Farm Profitability, Asset-Use Efficiency and Energy-Cost Intensity Across European Farming Systems

Abstract

European agriculture is increasingly required to remain economically viable while improving resource-use efficiency and reducing exposure to rising input and energy costs. This study examines the relationship between farm profitability, asset-use efficiency and energy-cost intensity across different European farming systems. The analysis is based on FADN/FSDN economic indicators reported by country and farm type for the period 2015–2023. Four types of farming were analysed separately: field crop farms, dairy farms, mixed farms and wine farms. Return on assets was used as the dependent variable, while asset turnover, economic size, fixed asset share, productivity indicators and energy intensity were included as explanatory variables. The empirical analysis combined farm-type-specific panel models with several robustness checks, including common time effects, first-difference specifications and a unified farm-type interaction model. The results show that asset turnover is the most consistently significant correlate of profitability, with a positive association across all the analysed farming systems. Energy intensity is negatively associated with profitability, with the most robust evidence identified in field crop and mixed farms. In dairy and wine farms, the association between energy intensity and profitability was not statistically significant at the 5% level in the main fixed-effects models. Fixed asset share was negatively and statistically significantly associated with profitability only in mixed farms. These findings indicate that profitability-related relationships differ across farming systems and that asset-use efficiency and energy-cost exposure should be interpreted within the structural and production characteristics of each farm type. The study does not aim to identify causal effects, but provides comparative evidence on conditional associations between profitability, accounting-based efficiency indicators and resource-use indicators in European agriculture.

1. Introduction

The European Union agri-food system is one of the largest and strategically most important economic systems in the world. It encompasses primary agricultural production, food processing, logistics, distribution, retail, and food consumption, thereby playing a fundamental role in ensuring food security, economic development, and social cohesion across Europe. According to the European Commission, the EU agri-food system generated approximately €900 billion in added value in 2022 and supported around 30 million jobs throughout the food value chain, highlighting its significant contribution to European economic and social stability [1]. Agricultural production in the European Union is carried out on approximately 157 million hectares of utilised agricultural land, representing around 38% of the EU’s total land area [1]. The sector consists of roughly 9 million agricultural holdings, of which more than 90% are family farms [2]. The European agri-food sector is a key component of the EU’s strategic autonomy, ensuring a stable supply of food for approximately 450 million citizens [1]. In addition, the European Union is one of the largest producers and exporters of agri-food products in the world, which makes the agri-food sector an important factor in economic competitiveness and international trade. Beyond its economic role, agriculture contributes to rural development, landscape preservation, biodiversity management, and the provision of public goods that are essential for the sustainability of rural communities.

However, despite ongoing efforts to improve circularity and resource efficiency, many segments of the EU agri-food system continue to exhibit linear patterns based on resource extraction, production, consumption, and waste generation [3]. These patterns contribute to intensive resource use, waste generation, and greenhouse gas emissions [3,4]. At the same time, European agriculture faces increasing pressure to maintain farm profitability while improving the efficiency with which production resources are used.

The European Union agri-food system is characterised by a high level of productivity, technological development and market integration. Supported by the Common Agricultural Policy (CAP), European agriculture has achieved substantial increases in productivity and efficiency over recent decades, ensuring food security and competitiveness in global markets.

However, this development model has also resulted in a growing dependence on natural resources and external inputs, including mineral fertilisers, pesticides, fossil fuels, and animal feed [5]. One of the key characteristics of the EU agri-food system is its resource-intensive nature. Modern agricultural production requires considerable amounts of water, energy, and chemical inputs, accounting for nearly a quarter of total water abstraction in the EU [6], while simultaneously facing increasing constraints related to resource scarcity and climate change. Furthermore, agriculture accounts for approximately 10% of total greenhouse gas emissions in the European Union, with the largest share originating from livestock production, fertiliser use, and land management practices [1]. These indicators highlight the need for improved resource efficiency and reduced environmental impacts.

Despite its high productivity and technological advancement, the EU agri-food system faces considerable sustainability challenges. European agriculture must reconcile economic viability with more efficient and environmentally responsible use of resources. This challenge is driven by the need to maintain agricultural productivity and food security while simultaneously reducing greenhouse gas emissions, biodiversity loss, soil degradation, and the overuse of natural resources [7,8]. This dual pressure highlights the need to examine how farm profitability is associated with resource-use patterns, asset-use efficiency and energy-cost exposure across different farming systems [9]. Against this background, this study examines how farm profitability is associated with asset-use efficiency and energy-cost intensity across field crop, dairy, mixed and wine farming systems in Europe.

Resource losses and food waste represent important sources of economic and environmental inefficiency across the agri-food value chain [10,11]. More than 58 million tonnes of food waste are generated annually in the European Union, while approximately 10% of the food made available to consumers is estimated to be wasted [12]. Although food waste is not directly analysed in the empirical model, it illustrates how the inefficient use of land, water, energy, fertilisers, and other inputs can increase economic costs and the environmental footprint of agricultural and food systems [4,13]. This broader context supports the study’s focus on asset-use efficiency and energy-cost intensity as indicators examined alongside farm profitability. At the same time, rising production costs, climate-related risks, market volatility, and the unequal distribution of value along the supply chain place considerable pressure on farm profitability and long-term resilience [14]. This economic vulnerability may be further reinforced by the concentration of CAP direct payments, with approximately 80% of income support received by the largest 20% of agricultural holdings [15].

These interconnected environmental, economic, and social challenges have intensified the need for agricultural and food systems that are economically viable, resource-efficient, climate-resilient, and environmentally responsible. Reducing food waste has consequently become an important international priority [16], alongside broader efforts to improve resource efficiency and the environmental sustainability of agricultural production [17].

Furthermore, European policy initiatives promote more sustainable and resource-efficient production and consumption patterns. In this context, circular economy principles have become increasingly relevant for agriculture and represent an important element of the European Union’s sustainable development strategies [3,18,19,20]. At the level of the wider economy, these principles seek to retain the value of resources, products, and materials through reuse, recycling, repair, and recovery [21,22,23]. In agriculture, they are reflected in approaches aimed at improving the cycling and efficient use of nutrients, water, energy, and biomass and reintegrating agricultural by-products and residues into production processes [24,25,26,27].

In the last decade, practices such as reducing food waste, valorising by-products, using renewable energy sources, and promoting eco-innovations have increasingly been recognised as drivers of resource efficiency and long-term competitiveness [28]. At the farm level, however, the distinction between linear and circular production should not be understood as absolute because agriculture is inherently based on biological cycles and already involves varying degrees of nutrient, biomass, and by-product reuse. Farming systems differ primarily in the extent to which they improve resource-use efficiency, reduce external input dependence and waste generation, recycle nutrients, and reuse agricultural residues and by-products [21]. The wider implementation of these practices may require cooperation across the entire agri-food value chain [29].

Accordingly, improving the sustainability and resource-use efficiency of the EU agri-food system represents both an environmental and an economic priority. Circular economy principles may support this process, particularly across the wider value chain, but they should not be treated as a single or universally applicable model of farm-level production.

The transition towards more sustainable and resource-efficient agriculture in the European Union is strongly embedded in the European policy framework and is supported by a series of interconnected strategies and policies. The European Green Deal [30], the Farm to Fork Strategy [31], the Circular Economy Action Plan [3], the EU Biodiversity Strategy for 2030 [32] and the Common Agricultural Policy for the period 2023–2027 [33] all emphasise the need to reduce environmental pressures, improve resource efficiency and support more sustainable farming systems.

The European Green Deal represents an umbrella strategic framework for the transition of the European agricultural and food system towards a climate-neutral and resource-efficient economy, thereby indirectly directing the agricultural sector towards more sustainable and circular models of production and consumption, contributing to circular agriculture through the reduction in greenhouse gas emissions, more efficient use of resources, the protection of soil and biodiversity, and encouraging sustainable food production [30]. The Farm to Fork Strategy translates these goals into objectives for agriculture and the food system, including reductions in pesticide and fertiliser use, nutrient losses, and food waste. Its targets include reducing pesticide use by 50%, nutrient losses by 50%, fertiliser use by at least 20%, and increasing organic farming to 25% of agricultural land [31]. The Circular Economy Action Plan further strengthens this framework by promoting waste reduction, promoting resource reuse, recycling and biowaste valorization, which also strengthens material flows in agriculture [3]. The Common Agricultural Policy provides financial support to farmers for the implementation of ecological and agroecological measures, which enables the practical implementation of circular principles [33]. In parallel, the EU Biodiversity Strategy for 2030 additionally supports this process through the preservation of ecosystems, the restoration of degraded land, the return of nutrients from biowaste to the soil and back to the agricultural system, and generally sustainable management of natural resources [32].

Together, these strategies indicate that future agricultural development must be assessed not only through production and income indicators, but also through indicators related to emissions, resource use, and environmental performance. This policy framework supports the study’s focus on farm profitability, asset-use efficiency and energy-cost intensity. Asset turnover reflects the productive use of the farm asset base, while energy intensity captures energy-cost exposure relative to output. These indicators do not directly measure the implementation or effects of individual EU policies, but they reflect economic and resource-use dimensions relevant to their broader objectives.

However, the economic implications of this transition remain complex [11]. On the one hand, circular and resource-efficient practices may improve farm profitability by reducing input costs, increasing energy and nutrient efficiency, improving soil fertility and creating additional value from residues and by-products [25]. On the other hand, despite these advantages, circular practices remain unevenly implemented across countries, sectors, and farm types, because the adoption of such practices often requires investment, knowledge, institutional support and access to appropriate technologies [22,33]. For many farms, especially small and medium-sized holdings, profitability is not only an outcome of sustainable transformation, but also a precondition for adopting circular and environmentally oriented practices. Also, their implementation may generate uneven economic effects across farmers, consumers, and supply-chain actors. Consequently, achieving a balance between competitiveness, resilience, and environmental stewardship has become a central objective for the future development of European agriculture and food systems [31]. Therefore, the relationship between farm profitability, asset-use efficiency and energy-cost intensity cannot be assumed in advance; it needs to be empirically examined across different farming systems.

The Farm Accountancy Data Network (FADN) provides an important basis for analysing this relationship. Traditionally, FADN has been used to assess farm income, productivity, profitability and the effects of agricultural policy. More recently, it has also been used to construct broader sustainability indicators, including energy intensity, input-use intensity, eco-efficiency and resilience indicators. The ongoing transition from FADN to the Farm Sustainability Data Network (FSDN) further strengthens the relevance of this approach, as future farm-level data systems are expected to provide a broader basis for analysing the economic, environmental and social dimensions of farm sustainability. Previous research has shown that FADN-based data can be used to examine the relationship between economic performance and resource-use indicators.

