Testing Baumol’s Cost Disease in Tourism: Productivity, Prices, and Labor Costs in Selected EU Countries Amid COVID-19 and the Russo–Ukrainian War

Abstract

This paper investigates the impact of the transition from manufacturing to tourism on sectoral productivity, output prices, and labor costs. Using panel data econometric models for 15 selected EU countries from 2011 to 2023, the study confirms key dynamics predicted by Baumol’s cost disease (BCD) hypothesis. The findings reveal that higher productivity is positively associated with both implied prices and hourly labor costs across sectors, supporting the wage equalization mechanism central to BCD. However, the relationship between productivity and wages or prices is weaker in labor-intensive sectors like tourism, underscoring their structural vulnerability to wage-driven cost pressures. Additionally, the analysis captures the impact of major external shocks, including the COVID-19 pandemic and the Russo–Ukrainian war, treated as jointly sourced super-shocks. The regression results indicate significant price disruptions following these shocks, whereas no statistically significant trend in labor costs was detected in the post-treatment period. These results highlight the differential effects of external shocks on wages versus prices, emphasizing the challenges faced by low-productivity, labor-intensive sectors in managing cost dynamics. The findings provide valuable insights for policymakers addressing sectoral imbalances in the context of BCD and navigating the economic consequences of global disruptions.

1. Introduction

Baumol’s cost disease (BCD), introduced by [1], explains why labor-intensive sectors like healthcare, education, and tourism experience rising real costs due to stagnant productivity. It highlights why the costs of labor-intensive services rise faster than those of goods-producing industries with higher productivity gains. This theory is particularly relevant to service sectors such as healthcare, education, tourism, and the arts, where productivity improvements are inherently constrained by the need for human interaction and attention. In these sectors, the value of the service is often tied to the quality and duration of the provider’s attention, making the adoption of productivity-enhancing technologies more challenging without sacrificing service quality. As a result, labor costs in these industries tend to increase faster than productivity, ultimately leading to rising prices and slower economic growth. This phenomenon is especially pronounced in developed economies, where services dominate the economic structure and contribute more significantly to GDP.

In most developed economies of the European Union, the service sector dominates economic activity, accounting for 60–80% of GDP [2]. For instance, the United Kingdom (73%), Germany (64%), France (70%), and Luxembourg (81%) exemplify this trend, reflecting the sector’s critical role in driving economic growth and employment. Key subsectors within the service economy include financial services (e.g., banking and insurance), professional services (e.g., legal, consulting, and IT), healthcare and education, and tourism and hospitality.

Tourism services, such as waitering, tour guiding, and concierge activities, rely heavily on human attention to deliver personalized, high-quality experiences to customers. These roles are labor-intensive and difficult to automate without compromising service standards. For instance, a waiter in a restaurant must dedicate their full attention to ensuring customer satisfaction, while a tour guide must engage with travelers, answer questions, and provide an interactive experience. The necessity of human interaction in these roles underscores the “attention economy” embedded in tourism, where the service provider’s time and focus are central to the value delivered. As in their Experience Economy framework [3] argue that such interactions are not merely transactional but are increasingly designed to create memorable, customized experiences that enhance the perceived value of the service. This reliance on attention not only limits productivity growth but also exacerbates the impact of rising labor costs, as argued by [4].

Furthermore, the interplay between labor costs, productivity, and prices in tourism versus manufacturing sectors has not been comprehensively analyzed within the framework of Baumol’s cost disease. This study addresses these gaps by testing Baumol’s hypothesis in EU tourism sectors during an era of unprecedented economic upheaval—a period that brought two jointly associated shocks: the COVID-19 pandemic (2020–2022) and the subsequent Russo–Ukrainian war shock (2022–2023).

While recent studies examine Baumol effects through cost-share decompositions [5], our approach identifies causal mechanisms using these exogenous sector-specific interventions. The super-shocks created a natural experiment to test Baumol’s rigidity hypothesis under extreme labor market conditions, where tourism’s vulnerability to external crises contrasts sharply with manufacturing’s adaptive capacity.

To guide this analysis, we propose three research hypotheses that frame the core issues explored in this study:

H1. 

Productivity growth in the tourism sector shows a weaker effect on wages and prices compared to other industries, consistent with Baumol’s cost disease.

H2. 

Exogenous shocks, such as the COVID-19 pandemic and the Russo–Ukrainian war, exacerbate price volatility in tourism, exposing structural vulnerabilities tied to labor and energy dependencies.

H3. 

Technological innovation and labor market reforms can mitigate the impact of Baumol’s cost disease by enhancing productivity and resilience in tourism.

By addressing these hypotheses, this study aims to provide theoretical insights into the applicability of Baumol’s cost disease in the tourism sector while offering practical recommendations for policymakers and industry stakeholders. The findings contribute to bridging the gap between macroeconomic theory and the operational realities of this labor-intensive industry.

This paper contributes to the literature by addressing a critical gap in understanding the relationship between productivity, labor costs, and prices in the tourism sector, offering insights into the structural challenges it poses for economic growth in the selected areas of the EU. Specifically, it examines the implications of relying on tourism as a central development strategy while neglecting the challenges of deindustrialization, which can undermine long-term economic stability.

The shift away from manufacturing in many developed economies has triggered significant structural changes. As industries deindustrialize, there is an increasing transition toward knowledge-based sectors, such as IT, R&D, and consulting, as highlighted by [6]. At the same time, political globalization has further reshaped economic priorities, with production increasingly moving to lower-cost regions. In this context, many economies have turned to tourism as a compensatory growth strategy to address deindustrialization’s socioeconomic effects. However, this reliance on tourism brings its own set of vulnerabilities that require careful analysis, particularly in light of globalization’s broader economic transformations.

This paper is structured as follows. After the introductory part, the next section provides a detailed literature review, highlighting the key studies and gaps related to BCD in the tourism sector, particularly in the context of the post-COVID era and the Russo–Ukrainian war, jointly observed as a compounded super-shock. This is followed by the Theoretical Framework of BCDH, which outlines the conceptual underpinnings of BCDH and its application to tourism. After this, the paper describes the data sources and datasets used, ensuring transparency in how the empirical evidence is derived. The subsequent section, Empirical Specification, presents the methodological approach, including the baseline technique and robustness strategies employed in the analysis. We examine sectoral asymmetries in prices and labor costs using productivity interaction terms, both without and with a Difference-in-Differences (DiD) framework to estimate the causal effect of super-shocks on these variables.

The empirical part of the paper then follows, where the findings are presented and interpreted. The discussion and conclusion section synthesizes the results, draws policy implications, and suggests avenues for future research. Finally, the paper concludes with a reference list, and the Appendix includes the Mathematical Framework of BCDH and a complete list of the tables, which are based on the empirical evidence generated through the analyses conducted in this study.

2. Literature Review

Despite the growing significance of Baumol’s cost disease hypothesis (BCD) for understanding service sector dynamics, its application to tourism remains notably underexplored—especially in light of recent global disruptions such as the COVID-19 pandemic and the Russo–Ukrainian war. This chapter reviews both the theoretical foundations and empirical studies of BCD in tourism, structuring the discussion across seven key themes to highlight major debates, methodological issues, and research gaps.

2.1. BCD and Tourism: Gaps and Challenges

Although Baumol’s cost disease (BCD) has long provided a valuable framework for analyzing cost and productivity trends in labor-intensive sectors [1], its application to tourism remains underdeveloped. This is particularly significant given that tourism, as a labor-intensive industry, faces unique structural barriers to productivity growth. For example, scholars identify key challenges such as labor intensity, price-wage asymmetry, and technological constraints, which hinder productivity improvements in the sector [7].

These barriers create a mismatch between rising wages and stagnant productivity, a phenomenon that our study empirically validates through an EU-sectoral analysis of post-pandemic price-wage decoupling.

Empirical research on BCD has traditionally focused on sectors like healthcare and education, where technological substitution and productivity measurement are relatively straightforward [5].

In contrast, tourism’s reliance on experiential value and human interaction complicates both productivity improvements and measurement, highlighting the need for further sector-specific investigations. By contrast, tourism’s unique features—seasonality, experiential value creation, and its dependence on emotional and interactive labor—complicate both empirical analysis and effective policy response [8,9].

One study emphasizes that productivity growth in tourism is structurally constrained by the nature of service delivery, which requires high levels of emotional labor and direct human participation in co-created experiences [10]. These aspects are difficult to automate or substitute with technology, a dynamic also observed in the context of music festivals [10]. Such parallels from the broader cultural sector reinforce the idea that tourism is a prime example of a field where service quality is fundamentally tied to labor, thereby amplifying the effects of BCD.

The vulnerabilities of tourism have been starkly exposed by the COVID-19 pandemic and the Russo–Ukrainian war. Travel restrictions, health concerns, and geopolitical instability have disproportionately affected tourism, resulting in sharp demand declines, increased operational costs, and heightened labor market precarity—especially in economies heavily reliant on tourism for employment and foreign exchange [11,12]. While some studies point to technological adaptation—such as digital booking platforms or virtual tourism—as potential mitigators of these constraints, the evidence remains mixed. Some scholars argue that such adaptations cannot fully overcome the inherent labor-intensity of tourism [13], while others note that hybrid models may offer partial solutions [14].

A further challenge is the methodological division between quantitative, cross-country panel analyses—often hampered by limited data and sectoral aggregation—and qualitative or case-study approaches that capture sectoral nuance but may lack generalizability [5,15]. This divergence complicates the development of robust, context-sensitive policy recommendations and underscores the need for mixed-methods research.

In summary, while BCD theory is highly relevant to tourism, the literature remains fragmented, methodologically diverse, and insufficiently attentive to the sector’s complex dynamics—particularly under recent global shocks. This creates an urgent need for new studies that rigorously test BCD in tourism-specific contexts and critically examine the interplay between productivity, prices, and labor costs during external crises. Addressing this gap is essential for informing effective, resilient, and sustainable tourism policy.

2.2. Empirical Evidence of BCD in Tourism

Empirical evidence from diverse contexts underscores both sectoral commonalities and global variability in BCD’s impact. For example, data from China’s performing arts sector demonstrate that while digital innovations—such as automated ticketing and customer management—can partially reduce operational costs, the fundamental dependence on human labor persists [16]. Similarly, research on the UK tourism sector shows that advances in the digital economy have stimulated sectoral growth, but parallel studies in non-OECD countries reveal more pronounced BCD effects due to limited technological diffusion and entrenched structural constraints [17]. These findings reinforce that the ability to mitigate BCD through technology is highly contingent on local infrastructure, labor market conditions, and policy environments.

Notably, the literature highlights a sharp contrast between OECD and developing countries. High-income economies generally benefit from greater technological adoption, more formalized labor markets, and supportive regulatory frameworks, which collectively help moderate BCD-related cost growth. In contrast, tourism industries in developing economies are often characterized by informal employment, limited access to capital, and weaker institutional support, resulting in heightened exposure to rising labor costs and greater vulnerability to external shocks as noted by [13,18].

Despite these important insights, significant gaps remain. Much of the existing empirical work fails to fully account for sector-specific nuances, such as the informal labor dynamics that can temporarily suppress wage growth, or the long-term effects of external disruptions like pandemics and geopolitical crises. Furthermore, there is a methodological divide between studies employing macro-level quantitative data, which may obscure sectoral variation, and those utilizing qualitative or case-based approaches, which offer richer context but less generalizability, addressed by [5,15].

