Tourism-Induced Data Envelopment Analysis (T-DEA): An Application in the Eurozone Economic Space

Abstract

This study investigates technical efficiency scores and performance change patterns by applying the tourism-induced Data Envelopment Analysis (DEA) to the Eurozone from 1996 to 2019. The study uses direct employment in tourism and capital investment spending directly related to the travel and tourism sector as input variables, whereas it considers the direct contribution of tourism to a country’s Gross Domestic Product (GDP) and arrivals as output variables. This set of tested variables is rarely found in the relevant literature, as many studies focus on hotel business-related proxies. After receiving the scores, we regress them on renewable energy sources using panel data. Based on the results, the Eurozone countries’ technical efficiency scores increase by approximately 1% per year on average. On the contrary, productivity growth declines slightly (−0.1% per year), signaling the need for additional effort in technological advancement. The error correction term is negative and significant in the tested models, whereas long-run coefficients are insignificant. Moreover, the empirical results indicate the absence of statistically significant short-run linkages, consistent with a neutrality-type outcome. Practical implications call for accelerating the adoption of renewables in the sector by simultaneously integrating additional measures to support innovation and sustainable investment plans.

1. Introduction

Due to its high-leverage nature, tourism can exacerbate global challenges and address issues related to economic growth patterns, investment initiatives, and employment, with long-term implications (Stoiljkovic et al., 2025).

By nature, each destination struggles to improve and/or optimize processes and resources to achieve greater efficiency and effectiveness, given the fierce competition the sector faces worldwide. Such a reality is challenging, demanding, and long-lasting, as it requires a responsive character to changes and shocks at the interface of supply and demand dynamics, as well as the establishment of a robust assessment mechanism to measure resource allocation efficiency and productivity patterns. Therefore, evidence-based practices can lead to result-oriented policies to identify inefficiencies and benchmarking gaps. This approach helps comprehend if resource allocations and financial resource distribution generate the desired outputs or yield anticipated benefits. Hence, planners and decision-makers can recognize whether the effort expended is “value-for-money” and “value-for-time” or not. Not surprisingly, critical and reasonable questions arise: Does the tourism sector deliver the benefits host economies expect, for instance, investment opportunities, sustainable growth, and employment? If inefficiencies are identified, how might resource allocation be optimized to better support the structural sectors of the receiving economies?

Are there any improvement efforts that can be implemented? Can other sectors benchmark the tourism sector as a leading example and become efficient and productive? Undoubtedly, the answers demand multilayer and multidisciplinary contributions. Interestingly, the tourism-induces Data Envelopment Analysis (T-DEA) approach can provide inputs and insights to decode technical efficiency dynamics and productivity improvements, enabling robust decision-making.

In the literature, researchers have identified the critical role of tourism’s performance rates (Barros et al., 2011, inter alia). Notwithstanding, there is a need to augment such research attempts by employing more comprehensive analyses in scope and methods (Assaf, 2012). Interestingly, DEA is a method for investigating inefficiencies and patterns of productivity change (Corne, 2015) and is used to determine the performance trajectory at destinations. Arguably, performance is recognized as a key determinant of tourism competitiveness (Croes & Kubickova, 2013).

As Assaf (2012) argues, in the tourism sector, DEA has been applied to the context of a single country, whereas many research efforts rely on the hotel-oriented DEA applications. Therefore, an opportunity arises to discuss T-DEA in the context of the travel and tourism sector for a group of countries that belong to the same economic space (e.g., the Eurozone).

In response to this research opportunity, this study’s purpose is to decode technical efficiency dynamics and productivity change patterns across the Eurozone countries, employing the T-DEA approach. In a panel data format, the period of interest is 1996–2019. Such an attempt aligns with resource allocation to optimize it and detail sector-specific policies. Achieving greater efficiency rates helps foster robust development plans, become more competitive, and experience long-term economic growth. Furthermore, this study aims to assess whether the adoption of renewable energy sources drives these efficiency and productivity patterns, thereby linking environmental sustainability with economic performance. In the current literature, previous studies often use microdata (beds/nights), whereas this study uses macroeconomic indicators (direct employment/investment) to provide a broader economic picture. This study contributes to the T-DEA discussion in the following ways: (i) it applies a sector-specific T-DEA framework that moves beyond hotel- or region-oriented analysis, by employing direct measures of tourism activity; (ii) it empirically examines whether renewable energy adoption is systematically associated with tourism efficiency and productivity outcomes, thereby providing insight into the role of energy transition patterns in shaping the Eurozone’s tourism sector performance.

As a result, this research attempts to provide well-justified answers for the following questions: (i) What are the average trends in technical efficiency and total factor productivity change between 1996 and 2019? (ii) Do Eurozone countries, on average, operate close to the efficiency frontier, and is an efficiency gap observable across countries over time? (iii) What dynamic patterns characterize tourism efficiency and productivity, particularly in terms of persistence and mean reversion toward the long-run equilibrium relationships following deviations from long-run equilibrium relationships? (iv) Is renewable energy consumption systematically associated with tourism’s technical efficiency and total factor productivity change in the long-run? (v) Do short-run changes in renewable energy consumption and tourism performance exhibit dynamic adjustment patterns? (vi) How quickly are deviations from equilibrium relationships corrected? To provide relevant answers, this study implements a T-DEA and a panel analysis following the methodology described in the next section.

The structure of this study is as follows: Section 2 addresses T-DEA as a crucial research field and the questions this study seeks to address. Section 3 is then presented, followed by Section 5. Finally, Section 6 presents the key findings and areas for future research.

2. Theoretical Background

2.1. Efficiency and Productivity in the Tourism Sector

DEA is one of the most widely applied methodologies for assessing efficiency and productivity, both globally (Emrouznejad & Yang, 2018) and in sectors related to hospitality and tourism (Assaf & Josiassen, 2012). Hadad et al. (2012) assert that in any industry, economic efficiency concerns a relative assessment of this industry’s effectiveness rates when processing inputs into outputs compared to the most efficient industry in the sample (e.g., best practice).

It is a nonparametric linear programming technique that allows researchers to be flexible with variable properties and distributions in datasets. Attention is required to the model specifications and the econometric approaches used to capture the model’s dynamics and impacts on the economic system. For instance, researchers often employ either an input-oriented or an output-oriented approach, frequently using the variance-to-scale method to capture technical efficiency for the Decision-Making Units (DMUs) of interest. In practice, in the real economy, these solutions rely on operational research to maximize benefits while keeping costs constant or minimizing them. The input variables in a DEA model correspond to the costs, and the output variables represent the benefits.

It should be noted that, in DEA (regardless of the sector or domain of interest), technical efficiency measures how well a decision-making unit (e.g., a country) transforms inputs into outputs relative to the “best-practice” frontier. Moreover, Total Factor Productivity Change (TFPC) reveals how both efficiency and the production frontier itself change over time, typically decomposed using the Malmquist Productivity Index (MI) into efficiency change (catch-up effect) and technological change (frontier shift effect) (Kallioras et al., 2021). Essentially, TFPC equals Efficiency Change × Technological Change. Consequently, it is influenced by how entities catch up to the frontier and how the frontier itself shifts. Many determinants can be used to investigate their impact on technical efficiency and changes in total factor productivity. An approach to facilitate the flow of research work is to divide these factors into internal and external. Internal factors can be either managerial or operational, whereas core external factors include macroeconomic, environmental, policy-related, and technological factors.

Tourism is recognized as a market where intense competition is common. As a direct consequence, identifying dynamics that affect efficiency status at destinations is critical to research (Radovanov et al., 2020). Also, the tourism sector has a multifaceted business character. Addressing its complex nature when measuring productivity patterns requires steps beyond regression analyses, ratio measures, and accounting measures (Nurmatov et al., 2021). However, DEA has been employed by researchers to decode macroeconomic dynamics in demand-oriented research (Nozick et al., 1998; Niavis & Kallioras, 2021, inter alia).