Despite these advances, important gaps remain. Many existing studies are limited to single countries, specific sectors or selected farm types. Resource-use performance is often measured through indirect indicators such as input costs, energy intensity, labour productivity or land productivity, while comparative evidence across structurally different farming systems remains limited. In addition, the relationship between farm economic viability, asset-use efficiency and energy-cost intensity remains insufficiently explored across different types of farming. This is particularly relevant in European agriculture, where improvements in environmental and resource-use performance must be compatible with the economic viability of farms.

In this study, profitability is not treated as a measure of circularity. Rather, it represents the economic viability dimension against which resource-use patterns are assessed. The conceptual relationship is potentially two-directional. More efficient use of energy, nutrients, materials, and agricultural by-products may reduce production costs or generate additional value, thereby supporting profitability. At the same time, the adoption of resource-efficient and environmentally oriented practices may require investment, knowledge, and technological capacity, making farm profitability and financial viability important preconditions for their implementation.

Accordingly, the present study does not measure farm circularity directly. Instead, it examines whether profitability is associated with asset-use efficiency and energy-cost intensity across different farming systems. These indicators capture selected economic and resource-use dimensions relevant to sustainable agricultural performance, but they should not be interpreted as a comprehensive measure of circularity. The study is therefore positioned within the broader framework of sustainable and resource-efficient agriculture rather than as a direct assessment of farm-level circularity.

This study examines the relationship between farm profitability, asset-use efficiency and energy-cost intensity in European agriculture using FADN/FSDN economic indicators for the period 2015–2023. The analysis focuses on field crop, dairy, mixed and wine farms and examines whether profitability is conditionally associated with more efficient use of assets, differences in farm structure, productivity indicators and energy costs relative to output. The study does not aim to identify causal effects. Instead, it provides comparative evidence on how profitability-related accounting and resource-use indicators are associated across structurally different European farming systems. Such evidence may support future policy and farm-level assessments aimed at maintaining economic viability while improving resource-use efficiency.

2. Literature Review

The relationship between farm profitability, resource-use efficiency and sustainability-oriented farm performance has become an increasingly important research topic in the context of sustainable agricultural transformation. Contemporary agriculture is expected to maintain economic viability while improving the efficiency of resource use and reducing dependence on external inputs. This dual objective has contributed to the growing use of concepts such as eco-efficiency, agro-ecological efficiency and sustainability-oriented farm performance.

Eco-efficiency in agriculture is commonly understood as the ability to achieve higher agricultural output with lower use of land, water, nutrients, energy, labour and capital. Keating et al. [34] conceptualised eco-efficient agriculture as a framework that integrates economic and ecological dimensions of performance, emphasising that productivity improvements should not be achieved through excessive use of environmentally burdensome inputs. Similarly, Lal [35] argued that sustainable agro-ecosystems should improve resource-use efficiency, increase agronomic productivity and enhance ecosystem services, particularly through soil carbon sequestration, conservation agriculture, residue retention and integrated nutrient management. These conceptual contributions suggest that farm performance should not be assessed only through profitability or productivity indicators, but also through indicators reflecting resource-use efficiency and input dependence.

Recent reviews confirm that agricultural eco-efficiency has been analysed using different methodological approaches, including data envelopment analysis, stochastic frontier analysis, life cycle assessment and composite indicators. Wang et al. [36] emphasised that agricultural eco-efficiency combines economic output, resource consumption and environmental impacts, but also noted that there is still no unified approach to the selection of indicators and quantitative methods. This methodological diversity reflects the multidimensional character of sustainable agricultural performance and supports the need for empirical studies that combine economic indicators with resource-use indicators such as energy use, input intensity, labour productivity and land productivity.

A substantial part of the existing empirical literature uses efficiency-based methods to assess whether farms can maintain economic output while reducing resource use and environmental pressure. Gołaś et al. [37], using Polish FADN data, showed that commercial farms could reduce inputs and environmental pressures without reducing output, while more eco-efficient farms were generally larger, more productive and more profitable. Bonfiglio et al. [38] analysed arable farms in Italy and found that farms had substantial potential to reduce nitrogen, phosphorus and pesticide-related pressures while maintaining the same level of value added. Godoy-Durán et al. [39] also confirmed that horticultural farms could reduce aggregate environmental pressures without sacrificing economic performance. These studies indicate that economic and environmental objectives may be compatible, but only when farms improve the efficiency with which resources and inputs are used.

However, other studies suggest that the relationship between economic performance and environmental sustainability is not always synergistic. Špička et al. [40] found a trade-off between economic and environmental sustainability in Czech agriculture, particularly when different farm types were analysed together. Their results showed that field-crop farms achieved high economic performance but weaker environmental sustainability, while extensive livestock systems showed the opposite pattern. Arru et al. [41] also identified trade-offs between economic and environmental sustainability in several Italian crop and livestock sectors. These findings underline the importance of analysing farm types separately, since aggregate results may hide important production-system-specific patterns.

The role of farm type is particularly important in FADN/FSDN-based research. Different farming systems differ in their production technology, asset structure, input requirements and energy dependence. Crop farms, dairy farms, mixed farms and permanent crop farms may therefore exhibit different relationships between profitability and sustainability indicators. For example, Martinsson & Hansson [42] showed that dairy farm eco-efficiency depends strongly on greenhouse gas emission targets, and that farms may appear relatively efficient while still failing to follow a trajectory consistent with absolute emission reduction objectives. Niedermayr et al. [43] demonstrated that organic and integrated dairy systems in Austria generally achieved better environmental performance than conventional systems, but their economic viability depended partly on public payments and market premiums. These studies illustrate the specific economic and environmental characteristics of dairy systems, but capital intensity is not limited to livestock production. Modern large-scale field crop farms may also have a substantial capital base comprising land, advanced machinery, precision-agriculture technologies, irrigation and drainage infrastructure, storage and drying facilities, and digital management systems. Differences among farming systems should therefore be understood primarily in terms of the composition, scale, and utilisation of their assets rather than through a simple distinction between capital-intensive livestock farms and less capital-intensive crop farms.

FADN has been widely used as a harmonised source for analysing farm income, productivity, efficiency and sustainability. Kelly et al. [44] emphasised that FADN provides a representative and harmonised platform for economic farm-level analysis across the EU, but also noted that its original structure is insufficient for a complete assessment of environmental and social sustainability. Similarly, Tomaš Simin et al. [45] argued that FADN can serve as a useful basis for developing economic, ecological and social sustainability indicators, particularly in the context of the transition toward the Farm Sustainability Data Network. Their framework includes profitability, productivity, liquidity and stability indicators, as well as environmental-pressure indicators related to livestock density, greenhouse gas emissions, fertiliser and pesticide use, energy use and biodiversity. This supports the use of FADN/FSDN data as a basis for linking farm profitability with resource-use and sustainability-related indicators.

Several studies have attempted to extend FADN-based analyses by introducing environmental or sustainability-related indicators. Dabkienė et al. [46] developed an agri-environmental footprint indicator using Lithuanian FADN data and included variables related to fertiliser use, crop protection, greenhouse gas emissions, energy intensity, water use, biodiversity and livestock density. Syp et al. [47] used FADN data to assess environmental performance in farms participating in agri-environmental schemes and showed that such farms used fewer synthetic fertilisers and crop protection products and generated lower greenhouse gas emissions per hectare. Uthes & Herrera [48], using FADN data combined with additional sustainability indicators from the FLINT project, showed that higher external input intensity was negatively associated with economic, environmental and overall sustainability indicators. These studies demonstrate the value of combining farm accountancy data with environmental indicators, while also showing the limitations of relying only on monetary or proxy variables.

The literature also highlights the importance of profitability indicators in sustainability-oriented farm analysis. Farm profitability has often been analysed through return on assets, farm net value added, gross farm income or composite economic sustainability indices. Coppola et al. [49] showed that economic sustainability cannot be captured by a single profitability indicator, since efficiency, family labour opportunity costs and comparable income provide different insights into farm viability. Prigoreanu et al. [50] similarly argued that economic sustainability should be assessed through multiple indicators, including profitability, capitalization, liquidity, subsidy dependence and technical-economic efficiency. Reziti [51] used DuPont analysis to decompose farm profitability into profit margin, asset turnover and financial leverage, confirming the importance of asset-use efficiency in explaining farm returns.

Asset turnover is particularly relevant for understanding farm profitability because it reflects the ability of farms to generate output from their asset base. Grzelak & Staniszewski [52] found that production scale and productivity of intermediate consumption were key determinants of achieving a market-comparable return on assets in EU farms. Their results also suggested that more profitable farms may be associated with higher livestock density, mineral fertiliser use and environmental pressure. Miljatović et al. [53], analysing Serbian family farms, found that asset turnover was the only determinant with a positive and statistically significant effect on economic viability across all farm types. These findings support the inclusion of asset turnover as a central explanatory variable in profitability modelling. Accordingly, asset turnover is expected to be positively associated with ROA in the present study.

Input intensity and energy intensity also represent important links between profitability and resource use efficiency. Uthes & Herrera [48] found that higher input intensity was negatively associated with farm sustainability indicators, while Oliveira et al. [54] showed that energy use and electricity consumption are important sources of environmental burden in dairy systems. Almeida et al. [55] demonstrated that circular and biorefinery-based solutions are not automatically more eco-efficient, since their performance depends on energy requirements, operating costs, avoided products and process yields. These findings indicate that higher input and energy use may reduce both economic performance and resource-use efficiency, particularly when additional resource use does not generate proportional increases in output. Accordingly, energy intensity is expected to be negatively associated with ROA when higher energy costs are not accompanied by proportional increases in output.

The connection between environmental performance and farm economic performance has also received increasing attention. Bazzani et al. [56] used FADN-compatible Italian data to estimate CO₂ abatement costs in arable crop systems and showed that emission reductions may imply income losses for farmers, although low-cost reductions were feasible in some regional and farm-size contexts. Stevanović et al. [57] found that higher emissions relative to revenues were associated with lower return on assets in Serbian agricultural enterprises. Novaković et al. [58] also showed that the relationship between farm profitability and sustainability-related performance indicators may differ across farm types, confirming the importance of analysing production systems separately rather than treating agriculture as a homogeneous sector. These findings indicate that the relationship between farm profitability and sustainability-oriented performance may depend on production scale and intensity, technology, input and output prices, subsidies, asset structure and the costs of environmental adjustment. Although the present study does not estimate farm-level emissions, this literature provides an important background for examining how profitability is associated with resource-use and energy-cost indicators.

