Tourism and Economic Growth in the EU-27: Mechanisms, Heterogeneity, and Threshold Effects

Abstract

This study revisits the tourism–economic growth nexus in the EU-27 over 2014–2024 by combining an integrative conceptual synthesis of transmission mechanisms with a multi-method empirical strategy. We move beyond a single fixed-effects regression by integrating: (i) a dynamic panel System Generalized Method of Moments (System GMM) estimator that addresses endogeneity and persistence in growth, (ii) interaction terms between tourism intensity and moderating factors (institutional quality, human capital, trade openness), (iii) sub-sample analyses across mature versus emerging tourism destinations and Mediterranean versus Continental economies, and (iv) a Hansen-type threshold model that identifies non-linearities in the tourism–growth relationship. The baseline fixed-effects results confirm a positive and statistically significant tourism elasticity, while the dynamic and threshold specifications reveal that this effect is highly heterogeneous: it is amplified in economies with stronger institutions and higher human capital, but becomes statistically indistinguishable from zero in destinations with very high tourism intensity, consistent with congestion and overtourism mechanisms. Evidence on heterogeneous post-pandemic resilience is treated as exploratory rather than as a confirmed empirical contribution, since the present specification relies on a COVID-19 dummy and sub-sample patterns rather than a dedicated COVID-19 interaction model. The findings reposition tourism not as a uniformly positive growth engine, but as a moderated and bounded driver whose contribution depends on absorptive capacity and management quality.

1. Introduction

Tourism is one of the most dynamic and resilient sectors of the global economy, accounting for more than 10% of world GDP and supporting over 370 million jobs worldwide. Within the European Union, the sector occupies a strategic position in the cohesion, recovery, and competitiveness agendas, with Member States allocating significant shares of NextGenerationEU and Cohesion Policy resources to tourism-related projects. The tourism-led growth hypothesis (TLGH) postulates that expansion in tourism activity stimulates aggregate output through investment, productivity spillovers, foreign-exchange earnings, and labour-market multipliers. While the TLGH has been extensively tested in emerging markets, evidence for advanced economies remains comparatively thin and inconclusive, particularly regarding the conditions under which tourism activity translates into sustained growth.

Three gaps in the existing literature motivate this study. First, much of the empirical work treats tourism as an undifferentiated growth input without articulating the specific channels—demand, investment, productivity, institutional, and externality—through which it operates. Second, the dominant econometric approach has been a single static specification (typically fixed-effects panel regression), which is unable to capture endogeneity, persistence, heterogeneity, or non-linearities in the tourism–growth relationship. Third, the policy discussion has tended to remain at a general level, offering little operational guidance for destination managers, tourism ministries, or EU funding agencies that must design differentiated strategies for very different types of destinations.

This paper addresses these gaps in four distinct ways. First, we synthesise an integrative conceptual framework that organises the tourism–growth nexus around five transmission mechanisms and three moderating factors. The framework is not presented as an entirely new theory, but as a structured synthesis of existing TLGH, institutional, human-capital and sustainability arguments that disciplines variable selection and interpretation. Second, we extend the standard fixed-effects panel model with a richer set of macroeconomic controls (trade openness, human capital, institutional quality, inflation) and re-estimate the relationship using a System GMM dynamic panel estimator that addresses endogeneity, persistence, and the dynamic nature of growth. Third, we test heterogeneity by interacting tourism with institutional quality and human capital, by splitting the sample into mature versus emerging tourism economies and Mediterranean versus Continental destinations, and by estimating a Hansen-type threshold model. Fourth, we translate the empirical findings into segmented, actionable policy recommendations linked directly to EU funding instruments and destination–management challenges, while treating the pandemic-resilience proposition as exploratory unless future research estimates formal COVID-19 interaction models.

Our results yield three substantive insights that refine, rather than simply confirm, the conventional TLGH literature. (a) The tourism–growth relationship in advanced European economies is non-linear: it is strongly positive at low and moderate levels of tourism intensity but flattens at very high intensities, consistent with congestion and overtourism mechanisms. (b) The magnitude of tourism’s contribution is conditional on absorptive capacity, with institutional quality and human capital acting as amplifying moderators. (c) The available country-level evidence suggests that exposure to the COVID-19 shock differed across destination types; however, because the manuscript does not estimate a formal COVID-19 × destination-type interaction model, this result is presented only as a descriptive and suggestive resilience pattern, not as a confirmed test of P5. Together, these findings reposition tourism in the policy debate as a bounded, moderated, and management-dependent growth driver rather than a uniform positive force.