A critical issue in T-DEA is selecting the inputs and outputs for the analysis. In the literature, there is no mutually agreed-upon selection process; rather, the choice depends on the research scope and the issues to address. Logically, one option is to conceptualize research aspects with selected proxies that are a good fit for the research purpose and mirror the phenomenon under investigation. The goal is to obtain statistically significant results and to justify their implications for high-impact decisions in a meaningful and interpretable way.

Thus, these contributions can create benefits, especially for receiving economies that are heavily dependent on tourism. In tourism research, DEA serves as a stimulus to identify areas for improvement and develop a resilient character in response to changes, both in space and over time. For instance, destinations struggle to increase visitation rates but often overlook or fail to observe pathways to optimize the current status of tourism systems, aiming to establish more stable, sustainable, resilient, and profitable outcomes (Gossling et al., 2016). This might be one reason why some destinations succeed while others fail in developing plans: they do not measure their efficiency in resource allocation during execution.

T-DEA has provided a wide range of inputs to inform tourism development plans. Nevertheless, current studies in the field contextualize tested variables in a hotel-oriented perspective or to elaborate on regional performance rates. Furthermore, existing studies (i) measure productivity rates employing aggregate indicators or (ii) do not use variables to elaborate on energy transitions and the use of renewables when exploring how energy patterns affect the underlying production frontier or the tourism’s technical efficiency scores. Hence, there is a need for additional research to mitigate the adverse environmental impacts that remain prevalent (Wu et al., 2023).

Many studies use hotel-related business proxies (Chiang et al., 2004; Barros & Mascarenhas, 2005; Sigala et al., 2005; Shang et al., 2009; Yu & Lee, 2009; Hsieh & Lin, 2010; Assaf & Josiassen, 2012; Huang et al., 2016; Li et al., 2022; Choi & Kim, 2024; inter alia). As stated by Pulina and Santoni (2018), the literature features largely high-impact studies that elaborate on tourism’s economic performance either from a destination perspective, at the regional and/or city level, or from a hotel perspective (e.g., Hathroubi et al., 2014; Perrigot et al., 2009; Ramanathan et al., 2016). For instance, Barros et al. (2011) and Bire et al. (2025) consider tourism attraction perspectives and efficiency, Cuccia et al. (2017) address the UNESCO World Heritage List inscription on the performance of tourism destinations in Italian regions, whereas Yan (2013) applies an advanced city tourism efficiency evaluation model in the DEA context. Although significant in the T-DEA stream, these approaches do not fully explore the potential of tourism as a sector-specific field of research, including its revenue and employment in hospitality, but rather only partially (e.g., a subsection or domain). Arguably, Avkiran (2006) states that the selection of output-oriented test variables should mirror the objectives and service set of the DMUs in question, whereas the adopted inputs should be traceable to these outputs.

In the literature, variations in results appear whether DMUs are efficient or not. As happens in empirical research studies, findings might vary and differ based on the analysis adopted, the methods used, the countries or group of countries under research (e.g., geographical coverage), the number of inputs and outputs, and contextual factors (e.g., proxies used to conceptualize technical efficiencies and productivity change) in the tourism sector. Such a reality shows the complex and heterogenous nature of tourism. Continuous research employing multiple approaches consistently contributes to the relevant literature, augmenting and improving research efforts in the field of interest.

2.2. Energy Transition and Economic Efficiency

In light of these concerns, an interesting research question is the role of renewables in improving the tourism sector’s technical efficiency. Notably, Porter’s hypothesis holds that well-planned, well-designed regulatory frameworks can lead to efficiency gains, innovation (e.g., eco-innovation), cost savings, and enhanced competitiveness (e.g., increased visitation rates). For instance, visitors (tourism demand) consider environmental quality as a core factor when selecting destinations (tourism supply) (José-Luis et al., 2021). In turn, energy consumption is closely related to levels of environmental degradation, as evidenced by empirical research on the energy–growth nexus, particularly regarding carbon dioxide and other greenhouse gas emissions.

This is a contemporary issue within the struggle to optimize processes and resources within viable and sustainable sector-specific (e.g., tourism) and national economies. A central issue Porter’s hypothesis addresses is that environmental regulations can drive innovation effectively by partially or fully offsetting the costs of compliance (Porter & van der Linde, 1995). Some neoclassical researchers argue that strict environmental regulations increase production costs and, consequently, reduce competitiveness, especially in international business settings. Notwithstanding, based on Porter’s hypothesis, the benefits from innovation initiatives can offset enterprises’ losses from environmental regulations and advance social development (Xu et al., 2022). The current literature and relevant research findings yield mixed results across countries, industries, and time periods, suggesting that different innovation capabilities respond differently to environmental regulations (Qiu et al., 2018). In parallel, competitiveness remains an issue that dynamic economic sectors (e.g., tourism) should take seriously to succeed in global markets. Supportively, Ekonomou and Halkos (2024) argue that energy-efficient destinations enhance their competitive position in international tourism markets. Indicatively, improvements in technical efficiency in the tourism sector can reduce operating costs via lower energy intensity while simultaneously strengthening green competitiveness (e.g., eco-friendly investments) by enhancing the sustainable destination’s image or profile. Then, tourism development or expansion plans can conceptualize or integrate technical efficiency as a potential key determinant of competitiveness. The argument behind this perspective is that more efficient destinations (e.g., countries) can deliver higher-quality services at lower unit costs. Consequently, efficiency gains obtained from T-DEA translate into enhanced competitiveness. Countries of interest can reallocate savings from renewables integration into innovation and differentiation, thereby strengthening their green market positioning. Yang et al. (2025) state that green finance and sustainable investment instruments are core issues in implementing efficient solutions to reduce the cost of capital for energy efficiency projects. Dirma et al. (2024) argue that renewable integration can lower energy prices for consumers, increase market competition, and limit dependence on imported resources. However, the author notes challenges concerning initial investment costs, technological barriers, and the need for political support. This shift to renewables is often characterized by long-term economic benefits amid high initial investment and costs, yet this fundamental trade-off between costs and benefits is central to integrating renewables globally (Levicky et al., 2022).

In this perspective, policymakers should consider reducing the trade-off between environmental protection and competitiveness through green energy innovations (Mia et al., 2025), especially considering that renewables drive green competitiveness by lowering energy costs. In all these concepts and theories, the long-term perspective should dominate efforts to become more efficient and productive without sacrificing business gains and objectives. In the long run, renewables reduce operational costs. Potential reasons include very low or near-zero fuel costs, lower maintenance costs once quality is assured, and greater cost competitiveness with fossil fuels, despite potential higher upfront investment capital requirements. For instance, implementing a dynamic panel threshold regression, Wang et al. (2024) found that the proportion of renewable energy generation significantly reduces extreme price fluctuations once certain thresholds are exceeded. Moreover, the European Union’s policy to establish a global clean and resilient transition argues that energy-intensive industries urgently need support to decarbonize, switch to clean energy, and tackle high costs, unfair global competition, and complex regulations (European Commission, 2025b). Importantly, recognizing the need for cost-effective energy transitions with broad applicability across economically intensive sectors, and as part of the Clean Industrial Deal, the European Commission defines four pillars to consider: lowering energy costs for all, completing the Energy Union, attracting investment and ensuring delivery, and being ready for potential energy crises (European Commission, 2025a). Successful transitions toward tourism energy efficiency concern determinants such as total factor productivity, capital–energy ratio, labor–energy ratio, energy supply composition, and output composition (He et al., 2020). Notably, overcoming challenges related to renewables’ initial costs, which might temporarily reduce efficiency, is expected to yield lower environmental degradation and decarbonized, or net-zero, economies. Consequently, what is expected is increased efficiency that brings greener economies closer to our reality. The main determinants of energy efficiency at the macroeconomic level are technological effects, capital investment, structural effects, and labor (He et al., 2020).

2.3. Criticism of DEA

DEA has faced some criticism, which has not diminished its applicability and comprehensiveness in research across multiple methodological schemes and econometric approaches. However, such criticism revealed some weaknesses that, in practice, increased the researcher’s interest in improving the method. Indicatively, the conceptualization method requires strong dependence when selecting input and output variables. Moreover, this selection process should be precise and context-variable oriented, whereas incorporating different types of variables might create problems (Ozbek et al., 2009). As claimed by Nurmatov et al. (2021), Rouse (1997), and Ramanathan (2003), due to its nonparametric nature, statistical hypotheses cannot be tested.