In the present study, the relationship between profitability and resource-use indicators is not interpreted as a direct causal link. Asset turnover, labour productivity, land productivity and energy intensity are accounting-based or resource-use indicators constructed from FADN/FSDN Standard Results. Their relationship with ROA may reflect differences in production technology, asset structure, cost exposure, scale, farm specialisation and country-specific conditions. Therefore, the estimated relationships are interpreted as conditional associations rather than as direct causal effects.

Despite the growing literature on agricultural eco-efficiency and farm sustainability, several gaps remain. First, many studies focus either on economic performance or on environmental efficiency, while fewer studies explicitly analyse how farm profitability is associated with asset-use efficiency and energy-cost intensity across different types of farming. Second, FADN-based studies often rely on broad input-cost indicators or single-country datasets, while comparative studies using harmonised European FADN/FSDN indicators across multiple production systems remain limited. Third, the literature shows that the relationship between profitability and sustainability-related indicators is strongly dependent on farm type, but comparative analyses across field crop, dairy, mixed and wine farming systems remain insufficiently developed.

The present study addresses these gaps by analysing the relationship between farm profitability, asset-use efficiency and energy-cost intensity across four different farm types: field crop farms, dairy farms, mixed farms and wine farms. By examining FADN/FSDN economic indicators across structurally different European farming systems, the study contributes to the literature on sustainable farm performance and provides comparative evidence on how profitability-related accounting and resource-use indicators differ across production systems.

3. Materials and Methods

3.1. Data Sources and Sample

The empirical analysis was based on data obtained from the Farm Accountancy Data Network/Farm Sustainability Data Network public database of the European Commission for the period 2015–2023 [57]. The FADN/FSDN database provides standard results describing the economic situation of farms by year, member state, type of farming and economic size class. In this study, the analysis was conducted separately for four types of farming: field crop farms, dairy farms, mixed farms and wine farms. This approach was used because different agricultural production systems differ in terms of production technology, asset structure, input use and energy-cost exposure.

Field crop farms and mixed farms include 241 observations each and cover 27 countries. Dairy farms include 223 observations and 25 countries, while wine farms include 135 observations and 15 countries. The field crop, dairy and mixed farm samples are unbalanced panels, whereas the wine farm sample is a balanced panel covering the complete period from 2015 to 2023. In the field crop, dairy and mixed-farm panels, the unbalanced structure mainly reflects the absence of Malta for 2022 and 2023. The United Kingdom was excluded from the analysis to ensure a consistent geographical scope focused on the current European Union member states throughout the analysed period. Differences in the size and country coverage of the remaining panels reflect the availability of FADN/FSDN Standard Results by country, year, and type of farming. No additional discretionary country-exclusion criterion was applied. Detailed country coverage is presented in Appendix A 

Table A1. Country coverage and panel structure by farm type.

Farm Type	Countries Included	Countries	Observations	Panel
Field crop farms	Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden	27	241	Unbalanced
Dairy farms	Austria, Belgium, Bulgaria, Croatia, Czechia, Denmark, Estonia, Finland, France, Germany, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden	25	223	Unbalanced
Mixed farms	Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden	27	241	Unbalanced
Wine farms	Austria, Bulgaria, Croatia, Cyprus, Czechia, France, Germany, Greece, Hungary, Italy, Luxembourg, Portugal, Romania, Slovenia, Spain	15	135	Balanced

. All monetary variables were used as reported in the FADN/FSDN Standard Results database and were not additionally adjusted for inflation or purchasing power parity. This approach was adopted because the study uses harmonised country–farm-type indicators and focuses on reported panel associations rather than on cross-country real-income comparisons. However, this is acknowledged as a limitation, particularly because the period 2015–2023 includes the COVID-19 pandemic, energy-price shocks, the Russia–Ukraine conflict and high inflation. To reduce the influence of common shocks during this period, the empirical analysis included year effects and additional robustness checks.

No winsorization procedure was applied. The analysis used the publicly reported FADN/FSDN Standard Results after excluding observations with missing values required for the calculation of the model variables. Robust covariance estimators, two-way fixed effects, first-difference models and sensitivity checks were used to assess the stability of the results across alternative specifications.

3.2. Variable Construction

The dependent variable in the econometric analysis was return on assets (ROA). ROA was selected as the main profitability indicator because it measures the ability of farms to generate income from their asset base. Total assets were measured using the FADN/FSDN indicator SE436, which represents the closing valuation of the assets owned by the agricultural holding. According to the FADN/FSDN Standard Results framework, total assets are calculated as the sum of total fixed assets (SE441) and total current assets (SE465). Total fixed assets comprise land, permanent crops and quotas (SE446), farm buildings (SE450), machinery and equipment (SE455), and breeding livestock (SE460). Total current assets comprise non-breeding livestock (SE470), stocks of agricultural products (SE475), and other circulating capital (SE480). Agricultural land is therefore included when it is owned by the holding, while breeding and non-breeding livestock are included in the corresponding fixed- and current-asset categories. Assets used by the holding but not owned by it, including rented agricultural land, are not included in SE436. All asset values refer to the closing valuation at the end of the accounting year. The explanatory variables were selected ex ante on the basis of the farm profitability and sustainability literature to represent conceptually distinct dimensions of farm economic performance: farm scale, asset-use efficiency, asset structure, factor productivity and energy-cost exposure. Economic size (ESS) captures differences in farm scale and potential economies or diseconomies of scale, while asset turnover (ATR) measures the ability of a farming system to generate output from its asset base. The inclusion of ATR is consistent with the DuPont framework and previous farm-level studies identifying asset-use efficiency as an important component of profitability and economic viability [51,52,53]. Fixed asset share (FAS) captures capital intensity and the composition of the asset base. A higher fixed asset share may support profitability through productive infrastructure, machinery, and long-term production capacity, but it may also increase fixed-cost exposure and reduce financial flexibility; consequently, its expected effect is ambiguous. Labour productivity and land productivity represent the efficiency with which labour and agricultural land are converted into output and were generally expected to be positively associated with profitability, although their effects may depend on production scale, technology, and the type of farming system [49,52]. Energy intensity (ENINT) measures energy expenditure relative to total output and was expected to be negatively associated with ROA when higher energy costs are not accompanied by a proportional increase in production. Its inclusion is supported by previous research showing that higher input and energy intensity may be associated with lower economic performance and reduced resource-use efficiency [48,54,55]. Debt ratio (DEBT) and input intensity (INPUTINT) were retained in the descriptive analysis to provide additional information on the financial structure and input dependence of the analysed farming systems, but they were not included in the final econometric specifications. INPUTINT is defined as total inputs relative to total output and is therefore closely related to the accounting structure of farm net income, which already reflects the relationship between farm output and production costs. Its inclusion could create an excessively direct and partly mechanical association with ROA and could also overlap with ENINT because energy costs are a component of total inputs. ENINT was therefore retained as the more specific variable directly related to energy dependence and resource-use efficiency.

DEBT is relevant for describing the financial structure of the analysed farm types but is not central to the research question, which focuses on profitability, asset-use efficiency and energy-cost intensity. In addition, both DEBT and ROA use total assets in the denominator, creating a potential definitional overlap. DEBT was therefore retained as a descriptive indicator but excluded from the final econometric specifications to maintain a focused model and avoid introducing an additional dimension of financial-structure interpretation.

The construction of the variables also requires careful consideration of possible accounting and definitional relationships. ROA is calculated as farm net income divided by total assets, whereas ATR is calculated as total output divided by total assets. Their common denominator creates an inherent accounting relationship consistent with the DuPont framework, in which asset turnover represents one component of profitability. Accordingly, ATR is not interpreted as an independent causal determinant of ROA, but as an asset-use efficiency indicator whose conditional association with profitability is examined after controlling for the remaining variables. Similarly, ENINT is calculated as energy costs relative to total output, while energy costs are part of the cost structure underlying farm net income. Therefore, the coefficient of ENINT is also interpreted as a conditional association rather than as evidence of a direct causal effect on farm profitability (

Table 1. Definition of variables used in the analysis.

Variable	Definition	Formula	Interpretation/Unit
ROA	Return on assets	ROA = SE420/SE436	Farm net income per EUR of total assets
ESS	Economic size	ESS = SE005	Economic size, thousand EUR per farm
ATR	Asset turnover	ATR = SE131/SE436	Total output generated per EUR of total assets
FAS	Fixed asset share	FAS = SE441/SE436	Share of fixed assets in total assets
DEBT	Debt ratio	DEBT = SE485/SE436	Share of liabilities in total assets
LABPROD	Labour productivity	LABPROD = SE131/SE010	Total output per annual work unit
LANDPROD	Land productivity	LANDPROD = SE131/SE025	Total output per hectare of utilised agricultural area
INPUTINT	Input intensity	INPUTINT = SE270/SE131	Total inputs relative to total output
ENINT	Energy intensity	ENINT = SE345/SE131	Energy costs relative to total output

Note: SE005 denotes economic size; SE010 total labour input; SE025 total utilised agricultural area; SE131 total output; SE270 total inputs; SE345 energy costs; SE420 farm net income; SE436 total assets; SE441 total fixed assets; SE485 total liabilities.

).

3.3. Econometric Analysis

The empirical analysis was conducted in several steps. First, descriptive statistics were calculated for all variables included in the study. In the next step, Pearson correlation coefficients were calculated in order to examine the initial bivariate relationships among the variables. The correlation analysis was used as a preliminary diagnostic tool, particularly for identifying strong associations among explanatory variables and potential multicollinearity issues. However, correlation coefficients were not used for causal interpretation, since they do not control for unobserved country-specific heterogeneity, common time effects or dynamic relationships among variables. Before estimating the panel models, the stationarity of the main variables was examined using the Im–Pesaran–Shin panel unit root test. This test is suitable for heterogeneous panels because it allows the autoregressive parameters to differ across cross-sectional units. The null hypothesis of the IPS test assumes the presence of a unit root, while rejection of the null hypothesis indicates stationarity [59]. The test was applied separately for each farm type because the panels differ in structure, number of countries and production characteristics. The IPS results were treated as diagnostic evidence rather than as an automatic rule for transforming or excluding every variable for which the unit-root null hypothesis could not be rejected. This approach was adopted because the time dimension of the panels is short and the statistical power of panel unit root tests is limited in panels covering only nine years.