The remainder of the paper is organised as follows. Section 2 develops the conceptual framework and derives testable propositions. Section 3 reviews the related literature. Section 4 describes the data and outlines the multi-method empirical strategy. Section 5 reports the empirical results, including baseline, dynamic, heterogeneity, threshold, and robustness analyses. Section 6 discusses the findings in relation to the conceptual mechanisms. Section 7 develops the policy implications. Section 8 concludes.

2. Conceptual Framework: How Tourism Influences Growth in Advanced European Economies

Although the tourism-led growth hypothesis has accumulated substantial empirical support, EU-specific empirical work rarely brings its mechanisms, moderators and negative externalities into a single operational structure. The conceptual contribution of this section is therefore integrative rather than theory-building in a strong sense: it consolidates established arguments from the TLGH, endogenous growth theory, institutional economics, human-capital theory and sustainable-destination research into a coherent framework that can guide empirical specification. A clear synthesis matters because it disciplines variable selection and connects coefficient signs, interaction effects and threshold results to specific mechanisms.

2.1. Five Transmission Mechanisms

Demand-multiplier channel: Inbound tourism expenditure represents an injection into the host economy. Following the standard Keynesian framework, this injection induces successive rounds of spending on intermediate inputs, wages, and induced household consumption; consequently, it generates an output multiplier whose magnitude depends on the local content of tourism supply chains and the marginal propensity to consume domestically (Brida et al., 2016). In the EU-27 single-market context, this multiplier is moderated by the openness of intra-EU supply chains: in destinations heavily dependent on imported intermediate goods and labour, leakages reduce the multiplier.

Investment-spillover channel: Tourism expansion stimulates complementary investment in transport infrastructure, accommodation, digital platforms, and ancillary services. These investments produce productivity-enhancing spillovers that benefit non-tourism sectors—particularly construction, retail, transport, and ICT—and accumulate over time. This channel typically operates with a lag, which justifies the inclusion of lagged gross fixed capital formation in the empirical model.

Foreign-exchange and balance-of-payments channel: International tourism receipts ease external financing constraints and support the import of capital goods that raise total factor productivity. Although less binding in the euro area than in emerging economies, this channel remains relevant for non-euro EU members and for sub-national regions whose specialisation pattern depends on foreign-currency earnings.

Productivity and knowledge-spillover channel: Building on endogenous growth theory (Romer, 1990; Lucas, 1988), tourism contributes to growth through service-sector innovation, the absorption of digital and managerial best practices from internationally exposed firms, and the demonstration effects of high-end international clients on local quality standards. Cultural and creative industries function as catalysts in this channel, generating reinforcing feedback loops between cultural investment and international arrivals.

Labour-market and human-capital channel: Tourism is a labour-intensive sector with low entry barriers, which has direct effects on employment and a particularly strong role in absorbing low- and medium-skill workers and reducing regional disparities. Over time, however, the sector also generates demand for higher-skilled labour in management, marketing, and digital services, contributing to human-capital deepening when properly supported by vocational training.

2.2. Three Moderating Factors

The strength of each transmission mechanism is conditioned on country-level absorptive capacity. We highlight three moderators that the empirical model can capture or proxy.

Institutional quality (rule of law, government effectiveness, regulatory quality) determines the share of tourism rents that are reinvested productively rather than dissipated through corruption or rent-seeking. Following North (1990) and the more recent governance literature, institutions act as complementary infrastructure for tourism’s growth contribution.

Human capital amplifies the productivity and knowledge-spillover channels by enabling the absorption of innovation, the diversification of the tourism product, and the upgrading of service quality. In the EU context, this is proxied by tertiary educational attainment and participation in lifelong learning.

Trade openness captures the broader external orientation of the economy. Higher openness typically reinforces the demand-multiplier and foreign-exchange channels but can also increase leakages if the local tourism supply chain is shallow.

2.3. Potential Negative Externalities

The framework does not assume a uniformly positive tourism–growth relationship. Three classes of negative externalities can attenuate or even reverse the relationship at high levels of tourism intensity. First, congestion costs (infrastructure strain, housing-cost inflation, public-service crowding) appear once visitor density exceeds local carrying capacity—the overtourism phenomenon documented in Barcelona, Venice, Dubrovnik, and Amsterdam. Second, a tourism-specific Dutch disease can arise when resource allocation tilts toward low-productivity tourism services at the expense of tradable manufacturing. Third, environmental externalities (emissions, biodiversity loss, water stress) impose costs that are not internalised by individual operators and that compromise the sector’s long-run sustainability. Taken together, these externalities motivate a non-linear specification with possible threshold effects.