To date, few studies [19,20], have explicitly examined how BCD mechanisms interact with acute external shocks, such as the COVID-19 pandemic and the Russo–Ukrainian war—events that have exacerbated structural vulnerabilities in tourism-dependent economies. By addressing these underexplored intersections, the present study seeks to advance the empirical literature and provide a more nuanced and comprehensive understanding of how BCD shapes cost dynamics, labor market outcomes, and resilience in the global tourism sector.

2.3. The Tourism-Led Growth Hypothesis and Structural Vulnerabilities

The tourism-led growth hypothesis has long posited that tourism development can serve as a powerful engine of economic growth, particularly for regions endowed with rich natural and cultural resources. This perspective has guided much policy and investment in both developed and developing economies. However, a growing strand of research has increasingly challenged this optimistic narrative by highlighting the structural vulnerabilities that are deeply embedded in tourism-dependent growth models.

For example, one study [18] utilizes a multisector growth model to demonstrate that the benefits of tourism specialization are not automatic: positive growth effects materialize only when the tourism sector achieves productivity gains that are at least comparable to those in other sectors. This insight stands in sharp contrast to the mechanism described by Baumol’s cost disease (BCD) in [1], and [7] that argues that tourism’s inherently labor-intensive nature places a ceiling on its productivity growth and, consequently, its capacity to drive sustainable economic expansion.

Building on this critique, the study of [18] introduces the “Beach Disease” hypothesis—an extension of the “Dutch Disease” concept—arguing that excessive reliance on tourism can lead to structural imbalances, especially in developing countries. These imbalances stem from tourism’s acute vulnerability to external shocks (e.g., pandemics [19], geopolitical conflicts, climate disasters) and from the sector’s limited ability to leverage technological innovation to offset rising costs or declining demand [13]. The COVID-19 pandemic and the Russo–Ukrainian war have provided stark empirical evidence of these risks, exposing the fragility of tourism-dependent economies and reinforcing the constraints identified by BCD.

Cross-country studies [18,19] further suggest that while some small, tourism-intensive economies can achieve short-term gains, over-reliance on tourism can crowd out more diversified, resilient forms of growth. This is particularly problematic in regions where tourism revenues are highly volatile and where local labor markets are susceptible to informalization and wage stagnation.

These findings collectively underscore the urgent need for diversified economic strategies that reduce excessive reliance on tourism and directly address its structural vulnerabilities. By systematically examining the interplay between BCD effects, tourism-led growth, and external shocks, this study aims to advance theoretical and empirical understanding of how tourism-dependent economies can pursue more resilient and sustainable growth trajectories.

2.4. Technological Mitigation of BCD Effects

Technological innovation is frequently cited as a potential remedy for the cost and productivity pressures described by Baumol’s cost disease (BCD). In various service sectors, such as healthcare, empirical evidence like [21] demonstrates that information and automation technologies can substantially reduce labor costs by automating routine tasks. In the tourism sector, parallel technological interventions—such as AI-driven customer support, online booking platforms, and digital payment systems—have shown promise in streamlining operations and improving efficiency [22].

However, the effectiveness of these technologies is far from uniform across the tourism landscape. Tourism remains fundamentally human-centric: personalized care, experiential services, and emotional labor are intrinsic to value creation in travel, hospitality, and cultural experiences [23]. While AI chatbots and automated systems can efficiently handle standardized or repetitive tasks [13], they are often inadequate for the nuanced, context-rich, and culturally sensitive interactions required in high-end or bespoke tourism offerings. This limits the potential for full automation in tourism without risking a decline in service quality and customer satisfaction.

Moreover, the adoption and impact of technological solutions are shaped by substantial global disparities. Developed economies, benefiting from advanced digital infrastructure and higher levels of digital literacy, are better positioned to leverage technology to mitigate BCD effects. In contrast, macroeconomic studies of [12,24] showed that developing economies often face barriers including limited access to technology, insufficient capital, and workforce skill gaps. These disparities not only hinder the scalability of technological solutions but may also exacerbate existing inequalities within the global tourism sector.

Critical viewpoints in the literature question whether digitalization in tourism is a panacea or, in some cases, a source of new vulnerabilities. For example, the study of [25] argues that while digital transformation can enhance operational efficiencies, it may also lead to job polarization and the marginalization of low-skilled workers, particularly in regions with fragile labor protections. Conversely, another study [14] highlights the potential of “hybrid” models that combine automation with high-value human input, suggesting that such balanced approaches can maximize efficiency without sacrificing the experiential quality central to tourism.

Despite these debates, empirical research on the impact of technological adoption in tourism remains limited, particularly in terms of cross-country comparisons and the long-term effects of digitalization on labor markets and service quality [23,26]. There is a clear need for future research to explore how technology can be integrated with human labor in ways that enhance both productivity and experience. Such work should also address the policy implications of uneven digital access and the need for workforce development to ensure inclusive and resilient sectoral transitions.

By advancing understanding of these complex dynamics, the tourism industry can better navigate the structural vulnerabilities associated with BCD, ensuring both operational efficiency and the preservation of the distinct human elements that drive customer satisfaction and sustainable growth.

2.5. Critiques of BCD Theory and Sector-Specific Challenges in Tourism

While Baumol’s cost disease (BCD) provides a valuable framework for understanding the persistent cost and productivity pressures in labor-intensive services, its application to tourism has drawn significant debate and critique. Scholars question the validity of several foundational assumptions, the adequacy of measurement approaches, and the relevance of BCD to the unique structural realities of the tourism sector.

2.5.1. Assumptions of Stagnancy

A central critique concerns BCD’s presumption of inherent stagnancy in certain sectors. Such an assumption has been contested for oversimplifying the true dynamics of productivity, especially in tourism by [27,28]. In practice, technological spillovers, managerial innovation, and evolving consumer demand can drive incremental improvements, challenging the classical BCD narrative. For example, the adoption of AI-driven tools, mobile applications, and automated customer service systems has demonstrated potential to reduce labor costs and enhance efficiency in some tourism subsectors. However, these innovations rarely eliminate the sector’s reliance on human labor altogether, suggesting that stagnancy may be mitigated but not fully overcome.

2.5.2. Measurement Challenges

Another significant debate centers on how productivity is measured in tourism. Traditional metrics, such as output per worker or revenue per employee, often fail to capture the experiential, intangible, and co-created dimensions that define tourism services. Much of tourism’s value derives from emotional labor, interaction quality, and knowledge transfer, all of which are difficult to quantify using standard benchmarks [29]. Consequently, studies relying solely on aggregate productivity metrics may systematically underestimate sectoral contributions and mischaracterize the nature of BCD in tourism.

2.5.3. Sector-Specific Structural Issues

Tourism’s distinct labor market dynamics further complicate the BCD framework. In developing economies, informal employment and wage precarity are widespread, suppressing visible labor costs and masking deeper productivity challenges as noted by [13,25]. The prevalence of informal and seasonal work creates hidden structural vulnerabilities and limits the effectiveness of standard policy interventions.

Moreover, the rise of platform-based tourism (e.g., Airbnb, Uber, digital tour operators) introduces both efficiencies and new risks. While such platforms can reduce transaction costs and increase market access, they often do so at the expense of labor protections, fostering precarious employment and amplifying economic insecurity among workers [25,30]. This dual dynamic complicates the traditional assumptions of labor rigidity and wage equalization that underpin the original BCD hypothesis.

2.5.4. Methodological Divergence and Research Gaps

The literature also exhibits a methodological divide: quantitative studies tend to focus on aggregate trends using formal sector data, while qualitative and mixed-methods research captures the lived realities of informal work, innovation, and crisis response. This divergence highlights the need for integrative research that bridges macro-level analysis addressed by [5] with the sector-specific, context-rich insight described by [15].

By directly addressing these critiques and sector-specific complexities, this study aims to advance a more nuanced and empirically grounded understanding of BCD’s applicability to tourism. In particular, it investigates how BCD dynamics interact with external disruptions—such as the COVID-19 pandemic and the Russo–Ukrainian war—revealing both persistent vulnerabilities and emerging opportunities for sectoral adaptation and resilience.

2.6. The Pandemic, War, and Environmental Implications of BCD

The COVID-19 pandemic and the Russo–Ukrainian war have profoundly intensified the structural challenges associated with Baumol’s cost disease (BCD) in global tourism. As a labor-intensive sector dependent on discretionary spending, tourism has experienced acute cost pressures and stagnant productivity growth during these crises, which have further exposed its inherent vulnerabilities and amplified the urgency for adaptive responses.

2.6.1. Economic and Structural Impacts

The pandemic’s disruption of international mobility, coupled with the geopolitical instability caused by the Russo–Ukrainian war, has had cascading effects on tourism demand, labor markets, and business models. In countries neighboring the conflict, such as Poland [20] and Romania [31], accommodation markets have been restructured: many hotels transitioned from serving tourists to housing refugees, yielding short-term economic relief but diminishing the supply for leisure travel and undermining sectoral recovery. Safety concerns, uncertainty, and shifting travel advisories have led to pronounced declines in tourist arrivals, especially in regions proximate to conflict zones. For example, ref. [31] noted that Romania’s Danube Delta and Bucovina regions faced drops in arrivals of up to 40–50% compared to pre-pandemic levels.

These disruptions have reinforced the core mechanisms of BCD by increasing cost pressures on already vulnerable, labor-intensive tourism businesses, while at the same time limiting opportunities for productivity-enhancing investments. The compounded impact of pandemic and conflict has exposed the limitations of traditional crisis management strategies described by [32,33] and underscored the need for more flexible, resilient, and context-specific tourism policies.

2.6.2. Environmental Implications

The shift from goods-based to service-based economies has often been associated with reduced CO₂ emissions, as service industries generally exhibit lower resource intensity [34]. In theory, this offers a potential environmental benefit of BCD-driven economic transformation. However, ref. [11] pointed out the reality in tourism-dependent economies is more complex: rising service prices, resource-intensive crisis responses (such as large-scale hotel conversions), and fluctuating demand patterns may offset or even reverse environmental gains.

Localized consumption—such as the provision of services to internally displaced people and refugees—can alter the carbon footprint of tourism, sometimes reducing international travel emissions but increasing local consumption of energy and resources. The persistent economic strain of maintaining tourism infrastructure under crisis conditions raises difficult trade-offs between economic sustainability and environmental goals. There remains a significant research gap in understanding how post-crisis tourism recovery strategies can be aligned with broader sustainability and carbon reduction objectives, especially in regions facing ongoing instability and migration pressures as shown in studies of [12,24].

2.6.3. Broader Implications for Tourism Policy

The intersection of pandemic and war highlights the necessity for tourism policies that balance economic resilience, environmental stewardship, and social responsibility. Addressing the structural challenges intensified by BCD requires a multi-pronged approach:

• Leveraging technological advancements to improve productivity and service delivery, especially in areas where automation can complement human labor without eroding experiential quality [23].
• Providing targeted support for tourism-dependent economies, including financial relief, workforce retraining, and incentives for business model adaptation, particularly in the face of prolonged shocks [24].
• Promoting sustainable practices that minimize environmental impact—such as energy-efficient operations, carbon offset programs, and responsible destination management—while ensuring that economic recovery does not come at the expense of long-term sustainability [11].

Crucially, future research and policy must move beyond short-term crisis response to embrace systemic approaches that enhance adaptability and resilience in tourism. This includes integrating environmental and social metrics into recovery planning, fostering cross-sectoral collaboration, and reimagining tourism’s role in regional economic development under conditions of uncertainty and change addressed by [32,33].