Because the IPS results indicated mixed stationarity patterns and because European country panels may be affected by common shocks, additional diagnostics were performed. Cross-sectional dependence was examined using the Pesaran CD test. Since cross-sectional dependence may reduce the reliability of first-generation panel unit root tests, the IPS tests were complemented with second-generation panel unit root tests based on the Pesaran CIPS/CADF approach. These tests were used to assess whether the level-form estimates should be interpreted cautiously and whether additional robustness checks were needed.

The main econometric analysis was based on panel regression models estimated separately for each farm type. The baseline fixed-effects model was used to examine the relationship between farm profitability and the selected explanatory variables. The dependent variable was return on assets, while the explanatory variables included economic size, asset turnover, fixed asset share, labour productivity, land productivity, and energy intensity. The baseline fixed-effects specification can be expressed as follows:

$${ROA}_{it}={\alpha }_i+{\beta }_1{ESS}_{it}+{\beta }_2{ATR}_{it}+{\beta }_3{FAS}_{it}+{\beta }_4{LABPROD}_{it}+{\beta }_5{LANDPROD}_{it}+{\beta }_6{ENINT}_{it}+{\epsilon }_{it}$$

where i denotes the country, t denotes the year, ${\alpha }_i$ represents unobserved time-invariant country-specific effects, and ${\epsilon }_{it}$ is the idiosyncratic error term. In the two-way fixed-effects robustness specification, year effects ${\lambda }_t$ were additionally included to control for shocks common to all countries in a particular year.

Given the accounting and definitional relationships among some of the indicators, the estimated coefficients are interpreted as conditional associations rather than as direct causal effects. The fixed-effects model controls for unobserved time-invariant country-specific characteristics, such as structural differences in agriculture, institutional conditions, production environment and long-term policy settings.

Random-effects models were also estimated and compared with fixed-effects models using the Hausman specification test. The Hausman test examines whether the unobserved individual effects are correlated with the explanatory variables. If the test is statistically significant, the random-effects estimator is inconsistent and the fixed-effects estimator is preferred [60]. In this study, the Hausman test was used as the main criterion for choosing between fixed-effects and random-effects specifications.

Cluster-robust standard errors were applied in the fixed-effects models in order to account for heteroskedasticity and within-country correlation.

Given the possibility of common shocks across European countries, two-way fixed-effects models were estimated as an important robustness specification:

$${ROA}_{it}={\alpha }_i+{\lambda }_t+{\beta }_1{ESS}_{it}+{\beta }_2{ATR}_{it}+{\beta }_3{FAS}_{it}+{\beta }_4{LABPROD}_{it}+{\beta }_5{LANDPROD}_{it}+{\beta }_6{ENINT}_{it}+{\epsilon }_{it}$$

where ${\lambda }_t$ denotes year fixed effects. The two-way fixed-effects specification was used as a common-time-effects specification and as a robustness check against cross-sectional dependence. These models include both country and year effects and therefore control not only for unobserved time-invariant country characteristics, but also for shocks common to all countries in a given year, such as changes in input prices, policy conditions, market disturbances, weather-related shocks or other macroeconomic factors.

Given the mixed stationarity properties of several variables and the short time dimension of the panels, first-difference models were estimated to examine whether the main findings were sensitive to estimation in levels. This was particularly important because the IPS results indicated that some explanatory variables were non-stationary in specific farm-type panels. First-difference model specification:

$${\Delta ROA}_{it}={\lambda }_t+{\beta }_1{\Delta ESS}_{it}+{\beta }_2{\Delta ATR}_{it}+{\beta }_3{\Delta FAS}_{it}+{\beta }_4{\Delta LABPROD}_{it}+{\beta }_5{\Delta LANDPROD}_{it}+{\beta }_6{\Delta ENINT}_{it}+{\Delta \epsilon }_{it}$$

where Δ denotes the first difference between two consecutive years. This specification was not used to replace the main models, but to assess whether the main conclusions were sensitive to the level specification and to reduce the risk that the results were driven by non-stationary variables in levels.

Variance inflation factors were calculated in order to assess multicollinearity among explanatory variables.

Finally, in order to formally compare coefficients across farm types, a unified farm type–country–year panel was estimated. This model was introduced because statistical significance in one farm-type-specific model and insignificance in another does not by itself imply that the corresponding coefficients are statistically different across farm types. The unified model included farm type × country fixed effects and year fixed effects, allowing the main coefficients to vary by farm type. Wald tests were then used to examine whether the farm-type-specific coefficients were statistically different from each other. As an additional sensitivity analysis, the unified interaction model was also estimated on the common-country sample, consisting only of countries observed in all four farm-type panels.

Because several variables are accounting-related and because the available data do not contain exogenous shocks, product prices, input prices, weather anomalies or detailed policy-payment controls, the estimated coefficients are interpreted as conditional associations rather than as direct causal effects. The analysis is therefore not intended to identify fully exogenous determinants of farm profitability. Instead, it examines how profitability is associated with asset-use efficiency and energy-cost intensity across different European farming systems. This distinction is particularly important because ROA and ATR share total assets as a denominator, while ENINT reflects energy costs relative to output and energy costs are part of the accounting structure underlying farm net income.

All statistical analyses were performed in R. Data preparation and econometric estimation were conducted using the packages readxl, dplyr, stringr, plm, lmtest, sandwich, car and fixest. The plm package was used for estimating fixed-effects, random-effects and first-difference panel models [61], while fixest was used for the unified farm-type interaction models and Wald tests.

4. Results

4.1. Descriptive Statistics

The descriptive statistics were calculated separately for four farm types: field crop farms, dairy farms, mixed farms and wine farms. This approach allows the comparison of profitability, farm structure, input use and energy-cost intensity across different agricultural production systems. The panel structure differs across farm types. Field crop and mixed farms include 241 observations each; dairy farms include 223 observations, while wine farms include 135 observations. The wine farm sample represents a balanced panel, whereas the remaining farm types are based on unbalanced panels.

Table 2. Descriptive statistics of selected variables by farm type.

Farm Type	ROA Mean ± SD	ESS Mean ± SD	ATR Mean ± SD	FAS Mean ± SD	DEBT Mean ± SD	INPUTINT Mean ± SD	ENINT Mean ± SD
Field crops	0.0691 ± 0.0518	100.61 ± 84.64	0.2521 ± 0.1455	0.8026 ± 0.1264	0.1633 ± 0.1371	0.9572 ± 0.2010	0.0920 ± 0.0314
Dairy farms	0.0834 ± 0.0578	265.23 ± 235.08	0.3517 ± 0.1491	0.7893 ± 0.0972	0.2280 ± 0.1817	0.9457 ± 0.1577	0.0648 ± 0.0208
Mixed farms	0.0674 ± 0.0473	166.82 ± 213.77	0.2734 ± 0.1381	0.7781 ± 0.1047	0.1751 ± 0.1550	0.9608 ± 0.1745	0.0797 ± 0.0232
Wine farms	0.0890 ± 0.0331	69.78 ± 67.70	0.2337 ± 0.0715	0.7340 ± 0.1391	0.1064 ± 0.0940	0.7393 ± 0.1499	0.0503 ± 0.0212

Note: ROA—return on assets; ESS—economic size; ATR—asset turnover; FAS—fixed asset share; DEBT—debt ratio/leverage; INPUTINT—input intensity; ENINT—energy intensity.

presents the descriptive statistics of the main variables used in the econometric analysis.

The results show clear differences in profitability across farm types. Wine farms recorded the highest average ROA, followed by dairy farms. Field crop farms and mixed farms had lower average profitability, with mixed farms showing the lowest mean ROA among the analysed groups. This indicates that specialised wine and dairy production systems were, on average, more profitable than field crop and mixed farming systems during the observed period.

The results also point to important structural differences. Dairy farms had the highest average economic size and the highest asset turnover, reflecting their relatively large production and asset base. Field crop farms had the highest average fixed asset share, while wine farms had the lowest debt ratio and the lowest input and energy intensity. These findings indicate that the analysed wine farms used lower amounts of inputs and energy relative to their total output than the other farm types. Overall, the descriptive statistics confirm that profitability, asset structure, input use and energy-cost intensity differ considerably across farm types, supporting the separate econometric analysis of each production system.

4.2. Correlation Analysis

Pearson correlation coefficients were calculated in order to examine the initial bivariate relationships among the main variables included in the econometric models. The full correlation matrices are presented in Appendix A 

Table A2. Pearson correlation matrix for field crop farms.

Variable	ROA	ESS	ATR	FAS	LABPROD	LANDPROD	ENINT
ROA	1.000	−0.317	0.543	−0.242	−0.356	−0.211	−0.035
ESS	−0.317	1.000	0.316	−0.324	0.600	0.298	−0.416
ATR	0.543	0.316	1.000	−0.746	−0.194	−0.380	0.012
FAS	−0.242	−0.324	−0.746	1.000	0.002	0.381	0.070
LABPROD	−0.356	0.600	−0.194	0.002	1.000	0.428	−0.567
LANDPROD	−0.211	0.298	−0.380	0.381	0.428	1.000	−0.297
ENINT	−0.035	−0.416	0.012	0.070	−0.567	−0.297	1.000

,

Table A3. Pearson correlation matrix for dairy farms.

Variable	ROA	ESS	ATR	FAS	LABPROD	LANDPROD	ENINT
ROA	1.000	−0.500	0.494	−0.312	−0.375	−0.060	0.029
ESS	−0.500	1.000	0.059	−0.028	0.666	0.033	−0.147
ATR	0.494	0.059	1.000	−0.655	−0.284	−0.038	0.451
FAS	−0.312	−0.028	−0.655	1.000	0.212	0.193	−0.346
LABPROD	−0.375	0.666	−0.284	0.212	1.000	0.143	−0.598
LANDPROD	−0.060	0.033	−0.038	0.193	0.143	1.000	−0.118
ENINT	0.029	−0.147	0.451	−0.346	−0.598	−0.118	1.000

,

Table A4. Pearson correlation matrix for mixed farms.