2.4. Testable Propositions

The integrative framework generates four directly testable propositions and one exploratory proposition that structure the empirical analysis:

P1 (baseline TLGH): Tourism activity is positively and significantly associated with real GDP growth in the EU-27 over 2014–2024, conditional on standard macroeconomic controls.

P2 (institutional amplification): The growth contribution of tourism is larger in countries with higher institutional quality.

P3 (human-capital amplification): The growth contribution of tourism is larger in countries with higher human capital endowments.

P4 (non-linearity/overtourism): The tourism–growth relationship is non-linear: positive at low and moderate tourism intensities, but weaker or statistically insignificant at very high intensities.

P5 (exploratory heterogeneous pandemic resilience): The negative impact of the COVID-19 shock may be more pronounced in mature, mono-thematic mass-tourism destinations than in diversified or innovation-oriented economies. This proposition is treated as exploratory in the present manuscript because the reported empirical models include a common COVID-19 dummy but do not yet estimate a formal COVID-19 × destination-type interaction.

3. Literature Review

Tourism has long been viewed as an important driver of economic development, particularly in advanced economies where it contributes to diversification, regional cohesion, and global competitiveness. Contemporary scholarship increasingly emphasises that tourism should be treated as a complementary growth factor rather than an exogenous source of expansion, opening space for more nuanced theoretical and empirical analysis (Sequeira & Nunes, 2008; Alqaralleh et al., 2025; Sun et al., 2025). We organise the review around the five mechanisms identified in Section 2.

The foundational contributions to the TLGH established the demand-multiplier and foreign-exchange channels. Balaguer and Cantavella-Jordá (2002), using Spain as a case study, demonstrated a long-run causal link between tourism and GDP. Dritsakis (2004) confirmed positive causality in Greece through cointegration and Granger causality tests. Sequeira and Nunes (2008) found that the positive impact of tourism is more pronounced in countries with strong institutional capacities, educated labour forces, and stable macroeconomic environments—evidence consistent with the moderating role of human capital and institutions developed in our framework.

Subsequent work has identified the productivity, knowledge-spillover, and labour-market channels. Brida et al. (2016) document that, in Western Europe, tourism fosters economic diversification and reduces regional disparities. Studies on OECD countries covering 1998–2022 demonstrate that cultural and creative industries serve as catalysts for tourism growth, generating reinforcing feedback loops between cultural investment and international arrivals. Evidence from OECD economies also suggests that diversifying the origin of tourist markets reduces the size of the shadow economy, reinforcing the formalisation of economic activity and amplifying the developmental effects of tourism.

A complementary stream warns of the negative externalities that motivate our non-linearity proposition. Cárdenas-García et al. (2015) emphasise that the growth contribution of tourism depends on the conditions under which tourism expansion is converted into broader economic development. Mihajlović et al. (2025) employ a dynamic panel threshold model for the EU and show that the tourism–growth–environment nexus is conditioned by development thresholds and carbon-emission dynamics, highlighting the ecological limits of the TLGH. Serio et al. (2025), applying a spatial econometric model to Italian provinces, show that environmental policies produce both direct and spatial spillover effects on tourism flows, underscoring the importance of coordinated green strategies.

Recent contributions have adopted more sophisticated econometric techniques that align closely with our multi-method strategy. Alqaralleh et al. (2025) apply a panel quantile ARDL framework to capture asymmetries, showing that tourism exerts stronger stabilising effects during economic downturns than in expansionary phases. Sun et al. (2025) demonstrate that institutional quality amplifies the positive contribution of tourism to GDP growth, supporting our Proposition P2. Nugroho and Verikios (2025) compare tourism-led and non-tourism-led growth through a dynamic computable general equilibrium framework and show that tourism can drive long-run growth when it induces productivity-enhancing sectoral shifts and higher capital accumulation.

The COVID-19 pandemic has prompted a renewed focus on resilience and adaptability. Gössling et al. (2021) stress that advanced economies, supported by diversified revenue sources and digital technologies, recovered more quickly. Rasool et al. (2021) provide panel cointegration evidence on the tourism–growth relationship in BRICS economies, while OECD work on resilient destinations emphasises crisis preparedness, digitalisation, green investment, and destination governance as key conditions for post-pandemic recovery.

International organisations also underline tourism’s continuing macroeconomic weight. According to UN Tourism (2025), international arrivals in Q1 2025 rose by 5% year-on-year, while WTTC (2025) forecasts international visitor spending of USD 2.1 trillion, a Travel and Tourism contribution of USD 11.7 trillion, or 10.3% of global GDP, and approximately 371 million jobs worldwide. OECD (2024, 2025) emphasises the importance of building resilient destinations through coordinated governance, digitalisation, green investment, and crisis preparedness.