2.7. Cross-Country Perspectives and Policy Implications

Comparative research reveals considerable heterogeneity in how Baumol’s cost disease (BCD) manifests across tourism sectors worldwide, shaped by factors including cultural proximity, technological adoption, labor market structures, and infrastructure development. For instance, the study of [15] underscores that cultural proximity can facilitate economic equilibration by attracting international visitors and stabilizing tourism demand, yet even these advantages may be insufficient to offset the underlying vulnerabilities associated with BCD in the long term.

In high-income OECD countries, greater access to digital infrastructure and higher rates of technology adoption have helped moderate labor cost growth and enhance productivity in tourism [24,26]. These economies are often better equipped to leverage innovations such as digital platforms, online marketing, and automation to address some of the structural pressures predicted by BCD. By contrast, developing economies face persistent barriers: informal labor markets, limited technological diffusion, and underinvestment in physical and digital infrastructure [13], that all exacerbate the cost, productivity, and resilience challenges associated with BCD. These disparities not only reflect differences in economic development but also highlight the importance of context-specific institutional and policy capacities.

Policy Implications

Given the persistent and uneven impact of BCD in tourism, policy intervention is critical—particularly where the sector’s public goods dimensions (e.g., cultural preservation, employment, regional development) justify government support [35]. Robust policy responses must go beyond short-term relief and address the deeper structural determinants of cost disease.

Key policy priorities include:

• Investing in Digital Infrastructure: Accelerating the adoption of digital technologies can drive productivity and efficiency in tourism, especially in developing economies and rural regions [24].
• Strengthening Labor Market Protections: Addressing wage precarity, informal employment, and weak labor rights is vital to reducing workforce vulnerabilities [25].
• Promoting Sustainability: Integrating environmental conservation goals with tourism development is essential for long-term sectoral viability [19].

The combined implementation of these strategies can help tourism-dependent economies not only address the structural challenges posed by BCD but also foster a more equitable, resilient, and sustainable trajectory for the sector’s future development. Importantly, the evidence underscores the need for adaptive policies that are responsive to local contexts and that integrate cross-sectoral collaboration, innovation, and capacity-building [24].

By drawing on cross-country lessons and tailoring interventions to local realities, policymakers can better navigate the dual imperatives of economic competitiveness and social and environmental responsibility in a post-pandemic, crisis-prone world.

3. Materials and Methods

3.1. Theoretical Framework of BCDH

Based on the insights provided by contemporary literature [1,27], we emphasize that the theoretical framework of this study aligns with ongoing debates about productivity in service economies, while also contributing unique insights into the application of Baumol’s cost disease hypothesis (BCDH) to tourism. The dominance of services in GDP and employment, as highlighted by [36], underscores the importance of examining the distinct challenges faced by labor-intensive sectors like tourism. Unlike manufacturing, where productivity improvements are more easily achieved through technological substitution [21], services—particularly those reliant on experiential value and human interaction—pose unique barriers to productivity growth [29]. These barriers are exacerbated by the constant quality assumption, where changes in input resources directly affect the perceived quality of outputs, making traditional productivity models unsuitable for services [5].

Furthermore, the heterogeneity within the service sector itself further complicates productivity dynamics, as demonstrated by [37]. Their work highlights that nontraditional services, which often involve higher levels of innovation and input–output interdependencies, exhibit greater productivity growth potential than traditional services. This distinction suggests that not all service subcategories contribute equally to economic performance, emphasizing the need for nuanced sectoral analysis such as [24].

At the same time, the rise of platform economies and advancements in artificial intelligence (AI) shown by [13] have introduced possibilities for technological substitution in some areas of service work. However, these advancements described by [8,10] remain limited in experiential sectors such as tourism, where personalized interactions and value creation are central to the customer experience. The critical role of disaggregation within service industries is further reinforced by [7], demonstrating that market and business services can fuel economic growth and productivity more effectively than other subcategories.

These findings highlight the importance of considering tourism not as a homogeneous sector but as one with significant internal diversity, requiring tailored approaches to understand its productivity trends [18]. These insights not only deepen our understanding of the structural challenges faced by service economies but also highlight the relevance of applying BCDH to labor-intensive sectors like tourism, where external shocks such as the COVID-19 pandemic [19] and the Russo–Ukrainian war [20] have amplified existing productivity constraints and cost dynamics.

BCDH built on intuition that goods-producing sectors tend to experience higher productivity growth over time, whereas service-producing sectors face slower productivity growth due to their labor-intensive nature. This disparity leads to rising relative prices for services, even when productivity in those sectors remains stagnant. As incomes rise, demand for services increases because services are income-elastic. Consequently, more labor shifts into the service sector despite its slower productivity growth. The result is an increase in the relative cost of services compared to goods over time.

Baumol’s theory emphasizes that productivity in services is constrained by the inherent need for human time and attention—factors that are difficult to replace or augment with technology. This constraint fundamentally differentiates services from goods in terms of productivity growth potential.

Thanks to [27] formalization of Baumol’s intuition—particularly the inflation differential in a two-sector economy—we can theoretically frame how external shocks (e.g., the combined impact of COVID-19 and the Russo–Ukrainian war as a unique “super shock”) amplify these dynamics in labor-intensive services such as tourism, where institutional rigidities further distort the relationship between productivity and prices.

To outline BCDH and its implications for inflation dynamics, labor costs, and productivity growth in a two-sector economy (goods and services), we draw on key insights derived from the Mathematical Framework of BCDH (see Appendix A.1 for details).

These insights are summarized as follows: Faster productivity growth in the goods sector ($g_G>g_S$) drives inflation dynamics. The inflation differential $\Delta \pi \left(t\right)=g_G-g_S$ explains why service sectors experience higher inflation. Rising wages and stagnant productivity in services lead to higher labor costs and prices.

While Baumol’s framework offers a robust explanation of sectoral cost dynamics, external shocks such as the COVID-19 pandemic or Russo–Ukrainian war can disrupt these theoretical predictions. The pandemic has introduced significant uncertainty by altering labor market structures, demand for services, and productivity patterns in ways that are difficult to predict. These disruptions make it harder to apply the theoretical framework to current and future economic conditions with certainty. However, we will assess these issues in the econometrics section of the paper. For now, we leave this as part of the theoretical discussion.

3.2. Empirical Model and Specification Rationality

This section outlines the empirical strategy used to test the Baumol’s cost disease (BCD) hypothesis and evaluate the impact of exogenous super-shocks—namely COVID-19 and the Russo–Ukrainian war—on sectoral outcomes. The analysis is divided into two parts. The first regression specification examines sectoral asymmetries in prices and labor costs using interaction terms, without employing a Difference-in-Differences (DiD) approach. The second specification explicitly uses a DiD framework to estimate the causal effects of the super-shocks.

Among the various econometric approaches suitable for this analysis—such as multilevel modeling, structural equation modeling, and other advanced techniques—we selected panel data modeling and a DiD decomposition to account for the three-dimensional structure of the data (country × sector × year) and the specific research objectives.

This choice is guided by the following considerations:

Suitability for Three-Dimensional Data: Panel data modeling is specifically designed to handle datasets with both cross-sectional and temporal dimensions. This makes it particularly well-suited to our three-dimensional structure, where observations are organized by country, sector, and year. Unlike structural equation modeling or multilevel modeling, panel data approaches allow us to explicitly model sectoral wage responses as endogenous outcomes over time while accounting for sectoral and country-specific heterogeneity.

Capturing Structural and Event-Driven Dynamics: The combination of panel data modeling and DiD decomposition enables a dual analysis of long-term structural divergence (via interaction terms) and short-term, causal impacts of major shocks (via DiD). This dual approach is more robust and targeted compared to purely cross-sectional or time-series methodologies, which cannot simultaneously capture both structural and event-driven factors.

Advantages Over Alternatives: While alternative methods, such as structural equation modeling or multilevel modeling, are effective for analyzing latent variables and hierarchical structures, they are less suited to addressing common challenges in three-dimensional panel data. These challenges include heteroskedasticity, autocorrelation, and cross-sectional dependence. Panel-based methods, particularly those combined with DiD, are designed to address these issues, ensuring more accurate and reliable estimates.

By leveraging these advantages, our approach ensures that the empirical strategy aligns with the data’s structure and the study’s objectives. This framework provides a robust foundation for analyzing both structural asymmetries in sectoral outcomes and the causal effects of exogenous super-shocks, such as COVID-19 and the Russo–Ukrainian war.

Our three-dimensional panel structure builds on Baumol’s framework by treating sectoral wage responses as endogenous outcomes shaped by:

(a)

structural asymmetries, modeled through productivity interaction terms, and

(b)

exogenous shocks, identified via a DiD estimation of the impacts of COVID-19 and the Russo–Ukrainian war.

This dual approach goes beyond aggregate cost-share decompositions by simultaneously capturing long-term sectoral divergence and short-term event-driven effects. Together, these two distinct yet complementary specifications provide a comprehensive analysis of both structural and event-driven factors influencing sectoral outcomes.

3.3. Panel Sectoral Asymmetries (PSA) as a Baseline Regression

The first specification focuses on testing Baumol’s hypothesis by analyzing sectoral asymmetries in the relationship between productivity and economic outcomes, such as prices and labor costs. Sectoral heterogeneity is introduced through a binary classification of sectors (e.g., tourism versus manufacturing), while the time dimension is held constant across sectors. Unlike the DiD approach, this specification leverages interaction terms between productivity and a sectoral dummy variable to identify differential productivity effects across sectors. The econometric model for the outcome variable—either the log of implied prices or the log of labor costs—is presented in Equation (1).

$$\begin{gathered} \log \left(Y_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)={\beta }_0+{\beta }_1\log \left({\mathit{Productivity}}_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)+{\beta }_2\mathrm{NACE\_R2}{\mathrm{I}}_{\mathrm{s}}+{\beta }_3\mathit{treatment\_pos}{t}_{\mathrm{ist}} \\ +{\beta }_4\left(\log \left({\mathit{Productivity}}_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)\times \mathrm{NACE\_R2}{\mathrm{I}}_{\mathrm{s}}\right)+{ϵ}_{\mathrm{i}\mathrm{s}\mathrm{t}} \end{gathered}$$

(1)

In this model, $Y_{\mathrm{i}\mathrm{s}\mathrm{t} }$ represents the dependent variable for country i in period t across sectors (tourism vs. manufacturing). The term ${\mathit{Productivity}}_{\mathrm{i}\mathrm{s}\mathrm{t}}$ is the natural logarithm of productivity, measured as output per worker or a similar metric. ${\mathrm{NACE\_R2}\mathrm{I}}_{\mathrm{s}}$ is a dummy variable equal to 1 for the tourism sector and 0 for manufacturing. Similarly, $\mathit{treatment\_pos}{t}_{\mathrm{i}\mathrm{s}\mathrm{t}}$ is a dummy variable that equals 1 for the post-COVID years (2020–2023) and 0 for the pre-treatment baseline period (2011–2019). Finally, the interaction term $\left({\mathit{Productivity}}_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)\times {\mathrm{NACE\_R2}\mathrm{I}}_{\mathrm{s}}$ captures sectoral asymmetry, highlighting how the productivity effects in tourism differ from those in manufacturing.

