessing the improvement in model fit (Power et al., 20 2013). In general, the AIC declines as the model fit increases (Kline, 2016). Where the model 21 17 fit was not significantly improved by the removal of a parameter, it remained in the final model. 1 This process was repeated until only two significant variables remained in the Model 2: number 2 of employees and farm size (expressed in hectares of utilised agricultural area, UAA). These 3 additions are theoretically relevant and empirically meaningful for explaining farmers' digital 4 adoption behaviour in a real-world context (McFadden, 2022; Okoye et al., 2025; Osrof et al., 5 2023). 6 Convergent validity of the model variables was assessed using average variance extracted 7 (AVE), Cronbach’s α coefficient, and composite reliability (CR). Discriminant validity was 8 tested by comparing the square root of the AVE of each construct with the inter-construct 9 correlation (Bagozzi and Yi, 2012). The goodness-of-fit of the models was assessed using the 10 χ2 and their degrees of freedom (df), the Tucker-Lewis Index (TLI), the comparative fit index 11 (CFI), the root mean square error of approximation (RMSEA) with a 90% confidence interval, 12 and the standardised root mean square residual (SRMR) (Kline, 2016). Since Model 1 and 13 Model 2 are not nested, due to additional parameters, we evaluated their relative performance 14 using multiple comparative indicators, as recommended in the SEM literature, such as CFI, 15 TLI, RMSEA, and AIC. The coefficient of determination (R2) was used to measure the 16 explained variance of the endogenous variables. We applied the Maximum Likelihood 17 estimation routine (Byrne, 2010). 18 19 4. Results 20 4.1. Descriptive statistics 21 18 The overall sample consisted of 251 respondents who were responsible for farm’s decisions 1 (78% always, 14% often, and 8% sometimes). Most respondents were male (92%), had 2 completed upper secondary education (53%), had an average age of 53 years, and a median of 3 30 years of experience in the agricultural sector (Table 1). Most farms were located in southern 4 Italy and on the islands (51.4), had a median utilised agricultural area (UAA) of 15 ha, 5 employed less than 10 people (68%), with a median turnover of €200.000. The most frequently 6 cultivated vegetables were tomatoes, both for fresh consumption (44%) and for the processing 7 industry (41%), followed by peppers (16%) and zucchinis (11%). 8 Most of the sampled farmers had not yet adopted any of the proposed technologies (n=175, 9 69.7%). Those who have deployed any of these technologies relied on Level 1 (i.e. soil moisture 10 sensors, n=43, 17.1%), and a few were already using automated irrigation systems (Level 2) or 11 predictive models (Level 3), accounting for 6.4% (n=16) and 6.8% (n=17) respectively (Table 12 1). In light of these findings, it is important to understand the motivation for the adoption of 13 new technologies and the factors that hamper their introduction. 14 15 Table 1. Description of the sample: farms characteristics and socio-demographics data of 16 farmers (n=251). 17 Variables Sample N % Age of the respondent Age (years, mean and SD) 52.8 (11.9) Gender Male 231 92.0 Female 20 8.0 Others or prefer not to answer 0 0.0 Educational level 19 Primary 8 3.2 Secondary lower 57 22.7 Secondary higher 132 52.6 Tertiary 54 21.5 Geographical area of the farm North-West 28 11.2 North-East 66 26.3 Center 28 11.2 South and Islands 129 51.4 Farm size UAA (ha, median and IQR) 15.0 (4.0-60.0) Farms by UAA classes < 2 ha 20 8.0 2 – 4.99 ha 47 18.7 5 – 19.99 ha 69 27.5 20 – 49.99 ha 45 17.9 > 50 ha 70 27.9 Most cultivated vegetables Tomato (for fresh consumption) 110 43.8 Tomato (for the processed industry) 104 41.4 Peppers 40 15.9 Zucchinis 27 10.8 Eggplants 13 5.2 Lettuce 13 5.2 Potatoes 12 4.8 Melons 9 3.6 Cauliflowers 8 3.2 Enterprise n. employee category Micro (1-9 employees) 171 68.1 Small (10-49) 64 25.5 Medium 1 (50-99) 12 4.8 Medium 2 (100-249) 4 1.6 Large (≥250) 0 0.0 Farm’s turnover Turnover (.000 euro, median and IQR) 200 (90-650) Farmer’s years of experience in agriculture Years of experience (median and IQR) 30 (20-40) Levels of water-smart technologies a No technological innovation 175 69.7 Level 1 43 17.1 Level 2 16 6.4 Level 3 17 6.8 Notes: Data are presented as the mean (SD) for continuous variables for which the hypothesis of normal 1 distribution cannot be rejected at p<0.05, as median (IQR) otherwise, or as number (%) for nominal variables. SD 2 = Standard Deviation. IQR = Interquartile Range. UAA = Utilised Agricultural Area. a Levels of water-smart 3 20 technologies: Level 1) – soil moisture sensors, Level 2) – a system combining sensors with an automatic irrigation 1 system, and Level 3) – sensors connected to an automated system, which in turn is connected to and interacts with 2 predictive models. 3 4 Table 2 shows the descriptive statistics of the latent and observable variables, as well as the 5 tests conducted on the constructs. The factor loadings of the variable items (λ) exceeded 0.50, 6 the Cronbach’s α and CR values were above 0.70, and the AVE values exceeded 0.50; these 7 results, with the only exception of perceived ease of use, demonstrated strong reliability, as well 8 as convergent and discriminant validity of all factors in the measurement model. We opted to 9 retain the perceived ease of use construct, first, because the construct is grounded in the TAM, 10 where plays a fundamental role. Secondly, while the Cronbach’s alpha is slightly low, the 11 construct shows acceptable composite reliability and standardized factor loadings. The AVE is 12 slightly below the conventional 0.50 threshold, but is still within an acceptable range, 13 particularly when supported by theoretical significance and other reliability indicators. 14 Discriminant validity was further confirmed by verifying that the square root of the AVE for 15 each construct, as shown in Table 3, was greater than the correlations between the constructs 16 (Bagozzi and Yi, 2012). 