Overall, the literature suggests that tourism in developed economies exerts a significant but complex impact on growth, shaped by institutional frameworks, sustainability policies, digitalisation, and market diversification. While the TLGH remains widely supported, contemporary research stresses the importance of mechanism-level analysis, heterogeneity, and non-linearities—the three dimensions on which the present paper makes its main contribution.

4. Data and Methodology

4.1. Sample, Variables, and Data Sources

The empirical analysis uses a balanced panel of the EU-27 Member States over 2014–2024 (T = 11; N × T = 297 country–year observations; 270 after the one-year lag on investment). The eleven-year window is sufficient to support reliable panel inference and spans both expansionary and contractionary phases, including the unprecedented COVID-19 shock and the subsequent recovery.

The dependent variable is the real GDP growth rate, the canonical indicator of macroeconomic dynamics. The key explanatory variable is the annual growth rate of nights spent at tourist accommodation establishments at the country level, consistent with NACE Rev. 2 coverage and aggregating nights in hotels, holiday and other short-stay accommodation, camping grounds, and recreational-vehicle and trailer parks. This measure is preferred over international arrivals because it captures actual length-of-stay intensity, which is more closely related to the economic footprint of tourism activity.

Relative to the baseline specification (general government final consumption, lagged gross fixed capital formation, COVID-19 dummy), we extend the variable set with four additional controls motivated by the conceptual framework. (i) Trade openness, defined as the sum of exports and imports as a share of GDP, captures the external orientation that conditions the demand-multiplier and foreign-exchange channels. (ii) Human capital is proxied by the share of the population aged 25–64 with tertiary educational attainment, capturing the absorptive capacity for productivity and knowledge spillovers. (iii) Institutional quality is the simple average of the World Bank Worldwide Governance Indicators (WGI) for government effectiveness, regulatory quality, and rule of law. (iv) Inflation (HICP, annual average rate of change) is included as a standard macroeconomic stability control. Macroeconomic indicators are drawn from Eurostat and the World Bank, tourism statistics from Eurostat tourism (NACE Rev. 2), and governance indicators from the World Bank WGI database. The COVID-19 dummy is therefore a common 2020 shock indicator: it takes the value 1 only in 2020 for all 27 Member States and 0 in all other years, including 2021. The recovery year 2021 remains fully included in the balanced panel but is not coded as part of the dummy.

4.2. Baseline Fixed-Effects Specification

The baseline specification follows the standard panel-data regression with country fixed effects:

y_it = α + x_itβ + c_i + u_it

(1)

where y_it is the dependent variable for country i in year t, α is the constant, x_it is the K-dimensional vector of regressors (tourism activity plus controls), β is the corresponding parameter vector, c_i captures unobserved country heterogeneity, and u_it is the idiosyncratic error. The choice between random and fixed effects is decisive: the random-effects estimator requires E [c_i|x_it] = 0, an assumption that is rarely defensible in cross-country macro panels. We therefore use fixed effects as the baseline and verify the choice through (i) the cluster-robust Hausman (Sargan–Hansen) specification test and (ii) the Mundlak–Chamberlain correlated random-effects (CRE) decomposition. Standard errors are clustered at the country level throughout.

4.3. Dynamic Panel Specification: System GMM

A static fixed-effects model is unable to address two features that are central to growth dynamics: (i) the persistence of real GDP growth—growth at t depends on growth at t − 1—and (ii) the potential endogeneity of tourism activity, which may itself respond to current and lagged growth. To address both issues, we estimate a dynamic panel specification:

y_it = α + ρy_i,t−1 + x_itβ + c_i + u_it

(2)