We expect ${\beta }_1$ > 0 because BCD posits that higher productivity growth in manufacturing (the “progressive” sector) should dampen output price inflation and labor cost growth. However, a positive ${\beta }_1$ > 0 for tourism would imply that productivity gains are associated with higher output prices or rising labor costs, potentially reflecting market power, wage bargaining, or input cost rigidities. The coefficient ${\beta }_2$ is also expected to be positive, as tourism, which inherently faces structural rigidities (e.g., labor-intensive operations), should have baseline output prices and labor costs higher than those in manufacturing. The coefficient ${\beta }_3$, associated with the treatment_post variable, is hypothesized to be positive ${\beta }_3$ > 0. This reflects the expectation that post-super-shock disruptions, such as supply-chain shocks and labor shortages, have increased output prices and labor costs across both sectors. Finally, the interaction term coefficient ${\beta }_4$ is expected to be negative (${\beta }_4$ < 0). This is a critical aspect of BCD, as it implies that productivity improvements in tourism, a “stagnant” sector, are less effective at curbing price or cost growth compared to manufacturing. A negative ${\beta }_4$ would indicate that productivity gains in tourism have weaker mitigation effects on costs and prices than in the progressive manufacturing sector. This interaction term is critical for identifying the sector-specific effects of productivity on prices or costs, particularly in tourism versus manufacturing.

3.4. Panel Difference-in-Differences (PDiD)

Next, we estimate a PDiD fixed-effects model as a second specification, building on foundational DiD methodologies developed by [38,39,40]. This approach is chosen because year-specific interactions introduce unobserved heterogeneity that varies across time and sectors, which fixed-effects models are better suited to control for.

$$\begin{array}{cc}\log \left(Y_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)={\beta }_0\hfill & +{\beta }_1\log \left({\mathit{Productivity}}_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)+{\beta }_2\mathrm{NACE\_R2}{\mathrm{I}}_{\mathrm{s}}\hfill \\ & +{\beta }_3\mathit{treatment\_pos}{t}_{\mathrm{t}}\hfill \\ & +{\beta }_4\left(\mathrm{NACE\_R2}{\mathrm{I}}_{\mathrm{s}}\times \mathit{treatment\_pos}{t}_{\mathrm{t}}\right)\hfill \\ & +{\beta }_5\left(\log \left({\mathit{Productivity}}_{\mathrm{i}\mathrm{s}\mathrm{t}}\right)\times \mathrm{NACE\_R2}{\mathrm{I}}_{\mathrm{s}}\right)\hfill \\ & +{\displaystyle \sum _{\mathrm{k}=2011}^{2023}}{\gamma }_{\mathrm{k}}\left(\mathrm{NACE\_R2}{\mathrm{I}}_{\mathrm{s}}\times I{\mathit{Year}}_{\mathrm{t}=\mathrm{k}}\right)+{\alpha }_{\mathrm{i}}+{ϵ}_{\mathrm{i}\mathrm{s}\mathrm{t}}\hfill \end{array}$$

(2)

Random-effects models, which assume that unobserved heterogeneity is uncorrelated with the regressors, are less appropriate in this context, particularly when capturing sector-specific trends over time.

As all other terms are the same as in the previous specification (1), we focus here on explaining the nature of the new interaction term $\left({\mathrm{NACE\_R2}\mathrm{I}}_{\mathrm{s}}\times {\mathit{treatment\_post}}_{\mathrm{t}}\right).$ This term captures the differential impact of the super-shock treatment period (2020–2023) on tourism relative to manufacturing. A significant coeficient ${\beta }_4$ indicates how the tourism sector, compared to manufacturing, was disproportionately affected by the 2020–2023 shocks (e.g., the pandemic and the Russo–Ukrainian war) in terms of the dependent variable (e.g., prices or labor costs).

Additionally, the term ${\gamma }_{\mathrm{k}}$ introduces year-specific sectoral fixed effects, allowing the model to account for unobserved heterogeneity in the tourism and manufacturing sectors during each year. Here, $\mathbb{I}\left({\mathrm{Year}}_{\mathrm{t}}=\mathrm{k}\right)$ is an indicator function that equals 1 if the year is k, and 0 otherwise. This interaction term ensures that the model captures sector-specific trends over time, isolating the unique dynamics in tourism and manufacturing during each year of the panel.

The coefficients ${\gamma }_{\mathrm{k}}$ provide a detailed year-by-year measure of the difference in the outcome variable (e.g., prices or labor costs) between tourism and manufacturing, relative to the baseline year. These coefficients are particularly important for identifying how sectoral disparities evolved over the entire study period, including during key treatment years such as 2020 and 2023, which correspond to the super-shock. A significant ${\gamma }_{\mathrm{k}}$ in these years would indicate that the tourism sector responded differently to external shocks compared to manufacturing, reflecting structural vulnerabilities or other sectoral asymmetries.

This term enriches the analysis by allowing for a flexible, time-varying assessment of sectoral differences in economic outcomes, ensuring that the model does not impose uniform effects across all years.

The second specification assumes parallel trends, meaning that productivity in tourism and manufacturing would have evolved similarly in the absence of the pandemic and the war-induced shocks. To validate this assumption, we perform robustness checks, including DiD models that estimate both pre-treatment and post-treatment trends. While placebo tests using pre-pandemic data were not explicitly validated beyond confirming parallel trends, the dynamic models ensure that the observed differences are attributable to the treatment period rather than pre-existing trends.

The second regression specification incorporates fixed effects at the country-sector-year level to control for unobserved heterogeneity. Country fixed effects account for time-invariant characteristics such as institutional quality or geographic factors, while year fixed effects capture global shocks or trends, including inflation or pandemic-related policy responses. Standard errors are clustered at the country-sector level to address within-panel correlation, ensuring reliable and robust statistical inference.

3.5. Testing Robustness of the Baseline Method

While the first regression specification focuses on sectoral asymmetry and long-term structural dynamics, the second specification isolates the causal effect of super-shock as an exogenous shock. Both regression analyses are based on a three-dimensional panel dataset comprising country, sector, and year observations, allowing for the inclusion of fixed effects to account for unobserved heterogeneity. This methodology will undergo a series of diagnostic tests to address potential econometric challenges, ensuring that the first specification is robust and well-suited to the complexities of the data.

The focus will be on refining the model to account for issues such as heteroskedasticity, cross-sectional dependence, and serial correlation. If serial correlation issues arise, we may need to reconsider our initial assumptions that random-effects (RE) or fixed-effects (FE) models suffice, as highlighted by [41,42,43] ensuring that our empirical analysis remains reliable and valid.

This multi-step approach addresses potential issues such as heteroskedasticity, cross-sectional dependence, and serial correlation. If serial correlation proves problematic, we will follow [44,45,46] by adjusting standard errors or applying block bootstrap methods. Together, these techniques ensure that the baseline method’s findings are rigorously validated against econometric challenges, enhancing the reliability of the analysis and mitigating risks of bias from statistical artifacts.

3.6. Data and Source

The dataset comprises data from 15 EU countries: Austria, Belgium, Czechia, Estonia, Greece, Finland, France, Italy, Lithuania, Luxembourg, Latvia, the Netherlands, Sweden, Slovenia, and Slovakia, covering the period from 2011 to 2023. The selection of these countries was constrained by data availability, as significant gaps in other EU countries rendered their inclusion unfeasible. For the selected countries, smaller data gaps were addressed using imputations similar to [47], performed with the mice package ensuring a complete and reliable dataset for analysis.

Table 1. Summary of data transformations.

Variables	Data	Transformation
Implicit Price	TFP2_VA	Logged
Labor Cost	Shares_Labour Income, TFP2_VA	Logged
Labor Cost per Hour	Computed from previous formula, TFP2_LPH	Logged
Labor Productivity	TFP2_VA, TFP2_LPH	Logged
Sector Indicator	NACE_R2I	Dummy (Tourism = 1)
Post-COVID Indicator	treatment_post	Dummy (Post-COVID = 1)

Source: Authors’ calculation. Note: The TFP0_VA dataset is a comprehensive panel dataset containing observations across various variables, providing information on value-added contributions, growth, and labor costs across EU countries, industries, and years. Key variables include Country (geopolitical entity), NACE_R2 (industry classification), year (time dimension), and economic metrics such as productivity and labor costs. The TFP2_LPH dataset complements this by focusing on labor-related metrics, including labor productivity and labor costs per hour, derived using variables like value-added and total hours worked, offering deeper insights into labor efficiency and cost dynamics across the EU.

presents a summary of the data transformations applied to the variables used in our analysis. The data used in this study underwent transformations to compute key indicators such as Implicit Price, Labor Costs per Hour, and Labor Productivity, specifically tailored for the manufacturing sector (C = Total Manufacturing) and the tourism sector (I = Accommodation and Food Service Activities). These transformations relied on growth accounting principles and sector-specific data from authoritative sources, including the wiiw-GDP Release 2024, produced by the Vienna Institute for International Economic Studies [48]. This dataset provides valuable insights into economic performance, offering data on value-added, growth accounting components, and capital deepening measures, which are critical for analyzing sectoral trends and labor income shares. To enhance interpretability, logarithmic transformations were applied to continuous variables, emphasizing percentage changes and growth rates. Dummy variables were introduced to capture sector and time-specific effects, such as the tourism sector (NACE_R2I) and the combined post-COVID and Russo–Ukrainian war period (treatment_post). Growth accounting and capital deepening serve as foundational inputs for deriving these key indicators, ensuring sector-specific nuances are adequately addressed.

These transformations enable a detailed decomposition of sectoral performance, facilitating comparisons between manufacturing and tourism, where tourism is proxied by the values inherited from Accommodation and Food Service Activities. For example, labor productivity is computed as the ratio of real value-added to total hours worked, providing insights into sectoral efficiency as reflected in wages. Similarly, the implicit price index (a proxy for sectoral inflation) and labor cost per hour offer critical perspectives on pricing and cost dynamics. The use of logged continuous variables highlights relative changes, while dummy variables capture key treatment effects, enabling robust and reliable analysis in the context of BCD. This approach assumes that tourism, proxied by Accommodation and Food Service Activities, offers a suitable framework for analyzing sectoral dynamics in the post-super-shock era.

Table 2. Descriptive statistics.

Sector	Manufacturing (“C”)	Tourism (“I”)
Mean_Productivity	2.24	1.16
Median_Productivity	2.02	1.05
Std_Productivity	2.38	1.93
Min_Productivity	−2.94	−3.99
Max_Productivity	9.23	6.32
Mean_Implied_Price	−2.66	−2.51
Median_Implied_Price	−2.38	−2.58
Std_Implied_Price	0.92	1.04
Min_Implied_Price	−7.62	−6.95
Max_Implied_Price	−0.96	0.13
Mean_Labour_Cost_per_Hour	6.91	3.32
Median_Labour_Cost_per_Hour	1.83	1.42
Std_Labor_Cost_per_Hour	43.55	23.15
Min_Labor_Cost_per_Hour	−88.45	−56.80
Max_Labor_Cost_per_Hour	390.93	307.57

Source: Authors’ calculation. Note: The significant difference between the mean and median of “Labor Cost per Hour” in the tourism sector reflects the skewed distribution of wages across countries. High labor costs in wealthier economies (e.g., Luxembourg, Austria) inflate the mean, while the median provides a more representative measure of central tendency. This variability is intrinsic to cross-country analyses and does not affect the robustness of the econometric results.

provides a summary of key descriptive statistics, including productivity, implied price, and labor costs per hour, for the manufacturing and tourism sectors.

We assess multicollinearity through two diagnostic measures: (1) pairwise correlation coefficients (reported in Appendix A.2,

Table A1. Correlation matrix of main study variables.