17 Overall, the results in Table 2 show a moderately positive perceived usefulness of water-smart 18 agricultural new technologies (mean score: 4.81), which means in particular that farmers 19 moderately agree that by using this technology they could reduce water consumption and 20 improve productivity by using this technology. The results also show a moderately positive 21 21 perceived ease of use (4.69) and output quality (4.59). Furthermore, important others had no 1 significant influence (3.63), and there was relatively low anxiety about applying new 2 technologies (3.16). The results indicated a positive evaluation of the sustainability aspects 3 related to the new technology (e.g., reduced water consumption, enhanced technical skills, etc., 4 mean score: 5.03), as well as positive technology self-efficacy (5.12). In particular, respondents 5 stated that they would use this innovation easily if they had technical support. Furthermore, 6 farmers exhibited a moderately positive intention to adopt innovative water-smart agricultural 7 technologies (4.58). 8 9 Table 2. Mean values (standard deviation, SD) and median values (interquartile range, IQR) of 10 single items and constructs, factor loadings (λ), composite reliability (CR), average variance 11 extracted (AVE) and Cronbach’s α of the sample (n=251). 12 Mean (SD) Median (IQR) λ CR AVE α Perceived Usefulness 4.81 (1.06) 5.00 (4.25-5.25) 0.84 0.56 0.84 PU1 4.80 (1.25) 5.00 (4.00-5.00) 0.78 PU2 4.61 (1.34) 5.00 (4.00-5.00) 0.75 PU3 4.86 (1.20) 5.00 (4.00-5.00) 0.79 PU4 4.95 (1.37) 5.00 (4.00-6.00) 0.67 Perceived Ease of Use 4.69 (0.92) 4.50 (4.00-5.00) 0.61 0.45 0.61 PEU1 4.98 (1.06) 5.00 (4.00-5.00) 0.59 PEU2 4.40 (1.10) 5.00 (3.00-5.00) 0.75 Output Quality 4.59 (0.98) 4.75 (4.00-5.00) 0.83 0.56 0.84 OQ1 4.57 (1.15) 5.00 (4.00-5.00) 0.70 OQ2 4.52 (1.25) 5.00 (4.00-5.00) 0.78 OQ3 4.80 (1.16) 5.00 (4.00-5.00) 0.85 22 Note: All items were measured on a 7-point scale (from ‘strongly disagree’ to ‘strongly agree’), where median 1 value 4 indicates ‘neither agree or disagree’. a Levels of water-smart technologies: Level 1) – soil moisture sensors, 2 Level 2) – a system combining sensors with an automatic irrigation system, and Level 3) – sensors connected to 3 an automated system, which in turn is connected to and interacts with predictive models. 4 5 4.2. Drivers of digital innovation 6 Model 1 showed a good fit with the collected data: χ2 (df) = 461.975 (280), CFI = 0.950, 7 RMSEA = 0.051 (90%CI 0.043 – 0.059), TLI = 0.942 and SRMR = 0.054. The standardized 8 OQ4 4.48 (1.20) 5.00 (4.00-5.00) 0.64 SAFA 5.03 (1.00) 5.00 (4.67-5.67) 0.78 0.55 0.79 SAFA1 5.12 (1.21) 5.00 (5.00-6.00) 0.68 SAFA2 4.87 (1.17) 5.00 (4.00-5.00) 0.72 SAFA3 5.10 (1.20) 5.00 (5.00-6.00) 0.81 Anxiety 3.16 (1.18) 3.00 (2.67-3.67) 0.85 0.66 0.85 ANX1 3.28 (1.33) 3.00 (3.00-4.00) 0.74 ANX2 3.10 (1.33) 3.00 (2.00-3.00) 0.88 ANX3 3.10 (1.37) 3.00 (2.00-3.00) 0.81 Technology Self-Efficacy 5.12 (1.11) 5.00 (4.67-6.00) 0.92 0.80 0.92 TSE1 5.07 (1.22) 5.00 (5.00-6.00) 0.87 TSE2 5.20 (1.18) 5.00 (5.00-6.00) 0.93 TSE3 5.08 (1.18) 5.00 (5.00-6.00) 0.88 Subjective Norms 3.63 (1.09) 3.67 (3.00-4.33) 0.76 0.53 0.74 SN1 3.84 (1.31) 4.00 (3.00-5.00) 0.84 SN2 3.52 (1.41) 3.00 (3.00-5.00) 0.51 SN3 3.53 (1.31) 3.00 (3.00-5.00) 0.78 Behavioural Intention 4.58 (1.35) 4.67 (4.00-5.33) 0.91 0.77 0.91 BI1 4.41 (1.51) 5.00 (4.00-5.00) 0.93 BI2 4.54 (1.47) 5.00 (4.00-5.00) 0.89 BI3 4.80 (1.41) 5.00 (4.00-6.00) 0.80 Behaviour a 1.50 (0.89) 1.00 (1.00-2.00) 23 path coefficients and their significance levels are shown in Table 4, whereas the unstandardized 1 coefficients and standard errors are shown in the Appendix Table A2. 2 3 Table 3. Spearman’s rank-order correlations (ρ) between the constructs including the squared 4 root of the AVE of each construct (reported in bold on the main diagonal). 5 PU PEU OQ SAFA ANX TSE SN BI BEH PU 0.75 0.31*** 0.63*** 0.52*** −0.28*** 0.40*** 0.36*** 0.57*** 0.36*** PEU 0.67 0.24*** 0.31*** −0.17** 0.34*** 0.14* 0.28*** 0.15* OQ 0.75 0.46*** −0.21*** 0.34*** 0.43*** 0.60*** 0.35*** SAFA 0.74 −0.23*** 0.63*** 0.16* 0.39*** 0.19** ANX 0.81 −0.16** −0.15* −0.29*** −0.17** TSE 0.89 n.s. 0.39*** n.s. SN 0.73 0.44*** 0.36*** BI 0.88 0.40*** Note: PU = Perceived Usefulness; PEU = Perceived Ease of Use; OQ = Output Quality; SAFA = Sustainability 6 Assessment of Food and Agriculture Systems; ANX = Anxiety; TSE = Technology Self-Efficacy; SN = subjective 7 norms; BI = Behavioural Intentions; BEH = behaviour; Sign.: *** p<0.001, ** p<0.01, * p<0.01, n.s. = not 8 significant. 9 10 Overall, Model 1 shows R2 values of 0.65 for the intention and 0.16 for the behaviour in 11 adopting a new water-smart technology. This means that, respectively, 65.1% of the variance 12 in intention and 16.4% of the variance in behaviour can be explained by the tested variables. 13 The results suggest that the intention to adopt an innovative water-smart technology 14 significantly influences the actual behaviour (i.e., the adoption of the technology itself), as 15 postulated by H6 (p<0.001). Behavioural intention, in turn, is positively influenced by 16 24 perceived usefulness with p<0.001, which is one of the two core variables of the TAM3 (H4 1 accepted). 2 3 Table 4. TAM3-extended models: coefficient of determination (R2), standardised coefficients 4 (β), p-values, research hypotheses and model fit indices (n=251). 