However, researchers argue that there is no need to assume the analytical form of the observed inputs and outputs as the method allows for different types of metrics (Radovanov et al., 2020). Another issue is that DEA does not evidence a causal link; rather, it only confirms if tested DMUs are efficient. It provides efficiency rates, not causal linkages, unless linked to regression analysis or causality tests.

2.4. Contribution to the Literature

In this process, a key question is which determinants affect the tourism sector’s performance, and what their magnitudes are. Consequently, a research gap exists in the literature that calls for insights into the determinants of tourism sector performance and their relative importance (Assaf & Josiassen, 2012). Hence, new or additional influential factors may emerge via robust research applications in T-DEA. However, Nurmatov et al. (2020) note that documenting inputs and outputs in the tourism sector remains an open question in DEA as the method is flexible enough to consider a wide range of tourism-related variables.

This study contributes to the relevant literature in the following ways: it adopts direct contributions of employment in tourism and capital investment spending directly related to the travel and tourism sector as inputs within an output-oriented variable returns to scale (VRS) framework. The authors then derive Malmquist-based Total Factor Productivity Change (TFPC) to capture the intertemporal dynamics of productivity. This set of input variables is less visible or less observed in the T-DEA literature. Furthermore, this study focuses on capital investment spending directly related to the travel and tourism sector rather than on foreign direct investment in a general sense. Given its importance when studying the tourism sector, it remains disconnected from T-DEA research attempts. Supportively, capital investment spending enhances competitiveness and advances a country’s destination attributes, which are considered as destination “pull” motivation factors (Ekonomou, 2022). Consequently, receiving tourism economies (e.g., Eurozone countries) should include such a proxy in performance-related research efforts.

Additionally, this research focuses on the direct contribution of tourism to a country’s Gross Domestic Product (GDP) as one of the output variables rather than using aggregate measures like the overall GDP of the countries of interest. If this dimension is left untested, research may not provide complete or holistic results for the tourism sector. For instance, this might be the case when the research focuses solely on revenues gathered or consumption from hotel or accommodation facilities. Significantly, this research focuses on macroeconomic dynamics (e.g., tourism’s direct contribution to a country’s GDP) rather than microeconomic measures (e.g., restaurants, accommodation capacity, food and beverage services, and the number of hotels at destinations). Notably, tourism is a significant contributor to the global economy, both to gross domestic product and to employment (Pulina & Santoni, 2018). DEA analysis in the context of the hotel sector concerns the number of hotel-type establishments (Nurmatov et al., 2021; Pardo Martínez & Poveda, 2024), accommodation capacity (e.g., beds, rooms), physical space (e.g., surface, areas), food and beverage subsidiaries, franchises, employees and labor cost, occupancy (e.g., overnight stays, rate, length of stay), reservations, and stays (Nurmatov et al., 2021). Moreover, using macro-variables in T-DEA analysis (e.g., tourism’s contribution to GDP) can be an advantageous way to measure economic efficiency as they capture value rather than mere tourism activity volume, unlike micro-tourism indicators. Also, macro-variables help with policy relevance and strategic interpretation by capturing spillover and multiplier effects.

The “direct” properties of this study’s variables help identify the sector’s heterogeneous nature rather than conceptualizing it as a homogeneous, generic bundle of activities, especially in benchmarking research efforts. Interestingly, the obtained efficiency measures and productivity change scores are then regressed on renewables to directly research whether they act as a driver of tourism efficiency and productivity benefits (gains) rather than merely as a correlate of economic output. Such an approach elaborates on how green transition patterns might impact tourism performance rates if supported by evidence. We can argue that tourism is efficient, but is it sustainable enough to make a measurable and “visible” difference in environmental benefits, or is there still a long way to go to achieve the desired levels of sustainability for net-zero economies and zero ecological footprint? Renewables play a dominant role in this process; thus, the authors investigate their potential impacts on technical efficiency and total factor productivity change.

Lastly, there is a need for greater geographical variability when applying T-DEA to overcome the potential geographical concentration of T-DEA studies (Nurmatov et al., 2021). In the field, many studies rely on a single country, which limits the scope for more effective benchmarking. Eurozone member states are often overlooked in this strand of literature, despite their significant contributions and importance in the global tourism sector.

3. Methodology

This study implements a tourism-induced DEA for member states of the Eurozone economic space from 1996 to 2019. In the context of T-DEA, these countries are referred to as DMUs that convert multiple inputs into multiple outputs. The adopted model orientation when applying the method (input-oriented or output-oriented) depends on whether the researcher wishes to improve the input or output levels (Radovanov et al., 2020). For this reason, an output-oriented technical efficiency variance-to-scale (VRS) analysis is performed. Then, the total factor productivity change is calculated. Furthermore, we conduct a panel data analysis to regress output-oriented technical efficiency and total factor productivity on the percentage of renewables in final energy consumption.

To perform the output-oriented T-DEA analysis, two input and two output variables were used. The input variables comprise the number of jobs generated directly in the travel and tourism sector (EMP_Tour) and capital investment spending by all industries directly involved in the travel and tourism sector (INVEST_TOUR). This also includes investment spending by other industries on specific tourism assets, such as new visitor accommodation, passenger transport equipment, and restaurants and leisure facilities for specific tourism use.

The output variables constitute the GDP generated by industries that deal directly with tourists, including hotels, travel agents, airlines, and other passenger transport services, as well as the activities of the restaurant and leisure industries that deal directly with tourists (GDP_TOUR) and the inbound arrivals of tourists (ARRIVALS). The data were obtained by the World Travel and Tourism Council. Specifically, tourism’s direct contribution to a country’s GDP concerns the GDP generated by industries that deal directly with tourists, including hotels, travel agents, airlines, and other passenger transport services, as well as the activities of restaurant and leisure industries that deal directly with tourists. It is equivalent to total internal travel and tourism spending within a country less the purchases made by those industries (including imports) (WTTC, 2019).

Table 1. Descriptive statistics.

	Input Variables		Output Variables
Country	EMP_Tour	INVEST_Tour	GDP_Tour	ARRIVALS
Austria	329,970	4890	28,772	22,181,042
Belgium	118,493	2417	11,318	7,196,000
Cyprus	26,569	0.420	1450	2,594,583
Estonia	24,500	0.410	0.840	2,120,958
Finland	59,346	1632	5780	3,011,292
France	1,182,597	30,996	94,290	78,394,417
Germany	2,847,325	26,335	130,610	25,551,542
Greece	21,259	6150	14,165	17,199,558
Ireland	38,475	5489	4408	7,776,917
Italy	1,189,943	14,296	96,469	44,352,467
Latvia	28,666	0.267	0.859	1,289,833
Lithuania	28,822	0.270	0.754	1,759,167
Luxembourg	13,069	0.692	2385	908,000
Malta	16,087	0.251	0.444	1,467,917
Netherlands	483,037	4422	15,315	11,731,458
Portugal	215,722	2865	9143	10,764,417
Slovakia	49,668	0.605	1669	2,408,667
Slovenia	29,958	0.640	1542	2,092,958
Spain	811,490	18,617	62,181	57,294,042
Mean	395,526	6403	25,389	15,794,486
Standard deviation	689,122	9050	38,674	21,151,420

Notes: EMP_Tour is measured in thousands of jobs, INVEST_TOUR and GDP_TOUR are measured in US$ billion (in real prices), and arrivals refer to inbound tourist arrivals.

presents the descriptive statistics of input and output variables included in T-DEA. Although the countries of interest are members of the Eurozone, this study’s monetary variables are expressed in US$ billion (in real prices). This fact does not bias any calculations, methodological steps, or results interpretation, whereas the data used were obtained from WTTC in this measurement unit. The adopted ARDL model focuses on dynamic adjustments. From a methodological point of view, technical efficiency scores obtained from DEA are bounded, meaning they cannot exceed 1 or fall below 0. Consequently, this study’s ARDL/ECM framework is chosen to capture dynamic adjustment and long-run equilibrium, which static Tobit models cannot.