Variable	ROA	ESS	ATR	FAS	LABPROD	LANDPROD	ENINT
ROA	1.000	−0.332	0.319	−0.361	−0.431	−0.112	0.069
ESS	−0.332	1.000	0.469	−0.166	0.505	0.199	−0.181
ATR	0.319	0.469	1.000	−0.710	−0.136	−0.172	0.278
FAS	−0.361	−0.166	−0.710	1.000	0.167	0.279	−0.268
LABPROD	−0.431	0.505	−0.136	0.167	1.000	0.362	−0.527
LANDPROD	−0.112	0.199	−0.172	0.279	0.362	1.000	−0.293
ENINT	0.069	−0.181	0.278	−0.268	−0.527	−0.293	1.000

and

Table A5. Pearson correlation matrix for wine farms.

Variable	ROA	ESS	ATR	FAS	LABPROD	LANDPROD	ENINT
ROA	1.000	0.040	0.523	−0.113	0.192	0.140	−0.360
ESS	0.040	1.000	0.610	−0.669	0.847	0.422	−0.434
ATR	0.523	0.610	1.000	−0.594	0.586	0.283	−0.505
FAS	−0.113	−0.669	−0.594	1.000	−0.649	−0.231	0.439
LABPROD	0.192	0.847	0.586	−0.649	1.000	0.710	−0.582
LANDPROD	0.140	0.422	0.283	−0.231	0.710	1.000	−0.553
ENINT	−0.360	−0.434	−0.505	0.439	−0.582	−0.553	1.000

, separately for each farm type. The results indicate that the relationship between profitability and explanatory variables differs across production systems. Asset turnover was positively correlated with ROA in all farm types, with the strongest association observed in field crop and wine farms. This indicates a positive bivariate association between asset-use efficiency and ROA. Input and energy-related indicators showed different patterns across farm types. In field crop, mixed and wine farms, ROA was negatively associated with input intensity, while the relationship between ROA and energy intensity was generally weak or negative. In wine farms, input intensity had the strongest negative correlation with ROA, indicating a negative bivariate association between input use relative to output and profitability in this production system. Correlation coefficients should be interpreted only as preliminary evidence, since they do not control for unobserved country-specific heterogeneity, common time shocks or dynamic relationships.

4.3. Panel Unit Root Tests and Additional Diagnostics

Before estimating the econometric models, the stationarity of the variables was examined using the Im–Pesaran–Shin (IPS) panel unit root test. The IPS test was applied separately for each farm type in order to assess whether the variables included in the models exhibit stationary behaviour over time. The null hypothesis of the IPS test assumes the presence of a unit root, while rejection of the null hypothesis indicates stationarity.

Table 3. IPS panel unit root test results by farm type.

Variable	Field Crops	Dairy Farms	Mixed Farms	Wine Farms
ROA	−3.4126 (0.0003)	−2.3839 (0.0086)	−3.1658 (0.0008)	−4.3410 (<0.0001)
ESS	−6.6332 (<0.0001)	−4.7152 (<0.0001)	−3.9528 (<0.0001)	−5.1798 (<0.0001)
ATR	−4.4146 (<0.0001)	1.0664 (0.8569)	−0.5563 (0.2890)	−6.2718 (<0.0001)
FAS	−0.5528 (0.2902)	2.3861 (0.9915)	1.7549 (0.9604)	−3.3695 (0.0004)
LABPROD	3.5039 (0.9998)	8.1370 (1.0000)	6.9452 (1.0000)	0.1398 (0.5556)
LANDPROD	−0.4307 (0.3333)	7.2161 (1.0000)	4.4859 (1.0000)	−4.6383 (<0.0001)
ENINT	−7.8152 (<0.0001)	−4.8420 (<0.0001)	−6.4532 (<0.0001)	−0.2839 (0.3882)

Note: IPS statistics are reported with p-values in parentheses. The null hypothesis assumes the presence of a unit root. Values with p < 0.05 indicate stationarity.

presents the IPS test statistics and corresponding p-values for the main variables included in the econometric analysis.

Because first-generation panel unit root tests may be affected by cross-sectional dependence, additional diagnostic tests were performed. The Pesaran CD test was used to examine whether residual cross-sectional dependence was present in the fixed-effects specifications. The results are presented in

Table 4. Pesaran CD test for cross-sectional dependence.

Farm Type	Country FE Model	p-Value	Two-Way FE Model	p-Value	First-Difference Model	p-Value
Field crops	9.0254	<0.001	−0.8812	0.3782	−0.7952	0.4265
Dairy farms	19.3696	<0.001	−0.8382	0.4019	0.487	0.6262
Mixed farms	9.3297	<0.001	−0.9759	0.3291	−1.0666	0.2862
Wine farms	−0.0085	0.9932	−1.7440	0.0812	−0.9564	0.3389

Note: The null hypothesis of the Pesaran CD test assumes cross-sectional independence. Country FE refers to the fixed-effects specification with country effects only. Two-way FE includes both country and year effects. First-difference models are reported as additional robustness checks.

.

The IPS test results show that ROA is stationary across all analysed farm types, which supports its use as the dependent variable in the panel models. Economic size is also stationary in all farm types, indicating stable time-series properties of this structural variable. Asset turnover is stationary in field crop and wine farms, but non-stationary in dairy and mixed farms. Fixed asset share is stationary only in wine farms, while labour productivity is non-stationary across all farm types. Land productivity shows mixed results. It is stationary only in wine farms, while in the other farm types the null hypothesis of a unit root cannot be rejected. Energy intensity is stationary in field crop, dairy and mixed farms, but not in wine farms.

Overall, the IPS results indicate that several explanatory variables exhibit non-stationary behaviour in specific farm-type panels. Therefore, the level-form estimates are interpreted cautiously as short-panel conditional associations rather than as long-run equilibrium relationships. Given the short time dimension of the panels, the IPS results were treated as diagnostic evidence rather than as an automatic rule for transforming or excluding variables.

The Pesaran CD test results show evidence of cross-sectional dependence in the country’s fixed-effects specifications for field crop, dairy and mixed farms. This indicates that these panels may be affected by common shocks or unobserved factors shared across European countries. However, after common year effects were included, cross-sectional dependence was no longer statistically significant in any farm-type panel at the 5% level. The first-difference specifications also showed no evidence of statistically significant cross-sectional dependence. These results support the use of two-way fixed-effects specifications and first-difference robustness checks in the subsequent analysis.

Second-generation panel unit root tests based on the Pesaran CIPS/CADF approach were also applied as complementary diagnostics. The detailed CIPS/CADF test statistics are reported in Appendix A 

Table A6. Pesaran CIPS/CADF panel unit root test statistics by farm type.

Variable	Field Crops	Dairy Farms	Mixed Farms	Wine Farms
ROA	−2.7640	−1.8518	−2.2119	−1.6443
ESS	−1.7874	−2.1628	−2.4295	−1.9833
ATR	−2.4304	−2.1366	−3.5279	−1.5611
FAS	−2.0439	−1.7186	−2.3314	−1.3457
LABPROD	−2.5714	−2.6574	−2.2688	−1.7432
LANDPROD	−2.7158	−1.5820	−2.0677	−1.4908
ENINT	−3.1332	−1.9017	−1.8885	−2.2429
INPUTINT	−3.0616	−1.6294	−2.3140	−1.6906
DEBT	−1.5415	−1.4497	−2.5229	−1.8444

. These tests confirmed that the stationarity properties of the variables are mixed and that the results should not be interpreted as long-run equilibrium relationships. Therefore, additional first-difference models were estimated to examine whether the main findings were sensitive to estimation in levels.

In summary, the diagnostic tests indicate that the empirical results should be interpreted cautiously. The subsequent analysis therefore relies on several complementary specifications: country fixed-effects models, two-way fixed-effects models with common time effects, first-difference robustness models and a unified farm-type interaction model. These procedures reduce specification dependence, although they do not fully eliminate the limitations arising from the short time dimension and mixed stationarity properties of the data.

4.4. Econometric Results

After the preliminary descriptive, correlation and stationarity, and cross-sectional dependence analyses, panel econometric models were estimated separately for each farm type. Given the mixed stationarity properties of several variables, the short time dimension of the panels and the relatively limited number of countries, the results are interpreted as conditional associations rather than as causal effects or long-run equilibrium relationships. Therefore, the interpretation does not rely on a single estimator only, but on the consistency of results across complementary specifications, including fixed-effects models, two-way fixed-effects models, first-difference robustness checks and the unified farm-type interaction model.

The econometric analysis focuses on the relationship between ROA, asset-use efficiency, farm structure, productivity indicators and energy-cost intensity.

Table 5. Farm-type-specific econometric results.

Variable/Diagnostic	Field Crops FE Robust	Dairy Farms FE Robust	Mixed Farms FE Robust	Wine Farms FE Robust
ESS	−0.000103 (0.4113)	0.000080 (0.5501)	−0.000087 (0.0873)	0.000086 (0.4266)
ATR	0.5980 ** (<0.001)	0.3399 ** (<0.001)	0.2995 ** (0.0016)	0.6271 ** (<0.001)
FAS	0.0048 (0.9403)	0.0166 (0.8310)	−0.1494 * (0.0342)	0.0493 (0.4703)
LABPROD	0.00000007 (0.4311)	0.00000007 (0.5126)	0.00000008 (0.3630)	−0.00000015 (0.7886)
LANDPROD	−0.00000301 (0.5940)	−0.00000040 (0.1539)	0.00000106 (0.6850)	−0.00000182 (0.5611)
ENINT	−0.8492 ** (<0.001)	−0.1873 (0.4663)	−0.8628 ** (<0.001)	−0.0714 (0.7763)
Valid country–year observations	241	223	241	135
Countries	27	25	27	15
Within R2	0.759	0.503	0.473	0.588

Note: Values are coefficients with p-values in parentheses. FE robust refers to fixed-effects models with cluster-robust standard errors. Valid country–year observations refer to the number of observations in the original farm-type panel. Within R² is reported for the fixed-effects specification. Significance levels: ** p < 0.01; * p < 0.05.

presents the farm-type-specific fixed-effects results. To ensure comparability across farming systems, fixed-effects models with cluster-robust standard errors are reported for all four farm types.

The results presented in Table 5 show that the associations between profitability and the selected explanatory variables differ across farm types, although some common patterns can be observed. The most consistent result refers to asset turnover (ATR), which has a positive and statistically significant coefficient in all analysed farm types. This pattern is consistent with a positive association between asset-use efficiency and ROA, although it should not be interpreted as evidence of a direct causal effect.