5 Model 1 Model 2 R2 β p Hypotheses R2 β p Hypotheses PU 0.791 0.822 PEU → PU 0.133 0.044 H3b accepted 0.121 0.060 H3b accepted OQ → PU 0.658 <0.001 H1a accepted 0.708 <0.001 H1a accepted SAFA → PU 0.199 0.027 H1b accepted 0.152 0.081 H1b accepted SN → PU 0.045 0.512 H1c rejected 0.035 0.607 H1c rejected EMP → PU 0.183 <0.001 UAA → PU 0.040 0.351 PEU 0.305 0.303 TSE → PEU 0.487 <0.001 H2a accepted 0.488 <0.001 H2a accepted ANX → PEU −0.169 0.041 H2b accepted −0.163 0.048 H2b accepted BI 0.651 0.652 PU → BI 0.601 <0.001 H4 accepted 0.602 <0.001 H4 accepted PEU → BI 0.069 0.295 H3a rejected 0.060 0.351 H3a rejected SN → BI 0.278 <0.001 H5 accepted 0.268 <0.001 H5 accepted EMP → BI 0.034 0.473 UAA → BI 0.086 0.065 BEH 0.164 0.179 BI → BEH 0.404 <0.001 H6 accepted 0.370 <0.001 H6 accepted EMP → BEH 0.037 0.534 UAA → BEH 0.153 0.009 Model fit indices χ2 (df) 461.975 (280) 500.298 (323) CFI 0.950 0.952 TLI 0.942 0.944 AIC 657.561 722.298 25 RMSEA (90% C.I.) 0.051 (0.043 – 0.059) 0.047 (0.039 – 0.055) SRMR 0.054 0.058 Note: PU = Perceived Usefulness; PEU = Perceived Ease of Use; OQ = Output Quality; SAFA = Sustainability 1 Assessment of Food and Agriculture Systems; ANX = Anxiety; TSE = Technology Self-Efficacy; SN = subjective 2 norms; BI = Behavioural Intentions; EMP = number of employees; UAA = average farm size (Utilised agricultural 3 area); BEH = Behaviour. 4 5 Perceived ease of use does not significantly affect the intention to adopt technologies, therefore 6 not supporting H3a; however, it positively affects perceived usefulness of new technologies 7 with p<0.05, confirming H3b. H5 is also supported since subjective norm has a positive effect 8 on the intention to adopt a technology (p<0.001), showing that perceived social pressure has an 9 influence on the farmers’ motivation to adopt a new technology. The construct of anxiety shows 10 a negative effect on the perceived ease of use (p<0.05, H2b accepted), a property that is 11 stimulating and that could open up new ways of designing and conceptualizing modern 12 technologies. Perceived ease of use, on the other hand, is positively influenced by the self-13 efficacy of the technology, with p<0.001, supporting H2a. In turn, perceived usefulness is 14 influenced by the quality of the output (i.e., the perceived quality of the outcomes achieved 15 through the use of the technology, p<0.001) and by the SAFA indicators (p<0.05), thereby 16 supporting H1a and H1b, respectively. However, H1c is not supported, as the effect of 17 subjective norms on perceived usefulness is not significant. 18 Overall, Model 2 shows good fit with the data (χ2 (df) = 500.298 (323), CFI = 0.952, RMSEA 19 = 0.047 (90%CI 0.039 – 0.055), TLI = 0.944 and SRMR = 0.058). Despite a higher AIC for 20 Model 2 as compared to Model 1, due to the additional parameters, the extended model shows 21 26 slight improvements in global fit (CFI, TLI, RMSEA), while also improving the explained 1 variance of dependent variables, in particular behavioural outcome (up to 17.8%) (Table 4). 2 The overall path and the tested hypotheses are confirmed, and the inclusion of structural 3 variables adds value to the explanatory power of the model without substantially compromising 4 parsimony. In particular, the number of employees positively influences respondents’ perceived 5 usefulness (p<0.001), indicating that decision-makers in larger farms, in terms of workforce, 6 find the innovative technology capable of enhancing farm performance. In turn, the average 7 farm size in UAA positively influences the behaviour (p<0.001) and behavioural intentions 8 (p<0.10). In other words, respondents working in larger farms are more willing to adopt the 9 new technologies or have already adopted them. 10 11 5. Discussion 12 This study provides empirical evidence on the factors influencing the intention and behaviour 13 of Italian farmers to adopt digital water-saving technologies, with a focus on the horticultural 14 sector. The results show that about 70% of the farmers surveyed have not adopted any of the 15 proposed technologies, emphasising the persistent gap between technological innovation and 16 practical implementation in the field. This finding confirms the limited uptake of smart farming 17 solutions in Italy and emphasises the importance of understanding the barriers to adoption and 18 the enablers of adoption. This is in line with previous research highlighting the slow pace of 19 digitalisation in Italian agriculture compared to other European countries (Addorisio, Spadoni, 20 et al., 2025; Cimino et al., 2024; Okoye et al., 2025; Timpanaro et al., 2024). 21 27 Using the extended TAM3 model, the study explains about 18% of the variance in the actual 1 adoption of water-smart technologies, and 65% in adoption intention. As expected, behavioural 2 intention is a significant predictor of the behaviour, indicating that farmers motivation in 3 adopting the innovative technologies affect the actual adoption. The applied model further 4 assumes that the effect of other variables (e.g., self-efficacy) on behavioural intention is 5 mediated by perceived usefulness and perceived ease of use. Consistent with prior research 6 (Davis, 1989; Venkatesh and Davis, 2000) perceived usefulness proved to be a key factor in 7 willingness to adopt, emphasising the importance farmers attach to tangible benefits such as 8 improved efficiency, performance or sustainability. Consistent with other studies, which 9 suggest that perceived usefulness significantly influences technology adoption when it does not 10 significantly increase production costs (Pierpaoli et al., 2013), our results show that economic 11 viability remains a key concern, especially in a sector dominated by small to medium sized 12 farms with limited investment capacity. The importance of perceived benefits is also supported 13 by studies of adoption in agriculture across Europe, including those by Paustian and Theuvsen 14 (2017) and Shang et al. (2021), who argue that demonstrable improvements in performance are 15 essential to overcome innovation scepticism. In general and specifically in the Italian context, 16 where high input costs and fragmented land ownership remain systemic barriers to 17 modernisation, benefits are often associated not only with profitability, but also with the ability 18 to ensure compliance with environmental standards and receive subsidies under the EU's 19 Common Agricultural Policy (CAP). The perceived benefits of digital tools by farmers are 20 28 therefore determined by a complex interplay of market incentives, legal requirements and 1 evolving sustainability expectations. 