The fractional programming problem is transformed into a linear programming problem by using the Charnes–Cooper transformation, rescaling the input and output weights, and converting the original efficiency ratio into a linear function that can be solved with linear programming techniques.

The input proxy “investment spending” is used to contextualize capital inputs, given data availability. It is assumed that investment is rapidly converted into operational tourism assets since investment in one year might not yield output until the following year, yet using investment (money flow) as a proxy for capital input is common when stock data are not available.

This study is a tourism-induced DEA analysis at the national level, differentiated from previous studies that mostly focus on hotel industry proxies, specific destinations, or a single country. The selection of variables in the present DEA framework is guided by both theoretical relevance and methodological approaches. In particular, the number of decision-making units should substantially exceed the number of inputs and outputs to ensure adequate discriminatory power. Given the cross-country panel structure and the number of observations, expanding the set of indicators could inflate dimensionality and yield spurious efficiency scores. Moreover, the adopted indicators reflect data availability and comparability across the countries of interest (the Eurozone) over the full sample period. The selected variables capture the core dimensions of tourism production. Expanding the indicator set would compromise the discriminatory power of the DEA model, given the sample size and the availability of data across countries and years. The selection of indicators was up to the year 2019. Given the research interest in conceptualizing high-impact input and output variables directly associated with tourism, and based on their magnitude and importance in exploring performance rates in the Eurozone’s tourism sectors, this was an issue that could be difficult to overcome. Moreover, the econometric model explicitly accounts for the 2019 structural break. The pre-pandemic focus, therefore, provides a stable benchmark against which post-pandemic dynamics can be examined in future research. For the DEA analysis, the MaxDEA X 12.1 was employed, whereas for the panel analysis, the STATA statistical package was used.

3.1. Output-Oriented Technical Efficiency T-DEA

The underlying premise of the output-oriented DEA is to maximize outputs while holding inputs constant. The term technical efficiency refers to a DMU’s ability to obtain the maximum possible output based on a given set of inputs. The orientation of DEA toward output-oriented technical efficiency helps identify potential areas for improvement by measuring the gap between current and potential outputs for a given set of inputs. Consequently, it measures how much a DMU can proportionally increase its outputs by keeping input levels unchanged relative to a best-practice frontier formed by the most efficient DMUs. Scores below 1 for output-oriented technical efficiency indicate that the DMU is operating inefficiently, essentially below the production frontier. This means the DMU could produce more output with the same level of inputs, so its observed output is below the maximum possible output for those inputs. Then we calculate the improvement effort to achieve its maximum output potential. In contrast, a score of 1 suggests that the DMU under research is efficient. Practically, this means that the DMU in question generates the maximum possible output given that input levels are the same in the process.

To meet this study’s research objectives, the Banker–Charnes–Cooper (BCC) DEA model (Equation (1)) is performed (Banker et al., 1984).

$$\begin{gathered} max={\sum }_{i=1}^m{v}_i{x}_{ik}+v_o\mathrm{subject}\mathrm{to} \\ \left\{\begin{array}{c}{\displaystyle \sum _{r=1}^q}{\mu }_r{y}_{rj}-{\displaystyle \sum _{i=1}^m}{v}_i{x}_{ij}+v_0\le 0 \left(j=1,\dots ,n\right),\\ {\displaystyle \sum _{r=1}^q}{\mu }_r{y}_{rk}=1,\\ {\mu }_r\ge 0 \left(r=1,\dots ,q\right), v_i\ge 0 \left(i=1,\dots ,m\right), v_0\in R\end{array}\right. \end{gathered}$$

(1)

Let there be a set of DMU_j (j = 1, …, n). Define (x_1j, …, x_mj) as the input vector of DMU_j, with the input weight vector (v₁, …, v_m), and define (y_ij, …, y_qj) as the output vector DMU_j with the output weight vector (u₁, …, u_q).

Assume that each DMU_j consumes x_ij amount of input i to produce y_rj amount of output r and that the input and output DMU_k (k = 1, …., n) being evaluated are respectively (x_ik, …, x_mk) and (y_ik, …, y_qk), where, x_ik ≥ 0 and y_rk ≥ 0. Let μ_r = tu_r and v_i = tv_i, where $t={(\sum _{i=1}^m{v}_i{x}_{ik})}^{-1}$. Also, in Equation (1), μ₀ and v₀ are two free variables.

3.2. TFPC

TFPC measures the efficiency with which a set of DMUs (e.g., Eurozone countries) improve their output relative to their combined inputs over time. The Malmquist productivity index is used to measure this change. Färe et al. (1994) specify an output-based Malmquist productivity change index (Equation (2)).

$$M_o(y^{t+1},x^{t+1},y^t,x^t)={\displaystyle \frac{D_o^{t+1}\left(x^{t+1}, y^{t+1}\right)}{D_o^t\left(x^t{,y}^t\right)}}\times \left[\right({\displaystyle \frac{D_o^t\left(x^{t+1},y^{t+1}\right)}{D_o^{t+1}\left(x^{t+1},y^{t+1}\right)}}\left)x\right({\displaystyle \frac{D_0^t(x^t,y^t)}{D_o^{t+1}(x^t,y^t)}}{\left)\right]}^{{\displaystyle \frac{1}{2}}}$$

(2)

$D_O^t(x,y)$: output distance function evaluated with $t$ period technology;

$D_O^{t+1}(x,y)$: output distance function evaluated with period $t+1$ technology;

$x^t$, $y^t$: inputs and outputs at time $t$;

$x^{t+1}$, $y^{t+1}$: inputs and outputs at time $t+1$.

The ratio outside the brackets measures the change in relative efficiency. Essentially, it measures the change in how far observed production deviates from the maximum potential production between years t and t + 1. The geometric mean of the two ratios inside the brackets captures the shift in technology (e.g., technical change). Scores greater than 1 indicate improved productivity for the years of reference, whereas scores equal to 1 suggest no productivity change (stagnation). In contrast, scores below 1 indicate a decline in productivity. In practice, higher TFPC scores indicate that more output is produced per unit of input.

3.3. Panel Data Analysis

A panel data analysis is conducted to investigate potential impacts and relationships between tourism-oriented variables and sustainability issues, focusing on renewables as part of total final energy consumption. To this effort, cross-sectional dependence tests are implemented adopting Pesaran’s (2004) methodology. The following step concerns a unit root test developed by Lee and Tieslau (2019) to define whether the panel variables are stationary. Additionally, this test can help detect potential abrupt structural breaks in the panel series.

The analysis then continues by processing a Common Correlated Effects (CCE)–augmented Autoregressive Distributed Lag with Error Correction Model (ARDL/ECM) using the Pooled Mean Group (PMG) estimator. Such an approach relies on the studies of Pesaran et al. (1999) and Pesaran (2006). The ARDL model, reparameterized in error-correction form, is used as a dynamic specification to examine persistence, mean reversion, and long-run equilibrium relationships among panel variables. In this context, the error-correction term captures the speed at which short-run deviations from a stable equilibrium relationship are corrected, while the long-run coefficients describe conditional associations. The efficiency scores and Malmquist productivity indices are relative and frontier-based measures rather than stochastic outcomes. Accordingly, policy implications are based on observed dynamic patterns and adjustment properties.

Specifically, the CCE augmentation includes cross-sectional averages of both dependent and independent variables. It is employed to capture unobserved common factors. The PMG estimator treats heterogeneous short-run dynamics and error-correction speeds across units. Furthermore, it imposes homogeneity on the long-run coefficients. Furthermore, the adopted methodology considers structural break dummies to account for major shifts. Additionally, it controls for unit- and time-specific effects to account for unobserved heterogeneity. This approach is well-suited for heterogeneous panel data, where cross-sectional dependence may arise from unobserved common shocks or spillover effects across panel units. This combined methodology thus provides robust and efficient estimates even in the presence of cross-sectional dependence and heterogeneous dynamics across panel units. Therefore, it is employed to decode the impacts of renewables on efficiency and productivity change scores obtained from T-DEA. Panel analysis is based on the following model specifications (Equations (3)–(5)). In

Table 2. Variable interpretation of Equations (1)–(3).