	Productivity	Implied _Price	Labor _Cost	Labor_Cost _per_Hour	NACE_R2I	Treatment _Post
Productivity	1.00	−0.01	0.10	0.17	−0.24	−0.21
Implied_Price	−0.01	1.00	0.15	0.01	0.08	0.36
Labour_Cost	0.10	0.15	1.00	0.16	0.01	0.06
Labour_Cost_ per_Hour	0.17	0.01	0.16	1.00	−0.09	−0.07
NACE_R2I	−0.24	0.08	0.01	−0.09	1.00	0.43
treatment_post	−0.21	0.36	0.06	−0.07	0.43	1.00

Note: All entries are Pearson correlation coefficients. Except for the last two variables (NACE_R2I and treatment_post), all variables are log-transformed.

) and (2) variance inflation factors (VIFs) (

Table A2. Multicollinearity assessment: VIFs for model variables.

Variable	VIF
Productivity	1.76
NACE_R2I	1.95
treatment_post	1.27
Productivity × NACE_R2I	2.00
Productivity	1.76
NACE_R2I	1.95

Note: Results are identical for both outcome variables. All VIF values are well below the standard cut-off of 5, indicating no problematic multicollinearity.

). All VIF values remain below 3.0, well under the conventional threshold of 10 described by [49], indicating acceptable levels of multicollinearity in our specifications.

4. Results

4.1. Evidence of BCDH’s Analysis

The results analysis is structured in two parts. First, it involves the selection and application of a suitable panel data DiD technique to evaluate the primary outcomes. Second, it entails the reassessment of these primary results using the baseline regression method selected through rigorous econometric testing. This two-step approach ensures the robustness and reliability of the findings, addressing both methodological and data-specific challenges.

4.1.1. Subsubsection Econometric Technique Selection: Justification for Panel-Corrected Standard Errors (PCSE)

The choice of PCSE as the baseline method was guided by extensive diagnostic testing to address key econometric challenges in the panel dataset.

4.1.2. Model Diagnostics

The diagnostics (presented in

Table 3. Model diagnostics supporting the use of Panel-Corrected Standard Errors (PCSE).

Test	Statistic	Degrees of Freedom (df)	p-Value	Conclusion
Heteroskedasticity	Chi-sq = 10.336	df = 4	p = 0.035	Heteroskedasticity detected
Cross-Sectional Dependence	z = 4.91	–	p = 0.000	Significant cross-sectional dependence
Serial Correlation	Chi-sq = 58.308	df = 26	p = 0.000	Serial correlation detected
F-Test for Individual Effects	F = 2.287	df1 = 14, df2 = 371	p = 0.005	Significant individual effects id

Source: Authors’ calculation.

) were conducted during the initial evaluation phase—primarily using fixed-effects (FE) and/or random-effects (RE) models—to address econometric challenges, ultimately leading to the selection of PCSE as the more appropriate method.

The evidence of heteroskedasticity suggested non-constant error variances across observations, risking inefficient estimates in traditional models. Additionally, significant cross-sectional dependence was identified, likely due to economic interdependence among countries, which standard fixed-effects (FE) and random-effects (RE) models cannot manage effectively. Serial correlation within panel units further complicated the analysis, necessitating a method that accounts for autocorrelated disturbances.

While FE and RE models can handle individual effects, they fall short in simultaneously addressing heteroskedasticity, serial correlation, and cross-sectional dependence. Panel-Corrected Standard Errors (PCSE) stands out in this context by correcting for non-constant variances, modeling contemporaneous error covariance, and accommodating autocorrelation. PCSE estimation follows procedure used by [50], implemented via the pcse package in R described by [51] with corrections for heteroskedasticity and AR(1) autocorrelation. For spatial dependence, we apply fixed-effects covariance estimator similar to [52] and the plm package used by [53] with kernel-weighted spatial lags.

Therefore, based on these diagnostic findings, the PCSE results presented in Table 3 were determined to be the most robust and appropriate econometric technique for this analysis.

4.1.3. Interpretation of PCSE Results: Log Implied Price as the Dependent Variable

In Model 1, the dependent variable is log_Implied_Price, and the results reveal several patterns consistent with the dynamics predicted by BCDH, as shown in

Table 4. Panel-Corrected Standard Errors (PCSE) results.

Variable/Dependent Variable	Log Implied Price	Log Labor Cost per Hour
	Model 1	Model 2
(Intercept)	−3.022 *** (−27.07)	−2.771 *** (−3.75)
log_Productivity	0.073 ** (2.56)	0.504 *** (3.47)
NACE_R2I	0.387 ** (2.34)	0.323 (0.64)
treatment_post	0.634 *** (3.60)	−0.048 (−0.04)
log_Productivity:NACE_R2I	−0.137 ** (−2.91)	−0.494 ** (−2.00)

Source: Authors’ calculation. Note: t-values are in parentheses. Significance levels: *** p < 0.01, ** p < 0.05.

. The coefficient for log_Productivity is positive and statistically significant (0.073, p = 0.011), indicating that higher productivity levels are associated with higher implied prices. This finding aligns with Baumol’s theory, which suggests that as productivity increases, wages also rise, driving up prices even in sectors where productivity gains are slower. However, it is important to note that the real impact of productivity on labor costs (or wages) is assumed but not explicitly extracted in this model, as labor cost is not directly included as a dependent variable in Model 1. The observed relationship between productivity and implied prices reflects the cost-pass-through mechanism, where higher wages—driven by productivity gains—translate into higher prices, but the direct link between productivity and wages remains implicit in the interpretation.

The variable NACE_R2I, identifying the tourism sector (accommodation and food services) relative to manufacturing, has a positive and significant coefficient (0.387, p = 0.020). This result suggests that implied prices in the tourism sector are generally higher than in manufacturing. This can be attributed to structural differences between the two sectors: tourism is notably labor-intensive and experiences slower productivity growth, making it more vulnerable to rising labor costs under wage equalization. While the model does not directly capture labor costs, the higher implied prices observed in tourism suggest that rising wages in this sector—driven by spillover effects from higher-productivity industries—contribute to the cost disease phenomenon.

The interaction term log_Productivity:NACE_R2I has a significant negative coefficient (−0.137, p = 0.003), indicating that the relationship between productivity and implied prices is weaker in the tourism sector compared to manufacturing. This finding underscores the challenges faced by low-productivity sectors like tourism, where modest productivity improvements do little to offset price pressures driven by rising wages. Again, while the model does not explicitly measure wages, the weaker link between productivity and prices in tourism highlights how labor-intensive sectors struggle to manage wage-driven cost pressures, a central feature of BCDH.

The coefficient for treatment_post, which captures the effect of the super-shock period (2020–2023), is positive and highly significant (0.634, p < 0.001). This result indicates that the pandemic significantly increased implied prices across both sectors. This surge in prices can be attributed to pandemic-induced disruptions, such as supply chain breakdowns, increased operational costs (e.g., health protocols and labor shortages), and shifts in consumer demand. The findings suggest that external shocks, such as COVID-19 and the European war, exacerbate cost disease dynamics by adding upward pressure on prices, particularly in sectors already vulnerable to wage-cost inflation.

In summary, the results provide strong evidence that implied prices increase with productivity, but the strength of this relationship varies significantly by sector. While productivity gains in manufacturing are more closely tied to price stability, slower productivity growth coupled with rising wages in the tourism sector leads to significant price increases. This finding aligns with Baumol’s cost disease hypothesis (BCDH), which predicts that labor-intensive sectors with slower productivity growth are more vulnerable to rising wages and price pressures. However, it is critical to emphasize that, although the model assumes a link between productivity and wages in driving these price changes, it does not directly measure labor costs or wages. Instead, the observed dynamics are inferred based on the theoretical framework of BCDH, where rising wages—assumed to follow productivity growth—are implicitly passed through to prices.

Nonetheless, the interpretation of these results should be nuanced, as the indirect calculation of productivity and implied price indicators may introduce limitations to the analysis. For example, the heterogeneity of the manufacturing sector across countries—encompassing a wide range of industries with differing levels of productivity and labor intensity—makes cross-country comparisons of average productivity more challenging, as aggregated measures may obscure important sectoral differences. In contrast, the tourism sector is relatively more homogeneous in its activities but remains structurally distinct from manufacturing in terms of labor intensity, wage dynamics, and cost structures. These structural differences mean that the application of BCDH, while relevant, may manifest differently across sectors and countries.

Finally, the pandemic and war amplified the cost disease dynamics by adding upward pressure on prices across both sectors, but their impact was likely uneven. For labor-intensive industries like tourism, disruptions such as supply chain breakdowns, increased operational costs, and labor shortages compounded the sector’s inherent vulnerability to wage-driven cost pressures. This underscores the importance of accounting for both sectoral and country-level heterogeneities when interpreting the results and applying the BCDH framework to explain price dynamics. Further research is necessary to explore these dynamics in greater depth, particularly with respect to the structural and country-specific factors that shape wage and price relationships in different sectors.

4.1.4. Interpretation of PSCE Results: Log Labor Cost per Hour as the Dependent Variable

In Model 2, where the dependent variable is log_Labor_Cost_per_Hour, the results provide valuable insights into the relationship between productivity, sectoral characteristics, and labor costs, as shown in Table 4. The analysis reveals a clear positive relationship between productivity and hourly labor costs, as indicated by the statistically significant coefficient for log_Productivity (0.504, p < 0.001). This suggests that higher productivity is strongly associated with higher wages per hour across sectors. This finding aligns with BCDH, where productivity growth drives wage increases. It also highlights wage equalization mechanisms, implying that productivity gains in one sector can influence wage levels across sectors, even those with slower productivity growth.

The coefficient for NACE_R2I, which distinguishes the tourism sector (accommodation and food services) from manufacturing, is positive (0.323) but statistically insignificant (p = 0.64). This result suggests that baseline log hourly labor costs in the tourism sector are not systematically different from those in manufacturing, on average, within the dataset. However, this should not be interpreted as evidence of structural alignment between the two sectors in labor terms. Labor costs are influenced by a wide range of factors, including fiscal treatment, remuneration policies, and technological intensity, which differ significantly between the sectors and across countries. The statistical insignificance of the NACE_R2I coefficient simply indicates that the model does not find robust evidence of a systematic difference in baseline log hourly labor costs between the two sectors, but it does not imply that the two sectors are homogeneous in terms of labor structure or cost components. Further research is required to explore these structural differences in greater detail.

However, the interaction term log_Productivity:NACE_R2I, which captures how the relationship between productivity and wages differs between sectors, is negative and statistically significant (−0.494, p = 0.046). This result suggests that the positive impact of productivity on hourly labor costs is smaller in the tourism sector than in manufacturing. In other words, while productivity gains in manufacturing are strongly associated with wage increases, the same gains in tourism have a weaker effect. This finding reflects the structural differences between the two sectors: tourism, being more labor-intensive and often characterized by lower-skill activities, struggles to translate productivity improvements into proportional wage increases. This limitation is a critical feature of Baumol’s hypothesis, where labor-intensive sectors face inherent constraints in achieving wage growth comparable to that of high-productivity sectors.