2 Furthermore, in horticulture, a labour-intensive and water-dependent sub-sector, the perceived 3 benefits of smart technologies are particularly relevant. With climate change exacerbating water 4 scarcity in regions such as southern Italy, precision irrigation and monitoring systems are 5 increasingly recognised as essential climate adaptation tools (Lakhiar et al., 2024). However, 6 despite their potential, the perceived benefits of these tools often depend on farmers' ability to 7 access relevant knowledge, trust the data provided and integrate new practises into existing 8 routines – all factors mediated by the socio-cultural and institutional dynamics of the Italian 9 agricultural landscape (Addorisio, Casolani, et al., 2025). 10 In contrast to the predictions of the original TAM3 model (Venkatesh and Bala, 2008), 11 including those of its specific application (Kühnemund and Recke, 2025), this study found that 12 perceived ease of use did not significantly influence Italian farmers’ intention to adopt digital 13 water-smart technologies. This deviation from theoretical expectations can be explained by the 14 structural and cultural characteristics of Italian agriculture. The sector is dominated by small 15 family farms, many of which, especially in the southern regions, face a pronounced digital 16 divide due to limited access to broadband infrastructure, digital skills and technical training 17 (Addorisio, Casolani, et al., 2025). Similar findings have been found in other studies (for an 18 overview, see Osrof et al., 2023), including research on the Italian fruit and wine sector, where 19 ease of use was also found to be insignificant (Canavari et al., 2021). As suggested by Schulze 20 Schwering et al. (2022), perceived ease of use may play a minor role when farmers rely more 21 29 on external support or adopt collaborative practises, suggesting that social norms may override 1 ease of use concerns. Indeed, social norms had a significant influence on adoption intentions in 2 this study, mirroring the findings of Senyolo et al. (2018) and Dissanayake et al. (2022). These 3 findings suggest that peer influence and visibility of successful cases may help to foster a 4 culture of innovation and support wider adoption. 5 The study also shows that sustainability-related factors — such as improved water management 6 and skills development— increase perceived usefulness. This supports the arguments of Montes 7 de Oca Munguia et al. (2021), who emphasise that the integration of sustainability goals can 8 increase the perceived value of smart technologies. This is particularly relevant in the Italian 9 Mediterranean climate context, where water scarcity, soil degradation and the effects of climate 10 change pose an acute challenge to horticultural productivity and environmental resilience. The 11 perceived alignment of digital technologies with sustainability goals therefore not only 12 improves acceptance, but also strengthens farmers’ motivation by linking environmental 13 benefits with economic performance (Paustian and Theuvsen, 2017). This last point is thought-14 provoking when it comes to examining the role of farmers and their commitment to 15 sustainability, as well as their awareness of the use of smart devices to promote more sustainable 16 practices. In the face of climate change and the pressure that agriculture is putting on 17 environmental resources, only the direct and committed involvement of farmers can promote a 18 more conscious and widespread use of smart technologies with the aim of reaping their benefits 19 (Menozzi et al., 2015). Linking sustainability to usefulness may also increase perceived 20 profitability, which strengthens farmers’ interest in adoption. 21 30 Technology self-efficacy was found to have a significant effect on perceived ease of use, 1 suggesting that farmers who have confidence in their technical abilities are more likely to 2 perceive smart technologies as manageable. This supports previous findings highlighting the 3 relevance of perceived behavioural control for the adoption of sustainable practises (Menozzi 4 et al., 2015). In the Italian context, where a significant part of the agricultural population 5 consists of small farmers or ageing operators with limited participation in formal ICT training, 6 self-efficacy becomes a crucial factor for the adoption of digital practises. Italian agriculture is 7 characterised by a dual structure: on the one hand, there are highly mechanised and export-8 oriented companies, and on the other, traditional small farms that often lack digital 9 infrastructure. Therefore, strengthening the confidence and skills of farmers through training is 10 crucial to bridging this technological gap. The results also shed light on important barriers, 11 including limited digital skills and poor access to information, which are in line with the 12 findings of other studies (Osrof et al., 2023; Sabbagh and Gutierrez, 2023; Yigezu et al., 2018). 13 The negative correlation between anxiety and perceived ease of use further underscores the 14 need for user-friendly technologies that reduce cognitive and operational complexity. Targeted 15 training initiatives and inclusive measures are needed to address these barriers, especially for 16 farmers with limited resources. 