Variables in Equations	Interpretation
$y$	Dependent variable
$x$	Explanatory variable
${\Delta y}_{it}$	Short-run change (associations) in the dependent variable
${\Delta x}_{it}$	Short-run changes (associations) in the explanatory variable
$a_i$	Unit-specific fixed effect/intercept.
${\lambda }_i$	Speed of adjustment—how quickly deviations from long-run equilibrium are corrected (ECM-PMG). Rate of mean reversion toward the long-run equilibrium relationship
$\left(y_{i,t-1}-{\beta x}_{i,t-1}\right)$	Error-Correction Term (ECT)—deviation from long-run equilibrium
${\gamma }_i$	Short-run coefficient of x (PMG)—captures short-run associations
${\displaystyle \sum _m}{\delta }_{m,i}{D}_{m,it}$	Short-run breaks
$D_{m,it}$	Structural break or policy dummy
${\overline{y}}_t$$,{\overline{x}}_t$	Cross-sectional averages of y and x (CCE)
${\psi }_{y,i }$$,{\psi }_{x,i}$	Loadings on the cross-sectional averages (CCE)
$\beta$	long-run coefficient parameter—can describe systematic levels linkage—conditional long-run associations
m	Index for multiple structural break dummies
${\delta }_{mi}$	Heterogeneous effects, the short-run impact of a structural break dummy
${\epsilon }_{it}$	Error term in Equation (1)
$a_i$	Unit-specific intercept
$u_{it}$	Error term in Equation (2)

the interpretation of the variables is presented.

$${\Delta y}_{it}=a_i+{\lambda }_i\left(y_{i,t-1}-{\beta x}_{i,t-1}\right)+{\gamma }_i{\Delta x}_{it}+\sum _m{\delta }_{m,i}{D}_{m,it}+{\psi }_{y,i}{\overline{y}}_t+{\psi }_{x,i}\overline{x_t}+{\epsilon }_{it}$$

(3)

$$\begin{gathered} y_{it}=a_i+{\beta x}_{it}+u_{it} \\ \left(\mathit{long-run}\mathit{relationships}\right) \end{gathered}$$

(4)

$$\begin{gathered} {\Delta y}_{it}=a_i+{\lambda }_i\left(y_{i,t-1}-{\beta }_{i,t-1}\right)+{\gamma }_i{\Delta x}_{it}+{\psi }_{y,i}{\overline{y}}_t+{\psi }_{x,i}\overline{x_t}+{\epsilon }_{it} \\ \left(\mathit{short-run}\mathit{relationships}\right) \end{gathered}$$

(5)

${ECM}_{i,t-1}=y_{i,t-1}-{\beta x}_{i,t-1}$: size of last period’s imbalance;

${\lambda }_i{ECM}_{i,t-1}$: amount of adjustment back toward equilibrium in period t;

${\lambda }_i$: how much of that imbalance is corrected each period.

The analysis concerns two baseline models and their two reverse versions as follows (

Table 3. Model specifications.

	Dependent Variable	Explanatory Variables
Baseline models
Model 1	Output-oriented technical efficiency	Renewable energy consumption
Model 2	Total factor productivity change	Renewable energy consumption
Reverse Models
Model 3	Renewable energy consumption	Output-oriented technical efficiency
Model 4	Renewable energy consumption	Total factor productivity change

).

4. Results

4.1. Results of Output-Oriented Technical Efficiency

Results of output-oriented technical efficiency (output) and Malmquist productivity index (MI) are presented in

Table 4. Mean values of output-oriented technical efficiency and Malmquist productivity index.

Country	Output	MI (TFPC)
Austria	0.94	0.94
Belgium	0.91	0.98
Cyprus	0.84	1.05
Estonia	0.69	1.07
Finland	0.82	1.00
France	0.96	0.97
Germany	0.96	0.93
Greece	0.57	1.00
Ireland	0.87	0.97
Italy	0.96	0.97
Latvia	0.61	1.04
Lithuania	0.76	1.02
Luxembourg	0.98	0.96
Malta	0.81	1.06
Netherlands	0.44	1.03
Portugal	0.69	0.98
Slovakia	0.52	1.03
Slovenia	0.62	1.08
Spain	0.90	0.97
Arithmetic mean	0.78
Geomean		1.00
Stand. Deviation	0.16	0.04

Notes: (i) Output stands for output-oriented technical efficiency, MI denotes the Malmquist productivity Index and provides values for Total Factor Productivity Change (TFPC). (ii) The output averages are arithmetic means, and the MI averages are geometric means.

. Arithmetic mean values of the output-oriented technical efficiency score and the geometric mean values of the Malmquist productivity index are also provided in Table 4.

On average, the value of the output-oriented technical efficiency (output) of tourism in the Eurozone is 0.78. This implies that, on average, Eurozone countries could increase their outputs by 22% while maintaining the same levels of inputs, operating on the efficiency frontier. This score provides, on average, the efficiency gap (distance from 1). The required output expansion (percentage of growth needed) to reach the frontier is approximately 28% ((1/0.78) − 1 = 0.28). It should be noted that the efficiency gap (e.g., the technical efficiency score obtained by DEA) is the proportion by which current performance falls short of the frontier. The required output expansion (improvement effort) is the proportional increase in output required to reach the frontier (

Table 5. Efficiency gap vs. improvement effort.

Measure	Formulas (Output Technical Efficiency Score = 0.78)	Values
Efficiency Gap (compared to a full efficiency score or efficiency score of output = 1)	1 − output technical efficiency score	0.22
Improvement Effort (relative to the efficiency frontier)	(1 − output technical efficiency score) × 100	22%
Improvement Effort (relative to current performance, measure % expansion of outputs needed)	((1/output technical efficiency score) − 1) × 100	28%
The efficiency gap (distance to frontier) is 22%, whereas the effort to cover this gap (required output expansion) is 28%.

).

Luxembourg, with a score of 0.98, operates at 98% of its potential. To reach the frontier, the country requires an output expansion of approximately 0.02%. The Netherlands, with a score of 0.44, operates at 44% of its potential. To reach the frontier, the country requires an output expansion of approximately 127%.

4.2. TFPC

To calculate the average scores for total factor productivity change, we used the geometric mean, which captures compound and multiplicative changes between 1996 and 2019. As indicated in Table 4 above, the average value of total factor productivity change (geomean) is 1.00. This result suggests that, by holding input quantities constant, the tourism system in the Eurozone, on average, experienced neither increased nor decreased productivity over the entire period of interest. On average, the total output remained unchanged over the reference period (1996–2019). An explanation could be that the cumulative benefits (gains) in some years were offset by declines in other years, resulting in zero net productivity growth (stagnation). Notably, the tourism sector maintained its productivity level in the long run, although it was unable to sustain the improvements.

Based on the Malmquist Productivity Index (MI), Slovenia has the highest total factor productivity change (1.08). In contrast, Germany has the lowest value for total factor productivity change (0.93). Given that TFPC measures the shift in the frontier (technology) and the catch-up (efficiency), Germany experienced an 7% decline in productivity between 1996 and 2019, whereas Slovenia experienced an 8% increase over the same period.

Such a result indicates a downward trend in productivity or a relative decline in efficiency levels. This means the country has become relatively less productive over time, either due to reduced efficiency in using inputs or a failure to keep pace with technological advancements. The following Figure 1 presents trends of technical efficiency and productivity change.

Given the period of interest (1996–2019) and input levels included in this study, the output in the Eurozone’s tourism sector shows a clear upward trend, increasing by approximately 1% per year on average. These results indicate that the tourism sector is experiencing steady growth. On the contrary, the total factor productivity change, on average, exhibits a slight downward trend, with scores decreasing by approximately 0.1% per year. This means that while output levels expanded from 1996 to 2019, productivity growth (improvements) relative to input levels did not keep pace and eroded slightly over time.

4.3. Results of Cross-Section Dependence (CD) Tests

This study employs Pesaran’s (2004) CD test under the null hypothesis of no cross-sectional dependence for the variables being tested. As indicated in

Table 6. Results of CD tests.