In field crop farms, the fixed-effects model with cluster-robust standard errors shows that ATR is positively and statistically significantly associated with ROA (p < 0.001). Energy intensity (ENINT) is negatively and statistically significantly associated with ROA (p < 0.001), indicating that higher energy costs relative to output are associated with lower profitability in field crop production. This result is particularly relevant because field crop farms are strongly dependent on fuel, machinery use, drying, irrigation and other energy-related operations. By contrast, economic size, fixed asset share, labour productivity and land productivity are not statistically significant in the robust fixed-effects model. Therefore, among the variables included in the field crop model, statistically significant associations with ROA were identified for asset turnover and energy intensity.

For dairy farms, ATR is positively and statistically significantly associated with ROA (p < 0.001), making asset turnover the main statistically significant correlate of profitability in the dairy-farm model. By contrast, economic size, fixed asset share, labour productivity, land productivity and energy intensity are not statistically significant in the fixed-effects specification. Therefore, the dairy-farm results primarily point to the relevance of asset-use efficiency rather than to a robust association between profitability and energy-cost intensity.

In mixed farms, ATR is positively and statistically significantly associated with ROA (p < 0.01). Energy intensity is negatively and statistically significantly associated with ROA (p < 0.001), indicating that higher energy costs relative to output are associated with lower profitability in mixed farming systems. Fixed asset share is also negatively and statistically significantly associated with ROA (p < 0.05), suggesting that a higher share of fixed assets in total assets is associated with lower profitability in this farm type. Economic size, labour productivity and land productivity are not statistically significant at the 5% level.

For wine farms, the fixed-effects model with cluster-robust standard errors shows that ATR is positively and statistically significantly associated with ROA (p < 0.001). It is the only explanatory variable with a statistically significant coefficient in the wine-farm specification. Other variables, including economic size, fixed asset share, labour productivity, land productivity and energy intensity, are not statistically significant in the robust fixed-effects model. Therefore, the wine-farm results indicate that asset turnover is the main correlate of ROA in this farming system.

Overall, the econometric results show that asset turnover is the most consistently significant correlate of farm profitability across all analysed production systems. Energy intensity is negatively and statistically significantly associated with profitability only in field crop and mixed farms, while this association is not statistically significant in dairy and wine farms in the fixed-effects models. These findings indicate that the relationship between profitability, asset use and energy-cost intensity differs across farm types and should be interpreted within the production and structural characteristics of each farming system.

4.5. Unified Farm-Type Interaction Model

To formally examine whether the associations between ROA and the main explanatory variables differ across farm types, a unified farm type–country–year panel model was estimated. This model included farm type × country fixed effects and year fixed effects, with coefficients allowed to vary by farm type. This specification is important because differences in statistical significance across separately estimated farm-type models do not necessarily imply that the coefficients are statistically different across farm types.

The results of the unified interaction model are presented in

Table 6. Unified farm-type interaction model.

Variable	Field Crops	Dairy Farms	Mixed Farms	Wine Farms
ATR	0.5356 (<0.001) ***	0.3360(<0.001) ***	0.2912 (<0.001) ***	0.6079 (<0.001) ***
ENINT	−0.8272 (<0.001) ***	−0.1492 (0.5478)	−0.8200 (<0.001) ***	−0.0095 (0.9695)

Note: Dependent variable: ROA. The model was estimated on the unified farm type–country–year panel and included farm type × country fixed effects and year fixed effects. Standard errors were clustered at the farm type × country level. Values are coefficients with p-values in parentheses. For readability, selected coefficients for ATR and ENINT are reported. Significance levels: *** p < 0.001.

. For readability, Table 6 reports the key coefficients for asset turnover and energy intensity.

The unified interaction model confirms the main findings from the farm-type-specific models. ATR remained positive and statistically significant across all four farm types. The coefficient was highest in wine farms and field crop farms, followed by dairy and mixed farms. This further supports the conclusion that ATR is the most consistent correlate of ROA across the analysed farming systems. Energy intensity showed a negative and statistically significant association with ROA in field crop and mixed farms, while it was not statistically significant in dairy and wine farms.

To test whether the farm-type-specific coefficients are statistically different from each other, Wald tests for equality of coefficients were performed. The results are presented in

Table 7. Wald tests for equality of coefficients across farm types.

Variable	Wald F	df	p-Value	Interpretation
ATR	3.536	3; 93	0.0177	Significant differences across farm types
ENINT	4.742	3; 93	0.004	Significant differences across farm types

Note: Wald tests examine whether the farm-type-specific coefficients are jointly equal across field crop, dairy, mixed and wine farms in the unified interaction model.

.

The Wald tests indicate statistically significant differences across farm types for ATR and ENINT. This suggests that the profitability associations of asset turnover and energy intensity are not identical across farming systems. Therefore, the comparative results should not be interpreted only on the basis of significance or insignificance in separately estimated farm-type models, but also through the formal cross-farm-type tests.

As an additional sensitivity analysis, the unified interaction model was re-estimated using only the common country sample, consisting of countries observed in all four farm-type panels. The common sample included 13 countries and 468 country–year observations. The detailed results of this sensitivity analysis are reported in Appendix A 

Table A7. Common-country sensitivity analysis for the unified interaction model.

Variable	Field Crops	Dairy Farms	Mixed Farms	Wine Farms
ATR	0.6150 (<0.001) **	0.3737 (0.0072) **	0.3563 (0.0366) *	0.7126 (<0.001) **
ENINT	−1.0328 (<0.001) **	0.0600 (0.8902)	−0.9375 (0.0293) *	−0.7608 (0.0546)

Note: The common-country sample includes 13 countries observed in all four farm-type panels and 468 country–year observations. Values are coefficients with p-values in parentheses. Significance levels: ** p < 0.01; * p < 0.05.

and

Table A8. Wald tests for equality of coefficients in the common-country sample.

Variable	Wald F	df	p-Value
ATR	2.301	3; 51	0.0883
ENINT	2.055	3; 51	0.1177

. In this specification, ATR remained positive and statistically significant across all four farm types. ENINT remained negative and statistically significant in field crop and mixed farms, while it was not statistically significant at the 5% level in dairy or wine farms. However, the Wald tests no longer indicated statistically significant differences across farm types at the 5% level. These results suggest that the positive association between ATR and ROA is highly robust, while formal differences across farm types become weaker when the comparison is restricted to the common country sample.

4.6. Robustness Checks

Robustness checks were performed to examine whether the main results remain stable across alternative specifications. The robustness analysis included fixed-effects models with cluster-robust standard errors, two-way fixed-effects models and first-difference models. In addition, VIF values were calculated to assess multicollinearity among explanatory variables. The detailed robustness results are presented in Appendix A 

Table A9. Robustness checks for field crop farms.

Variable	FE Cluster-Robust	Two-Way FE Cluster-Robust	First-Difference Cluster-Robust
ESS	−0.000103 (0.4113)	−0.000107 (0.3542)	−0.000229 (0.0046) **
ATR	0.5980 (<0.001) **	0.5754 (<0.001) **	0.6840 (<0.001) **
FAS	0.0048 (0.9403)	−0.0287 (0.6255)	−0.0279 (0.7309)
LABPROD	0.00000007 (0.4311)	0.00000009 (0.2820)	0.00000015 (0.0401) *
LANDPROD	−0.00000301 (0.5940)	0.00000570 (0.2756)	−0.00000505 (0.2438)
ENINT	−0.8492 (<0.001) **	−0.6207 (0.0032) **	−0.5606 (0.0059) **

Note: Values are coefficients with p-values in parentheses. FE—fixed effects. Two-way FE includes country and year effects. First-difference models are reported as additional robustness checks. Significance levels: ** p < 0.01; * p < 0.05.

,

Table A10. Robustness checks for dairy farms.

Variable	FE Cluster-Robust	Two-Way FE Cluster-Robust	First-Difference Cluster-Robust
ESS	0.000080 (0.5501)	0.000157 (0.1815)	0.000080 (0.4326)
ATR	0.3399 (<0.001) **	0.3895 (<0.001) **	0.4542 (<0.001) **
FAS	0.0166 (0.8310)	−0.0062 (0.9053)	−0.0665 (0.1583)
LABPROD	0.00000007 (0.5126)	0.00000019 (0.1206)	0.00000013 (0.2041)
LANDPROD	−0.00000040 (0.1539)	0.00000001 (0.9847)	0.00000018 (0.7257)
ENINT	−0.1873 (0.4663)	0.0378 (0.9257)	−0.0776 (0.8325)

Note: Values are coefficients with p-values in parentheses. FE—fixed effects. Two-way FE includes country and year effects. First-difference models are reported as additional robustness checks. Significance levels: ** p < 0.01.

,

Table A11. Robustness checks for mixed farms.

Variable	FE Cluster-Robust	Two-Way FE Cluster-Robust	First-Difference Cluster-Robust
ESS	−0.000087 (0.0873)	−0.000071 (0.2048)	−0.000047 (0.4529)
ATR	0.2995 (0.0016) **	0.2840 (<0.001) **	0.4311 (<0.001) **
FAS	−0.1494 (0.0342) *	−0.1693 (0.0141) *	−0.1475 (0.0516)
LABPROD	0.00000008 (0.3630)	0.00000013 (0.2336)	0.00000015 (0.2566)
LANDPROD	0.00000106 (0.6850)	0.00000153 (0.6283)	−0.00000162 (0.5382)
ENINT	−0.8628 (<0.001) **	−0.8629 (<0.001) **	−0.8696 (<0.001) **

Note: Values are coefficients with p-values in parentheses. FE—fixed effects. Two-way FE includes country and year effects. First-difference models are reported as additional robustness checks. Significance levels: ** p < 0.01; * p < 0.05.

and

Table A12. Robustness checks for wine farms.

Variable	FE Cluster-Robust	Two-Way FE Cluster-Robust	First-Difference Cluster-Robust
ESS	0.000086 (0.4266)	−0.000121 (0.4177)	−0.00000174 (0.9847)
ATR	0.6271 (<0.001) **	0.5402 (<0.001) **	0.6083 (<0.001) **
FAS	0.0493 (0.4703)	0.1158 (0.0851)	0.0546 (0.5779)
LABPROD	−0.00000015 (0.7886)	−0.00000080 (0.1989)	−0.00000014 (0.7497)
LANDPROD	−0.00000182 (0.5611)	−0.00000138 (0.5806)	−0.00000190 (0.1996)
ENINT	−0.0714 (0.7763)	−0.5632 (0.0943)	−0.6308 (0.1182)

Note: Values are coefficients with p-values in parentheses. FE—fixed effects. Two-way FE includes country and year effects. First-difference models are reported as additional robustness checks. Significance levels: ** p < 0.01.