17 The structural model also examined the influence of individual and organizational 18 characteristics. Among these, only farm size and number of employees significantly affected 19 the model’s endogenous variables, while other factors such as education level, age, and farm 20 location showed no consistent impact. These findings reflect the mixed evidence in the literature 21 31 (Osrof et al., 2023) and may be partially explained by the evolving role of education and age in 1 farming. For instance, highly educated individuals may pursue careers outside agriculture 2 (Michels et al., 2020), while smart technologies are increasingly designed to be accessible 3 regardless of users’ educational background (Wachenheim et al., 2021). Similarly, although 4 older age is often associated with lower adoption rates, several studies have shown no 5 significant age effect (García-Jiménez et al., 2022). Larger farms—measured by UAA 6 acreage—were more likely to adopt or intend to adopt digital water-smart solutions. This 7 confirms the role of scale economies in overcoming the high investment costs associated with 8 advanced technologies (Osrof et al., 2023). The significant effect of the number of employees 9 on perceived usefulness may be interpreted in two ways: first, these technologies may be seen 10 as reducing labour needs and costs; second, they may be valued for their potential to enhance 11 employees’ technical capabilities and improve workforce productivity. The latter interpretation 12 aligns more closely with the observed positive relationship between sustainability orientation 13 and perceived usefulness. 14 In summary, this study contributes to the understanding of digital technology adoption in Italian 15 agriculture by validating core constructs of the TAM3 model while highlighting context-16 specific dynamics. This study not only confirms existing theoretical frameworks, but also 17 highlights the importance of adapting digital transformation strategies to the Italian agricultural 18 reality, characterized by structural constraints, strong social networks and a growing, yet 19 uneven, momentum towards sustainable innovation. By identifying key drivers and barriers to 20 adoption, the findings provide valuable guidance for the development of tailored policies, 21 32 training programs, and user-oriented technologies that can support the digital transformation of 1 the horticultural sector. Promoting the integration of sustainability goals, addressing capacity 2 gaps, and leveraging social norms can collectively enhance adoption and support the transition 3 toward a more sustainable and resilient agricultural system. 4 5 6. Conclusion 6 The adoption of digital technologies in the Italian horticultural sector is a multifaceted challenge 7 influenced by a variety of individual, technological, social and contextual factors. These same 8 factors are of utmost importance for policy makers that needs to design appropriate measures 9 for promoting the adoption of water-saving solutions. Identifying both the motivations and the 10 barriers to innovation might indeed inform policies by suggesting models that are suited to the 11 challenges faced by farmers. Likewise, these findings can inform all interested stakeholders — 12 from solution and technology developers to producer organizations — about the key aspects 13 needed to promote the adoption of digital technologies within the agricultural sector. 14 This study shows that individual intention is an important determinant of the actual adoption of 15 innovative water-saving technologies and highlights the crucial role of farmer motivation in 16 decision-making. Perceived usefulness of these technologies has a significant effect on 17 adoption intention, while perceived ease of use requires further investigation due to its limited 18 relevance in the current context. Social norms were identified as an important determinant of 19 farmers’ intentions, highlighting the importance of community influence and external support 20 in promoting the adoption of digital technologies. To close the observed adoption gap, targeted 21 33 interventions should be developed to address barriers such as digital literacy, infrastructure and 1 accessibility of technology. Furthermore, the regional and culture-specific nuances observed in 2 this study should be taken into account when developing customised strategies. 3 The results highlight important policy and business implications, suggesting that government 4 agencies, agricultural cooperatives, and technology developers should emphasize the economic 5 and environmental benefits of digital irrigation technologies. The identification of appropriate 6 measures, such as the creation of networks, outreach activities, knowledge-sharing initiatives 7 and technical assistance, can also accelerate adoption. By addressing these research and 8 knowledge gaps, this study contributes to both the academic literature and practical policy 9 making. It provides a refined theoretical model to understand technology adoption in small- and 10 medium-sized farms and offers practical insights to promote sustainable and efficient water 11 management in agriculture. Further exploration of constructs that have negative correlates, such 12 as anxiety, could lead to more user-centred technology design that reduces barriers to 13 technology adoption and improves usability. Policy, and in particular the Agricultural 14 Knowledge and Innovation Systems (AKIS) under the CAP, can play a crucial role in 15 addressing these aspects by facilitating multi-actor collaborations between farmers, researchers, 16 advisors, and technology developers, as well as strengthening the dissemination of best 17 practices and provide continuous support (Esposti, 2012). 18 Some limitations of this study should be mentioned. First, the study reflects not only a specific 19 context, such as the horticultural sector, but also national characteristics, which can vary greatly 20 from country to country due to different regulatory and incentive frameworks, cultural practises 21 34 and, most importantly, technological infrastructures. Nevertheless, the sample is not 1 representative of Italian farmers. This must be taken into account when interpreting the results 2 and deriving consequences for corporate management. An extension of the sample and a 3 repetition of the study in other countries could therefore be interesting to test the validity of all 4 the hypotheses put forward in the original theory. Extending the analysis to other crops farming, 5 as well as to livestock farming, and expanding the sample size could provide more accurate 6 insights over the phenomenon. EU Cross-regional studies might also be very interesting to point 7 out differences in the adoption rate to be linked to cultural specificities, but also to normative, 8 political and infrastructural settings. The results might in this way serve policy makers and 9 institutions to design ad hoc policies to meet the desired technology level. 