Variable	Test	Statistic	p-Values
output	Pesaran CD test	19.015	0.000
TFPC		0.1052	0.9162
renew		54.704	0.000

Note: p-values indicate significance at at 1% level.

, results of CD tests reveal that output-oriented technical efficiency (output) and renewables (renew) are cross-sectionally dependent (p-values < 0.001). In contrast, total factor productivity change exhibits cross-sectional independence (p-value = 0.9162). When justified, the CD phenomenon is attributed to common shocks and spillover effects (e.g., oil price changes, energy price fluctuations, financial crises, climate crisis and shocks, environmental agreements), unobserved (latent) common factors (e.g., correlated residuals), economic and financial linkages, and interdependencies (e.g., trade openness, investment capital, and financial integration), and spatial and geographical effects (e.g., neighboring countries, abolition of artificial borders). If left untested or ignored, CD can bias panel tests and lead to misleading results (Pesaran, 2004; De Hoyos & Sarafidis, 2006; Chudik & Pesaran, 2015). Interestingly, total factor productivity change exhibits no CD. One explanation is that domestic factors such as local infrastructure quality, national policy frameworks, labor market dynamics, and country-specific demand conditions generate the “independence” property of this variable. Furthermore, these drivers of variables’ independence are highly idiosyncratic and not systematically synchronized across Eurozone members. As a result, total factor productivity change exhibits no significant cross-sectional dependence in the diagnostic tests employed. Whether panel data exhibit CD or not is crucial for the analysis the researcher should follow to obtain unbiased results (e.g., regression coefficients, evidence of relationships and directions, and interpretations).

4.4. Results of Unit Root Tests

Unit root tests were based on the work of Lee and Tieslau (2019) under the null hypothesis of non-stationarity. Results indicate that all tested variables are stationary as p-values reject the null hypothesis (p-values < 0.001) (

Table 7. Results of unit root tests.

Variables Tested	Individual Value	Break 1	Break 2	Lags	PDLM	p-Values
output	−7.728	1996	2011	5	−8.266	0.000
TFPC	−19.056	2006	2001	0	−27.654	0.000
renew	−8.748	2019	2006	1	−9.842	0.000

Notes: (i) PDLM stands for Panel Lagrange Multiplier with level and trend shifts; (ii) p-values lower than 0.01 indicate significance at 1% level; (iii) Break 1996: Energy market liberalization and restructuring; (iv) Break 2011: Aftermath of global financial crisis; (v) Break 2006: EU2020 climate and energy package; global oil price spike, energy price shock before 2008 financial crisis; (vi) Break 2019: European Green Deal; (vii) Break 2001: Dot-com crisis (stock market) and global productivity slowdown.

). Consequently, variables under consideration carry no unit roots. This test treats the CD phenomena justified in the previous step.

4.5. Results of Regression Analysis

This study employs the panel Autoregressive Distributed Lag (ARDL) model, reparameterized into an error-correction form (ECM) (Pesaran et al., 1999). Such an approach enables the detection of long-run equilibrium and captures short-run dynamics in a panel series of interest. Since panels are cross-sectionally dependent, the adopted model is augmented using the Common Correlated Effects (CCE) approach indicated by Pesaran (2006). This method concerns cross-sectional averages of the variables being tested. The employed methodology considers common shocks, country heterogeneity, and abrupt structural breaks. Results of the baseline Model 1 are provided in

Table 8. Results of the CCE-ARDL/ECM baseline model where the dependent variable is output-oriented technical efficiency (y: output, x: renewables) (Model 1).

Dependent Variable: D.y	Coeff.	Std. Error	z	P > |z|	[95% Conf. Interval]
Short-Run Estimates
Mean Group:
D.x	0.0023	0.0047	0.48	0.629	−0.006929	0.01146
D_posty2011	0.0039	0.0128	0.31	0.759	−0.021203	0.02909
_postx2006|	0.0113	0.0191	0.59	0.554	−0.026114	0.04868
D_postx2019	−0.0119	0.0101	−1.18	0.238	−0.031818	0.00789
Long-Run Estimates
Pooled:
L.y	−0.6221	0.07895	−7.88	0.000	−0.77681	−0.46731
x	0.0047	0.00716	0.65	0.513	−0.00935	0.01872
Cross-sectional averaged variables: y (output) × (renew)
Number of lags used: 1

, and the results of the baseline Model 2 are presented in

Table 9. Results of the CCE-ARDL/ECM model where the dependent variable is total factor productivity change (y: TFPC, x: renewables) (Model 2).

Dependent Variable: D.y	Coeff.	Std. Error	z	P > |z|	[95% Conf. Interval]
Short-Run Estimates
Mean Group:
D.x	−0.0075	0.0113	−0.67	0.506	−0.0298	0.0146
D_posty2001	−0.0071	0.0656	−0.11	0.914	−0.1357	0.1215
D_posty2006	0.0184	0.0478	0.38	0.700	−0.0753	0.1121
D_postx2019	−0.0042	0.0480	−0.09	0.930	−0.0983	0.0898
Long-Run Estimates
Pooled:
L.x	−1.027	0.0873	−11.77	0.000	−1.198	−0.8565
x	−0.0110	0.0070	−1.57	0.117	−0.0247	0.0027
Cross-sectional averaged variables: y (TFPC) × (renew)
Number of lags used: 1

, whereas

Table 10. Results of the CCE-ARDL/ECM baseline model where the independent variable is output-oriented technical efficiency (y: renewables, x: output) (Model 3).

Dependent Variable: D.y	Coeff.	Std. Error	z	P > |z|	[95% Conf. Interval]
Short-Run Estimates
Mean Group:
D.x	1.0234	1.4663	0.70	0.485	−1.8505	3.8974
D_posty2011	−0.2091	0.1946	−1.07	0.283	−0.5905	0.1722
D_postx2006	−0.0151	0.2613	−0.06	0.954	−0.5274	0.4971
D_postx2019	0.1454	0.2969	0.49	0.624	−0.4366	0.7274
Long-Run Estimates
Pooled:
L.y	−0.4506	0.2729	−1.65	0.099	−0.9855	0.0842
x	−3.1979	13.6878	−0.23	0.815	−30.025	23.6296
Cross-sectional averaged variables: y (renew) × (output-oriented technical efficiency)
Number of lags used: 1

and

Table 11. Results of the CCE-ARDL/ECM model where the independent variable is total factor productivity change (y: renewables, x: TFPC) (Model 4).

Dependent Variable: D.y	Coeff.	Std. Error	z	P > |z|	[95% Conf. Interval]
Short-Run Estimates
Mean Group:
D.x	0.6824	0.4154	1.64	0.100	−0.1318	1.4967
D_posty2001	−0.4586	0.3094	−1.48	0.138	−1.065	0.14776
D_postx2006	−0.1637	0.2692	−0.61	0.543	−0.6915	0.36397
D_postx2019	0.0410	0.3380	0.12	0.903	0.6214	0.70354
Long-Run Estimates
Pooled:
L.y	−0.4136	0.2030	−2.04	0.042	−0.8116	−0.01559
x	−1.7179	2.6350	−0.65	0.514	−6.8825	3.4467
Cross-sectional Averaged Variables: y (renew) × (total factor productivity change)
Number of lags used: 1

present the results of the reverse models, respectively.

In the following regression models, D denotes the first-difference operator, capturing short-term changes (associations), while L denotes the lag operator. Moreover, dummy break variables D_posty2011 denote the short-run shift after 2011, D_postx2006 denote the short-run shift after 2006, D_posty2001 denote the short-run shift after 2001, and D_postx2019 denote the short-run shift after 2019.