The coefficient for treatment_post, representing the super-shock period (2020–2023), is negative (−0.048) and statistically insignificant, indicating no meaningful change in hourly labor costs during this time. This result suggests that the pandemic did not directly affect wage dynamics in a statistically significant way. The stability of labor costs during the treatment period likely reflects two key factors: wage rigidity and policy safeguards. Wage rigidity, often observed in labor-intensive sectors like tourism, stems from institu-tional factors such as union agreements, minimum wage laws, and contractual obliga-tions that prevent rapid wage adjustments. Additionally, policy safeguards implemented during the pandemic—such as wage subsidies, furlough schemes, and employment pro-tection policies—helped stabilize wages despite the severe disruptions faced by the tour-ism sector.

Interestingly, this lack of a significant treatment effect contrasts with the findings for implied prices, which were significantly affected during the pandemic. This distinction underscores the decoupling of price changes from wage dynamics, reflecting the structur-al constraints of labor-intensive sectors in responding to external shocks. These insights align with Baumol’s hypothesis, further highlighting the challenges faced by tourism in managing wage pressures amidst slower productivity growth.

Overall, Model 2 demonstrates that productivity is a key driver of hourly labor costs, consistent with BCDH. However, the weaker relationship between productivity and wages in the tourism sector highlights the structural challenges faced by labor-intensive industries. These sectors are less able to translate productivity gains into wage increases, leaving them more vulnerable to wage pressures driven by spillover effects from high-productivity industries. While the super-shock period did not significantly alter hourly labor costs, the results emphasize the importance of sectoral dynamics and structural constraints in shaping wage outcomes, particularly in the context of productivity growth. These findings reinforce the broader narrative of Baumol’s hypothesis, with tourism exemplifying the challenges faced by labor-intensive sectors in managing wage growth amidst slower productivity gains.

4.2. Robustness of PSCE Results

In this section, we evaluate the robustness of the results presented in earlier models, particularly focusing on the dynamics predicted by BCDH. To ensure the consistency and reliability of our findings, we compare results across multiple estimation techniques, including OLS with robust and clustered standard errors, PCSE, and linear models estimated using FE and with random slopes. These results are presented in

Table 5. Linear models with robust and clustered standard errors.

Variable/Dependent Variable	Log Implied Price	Log Labor Cost per Hour
	Model 1	Model 2
log_Productivity	0.073 (*** 2.82/** 2.19)	0.504 (*** 3.03/*** 3.24)
NACE_R2I	0.387 (*** 2.81/** 2.18)	0.323 (0.62/0.68)
treatment_post	0.634 (*** 6.13 /*** 4.01)	−0.048 (−0.10/−0.11)
log_Productivity:NACE_R2I	−0.137 (*** −3.44/** −2.55)	−0.494 (** −2.15/* −1.67)

Source: Authors’ calculation. Notes: t-Values: The first t-value corresponds to heteroskedasticity-robust standard errors; the second t-value corresponds to clustered standard errors. Interpretation: The inclusion of both t-values allows for comparison of the robustness of results across different error structures. Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

,

Table 6. Robustness check with clustered standard errors using FE-OLS.

Variable/Dependent Variable	Log Implied Price	Log Labor Cost per Hour
	Model 1	Model 2
log_Productivity	0.068 (2.525 *)	0.406 (2.428 *)
NACE_R2I	0.389 (2.829 **)	0.088 (0.178)
treatment_post	0.628 (6.039 ***)	−0.060 (−0.130)
log_Productivity:NACE_R2I	−0.144 (−3.328 ***)	−0.382 (−2.402 *)

Source: Authors’ calculation. Notes: Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

,

Table 7. Robustness check with Linear Mixed-Effects Model with Random Slopes.

Variable/Dependent Variable	Log Implied Price	Log Labor Cost per Hour
	Model 1	Model 2
(Intercept)	−3.013 (−26.117 ***)	−2.588 (−4.368 ***)
log_Productivity	0.069 (2.468 *)	0.431 (3.333 **)
NACE_R2I	0.384 (3.284 **)	0.135 (0.249)
treatment_post	0.635 (6.358 ***)	−0.060 (−0.130)
log_Productivity:NACE_R2I	−0.138 (−3.139 **)	−0.413 (−2.030 *)

Source: Authors’ calculation. Note: Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

and

Table 8. Panel Difference-in-Differences (Log Implied Price as Dependent Variable).

Variable	Estimate (Std. Error)
log_Productivity	0.060 ** (0.030)
factor(year)2021	0.803 ** (0.336)
treatment_post:factor(year)2020	−0.630 * (0.331)
treatment_post:factor(year)2022	1.020 *** (0.357)
log_Productivity:dummy_service	−0.075 * (0.041)

Source: Authors’ calculation. Note: The DiD effect for 2023 could not be estimated due to collinearity, as all countries had fully transitioned into or out of the treatment state, leaving no cross-sectional variation. Note: Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

.

Each approach offers unique insights into the relationships between productivity, sectoral characteristics, wages, and prices, while addressing potential econometric concerns. Below, we confirm and enhance the earlier baseline evidence from PCSE results.

4.2.1. Log Implied Price as the Dependent Variable

The robustness of the relationship between log_Productivity and log Implied Price is evident across all models. In the baseline PCSE results, the coefficient for log_Productivity is positive and significant (0.073, p < 0.05), indicating that higher productivity levels are associated with higher implied prices. This finding remains consistent when examining results from OLS with robust and clustered standard errors (Table 4: 0.073, significant under heteroskedasticity-robust standard errors; significant but weaker under clustered standard errors) and when using FE-OLS (Table 6: 0.068, marginally significant). Additionally, the mixed-effects model (lmer) confirms this relationship (Table 7: 0.069, p < 0.1). These consistent results across estimation techniques reinforce the conclusion that productivity gains are associated with higher implied prices, in line with Baumol’s hypothesis.

Similarly, the baseline PCSE results show that NACE_R2I, identifying the tourism sector, has a positive and significant coefficient (0.387, p < 0.05), indicating that implied prices in tourism are generally higher than in manufacturing. This pattern is confirmed across other models, particularly in the FE-OLS estimation (Table 6: 0.389, p < 0.1) and the mixed-effects model (Table 7: 0.384, p < 0.05). These findings underscore the structural differences between the tourism and manufacturing sectors, with tourism being more vulnerable to wage-driven cost pressures due to its labor-intensive nature.

The interaction term log_Productivity:NACE_R2I, which captures how the relationship between productivity and implied prices varies by sector, is consistently negative and significant across all models. In the PCSE results, the coefficient is −0.137 (p < 0.01), indicating that the relationship between productivity and prices is weaker in the tourism sector. This finding is corroborated in OLS (Table 5: −0.137, significant under robust and clustered standard errors), FE-OLS (Table 6: −0.144, p < 0.1), and the mixed-effects model (Table 7: −0.138, p < 0.05). These consistent results highlight the structural challenges faced by labor-intensive sectors like tourism in managing wage-driven cost pressures.

Verdoorn’s Law posits that productivity growth is positively correlated with output growth, suggesting that higher demand fuels productivity improvements through econo-mies of scale and technological advancements [54]. This principle has traditionally been applied to material production sectors, where increased production levels lead to a virtu-ous cycle of higher efficiency and cost reductions.

However, our findings directly contradict this prediction in the case of tourism. The negative coefficient (−0.137***) indicates weaker price adjustments in response to produc-tivity gains and provides no evidence of demand-driven productivity growth. This rigidity in the tourism sector appears to stem from institutional factors—such as union contracts and skill gaps—rather than the virtuous cycles commonly associated with Verdoorn’s Law in manufacturing sectors.

Furthermore, while Verdoorn’s Law theoretically links prices to productivity, our sector-interaction terms test this directly. The negative coefficient for tourism (−0.137***) rejects such feedback loops in stagnant sectors, underscoring tourism’s structural con-straints and limited capacity for productivity-driven growth.

Finally, the treatment variable treatment_post, capturing the effect of the super-shock period, is positive and highly significant in all models. In the PCSE results, the coefficient is 0.634 (p < 0.01), indicating a substantial pandemic-induced increase in implied prices. This finding is confirmed in OLS (Table 5: 0.634, significant under both robust and clustered standard errors), FE-OLS (Table 6: 0.628, significant), and the mixed-effects model (Table 7: 0.635, p < 0.01). The consistent significance of this variable across estimation techniques emphasizes the pandemic’s role as an external shock that exacerbated cost disease dynamics.

4.2.2. Log Labor Cost per Hour as the Dependent Variable

When examining log Labor Cost per Hour as the dependent variable, the robustness of the relationship between log_Productivity and hourly labor costs is similarly confirmed. In the PCSE results, the coefficient for log_Productivity is 0.504 (p < 0.01), indicating that higher productivity is strongly associated with higher wages per hour. This finding holds across OLS (Table 5: 0.504, significant under both robust and clustered standard errors), FE-OLS (Table 6: 0.406, p < 0.1), and the mixed-effects model (Table 7: 0.431, p < 0.01). These results consistently align with Baumol’s hypothesis, where productivity growth drives wage increases.

The coefficient for NACE_R2I is positive but statistically insignificant in the PCSE results (0.323, p > 0.1), suggesting no meaningful difference in baseline labor costs between the tourism and manufacturing sectors. This insignificance is also observed in OLS (Table 5: 0.323, insignificant) and FE-OLS (Table 6: 0.088, insignificant). However, the interaction term log_Productivity:NACE_R2I is negative and significant in the PCSE results (−0.494, p < 0.05), indicating that the impact of productivity on wages is weaker in the tourism sector. This finding is similarly observed in OLS (Table 5: −0.494, marginally significant) and the mixed-effects model (Table 7: −0.413, p < 0.05), reinforcing the structural challenges faced by labor-intensive sectors in achieving proportional wage growth from productivity gains.

Interestingly, the treatment variable treatment_post is negative and insignificant in the PCSE results (−0.048, p > 0.1), suggesting no meaningful change in hourly labor costs during the shock period. This finding is consistent across other models, including OLS (Table 5: −0.048, insignificant) and FE-OLS (Table 6: −0.060, insignificant). The lack of a significant treatment effect contrasts with the results for implied prices, indicating that the pandemic’s impact on wages was likely mitigated by labor protections or wage stability policies.

4.2.3. Enhancing Earlier Results

The robustness checks provide strong support for the baseline PCSE evidence while enhancing the interpretation of sectoral dynamics and external shocks.

The key relationships between productivity, sectoral characteristics, and implied prices or labor costs are consistent across estimation techniques, confirming the validity of the baseline results.

The weaker link between productivity and wages or prices in tourism, as captured by the interaction term log_Productivity:NACE_R2I, underscores the structural vulnerabilities of labor-intensive sectors in managing wage-driven cost pressures.

Additionally, the significant impact of the shock period on implied prices, but not on labor costs, highlights the differential effects of external shocks on prices versus wages. Overall, the robustness checks confirm and strengthen the earlier findings, providing a comprehensive and reliable picture of the dynamics underpinning BCDH.

4.2.4. Log Implied Price as the Dependent in the PDiD Model

The key coefficient, log_Productivity:dummy_service = −0.075 (p = 0.085), indicates that tourism prices are less responsive to productivity gains than manufacturing prices, as shown in Table 8.

This finding aligns with BCD theory, which posits that labor-intensive sectors like tourism experience slower productivity growth, leading to rising relative prices over time. Despite being marginally significant at the 10% level, this coefficient supports the hypothesis that tourism suffers from BCD.

The joint χ² test for pre-treatment parallel trends (p = 0.235) and the absence of significant individual pre-trends (p > 0.1) confirm that the tourism and manufacturing sectors followed similar trends before 2020, which delimits the post- and treatment periods (see

Table 9. Pre-treatment parallel trends—joint test of pre-period interactions (log implied price).