10 Second, we did not consider prospective behaviour, i.e., we did not measure actual behaviour 11 in the future (i.e., future adoption of the innovative technologies), but only current behaviour. 12 Although this approach is quite common in similar cross-sectional studies (see, e.g., (Li et al., 13 2023) on smart agricultural technology adoption), it might have limited the compatibility of 14 behaviour with its antecedents (Fishbein and Ajzen, 2010). For instance, it has been noticed 15 that relations between intention and behaviour might be stronger for studies measuring 16 concurrent behaviour than those measuring behaviour prospectively (McEachan et al., 2011). 17 However, we decided to use this multiple-level measurement of behaviour in order to capture, 18 better than a binary variable, the progressive adoption from basic systems (in our case soil 19 moisture sensors) to more complex systems (such as predictive ones). Thus, the intention-20 behaviour link in this study should be interpreted as a sequential progression in the adoption 21 35 pathway, rather than a direct action-intention relation. Future studies employing longitudinal 1 data collection could better report this linkage between intention and prospective adoption. 2 Finally, this study used self-report measures about the behaviour which may be subject to 3 response biases. However, the CATI method can help with complex or sensitive questions 4 by allowing the interviewer to clarify questions and guide the respondent, thus reducing 5 misinterpretation and encouraging more accurate responses (Dillman et al., 2014). 6 Despite these limitations, this study is, to our knowledge, one of the first aimed at 7 investigating the relative importance of behavioural precursors in explaining the intention 8 to adopt innovative water-smart technologies in Italian horticultural farms. 9 10 Acknowledgements 11 This study was carried out within the Agritech National Research Center and received funding 12 from the European Union Next-GenerationEU (Piano Nazionale di Ripresa e Resilienza 13 (PNRR) – Missione 4 Componente 2, Investimento 1.4 – D.D. 1032 17/06/2022, CN00000022). 14 This paper reflects only the authors' views and opinions, neither the European Union nor the 15 European Commission can be considered responsible for them. 16 17 References 18 Addorisio, R., Casolani, N., Maesano, G., Coderoni, S., Perito, M. A., Mattetti, M., and 19 Canavari, M. (2025). Barriers and drivers of digital agriculture adoption: Insights from 20 Italian farming stakeholders. International Journal on Food System Dynamics 16: 1–12. 21 36 Addorisio, R., Spadoni, R., and Maesano, G. (2025). Adoption of Innovative Technologies for 1 Sustainable Agriculture: A Scoping Review of the System Domain. In Sustainability 2 (Vol. 17, Issue 9). 3 Adeyemi, O., Grove, I., Peets, S., and Norton, T. (2017). Advanced Monitoring and 4 Management Systems for Improving Sustainability in Precision Irrigation. In 5 Sustainability (Vol. 9, Issue 3). 6 Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human 7 Decision Processes . 8 Alexandrova-Stefanova, N., Nosarzewski, K., Mroczek, Z. K., Audouin, S., Djamen, P., 9 Kolos, N., and Wan, J. (2024). Shaping sustainable agrifood futures: pre-emerging and 10 emerging technologies and innovations for impact. Rome.: 11 Asadi, E., Isazadeh, M., Samadianfard, S., Ramli, M. F., Mosavi, A., Nabipour, N., 12 Shamshirband, S., Hajnal, E., and Chau, K.-W. (2020). Groundwater Quality Assessment 13 for Sustainable Drinking and Irrigation. In Sustainability (Vol. 12, Issue 1). 14 Bagozzi, R. P., and Yi, Y. (2012). Specification, evaluation, and interpretation of structural 15 equation models. Journal of the Academy of Marketing Science . 16 Baldoni, E., Coderoni, S., and Esposti, R. (2018). Immigrant workforce and labour 17 productivity in Italian agriculture: a farm-level analysis. Bio-Based and Applied 18 Economics 6: 259–278. 19 Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. In 20 Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, 21 37 NJ, US: Prentice-Hall, Inc , xiii, 617– xiii, 617. 1 Blasch, J., van der Kroon, B., van Beukering, P., Munster, R., Fabiani, S., Nino, P., and 2 Vanino, S. (2022). Farmer preferences for adopting precision farming technologies: a 3 case study from Italy. European Review of Agricultural Economics 49: 33–81. 4 Byrne, B. M. (2010). Structural equation modeling with AMOS: Basic concepts, applications, 5 and programming, 2nd ed. In Structural equation modeling with AMOS: Basic concepts, 6 applications, and programming, 2nd ed. New York, NY, US: Routledge/Taylor & 7 Francis Group. 8 Canavari, M., Medici, M., Wongprawmas, R., Xhakollari, V., and Russo, S. (2021). A Path 9 Model of the Intention to Adopt Variable Rate Irrigation in Northeast Italy. In 10 Sustainability (Vol. 13, Issue 4). 11 Cimino, A., Coniglio, I. M., Corvello, V., Longo, F., Sagawa, J. K., and Solina, V. (2024). 12 Exploring small farmers behavioral intention to adopt digital platforms for sustainable 13 and successful agricultural ecosystems. Technological Forecasting and Social Change 14 204: 123436. 15 Davis, F. D. (1989). Perceived Usefulness, Perceived Ease of Use, and User Acceptance of 16 Information Technology. MIS Quarterly 13: 319–340. 