In Table 8, the estimated adjustment coefficient is −0.622 (p-value < 0.01), indicating statistically significant mean-reverting dynamics in output-oriented technical efficiency. Approximately 62% of any deviation from the long-run equilibrium relationship is corrected each year, implying convergence within about 1.6 years. This result reflects strong persistence properties in tourism efficiency. In contrast, the estimated long-run association between renewable energy consumption and tourism efficiency is positive but statistically insignificant (p-value > 0.10). This indicates that renewables are not systematically associated with long-run differences in tourism efficiency. Similarly, short-run changes in renewables do not exert a statistically significant effect on efficiency (coefficient = 0.0023, p-value = 0.629). Taken together, these results suggest that while tourism efficiency exhibits strong internal adjustment and stability, renewable energy adoption does not constitute a dominant long-run determinant of relative efficiency. It should be noted that the structural break in 1996 was not included in the model because this year (1996) is the first year of this study’s data set. This dummy, defined as year ≥ 1996 in the model specification, has no within-sample variation, whereas its first difference is zero for all observations. The PMG/CCE estimator drops differenced break dummies that exhibit no variation. Technically, level dummies at the initial sample point of the data set are not econometrically identifiable in the ECM specification.

In Table 9, the total factor productivity change adjustment coefficient is −1.027 and is highly significant (p-value < 0.001), indicating a very rapid mean reversion toward the long-run equilibrium relationship. Deviations from equilibrium are absorbed within approximately one year, highlighting the limited persistence of productivity shocks in the tourism sector. Long-run associations and short-run changes are not statistically significant. These findings imply that renewable energy adoption does not translate into systemic long-run productivity gains in the Eurozone’s tourism sector during the period under study. Renewables do not systematically improve or worsen tourism’s productivity over the long run. Moreover, this result is consistent with the output-oriented technical efficiency (dependent)-renew (explanatory) model, in which renewables also exhibited no long-run association. These results confirm that while productivity is mean-reverting and adjusts quickly after shocks, renewables adoption does not drive productivity changes in the tourism sector across the Eurozone countries.

In Table 10, the estimated adjusted coefficient in the reverse specification is negative and marginally significant (coefficient = −0.4506, p-value = 0.099), indicating moderate mean-reverting behavior in renewable energy consumption. Approximately 45% of deviations from the long-run equilibrium relationship are corrected each year, implying convergence within roughly two years. The long-run associations and short-run changes in renewables are statistically insignificant. These results suggest that renewable energy dynamics are driven by other factors such as national energy policies, regulatory frameworks, and broader macroeconomic conditions rather than by efficiency performance in the Eurozone.

In Table 11, where the renewable energy consumption is regressed on total factor productivity change, the estimated adjustment coefficient is −0.4136 and statistically significant (p-value = 0.042), indicating stable but relatively slow mean reversion. Approximately 41% of deviations from the long-run equilibrium relationship are corrected annually, implying convergence within about 2.4 years. The estimated long-run associations and short-run changes are statistically insignificant.

The absence of statistically significant long-run associations between renewable energy consumption and efficiency scores, and between renewable energy consumption and productivity indices, suggests that renewable energy consumption is insufficient to drive sustained performance improvements in the tourism sector. Policy relevance, therefore, lies not in direct efficiency gains but in complementary roles or renewables within innovation, investments, and technological upgrading strategies. The results indicate that without parallel advances in technology, infrastructure, and sector-specific innovation, energy transitions are unlikely to generate measurable long-run productivity or efficiency gains. This evidence illustrates the need for integrated and innovative solutions and well-defined and applicable policy frameworks across Eurozone countries.

5. Discussion

Advanced efficiency and productivity patterns remain targets for all dynamic systems, including the tourism system. The outcome attained, then measured and compared with competitors or other performers, remains an ongoing area of research. Shaping efficiency outcomes is not a static approach but rather a dynamic, long-lasting process. This path enables the identification of gaps for improvement and the adjustment of inputs (e.g., costs) and outputs (e.g., benefits) to experience expansion and development (e.g., growth patterns) aligned with the goals and benchmarks established (e.g., progress achieved) in the sector of interest (e.g., tourism). The key issue is operational efficiency gains. Research findings from T-DEA indicate that, on average, Eurozone countries exhibit high efficiency scores (0.78), meaning they generate 78% of the output they could with unchanged inputs. Such results suggest that, relative to other countries, Eurozone destinations are close to maximizing tourism outputs given their existing input endowments of tourism-related employment and capital investment spending at destinations. This high score serves as a reference benchmark for less-efficient destinations (e.g., countries) to improve their tourism sector performance. Interestingly, in their work, Stoiljkovic et al. (2025) assessed the tourism efficiency of European countries using DEA. They found that Lithuania and Estonia have the lowest efficiencies, below 50%, and Hungary, Albania, and Bulgaria have efficiencies between 50 and 60%, whereas the average efficiency is above 80% (84%). Furthermore, Assaf (2012) benchmarked the Asia Pacific tourism industry. In this study, efficiency scores range from 0.72 (India) to 0.84 (Singapore). Indicatively, compared to these studies, the Eurozone demonstrates a lower average tourism efficiency (0.78) than the European-wide average reported by Stoiljkovic et al. (2025), while showing a broader dispersion of efficiency scores (0.44 to 0.99) than the narrower range observed in Assaf’s (2012) study. This comparison suggests greater heterogeneity in tourism performance across Eurozone countries and indicates a relatively moderate score by international standards. It should be noted that results are sensitive to methodological approaches and might vary depending on the number of input–output pairs used, the scale (micro or macro data), the DEA orientation used to obtain technical efficiency scores, the number of countries under investigation, and the time span of reference. As such, cross-study comparisons should be interpreted as indicative rather than definitive, stimulating additional research opportunities and methodological refinements or improvements.

Adjustment speeds (mean reversions) for both output-oriented technical efficiency (baseline Model 1) and total factor productivity change (baseline Model 2) evidence that the tourism sector in the Eurozone corrects deviations from its long-run equilibrium relatively quickly. In baseline Model 1, approximately 62% of any disequilibrium in tourism technical efficiency is corrected each year, meaning full adjustment occurs in about 1.6 years. In baseline Model 2, an even stronger adjustment mechanism confirms that productivity shocks are fully absorbed within roughly one year. This means they quickly restore efficiency and productivity levels without producing persistent long-run distortions, exhibiting structural resilience.

In both the reverse model, where renew was employed as the dependent variable, the error correction term is negative and significant. In the reverse version of Model 1, results indicate that 45% of any deviation from the long run is closed each year. This means that renewable energy use in the Eurozone’s tourism system tends to drift back toward its long-run path after experiencing shocks in tourism output-oriented technical efficiency. However, the mean reversion toward the long-run equilibrium relationship is weaker than in the baseline model, where tourism output-oriented technical efficiency was the dependent variable.

For the reverse Model 2, where total factor productivity change was the independent variable, the adjustment coefficient implies that about 42% of disequilibrium is corrected annually. This means that renewable energy use returns to its long-run path in roughly 2.4 years. In practice, the insignificance of the long-run coefficient of total factor productivity change implies that this correction is driven more by the internal dynamics of the energy system than by tourism productivity itself.

The lack of statistically significant long-run coefficients indicates that renewable energy adoption does not decisively shape tourism efficiency or productivity outcomes across Eurozone countries. This finding suggests that their effects on sectoral performance are indirect and mediated by broader structural conditions. In this context, policy relevance arises from the need to align renewable energy initiatives with complementary measures such as technological upgrading, innovation endeavors, and infrastructure development. Without such coordination, energy transitions alone will hardly produce sustained long-run performance improvements in the tourism sector. Consequently, this study’s policy discussion does not rely on statistically significant long-run effects; instead, it draws attention to the absence of systematic long-run associations. In practice, this highlights that renewable energy adoption alone is insufficient to drive efficiency and productivity improvements and that it must be complemented by broader structural and innovation-oriented policies. This perspective shifts the focus from direct performance impacts to the conditions under which energy transitions can become economically meaningful for the tourism sector. In particular, it underscores the importance of coordinated policy frameworks that integrate energy transitions under sector-specific investment accomplishments, technological upgrading, and institutional support.