.

Overall, the robustness checks show that asset turnover has the most stable and consistently significant association with farm profitability across production systems. Its coefficient remained positive and statistically significant in the robustness specifications for field crop, dairy, mixed and wine farms. This confirms that the association between asset-use efficiency and ROA is the most robust empirical pattern in the study.

Energy intensity also showed a negative association with ROA in several specifications, although its robustness differed across farm types. The negative association was most consistent in field crop and mixed farms, where ENINT remained statistically significant in the two-way fixed-effects and first-difference specifications. In dairy farms, ENINT was not statistically significant in the fixed-effects, two-way fixed-effects or first-difference specifications In wine farms, the association between ENINT and ROA was not statistically significant at the 5% level across the main robustness specifications. Therefore, the results support a cautious interpretation in which higher energy costs relative to output are generally associated with lower profitability, but this relationship is more robust in field crop and mixed farms than in dairy and wine farms.

The VIF results generally indicated that multicollinearity was not a serious problem in the estimated models. The maximum VIF values were 3.04 for field crop farms, 3.50 for dairy farms, 3.34 for mixed farms and 9.06 for wine farms. The maximum values for field crop, dairy and mixed farms did not indicate substantial multicollinearity. In the wine-farm model, labour productivity had the highest VIF value, indicating a potential multicollinearity concern. Therefore, the coefficient of labour productivity in wine farms was interpreted cautiously, while the substantive interpretation of the robustness results focused on variables with more stable coefficients, particularly asset turnover.

Because the wine-farm panel included only 15 countries, additional caution was applied when interpreting the wine-farm results. The robustness checks confirmed the positive association between ATR and ROA, while the remaining explanatory variables were not consistently significant. This supports the conclusion that the wine-farm results are primarily driven by asset-use efficiency.

The robustness analysis assesses the sensitivity of the findings to alternative econometric specifications and covariance estimators, focusing on profitability, asset-use efficiency and energy-cost intensity.

5. Discussion

The results of this study indicate that the associations between farm profitability, resource-use efficiency, and macro-level environmental pressure differ across the analysed farm types. This finding is consistent with previous research showing that economic and environmental performance should not be analysed only at the aggregate agricultural level, because farm specialisation, production technology, asset structure and input requirements may substantially shape the direction and strength of these relationships [40,41]. By estimating separate models for field crop, dairy, mixed, and wine farms, this study shows that the direction, statistical significance, and robustness of the associations between ROA and the selected explanatory variables differ across production systems.

The most consistent finding refers to asset turnover. Asset turnover was positively associated with ROA across all analysed farm types. This association was statistically significant in field crop, dairy and wine farms in the farm-type-specific models, while in mixed farms the coefficient was positive and statistically significant in the fixed-effects robustness checks and in the unified interaction model. The unified farm-type interaction model further confirmed that ATR remained positive and statistically significant across all four farming systems, while the Wald tests showed that the strength of this association differed significantly across farm types. This indicates that asset turnover is the most consistently significant profitability correlate among the variables included in the estimated models. Farming systems with higher output relative to their asset base also tend to exhibit higher ROA, although this association should be interpreted in light of the common total-assets denominator used in both indicators. This pattern is consistent with previous studies reporting a positive relationship between asset-use efficiency, farm viability, and profitability. Reziti [51] showed that asset turnover represents an important component of profitability decomposition, while Miljatović et al. [53] found that asset turnover was the only determinant with a positive and significant effect on economic viability across all farm types. Similarly, Grzelak & Staniszewski [52] highlighted the importance of production scale and intermediate consumption productivity in achieving market-comparable returns on assets. The consistently significant association of asset turnover with ROA is particularly relevant when interpreting capital-intensive farming systems. In dairy farms, the observed association may reflect the importance of efficiently using buildings, milking equipment, machinery, and other long-term assets alongside herd management and milk production. In wine farms, the positive association between asset turnover and ROA is consistent with the interpretation that the utilisation of vineyards, equipment, and processing-related assets is relevant to economic performance. In field crop farms, where machinery and land-related capital are important components of production, higher asset turnover was likewise associated with higher ROA. In mixed farms, the positive association between ATR and ROA suggests that the efficient use of a heterogeneous asset base is important in production systems that combine crop and livestock activities. However, because ATR and ROA share total assets as a denominator, this finding should be interpreted as a robust accounting and economic association rather than as evidence that increasing asset turnover alone will necessarily raise profitability. The contribution of this finding does not lie in demonstrating the generally expected positive relationship between asset turnover and ROA. Rather, the additional empirical insight is that this association remains positive and statistically significant across structurally different farming systems and across several robustness specifications. At the same time, the unified interaction model shows that the magnitude of this association differs across farm types, indicating that asset-use efficiency is not equally related to profitability in all production systems. The associations of ROA with energy intensity and macro-level GHG pressure are considerably less uniform. Thus, the comparative contribution of the study lies in distinguishing the association that remains most consistent across farm types from those that exhibit production-system-specific and specification-sensitive patterns.

Energy intensity had a negative coefficient in most estimated specifications, although its statistical significance differed across farm types. The negative association was most consistent in field crop and mixed farms, where ENINT remained statistically significant in the fixed-effects, two-way fixed-effects and first-difference models. In dairy farms, ENINT was not statistically significant in the fixed-effects, two-way fixed-effects or first-difference specifications. In wine farms, the association between ENINT and ROA was not statistically significant at the 5% level across the main robustness specifications. This indicates that higher energy costs relative to output are associated with lower ROA, particularly in production systems that rely heavily on fuel, electricity, machinery operations, drying, cooling, irrigation, or livestock-related energy use. This finding is consistent with previous studies emphasising that energy use is both an economic cost and an environmental burden. Uthes & Herrera [48] showed that higher input intensity is negatively associated with economic, environmental and overall sustainability indicators, while Oliveira et al. [54] demonstrated that electricity and energy use are important sources of environmental burdens in dairy production. Almeida et al. [55] also showed that circular and biorefinery-based solutions are not automatically eco-efficient when they involve high energy requirements and operating costs. The present results add to this literature by showing that the profitability relevance of energy-cost intensity differs across farming systems and is most robustly identified in field crop and mixed farms. The negative association between energy intensity and ROA has potential implications for sustainable farm management. Because ENINT measures energy costs relative to total output, farms with higher ENINT may also be more exposed to fluctuations in fuel and electricity prices, which can increase production-cost volatility and weaken their financial resilience. However, the models do not directly estimate the effects of energy-price shocks or measure farm resilience, so this interpretation remains exploratory. The results suggest that measures aimed at reducing energy costs relative to output may be economically relevant, particularly in field crop and mixed farming systems; however, the models do not directly evaluate the effects of specific energy-efficiency interventions. The specific technologies and management practices through which energy-cost exposure could be reduced were not examined in the present analysis and require separate farm-level evaluation. The results for fixed asset share provide additional evidence that the relationship between asset structure and profitability is farm-type specific. In mixed farms, FAS was negatively and statistically significantly associated with ROA in the fixed-effects and two-way fixed-effects specifications. This suggests that, in mixed farming systems, a higher share of fixed assets in total assets may be associated with lower profitability, possibly because capital tied in long-term assets does not necessarily generate proportional output or income increases. This interpretation is consistent with the ambiguous theoretical role of fixed assets: productive infrastructure and machinery may support production capacity, but they may also increase fixed-cost exposure and reduce financial flexibility. In field crop, dairy and wine farms, FAS was not statistically significant in the main fixed-effects specifications, indicating that the profitability association of asset structure is not uniform across production systems.

Economic size, labour productivity and land productivity did not show a consistently significant association with ROA across the main farm-type-specific models. This does not imply that scale or productivity is irrelevant for farm performance. Rather, their relationship with ROA appears to be more dependent on farm type, model specification and the way these indicators are measured in aggregated FADN/FSDN data. Labour productivity and land productivity may capture important dimensions of technical and production performance, but their association with profitability can be affected by output prices, input costs, labour structure, land quality, technology and farm specialisation. Therefore, the absence of consistent statistical significance should be interpreted cautiously and not as evidence that these dimensions are economically unimportant.

The unified farm-type interaction model provides an important contribution because it formally tests whether selected coefficients differ across farming systems. The results show that the coefficients of ATR and ENINT differ significantly across farm types in the full unified sample. This confirms that the relationship between profitability, asset-use efficiency and energy-cost intensity is not identical across field crop, dairy, mixed and wine farms. However, the common-country sensitivity analysis shows that these formal differences become weaker when the comparison is restricted to the countries observed in all four farm-type panels. In that specification, ATR remained positive and statistically significant across all farm types, while ENINT remained statistically significant only in field crop and mixed farms. The Wald tests in the common-country sample did not indicate statistically significant differences across farm types at the 5% level. These findings suggest that the positive association between ATR and ROA is highly robust, while cross-farm-type differences in coefficient magnitudes should be interpreted with caution.

Compared with previous studies, the findings partly confirm and partly extend existing knowledge. Earlier FADN-based research has shown that farms with higher productivity and better resource-use efficiency are often more profitable and more eco-efficient [37,38,39]. The present study is consistent with this view because asset turnover showed a positive and statistically significant association with ROA in several farming systems. At the same time, the negative association between energy intensity and ROA in field crop and mixed farms is consistent with studies suggesting that higher input and energy dependence may reduce economic performance when additional costs are not matched by proportional output gains [48,54,55]. The study therefore contributes by showing that asset-use efficiency is the most stable profitability-related indicator across farm types, while energy-cost intensity has a more production-system-specific association with profitability.

The results also contribute to the debate on the role of FADN/FSDN data in sustainability assessment. Several studies have emphasised that FADN provides a harmonised and representative basis for farm economic analysis, but that environmental and social dimensions require additional indicators or linked data sources [44,45]. The present study uses FADN/FSDN Standard Results to examine profitability, asset-use efficiency, factor productivity and energy-cost intensity across several European farming systems. This approach provides comparative evidence based on harmonised indicators, but it also remains limited by the aggregate country–farm-type level of the available data. The results should therefore be interpreted as associations between reported economic and resource-use indicators, not as farm-level causal mechanisms.