Test	Statistic
Joint χ2 Test	χ2 = 10.453
Largest Pre-Gap	−0.68 (2018)
Test	Statistic

Source: Authors’ calculation.

and Appendix A.3: Figure A1). This validates the DiD approach, which assumes that any divergence between the two sectors during the treatment period (2020–2023) can be attributed to the COVID-19 pandemic and the onset of the Russo–Ukrainian war, as shown in Table 9.

The year-specific interaction term γk captures how the super-shock treatment period (2020–2023) differentially impacted tourism relative to manufacturing across years. The coefficient for 2020 (−0.63) is marginally significant at the 10% level (p = 0.073), reflecting a sharp decline in tourism prices due to the collapse in demand caused by travel restrictions. By contrast, the coefficient for 2022 (+1.02, p = 0.013) is highly significant, indicating a strong recovery in tourism prices, surpassing manufacturing, as pent-up demand for travel and leisure services surged.

These results demonstrate that the pandemic caused a temporary shock to tourism prices in 2020, followed by a significant and robust rebound in 2022, as the energy crisis proliferated due to the embargo on Russian exports to the EU. The interaction term highlights the unique dynamics of the tourism sector during the treatment period, underscoring its vulnerability to external shocks and its capacity for rapid recovery, albeit at the cost of diminishing tourism consumer surplus due to booming prices.

The use of fixed effects is appropriate for controlling unobserved heterogeneity across sectors, countries, and years, particularly when accounting for year-specific interactions during the treatment period. The negative log_Productivity: dummy_service coefficient confirms that tourism prices are less responsive to productivity gains compared to manufacturing, consistent with the BCD hypothesis. Additionally, the observed trends in 2020 and 2022 align with economic expectations, reflecting demand shocks and subsequent rebounds in the tourism sector during and after the pandemic.

The results provide strong and robust evidence supporting BCD and the differential impact of the super-shock on tourism prices. The observed dynamics in 2020 and 2022 underscore the sector-specific characteristics of tourism, particularly its vulnerability to external shocks and its capacity for recovery. Given the rigorous application of the DiD model with fixed effects, supported by the validation of pre-treatment parallel trends and the significance of key coefficients, there is no need to employ additional methods. The current approach sufficiently captures the causal relationships and sectoral dynamics under investigation.

4.2.5. Log Labor Cost per Hour as the Dependent Variable in the PDiD Model

Attempts to use the uncentered model were unsuccessful due to pre-trend violations, particularly the statistically significant 2014 anomaly (p = 0.019). Since the uncentered model failed to satisfy the fundamental parallel trends assumption, it was deemed invalid and abandoned. To address this issue, a centered specification was employed in the PDiD model. This approach resolved the pre-trend violations, including the 2014 anomaly, and yielded more interpretable results, as shown in

Table 10. Panel difference-in-differences (centered log Labor Cost per Hour).

Variable	Estimate (Std. Error)
c_log_Productivity	0.327 * (0.162)
dummy_service	0.785 * (0.428)

Source: Authors’ calculation. Note: Significance levels: * p < 0.1.

and

Table 11. Pre-treatment parallel trends—joint test of pre-period interactions (centered log Labor Cost per Hour).

Test	Statistic	p-Value	Conclusion
Joint χ2 Test	χ2 = 10.453	0.235	Parallel trends hold
Largest Pre-Gap	−0.68 (2018)	0.190	No single year significant

Source: Authors’ calculation.

. Centering the dependent variable isolates within-cluster variation by subtracting group means (by country and year), improving causal inference while preserving within-sector economic interpretations, as outlined by [55]. This clarification should resolve any confusion regarding the methodological adjustments made to ensure the robustness of the analysis.

The centered specification leads to significant improvements in parallel trends. The joint F-test for pre-treatment parallel trends improved from p = 0.019 (uncentered) to p = 0.139 (centered), as shown in Table 11 and Appendix A.3: Figure A2. Furthermore, the largest pre-treatment deviation (2013, β = −2.10) became statistically insignificant (p = 0.100).

The economic interpretability also improved, as coefficients now represent deviations from sector-specific baselines, allowing for more precise interpretations. For example, the 0.79 log-point service premium (p = 0.088) reflects structural labor cost differences net of productivity trends, as shown in

Table 12. Key Findings: PDiD Model (Centered Log Labour Cost per Hour).

Variable	Coefficient	p-Value	Interpretation
Service Premium	0.785	0.088	Confirms structural labor cost gap consistent with Baumol’s cost disease hypothesis.
COVID × Services (2020)	−1.05	0.575	No differential shock, supporting nominal wage rigidity.
Productivity Effect	0.327	0.063	Sector-neutral wage responsiveness; weaker productivity transmission in services.

Source: Authors’ calculation.

.

Additionally, the model’s explanatory power improved substantially (although in such a model this is of less importance), with the within-R² increasing from 0.19 to 0.28, as shown in Table 10. The results strongly support Baumol’s cost disease hypothesis. The persistent service-sector premium (0.785, p = 0.088) confirms significant structural labor cost differences between sectors, as shown in Table 12. The null post-treatment effects (p = 0.575) suggest that symmetric shock absorption occurred across sectors during this period, consistent with nominal wage rigidity in services. Moreover, the stability of the service premium across specifications (0.79 centered vs. 0.82 uncentered) indicates that the cost gap reflects level differences, rather than growth rate divergences, aligning with Baumol’s theoretical predictions (Table 10). The centered PDiD model reveals that labor costs in the tourism sector remained largely unaffected during the treatment period, as evidenced by the null post-treatment effects (COVID × Services: p = 0.575). This strongly supports the presence of nominal wage rigidity in services, where wages are less responsive to cyclical shocks. Wage rigidity in tourism stems from structural characteristics such as institutional constraints, skill mis-matches, and the predominance of low-wage, low-skill jobs, which limit the sector’s capacity for rapid wage adjustments. Furthermore, the stability of labor costs during the shock period can be attributed to policy safeguards implemented across the EU, including wage subsidies, furlough schemes, and employment protections, which mitigated the pandemic’s impact on wages. The resilience of labor costs in services, compared to the significant price adjustments observed during the same period, highlights the decoupling of wage and price dynamics in labor-intensive sectors. These findings confirm the structural labor cost asymmetries predicted by Baumol’s hypothesis, where labor-intensive sectors face significant challenges in achieving proportional wage adjustments to external shocks or productivity gains.

The centered specification also provides more reliable estimates of labor cost dynamics. The productivity coefficient (0.327, p = 0.063) indicates sector-neutral wage responsiveness, though this likely reflects offsetting mechanisms: competitive wage-setting in manufacturing versus institutional constraints in tourism. Although the interaction term was insignificant (p = 0.255), the negative point estimate (−0.295) aligns with Baumol’s prediction of weaker productivity transmission in services, as shown in

Table 13. Sectoral asymmetries in productivity transmission.

Test	Statistic	p-Value	Conclusion
log_Productivity × NACE_R2I	Value	Interpretation	log_Productivity × NACE_R2I
Price response	0.073 + (−0.137) = −0.064	Negative net effect: Productivity gains fail to lower prices	Price response
Labor cost response	0.504 + (−0.494) ≈ 0.010	Near-zero passthrough: Wages decoupled from productivity	Labor cost response

Source: Authors’ calculation.

. These results highlight sectoral asymmetries in productivity transmission. For instance, the near-zero labor cost passthrough (0.504 + −0.494 ≈ 0.010) and the negative price response (0.073 + −0.137 = −0.064) both suggest that productivity gains do little to lower prices or wages in tourism, as shown in Table 13.

The centered PDiD model effectively resolves pre-trend violations, improves model fit, and enhances interpretability, as shown in Table 10 and Table 11. The findings robustly support Baumol’s cost disease hypothesis, emphasizing the persistent structural labor cost gap between service and manufacturing sectors while confirming the resilience of labor costs in services to cyclical shocks. The null post-treatment effects validate the presence of wage stickiness, while the service-sector premium reflects long-term structural productivity asymmetries, as shown in Table 12. This analysis underscores the robustness and theoretical relevance of centered specifications in studying sectoral labor cost dynamics.

5. Discussion

Our analysis provides strong empirical support for Baumol’s cost disease hypothesis (BCDH), demonstrating how productivity growth and sectoral characteristics shape price and wage dynamics in tourism. Both models confirm that productivity gains lead to higher prices and wages overall, consistent with Baumol’s original thesis. However, the tourism sector—characterized by its reliance on human interaction and labor-intensive services—struggles to translate productivity improvements into proportional wage growth or price stability. This asymmetry echoes findings on tourism’s stagnating productivity of [7] and aligns with warnings of [18] about the economic risks of overreliance on tourism-led growth. Their “Beach Disease” hypothesis highlights tourism’s structural vulnerabilities, which are compounded by sector-specific constraints. Furthermore, the weaker productivity-price interaction observed in [27] reflects structural constraints, where labor-intensive services face significant challenges in mitigating wage-driven inflation.

• Impact of Exogenous Shocks

The COVID-19 pandemic and the Russo-Ukrainian war amplified these dynamics in several significant ways. First, these crises caused substantial price surges across the tourism sector, which were not matched by corresponding wage increases. This price-wage divergence underscores the sector’s vulnerability to external shocks, as noted by [19].

Second, despite the economic disruptions caused by these shocks, wage levels in the tourism sector remained largely stable. This stability reflects two factors: nominal wage rigidity and policy safeguards. Nominal wage rigidity arises from institutional factors, such as fixed-term contracts, union agreements, and minimum wage policies, which limit wage responsiveness to external shocks. Simultaneously, policy safeguards—including wage subsidies, furlough schemes, and employment protections—played a critical role in stabilizing wages during the pandemic.

These mechanisms mitigated wage volatility even as prices surged, consistent with findings of [24] and [56]. These opposing dynamics—price inflation without wage growth—highlight the structural fragility of the tourism sector and its limited ability to effectively absorb external shocks. Robustness checks across multiple statistical methods further validate these pat-terns, addressing concerns raised by [5] about measurement biases in cost-disease studies. The super-shock period’s distinctive effects—substantial price surges but stagnant wag-es—serve as a stark example of the sector’s structural challenges. Overall, the pandemic revealed the acute susceptibility of the tourism sector to cost-disease dynamics and rein-forced the need for targeted policy interventions.

• Structural Constraints in Tourism

Our results show that productivity broadly raises wages, consistent with Baumol’s thesis. However, tourism’s muted response underscores its structural inability to translate productivity gains into wage growth, reflecting a persistent productivity–wage decoupling also observed by [57] in Italian accommodation firms. This is further supported by the consistently negative interaction term across models (−0.137**), which signals tourism’s structural lag in price and wage adjustments. These findings challenge tourism-led growth claims of [13] and instead support [26] view of manufacturing’s superior productivity–wage alignment.

This lag is best understood through the Experience Economy framework described by [3], which emphasizes tourism’s shift toward delivering personalized, memorable experiences to justify rising costs. Such offerings enhance customer satisfaction but do not generate proportional productivity gains, thereby constraining wage growth. This experience-productivity paradox reflects structural characteristics of tourism that differ from manufacturing.