The trend analysis in Figure 1 states that tourism’s technical efficiency (output) increased steadily over time. In practice, this means that Eurozone countries can efficiently transform tested inputs into outputs. On the contrary, productivity change patterns decline at a low pace in the period 1996–2019. Such a divergence confirms that productivity change patterns showed a slight inefficiency in improving the system’s performance, practically in terms of innovation and technological upgrading. This result indicates that the Eurozone countries did not experience significant productivity gains, disclosing potential underlying structural constraints (e.g., technological). Consequently, renewables evolve in tandem with broader macroeconomic and policy shifts rather than independently shaping tourism efficiency. Also, a possible explanation is that countries under research due to constraints (e.g., technological constraints) did not manage to keep pace with long-lasting productivity gains (improvements). To reverse this slight decline in productivity, it is crucial to leverage European funding effectively and to promote sustainable development and investment plans. In this way, Eurozone countries will benefit from structural reforms and reverse the underlying decline in productivity. These investment plans might consider integrating innovation into the tourism sector, digitalization, and promoting an “eco-friendly” business character to accelerate the energy transition toward increased renewable energy consumption in the energy mix.

Results indicate that the energy transition operates primarily as a cost-absorbing adjustment rather than a productivity-enhancing mechanism. For instance, renewable energy investments often entail high upfront capital costs and transitional inefficiencies. Also, their productivity benefits may materialize with complementary technological adoption. Moreover, potential “free-riding” phenomena on national energy grids might be due to the fact that renewables have not yet been sufficiently embedded at the firm or destination level, indicatively in green infrastructure, energy-efficient accommodation establishments, and digitalization, to alter the sector’s production function in a measurable way. One additional explanation is that the renewables integration is observed primarily at the macro level rather than being structurally integrated into tourism-specific operations, thereby limiting the ability to reshape the tourism sector’s production function.

The observed decline in total factor productivity change, despite rising technical efficiency, is consistent with the structural characteristics of the tourism sector. One explanation is that, because tourism is a labor-intensive industry, it faces high labor costs without commensurate productivity gains. Wages frequently outpace technological progress. In essence, wages tend to rise, leading to rising costs without proportional productivity gains, as in Baumol’s cost disease, where wages rise despite low productivity growth. Moreover, the sector relies on low- and medium-skilled labor and exhibits relatively slow diffusion of automation, digital platforms, and process innovation beyond front-end services. Furthermore, upskilling and reskilling labor might drive productivity change. As a result, even as countries improve technical efficiency, the absence of technology-driven frontier shifts results in stagnant or declining productivity growth, consistent with the empirical patterns identified in this study.

By elaborating on the operational models and tourism policies adopted by Eurozone member states, other countries or groups of countries can decode, transfer, adapt, and adopt these best practices in their own contexts. Moreover, the relatively small gap from the efficiency frontier suggests that additional improvements in the Eurozone might depend less on traditional input expansion and development and more on innovation-driven strategies, such as digitalization and the adoption of renewable energy. Not surprisingly, productivity is considered one of the most comprehensive and dependable benchmarks for performance comparisons in many industries, including tourism (Liu & Wu, 2019). For instance, the Green Deal calls for resource efficiency and a competitive economy (e.g., a decarbonized economic system, climate neutrality, reduced emissions), whereas the European Development Regional Funds enhance “green” and resilient entrepreneurship. Funding should not be used to expand capacity (as efficiency is already high) but should be specifically targeted at labor-saving technologies or digitalization to reverse the productivity decline. In practice, given the coexistence of high technical efficiency and declining productivity growth, tourism policy in the Eurozone should not only set capacity-expansion goals. Instead, it should target frontier-shifting investments to stimulate sustained productivity improvements.

Renewable energy adoption may influence efficiency and productivity through several indirect channels. Economically, it can lower long-run operating costs and reduce exposure to energy price volatility for tourism-related activities; technologically, it is often associated with infrastructure improvements and efficiency-enhancing investments; and institutionally, it reflects regulatory quality, policy commitment to sustainability, and a destination’s orientation toward environmental standards. Together, these factors may shape the production environment and the competitiveness of the tourism sector, thereby affecting performance rates (e.g., efficiency and productivity outcomes). Furthermore, the economic, technological, and institutional effects of renewable energy adoption operate primarily at the systemic rather than the firm level. Energy infrastructure, regulatory frameworks, and technological upgrading associated with renewable energy transitions affect the overall production environment in which tourism activities operate, influencing sector-wide resource allocation, cost structures, and productivity conditions. These broad effects are more accurately reflected in measures such as employment and investment directly related to the travel and tourism sector, or in tourism’s direct contribution to a nation’s GDP, than in micro-level indicators. Consequently, macro-level T-DEA provides a suitable framework for assessing how economy-wide sustainability transitions may be associated with the tourism sector’s efficiency and productivity outcomes.

Tourism is profoundly linked to environmental concerns as it can impact the ecological status of natural resources (Sharpley & Telfer, 2015; Destek & Aydın, 2022). From this perspective, if the case is to “decarbonize” the tourism economy, it is crucial to examine energy-efficiency concepts to promote eco-friendly business practices without compromising growth (Marques et al., 2019). Importantly, Destination Management Organizations (DMOs) should promote and enhance effective destination management based on the principles of sustainable development, considering both supply and demand perspectives (Foris, 2020). Notably, Bano et al. (2021) emphasized the crucial role of renewables in promoting tourism and fostering economic growth within the context of sustainability, for instance, by adopting clean, sustainable energy sources as indicated by the United Nations Sustainable Development Goals (Leitao & Balsalobre-Lorente, 2021).

Identification of driving forces affecting national competitiveness is fundamental to establishing long-lasting growth (Su et al., 2023). However, globalization and the free economy signal that it is time to shift from the “competitive advantage” toward the “collaborative advantage” in the long journey of success (Fyall et al., 2012).

6. Conclusions

The present study measures the output-oriented technical efficiency and total factor productivity change in the context of tourism. Then, a panel data analysis is performed to capture the impacts and relationships of renewables on these efficiency- and productivity-related variables, and vice versa. For all analyses, we refer to the Eurozone economic space between 1996 and 2019.

Test results reveal that, on average, Eurozone countries exhibit a high technical efficiency score (0.78), while total factor productivity changes slightly, declining over the years of reference (1996–2019).

This study’s model specifications exhibit stable, rapid convergence to the long-run equilibrium when renewable energy consumption is the independent variable. The obtained mean reversion toward the long-run equilibrium relationships implies that deviations in output-oriented technical efficiency and productivity change patterns are largely corrected within approximately 1 to 2 years. In both reverse specifications, the error-correction terms show that renewables still adjust back toward their long-run equilibrium aftershocks in output-oriented technical efficiency and productivity change, albeit at a much slower and less precise rate than in the baseline models. Furthermore, trend analysis confirmed a steady and sustained increase in output-oriented technical efficiency over time. In contrast, a decline in total factor productivity change highlights diverging performance dynamics, as it shows a slight downward trend, with scores decreasing by approximately 0.1% per year.

This study addresses the stated research questions by examining the evolution of tourism efficiency and total factor productivity change by assessing whether systematic long-run relationships exist. The findings indicate that while tourism efficiency and productivity exhibit stable mean-reverting dynamics, renewable energy consumption is not systematically associated with long-run improvements. Furthermore, the estimated adjustment dynamics reflect internal persistence rather than structural interdependence.

A limitation of this research is the limited set of indicators used to represent tourism activity, driven by data availability and the need to preserve the discriminatory power of the DEA framework. While the selected variables capture core aspects of tourism performance, additional research could extend the analysis by incorporating disaggregated indicators reflecting tourism market segmentation, for instance, business and leisure tourism spending. In addition, the analysis focuses on the pre-COVID-19 period due to data availability. However, this approach helps isolate long-run dynamics from pandemic-driven structural breaks, while offering research opportunities for future endeavors.

This study can facilitate future research by structuring model specifications, including additional environmental variables such as primary energy consumption as an energy-efficiency proxy. Furthermore, researchers can further examine the heterogeneous nature of the tourism sector by testing how high-impact market segments, such as business and leisure tourism, affect technical efficiency and productivity patterns. For this reason, they can contextualize inputs and outputs in DEA under a different logic. The critical issue that remains is how to achieve sustainable growth by applying high-leverage practical implications in real economy conditions.