17 de Oca Munguia, O. M., and Llewellyn, R. (2020). The Adopters versus the Technology: 18 Which Matters More when Predicting or Explaining Adoption? Applied Economic 19 Perspectives and Policy 42: 80–91. 20 Dentoni, D., Cucchi, C., Roglic, M., Lubberink, R., Bender-Salazar, R., and Manyise, T. 21 38 (2023). Systems Thinking, Mapping and Change in Food and Agriculture. Bio-Based 1 and Applied Economics 11: 277–301. 2 Dillman, D. A., Smyth, J. D., and Christian, L. M. (2014). Internet, phone, mail, and mixed 3 mode surveys: The tailored design method, 4th ed. In Internet, phone, mail, and mixed 4 mode surveys: The tailored design method, 4th ed. Hoboken, NJ, US: John Wiley & 5 Sons Inc. 6 Dissanayake, C. A. K., Jayathilake, W., Wickramasuriya, H. V. A., Dissanayake, U., 7 Kopiyawattage, K. P. P., and Wasala, W. M. C. B. (2022). Theories and Models of 8 Technology Adoption in Agricultural Sector. Human Behavior and Emerging 9 Technologies 2022: 9258317. 10 Ermolieva, T., Havlik, P., Frank, S., Kahil, T., Balkovic, J., Skalsky, R., Ermoliev, Y., 11 Knopov, P. S., Borodina, O. M., and Gorbachuk, V. M. (2022). A Risk-Informed 12 Decision-Making Framework for Climate Change Adaptation through Robust Land Use 13 and Irrigation Planning. In Sustainability (Vol. 14, Issue 3). 14 Esposti, R. (2012). Knowledge, Technology and Innovations for a Bio-based Economy: 15 Lessons from the Past, Challenges for the Future. Bio-Based and Applied Economics 1: 16 235–268. 17 FAO. (2014). SAFA (Sustainability Assessment of Food and Agriculture systems) Guidelines. 18 Rome: FAO. 19 Finger, R. (2023). Digital innovations for sustainable and resilient agricultural systems. 20 European Review of Agricultural Economics 50: 1277–1309. 21 39 Fishbein, M., and Ajzen, I. (2010). Predicting and changing behavior: The reasoned action 1 approach. In Predicting and changing behavior: The reasoned action approach. New 2 York, NY, US: Psychology Press. 3 García-Jiménez, C. I., Velandia, M., Lambert, D. M., and Mishra, A. K. (2022). Information 4 sources impact on the adoption of precision technology by cotton producers in the 5 United States. Agrociencia . 6 Gemtou, M., Guillén, B. C., and Anastasiou, E. (2024). Smart Farming Technologies and 7 Sustainability BT - Digital Sustainability: Leveraging Digital Technology to Combat 8 Climate Change , T. Lynn, P. Rosati, D. Kreps, & K. Conboy (eds.). Cham: Springer 9 Nature Switzerland , 99–120. 10 Huffman, W. E. B. T.-H. of A. E. (2001). Chapter 7 Human capital: Education and 11 agriculture. In Agricultural Production (Vol. 1). Elsevier , 333–381. 12 Kapsdorferová, Z. (2024). Key Drivers and Innovative Approaches to Sustainable 13 Management in the Agricultural and Food Sector BT - Consumer Perceptions and Food 14 , D. Bogueva (ed.). Singapore: Springer Nature Singapore , 349–362. 15 Khanna, M. (2021). Digital Transformation of the Agricultural Sector: Pathways, Drivers and 16 Policy Implications. Applied Economic Perspectives and Policy 43: 1221–1242. 17 Kline, R. B. (2016). Principles and practice of structural equation modeling, 4th ed. In 18 Principles and practice of structural equation modeling, 4th ed. New York, NY, US: 19 Guilford Press. 20 Kühnemund, A., and Recke, G. (2025). Intention to use AI-Based Camera Systems in German 21 40 Pig Farming: An Extended Technology Acceptance Model . Bio-Based and Applied 1 Economics SE-Economic and policy analysis of smart agricultural systems. 2 Kutter, T., Tiemann, S., Siebert, R., and Fountas, S. (2011). The role of communication and 3 co-operation in the adoption of precision farming. Precision Agriculture 12: 2–17. 4 Lakhiar, I. A., Yan, H., Zhang, C., Wang, G., He, B., Hao, B., Han, Y., Wang, B., Bao, R., 5 Syed, T. N., Chauhdary, J. N., and Rakibuzzaman, M. (2024). A Review of Precision 6 Irrigation Water-Saving Technology under Changing Climate for Enhancing Water Use 7 Efficiency, Crop Yield, and Environmental Footprints. In Agriculture (Vol. 14, Issue 7). 8 Li, J., Liu, G., Chen, Y., and Li, R. (2023). Study on the influence mechanism of adoption of 9 smart agriculture technology behavior. Scientific Reports 13: 8554. 10 McEachan, R. R. C., Conner, M., Taylor, N. J., and Lawton, R. J. (2011). Prospective 11 prediction of health-related behaviours with the theory of planned behaviour: A meta-12 analysis. In Health Psychology Review. 13 McFadden, J. (2022). The digitalisation of agriculture: A literature review and emerging 14 policy issues. Paris.: 15 Menozzi, D., Fioravanzi, M., and Donati, M. (2015). Farmer’s motivation to adopt sustainable 16 agricultural practices. Bio-Based and Applied Economics 4. 17 Michels, M., von Hobe, C.-F., and Musshoff, O. (2020). A trans-theoretical model for the 18 adoption of drones by large-scale German farmers. Journal of Rural Studies 75: 80–88. 19 Montanarella, L., and Panagos, P. (2021). The relevance of sustainable soil management 20 within the European Green Deal. Land Use Policy 100: 104950. 21 41 Montes de Oca Munguia, O., Pannell, D. J., and Llewellyn, R. (2021). Understanding the 1 Adoption of Innovations in Agriculture: A Review of Selected Conceptual Models. In 2 Agronomy (Vol. 11, Issue 1). 3 Okoye, O. F., Righi, S., Sermoneta, C., Brunori, G., and Moretti, M. (2025). Towards Digital 4 Farming: Exploring Technological Integration in Agricultural Practices of a sample of 5 Italian livestock farms. Bio-Based and Applied Economics SE-Special Issue 13th AIEAA 6 Conference. 7 Osrof, H. Y., Tan, C. L., Angappa, G., Yeo, S. F., and Tan, K. H. (2023). Adoption of smart 8 farming technologies in field operations: A systematic review and future research 9 agenda. Technology in Society 75: 102400. 10 Patle, G. T., Kumar, M., and Khanna, M. (2019). Climate-smart water technologies for 11 sustainable agriculture: a review. Journal of Water and Climate Change 11: 1455–1466. 12 Paustian, M., and Theuvsen, L. (2017). Adoption of precision agriculture technologies by 13 German crop farmers. Precision Agriculture 18: 701–716. 