We identify three key structural constraints that underpin tourism’s productivity challenge:

• Labor-Intensive Service Enhancements: Tourism depends on service quality that is difficult to scale without increasing labor inputs. Personalized experiences, such as guided tours or tailored guest services, remain difficult to automate, as emphasized by [11].
• Future research should explore how AI-driven tools, robotics, and automation can be applied to labor-intensive roles such as catering, cleaning, and guided activities similar to findings of [23]. Examining barriers to adoption, cost-effectiveness, and scalability of these technologies across different tourism subsectors will help identify practical pathways to enhance productivity while addressing the sector’s structural constraints.

2.

Labor Market Rigidities: Institutional factors—including seasonal employment, SME dominance, and weak unionization—create wage stickiness [25]. These dynamics sustain price–wage spirals and reduce responsiveness to productivity improvements, as seen in near-zero labor cost elasticity (≈0.010) and stagnant wages during COVID-19 despite rising prices (treatment_post = 0.634***; wage = −0.048) similar to [55].

3.

Technology Asymmetry: Despite increased adoption of AI and IoT in marketing or booking systems indicated by [26], these tools do not significantly impact core, labor-intensive functions like housekeeping or frontline services. This explains the weak net effect of technology on prices (−0.064*), confirming the limited potential of automation to alleviate Baumol’s cost disease in tourism. Furthermore, as noted in [22], the fragmented and service-specific implementation of smart tourism initiatives prevents the systemic coordination needed to drive broad productivity gains, leaving structural constraints largely unaddressed, particularly for SMEs with limited capacity to adopt such technologies effectively.

These constraints create a macroeconomic decoupling between productivity and wage/price dynamics, consistent with Baumol’s original hypothesis and supported empirically in our findings (e.g., −0.137***). This stagnation limits tourism’s ability to absorb external shocks, as shown during the COVID-19 pandemic, which exposed resilience gaps and amplified imbalances noted by [27,35,58].

• Price Dynamics and Future Research

Our price-based analysis reveals how exogenous shocks intensify volatility in the tourism sector, underscoring its structural vulnerabilities. Specifically:

• COVID-19 collapse: Prices declined by 0.63 log-points (p < 0.10) due to demand destruction. The pandemic caused an unprecedented drop in global tourism demand, as widespread travel restrictions, border closures, and public health concerns led to a sharp contraction in tourism activity. The study of [12] highlighted how international tourist arrivals fell by 74% in 2020, driving price collapses across key tourism destinations. This demand shock exposed the sector’s dependency on discretionary spending and global mobility, amplifying its vulnerability to external disruptions.
• The 2022 surge: Prices rebounded by +1.02 log-points (p = 0.013) as the energy crisis triggered by the Russo–Ukrainian war transmitted cost pressures into tourism output prices. The war caused significant disruptions to global energy supplies, resulting in skyrocketing fuel prices and increased operational costs for tourism businesses. As noted by [59], the energy crisis disproportionately affected energy-intensive industries, including transportation and hospitality, forcing businesses to pass on rising costs to consumers. This price surge highlights tourism’s sensitivity to macroeconomic shocks and its reliance on stable energy markets.

These findings illustrate tourism’s unique price dynamics compared to other service industries. Unlike healthcare or education, where prices are often insulated by public funding or institutional structures, tourism’s exposure to global markets and reliance on consumer demand make it particularly susceptible to exogenous shocks [35]. Such volatility underscores the need for sector-specific resilience strategies, such as targeted subsidies during crises or investments in energy efficiency, to mitigate the long-term effects of these disruptions.

• Synthesis of Future Research Directions

To synthesize the relevance of future research on BCD mitigation, we propose a multi-level research framework (see Appendix A.4:

Table A3. Micro-level: firm/workforce-level dynamics. Focus: Individual enterprises, frontline labor, job functions.

Research Area	Description	Relevance to BCD Mitigation
Cost-effectiveness of automation	Assess how specific roles (e.g., housekeeping, reception, food service) can be automated to reduce reliance on labor that does not experience productivity growth.	High
Adoption barriers	Investigate skill gaps, resistance to automation, and cost constraints in SMEs and family-run tourism businesses.	High
AI and digital tools	Analyze productivity effects of AI-based systems (chatbots, booking automation, personalization engines) in hospitality and entertainment subsectors.	High
Green workforce transformation	Study how low-carbon job redesign and green upskilling impact labor costs and service quality.	Medium

,

Table A4. Meso-level: regional/institutional coordination. Focus: Destinations, DMOs, tourism clusters.

Research Area	Description	Relevance to BCD Mitigation
DMO-led innovation hubs	Evaluate the role of Destination Management Organizations in coordinating digital adoption, sustainability, and workforce modernization.	Medium–High
Regional cooperation models	Explore collaborative models across tourism-dependent regions to share best practices and reduce duplication of high-cost services.	Medium
Decarbonization initiatives	Assess regional incentives for sustainable infrastructure and how they interact with labor demand in tourism services.

and

Table A5. Macro-level: national/global policy and systemic reform. Focus: Institutional frameworks, international comparison, systemic resilience.

Research Area	Description	Relevance to BCD Mitigation
Digital tourism policy integration	Propose national strategies that blend innovation support, labor reform, and tourism incentives.	High
Cross-country comparative analysis	Identify how institutional and cultural settings modulate the effects of BCD across nations and derive policy implications.	High
Structural reforms for labor productivity	Examine the macroeconomic feasibility of redirecting subsidies or taxes toward automation and skill development.	High
Green transition and BCD	Investigate how environmental policy (e.g., carbon pricing, ESG standards) interacts with BCD in tourism-heavy economies.	Medium–High
Systemic shock resilience	Model the responsiveness of labor-intensive tourism systems to pandemics, climate shocks, or economic downturns.

). This framework highlights promising research topics at the micro, meso, and macro levels.

6. Conclusions

This study provides robust empirical evidence supporting Baumol’s cost disease hypothesis (BCDH) in the context of tourism, with important implications for understanding sectoral dynamics, wage-price relationships, and the impact of two major external shocks: the COVID-19 pandemic and the Russo–Ukrainian war. By examining productivity, prices, and labor costs across 15 selected EU countries from 2011 to 2023, the analysis reveals that productivity gains broadly drive wage increases and price inflation, consistent with BCD. However, the weaker link between productivity and wages or prices in labor-intensive sectors like tourism highlights the structural vulnerabilities of these industries in managing wage-driven cost pressures.

This study directly addressed the research questions and hypotheses outlined in the introduction, as follows:

H1: Productivity Growth in Tourism Shows a Weaker Effect on Wages and Prices.

Our findings strongly confirm this hypothesis. We observed weaker productivity–wage linkages in tourism across all models. For instance, the interaction term (−0.137***) consistently indicated a structural lag in tourism’s wage and price adjustments. Labor-intensive services like tourism are constrained by their reliance on human interaction, which limits their ability to achieve proportional productivity-driven wage growth. These structural constraints are consistent with Baumol’s cost disease hypothesis, highlighting tourism’s challenges in managing wage-driven cost pressures.

H2: Exogenous Shocks Exacerbate Price Volatility in Tourism.

This hypothesis is supported by the distinct price dynamics observed during the COVID-19 pandemic and the RussoUkrainian war. During the pandemic, prices collapsed by −46.8% due to demand destruction, whereas they surged by +177.4% during the energy crisis. These findings illustrate the tourism sector’s heightened sensitivity to external shocks, particularly its reliance on energy markets and discretionary consumer spending. The price volatility underscores the sector’s structural vulnerabilities and its limited resilience to economic disruptions.

H3: Technological Innovation and Labor Market Reforms Can Mitigate BCD.

Our results partially confirm this hypothesis. While AI and digital tools have enhanced peripheral functions like booking and marketing, their impact on core labor-intensive operations remains limited. For example, the weak technological passthrough to labor costs (≈0.010 elasticity) underscores the need for targeted investments in productivity-enhancing technologies. Similarly, labor market reforms are essential to address structural constraints, such as wage rigidity and skill mismatches, which hinder tourism’s ability to adapt to external shocks and achieve sustained productivity gains.

The tourism sector’s reliance on human interaction and labor-intensive services constrains its ability to translate productivity improvements into proportional wage growth or price stability. This structural lag is further exacerbated by external super-shocks, which caused significant price surges in the tourism sector without corresponding wage increases. These findings underscore the inherent fragility of the sector, where price inflation and stagnant wages create challenges for sustainable recovery and long-term growth.

Robustness checks confirm the validity of these findings, enhancing the reliability of the results. The study also highlights the pandemic’s role in accelerating the adoption of digital technologies and AI tools in tourism, offering a potential pathway for addressing the structural limitations of labor-intensive sectors. Policymakers are urged to leverage these insights to mitigate sectoral imbalances through hybrid models, such as tech-enhanced tourism, and targeted EU cohesion funds aimed at fostering innovation and resilience in tourism-dependent economies.

These findings reflect significant transformations in global labor markets, particularly in tourism. The increasing internationalization of the workforce, evident in the presence of workers from diverse backgrounds in tourism hubs, illustrates how globalization has reshaped wage distribution and employment structures. Such trends highlight the growing reliance on mobile, global labor forces to sustain tourism economies, even in traditionally localized markets.

While these changes have created opportunities for global workers and fostered greater mobility, they also bring persistent challenges, such as wage stagnation and inequality within labor-intensive sectors. These developments underscore the structural vulnerabilities of tourism, where globalization-driven labor mobility coexists with systemic issues in wage distribution. These trends reflect a broader struggle to create equitable labor systems amidst the pressures of globalization and neoliberal capitalism, as highlighted by [60].

Beyond its empirical results, this study makes an important theoretical contribution by extending Baumol’s cost disease hypothesis to the specific context of tourism, particularly under the unprecedented conditions created by recent global shocks. By integrating sectoral wage–price dynamics, exogenous crisis impacts, and the partial role of technological innovation, our work refines and updates the classical BCD framework for 21st-century service economies. This approach clarifies how and why labor-intensive sectors like tourism display unique vulnerabilities and adjustment patterns, and it highlights the need for more nuanced theoretical models that account for both structural inertia and adaptive responses to crisis. In doing so, the study advances the broader literature on cost disease and service sector transformation and offers a conceptual foundation for future research on tourism resilience and policy innovation.

Limitations and Future Directions

This study has certain limitations that should be addressed in future research. First, it treats tourism as a monolithic sector, providing a bird’s-eye view that overlooks the diversity of its subsectors—such as hospitality, transportation, and entertainment—which operate under distinct economic conditions and may not align uniformly with the assumptions of Baumol’s cost disease. Future research should examine these subsectors individually to better understand the unique dynamics and challenges they face.

Second, this study does not fully account for the role of technological advancements, such as AI-driven tools, digital booking systems, and automation, which have substantially improved productivity in certain segments of tourism. These advancements challenge the classical assumptions of stagnant productivity in labor-intensive sectors and warrant further investigation into how they reshape productivity–wage relationships.

Third, the role of government interventions during the COVID-19 pandemic—such as wage subsidies, stimulus programs, and temporary layoffs—is not explicitly addressed in this analysis, despite their likely influence on labor costs and sectoral resilience. Future studies could explore how such policies interacted with BCD dynamics and affected recovery trajectories in tourism-dependent economies.

Lastly, while this analysis of 15 EU countries (2011–2023) provides robust evidence of Baumol effects in tourism, expanding geographic coverage—particularly to tourism-dependent economies like Spain and Croatia—would enhance our understanding of these dynamics in broader institutional contexts. Such research could also explore how the tourism sector adapts to structural challenges while balancing sustainability and economic growth. These insights will be critical for fostering more resilient, equitable, and sustainable development in tourism-dependent regions.