From a policy perspective, the findings caution against assuming that the same economic–environmental relationships apply uniformly across farming systems. The consistently positive association between ATR and ROA suggests that policies and advisory measures aimed at improving the productive use of assets may be relevant across different farm types. However, this does not imply that the same asset-management strategy is appropriate for all farming systems. For dairy farms, asset-use efficiency may be related to the utilisation of buildings, livestock-related infrastructure and equipment. For field crop farms, it may depend more strongly on machinery use, land-related capital and the organisation of field operations. For wine farms, vineyard use and processing-related assets may be particularly relevant. In mixed farms, the challenge lies in managing a more heterogeneous asset base that combines crop and livestock production. For field crops and mixed farms, the negative association between ENINT and ROA identifies energy-cost exposure as an economically relevant issue. This finding suggests that improvements in energy efficiency, fuel use, machinery operations, irrigation, drying, cooling and other energy-related processes may be important for maintaining profitability. Nevertheless, the present models do not identify which specific technologies or management interventions would be most effective. Therefore, policy conclusions should be limited to the general importance of energy-cost management and should not be interpreted as evidence for the profitability effect of any particular intervention.

The study has several limitations that should be acknowledged. First, the analysis is based on aggregate FADN/FSDN Standard Results by country, year and farm type, rather than individual farm-level microdata. Therefore, the results reflect average patterns within farm types and countries. Although the analysis distinguishes four farming systems, these categories remain broad and internally heterogeneous. Farms classified within the same type may differ considerably in production scale, technological intensity, regional and climatic conditions, management practices, degree of specialisation, asset structure, and dependence on external inputs. Consequently, the estimated associations represent average relationships at the country–farm-type level and should not be interpreted as applying uniformly to all farms within a particular category. This level of aggregation may conceal important within-type differences and limits the generalisation of the results to individual holdings or more narrowly defined production systems. Second, the study does not estimate farm-level environmental impacts or farm-type-specific greenhouse gas emissions. Although environmental sustainability provides the broader context of the analysis, the econometric models focus on profitability, asset-use efficiency, productivity indicators and energy-cost intensity. Therefore, the results should not be interpreted as evidence of differences in greenhouse gas pressure or environmental performance across field crop, dairy, mixed and wine farms. Future research should integrate farm-level or farm-type-specific physical indicators, such as fertiliser use, livestock units, energy quantities, land-use practices, nutrient balances, pesticide use, biodiversity indicators and directly estimated greenhouse gas emissions. Third, the analysed period covers 2015–2023, which limits the time dimension of the panel. This period also includes several major shocks, including the COVID-19 pandemic, energy-price shocks, the Russia–Ukraine conflict and high inflation. Although year effects and robustness checks were used to reduce the influence of common shocks, the models cannot fully isolate their separate effects. Fourth, all monetary variables were used as reported in the FADN/FSDN Standard Results database and were not additionally adjusted for inflation or purchasing power parity. This may affect the interpretation of cross-country comparisons, especially in a period marked by high inflation and changing input prices. Fifth, no winsorization procedure was applied, so the results may remain sensitive to extreme reported values, although robust covariance estimators and sensitivity checks were used to assess the stability of the findings.

Finally, because the available data do not contain exogenous shocks, detailed product prices, input prices, weather anomalies, policy-payment controls or farm-level management practices, the estimated coefficients should be interpreted as conditional associations rather than as direct causal effects. Despite these limitations, the study provides comparative evidence on the relationship between farm profitability, asset-use efficiency and energy-cost intensity across different European farming systems. The results show a consistent positive association between ROA and asset turnover across all analysed farm types, alongside a negative association between ROA and energy intensity mainly in field crop and mixed farms. Therefore, the assessment of farm economic sustainability should consider not only profitability outcomes, but also the asset-use and energy-cost patterns with which those outcomes are associated.

6. Conclusions

This study analysed the relationship between farm profitability, asset-use efficiency, productivity indicators, and energy-cost intensity across European farming systems. The analysis was based on FADN/FSDN data for the period 2015–2023. Four types of farming were analysed separately: field crop farms, dairy farms, mixed farms, and wine farms. This approach made it possible to examine whether the associations between profitability and selected indicators of farm structure, asset-use efficiency, productivity, and energy-cost exposure differ across production systems with distinct structural and technological characteristics.

The main conclusion of the study is that asset turnover was the most consistently significant correlate of ROA across all analysed farming systems. Asset turnover showed positive and statistically significant associations in field crop, dairy, mixed and wine farms in the farm-type-specific fixed-effects models, and it remained positive and statistically significant across all four farm types in the unified interaction model. This finding should be interpreted cautiously because ATR and ROA share total assets as a denominator. Therefore, the result should be understood as a robust accounting and economic association between asset-use efficiency and profitability, rather than as evidence of a direct causal effect.

Energy intensity was negatively associated with ROA, with the most robust evidence identified in field crop and mixed farms. In these two farming systems, ENINT remained statistically significant across the main fixed-effects and robustness specifications. This indicates that higher energy costs relative to output are associated with lower profitability particularly in field crop and mixed farms. In dairy and wine farms, the association between ENINT and ROA was not statistically significant at the 5% level in the main fixed-effects models. These findings show that the profitability relevance of energy-cost intensity differs across farming systems.

The results also indicate that fixed asset share has a farm-type-specific association with profitability. In mixed farms, FAS was negatively and statistically significantly associated with ROA in the fixed-effects specification, suggesting that a higher share of fixed assets in total assets may be associated with lower profitability in this production system. However, this relationship was not consistently identified across all farm types. Economic size, labour productivity and land productivity did not show a consistently significant association with ROA in the main farm-type-specific models. This does not imply that scale and productivity are irrelevant for farm performance, but rather that their relationship with profitability depends on farm type, production structure and model specification.

Overall, the findings show that the relationship between farm profitability, asset-use efficiency and energy-cost intensity differs across farming systems. The unified farm-type interaction model confirmed statistically significant differences across farm types for ATR and ENINT in the full sample. However, the common-country sensitivity analysis showed that these formal differences became weaker when the comparison was restricted to countries observed in all four farm-type panels. This indicates that the positive association between ATR and ROA is highly robust, while cross-farm-type differences in coefficient magnitudes should be interpreted with caution.

These differences indicate that policy assessments should account for the structural and production characteristics of different farming systems rather than assume that the same profitability relationships apply uniformly across them. The consistently positive association between ATR and ROA suggests that improving the productive use of assets may be relevant across different farm types. However, the specific meaning of asset-use efficiency differs across farming systems. In field crop farms, it may be related to the utilisation of machinery, land-related capital, storage and drying facilities. In dairy farms, it may depend on the efficient use of buildings, milking equipment, livestock-related infrastructure and herd management. In wine farms, it may be linked to vineyard use and processing-related assets. In mixed farms, it reflects the challenge of managing a more heterogeneous asset base that combines crop and livestock production.

For field crop and mixed farms, the negative association between ENINT and ROA identifies energy-cost intensity as an economically relevant issue. Potential areas for further policy and managerial assessment include energy audits, advisory support, investment in energy-efficient machinery, improved organisation of field operations, irrigation efficiency, drying and cooling efficiency, and the use of renewable energy where economically feasible. However, because the present models do not evaluate specific technologies or policy interventions, these instruments should be regarded as options for further assessment rather than as directly tested recommendations.

Several limitations should be acknowledged. First, the analysis was based on aggregate FADN/FSDN Standard Results by country, year and farm type, rather than individual farm-level microdata. Therefore, the estimated relationships represent average patterns for broad farming systems and may not apply uniformly to all farms classified within the same type. The four analysed farming categories remain internally heterogeneous with respect to production scale, technology, production intensity, regional and climatic conditions, management practices, degree of specialisation and asset structure. This level of aggregation may conceal important within-type differences and limits the generalisation of the findings to individual holdings or more narrowly defined production systems.

Second, the study does not estimate farm-level environmental impacts or farm-type-specific greenhouse gas emissions. Although environmental sustainability provides the broader context of the analysis, the econometric models focus on profitability, asset-use efficiency, productivity indicators and energy-cost intensity. Therefore, the results should not be interpreted as evidence of differences in greenhouse gas pressure or environmental performance across field crop, dairy, mixed and wine farms.

Third, the relatively short time dimension of the panel limits the possibility of examining long-term dynamic relationships. The analysed period also includes several major shocks, including the COVID-19 pandemic, energy-price shocks, the Russia–Ukraine conflict and high inflation. Although year effects and robustness checks were used to reduce the influence of common shocks, the models cannot fully isolate their separate effects. Fourth, all monetary variables were used as reported in the FADN/FSDN Standard Results database and were not additionally adjusted for inflation or purchasing power parity. This may affect the interpretation of cross-country comparisons, especially in a period marked by high inflation and changing input prices. Fifth, no winsorization procedure was applied, so the results may remain sensitive to extreme reported values, although robust covariance estimators and sensitivity checks were used to assess the stability of the findings.

Finally, because the available data do not contain exogenous shocks, detailed product prices, input prices, weather anomalies, policy-payment controls or farm-level management practices, the estimated coefficients should be interpreted as conditional associations rather than as direct causal effects. Despite these limitations, the study provides comparative evidence on the relationship between farm profitability, asset-use efficiency and energy-cost intensity across different European farming systems. The results show a consistent positive association between ROA and asset turnover across all analysed farm types, alongside a negative association between ROA and energy intensity mainly in field crop and mixed farms. Therefore, the assessment of farm economic sustainability should consider not only profitability outcomes, but also the asset-use and energy-cost patterns with which those outcomes are associated.

Future research should use farm-level microdata and more narrowly defined production categories to examine whether the identified associations vary according to farm size, production intensity, technology, management practices and regional conditions. As more detailed farm-level environmental and sustainability data become available through the developing FSDN, future studies could integrate physical indicators such as fertiliser use, livestock units, energy quantities, land-use practices, nutrient balances, pesticide use, biodiversity indicators, water use and directly estimated greenhouse gas emissions. Additional research could also examine the role of subsidies, circular practices, energy-efficiency investments and technological innovation in strengthening farm profitability while improving resource-use efficiency. Such research would provide a more comprehensive understanding of how European farms can remain economically viable while reducing input dependence and improving sustainable resource management.