14 Paxton, K. W., Mishra, A. K., Chintawar, S., Roberts, R. K., Larson, J. A., English, B. C., 15 Lambert, D. M., Marra, M. C., Larkin, S. L., Reeves, J. M., and Martin, S. W. (2011). 16 Intensity of Precision Agriculture Technology Adoption by Cotton Producers. 17 Agricultural and Resource Economics Review 40: 133–144. 18 Pierpaoli, E., Carli, G., Pignatti, E., and Canavari, M. (2013). Drivers of Precision Agriculture 19 Technologies Adoption: A Literature Review. Procedia Technology 8: 61–69. 20 Power, E. F., Kelly, D. L., and Stout, J. C. (2013). Impacts of organic and conventional dairy 21 42 farmer attitude, behaviour and knowledge on farm biodiversity in Ireland. Journal for 1 Nature Conservation 21: 272–278. 2 Sabbagh, M., and Gutierrez, L. (2023). Farmers’ acceptance of a micro-irrigation system: A 3 focus group study. Bio-Based and Applied Economics 12: 221–242. 4 Schulze Schwering, D., Bergmann, L., and Isabel Sonntag, W. (2022). How to encourage 5 farmers to digitize? A study on user typologies and motivations of farm management 6 information systems. Computers and Electronics in Agriculture 199: 107133. 7 Senyolo, M. P., Long, T. B., Blok, V., and Omta, O. (2018). How the characteristics of 8 innovations impact their adoption: An exploration of climate-smart agricultural 9 innovations in South Africa. Journal of Cleaner Production 172: 3825–3840. 10 Shang, L., Heckelei, T., Gerullis, M. K., Börner, J., and Rasch, S. (2021). Adoption and 11 diffusion of digital farming technologies - integrating farm-level evidence and system 12 interaction. Agricultural Systems 190: 103074. 13 Timpanaro, G., Foti, V. T., Cascone, G., Trovato, M., Grasso, A., and Vindigni, G. (2024). 14 Living Lab for the Diffusion of Enabling Technologies in Agriculture: The Case of 15 Sicily in the Mediterranean Context. In Agriculture (Vol. 14, Issue 12). 16 Ungureanu, N., Vlăduț, V., and Voicu, G. (2020). Water Scarcity and Wastewater Reuse in 17 Crop Irrigation. In Sustainability (Vol. 12, Issue 21). 18 Venkatesh, V., and Bala, H. (2008). Technology Acceptance Model 3 and a Research Agenda 19 on Interventions. Decision Sciences 39: 273–315. 20 Venkatesh, V., and Davis, F. D. (2000). A Theoretical Extension of the Technology 21 43 Acceptance Model: Four Longitudinal Field Studies. Management Science 46: 186–204. 1 Venkatesh, V., Morris, M. G., Davis, G. B., and Davis, F. D. (2003). User Acceptance of 2 Information Technology: Toward a Unified View. MIS Quarterly 27: 425–478. 3 Wachenheim, C., Fan, L., and Zheng, S. (2021). Adoption of unmanned aerial vehicles for 4 pesticide application: Role of social network, resource endowment, and perceptions. 5 Technology in Society 64: 101470. 6 Yigezu, Y. A., Mugera, A., El-Shater, T., Aw-Hassan, A., Piggin, C., Haddad, A., Khalil, Y., 7 and Loss, S. (2018). Enhancing adoption of agricultural technologies requiring high 8 initial investment among smallholders. Technological Forecasting and Social Change 9 134: 199–206. 10 44 Appendices 1 2 Table A1. Constructs and Items. 3 Codes Items Perceived Usefulness PU1 This innovation would make my work easier PU2 This technology would make my work faster PU3 This innovation could improve my productivity PU4 By using this technology, I could reduce water consumption in my company Perceived Ease of Use PEU1 This technology should be easy to use PEU2 Using this technology will not require much effort Output Quality OQ1 I expect that the results of using this technology will be excellent OQ2 Using this technology will improve the quality of my products OQ3 By using this system, I would increase the efficiency of my work OQ4 By using this innovation, I would increase the quantity of product in the field SAFA SAFA1 By using this innovation, I could help reduce water consumption SAFA2 With the introduction of this technology, employees could receive training and enhance their knowledge and technical skills SAFA3 By introducing this innovation, I could receive training and improve my technical skills Anxiety ANX1 I get nervous when working with new technologies ANX2 New technologies make me feel uncomfortable ANX3 I am afraid of applying new technologies Technology Self-Efficacy TSE1 I would use this technology easily if someone showed me how to use it TSE2 I would use this innovation easily if I had technical support TSE3 I would use this innovation easily if I were familiar with the system 45 Note: All items were measured on a 7-point scale (from ‘strongly disagree’ to ‘strongly agree’). 1 2 Subjective Norms SN1 People whose opinions matter to me think that I should use this technology SN2 Many producers I know have already adopted this innovation SN3 My customers think that I should use this technology Behavioural Intention BI1 I will definitely use this technology in the near future BI2 I intend to use this technology in the near future BI3 If there were no significant barriers, I would use this system in the near future 46 Table A2. TAM3-extended model: unstandardized beta coefficients and standard errors (S.E.) 1 (n=251). 2 Model 1 Model 2 Beta S.E. Beta S.E. PEU → PU 0.202* 0.100 0.180# 0.096 OQ → PU 0.796*** 0.136 0.848*** 0.136 SAFA → PU 0.236* 0.106 0.179# 0.102 SN → PU 0.040 0.060 0.030 0.059 EMP → PU 0.267*** 0.065 UAA → PU 0.000 0.000 TSE → PEU 0.308*** 0.064 0.308*** 0.064 ANX → PEU -0.098* 0.048 -0.095* 0.048 PU → BI 0.862*** 0.110 0.866*** 0.112 PEU → BI 0.149 0.142 0.130 0.139 SN → BI 0.351*** 0.082 0.334*** 0.082 EMP → BI 0.072 0.101 UAA → BI 0.000# 0.000 BI → BEH 0.256*** 0.039 0.235*** 0.039 EMP → BEH 0.049 0.080 UAA → BEH 0.000** 0.000 Note: PU = Perceived Usefulness; PEU = Perceived Ease of Use; OQ = Output Quality; SAFA = Sustainability 3 Assessment of Food and Agriculture Systems; ANX = Anxiety; TSE = Technology Self-Efficacy; SN = subjective 4 norms; BI = Behavioural Intentions; EMP = number of employees; UAA = average farm size (Utilised agricultural 5 area); BEH = Behaviour. Sign.: *** p<0.001, ** p<0.01, ** p<0.05, # p < 0.10. 6 7