Equation (2) is estimated using the two-step System GMM estimator of Arellano and Bover (1995) and Blundell and Bond (1998) with Windmeijer-corrected standard errors. The System GMM estimator combines the equation in first differences (which removes country fixed effects) with the equation in levels, using lagged differences as instruments for the levels equation and lagged levels as instruments for the differenced equation. Tourism activity, government consumption, and lagged investment are treated as predetermined. We collapse the instrument matrix and restrict the lag length (typically t − 2 to t − 4) to reduce instrument proliferation. In the preferred parsimonious GMM specifications, the instrument count is safely below the number of groups (21–23 instruments versus N = 27); the full specification reports 25 instruments, which is only slightly below the number of groups and is therefore interpreted as a supplementary sensitivity check rather than as the primary dynamic estimate. Specification validity is assessed via (i) the Arellano–Bond test for AR(2) in residuals, (ii) the Hansen J test of over-identifying restrictions, and (iii) the Diff-in-Hansen test for the validity of the additional level-instrument set. Two qualifications are important for interpreting the dynamic panel results. First, the EU-27 setting implies a small cross-sectional dimension (N = 27) and a short time dimension (T = 11). Although this is a common structure in macro-panel studies of the European Union, it is not the ideal large-N/small-T environment for which System GMM was originally designed. Second, finite-sample bias, weak-instrument concerns, and instrument-count sensitivity may remain even after collapsing the instrument matrix and restricting the lag depth. For this reason, the GMM estimates are interpreted as a robustness and endogeneity-adjustment exercise rather than as a stand-alone source of causal identification. The key inference is therefore based on the consistency of the sign and significance of the tourism coefficient across FE, two-way FE, interaction, threshold and dynamic specifications, not on any single estimator.

4.4. Interaction Effects and Heterogeneity

To test Propositions P2 and P3, we augment the baseline FE model with two interaction terms: tourism × institutional quality and tourism × human capital. A positive and significant coefficient on the interaction term indicates that the marginal contribution of tourism increases with the moderator. To complement the interaction analysis, we estimate separate regressions on three sub-samples motivated by structural heterogeneity:

(a)

Mature tourism economies (Spain, Italy, France, Greece, Portugal, Austria, Croatia, Cyprus, Malta) versus emerging tourism destinations (the remaining EU-27 members).

(b)

Mediterranean coastal destinations (Spain, Italy, France, Greece, Portugal, Croatia, Cyprus, Malta, Slovenia) versus Continental destinations (the remainder).

(c)

High tourism intensity (top tercile of nights per capita) versus low tourism intensity (bottom two terciles).

4.5. Threshold Effects

To test Proposition P4 (non-linearity), we estimate a Hansen-type panel threshold model (Hansen, 1999, 2017) in which the marginal effect of tourism on growth differs across regimes defined by a threshold variable—either tourism intensity (nights per inhabitant) or institutional quality. The model takes the form:

y_it = α + β₁·TOUR_it·I(q_it ≤ γ) + β₂·TOUR_it·I(q_it > γ) + Z_itδ + c_i + u_it

(3)

where I(·) is the indicator function, q is the threshold variable, γ is the (estimated) threshold value, and Z is the set of controls. The threshold γ is estimated by minimising the concentrated sum of squared residuals; statistical significance is assessed by bootstrap (1000 replications).

4.6. Robustness Strategy

We assess robustness along six dimensions: (i) an alternative tourism measure (international tourist arrivals per capita), (ii) an alternative dependent variable (real GDP per capita growth), (iii) sample exclusion (dropping 2020 and 2021 to verify that results are not driven by the pandemic shock), (iv) two-way fixed effects (country and year), (v) formal multicollinearity diagnostics using VIF and condition-index statistics, and (vi) sensitivity of the threshold result to alternative threshold variables and sample restrictions. Each robustness check is documented in Section 5.7.

5. Empirical Results

5.1. Descriptive Statistics

Table 1 reports descriptive statistics for the baseline and extended variable sets. Real GDP growth displays substantial dispersion across the EU-27 over 2014–2024 (cross-sectional standard deviation approximately 1.5 times the mean), reflecting pronounced heterogeneity in macroeconomic performance, including the contractionary 2020 episode and the post-pandemic recovery. Tourism activity, proxied by the growth rate of nights spent, also varies considerably, with a coefficient of variation that supports a formal econometric assessment of whether tourism fluctuations are associated with growth differences once macroeconomic controls are taken into account.

Table 1. Descriptive statistics for the extended variable set.

Variable	Obs.	Mean	Std. Dev.	Min	Max
Real GDP growth rate	297	0.026	0.037	−0.109	0.246
Tourism nights growth rate	297	0.068	0.274	−0.788	1.615
General government consumption (% GDP)	297	0.199	0.032	0.110	0.266
Gross fixed capital formation (% GDP, t − 1)	270	0.214	0.042	0.110	0.532
COVID-19 dummy	297	0.091	0.287	0	1
Trade openness (% GDP)	297	1.404	0.677	0.561	3.808
Tertiary education (% pop. 25–64)	297	0.341	0.075	0.181	0.531
Institutional quality (WGI avg.)	297	1.043	0.625	−0.180	2.124
Inflation (HICP, annual %)	297	0.021	0.030	−0.013	0.190

Notes: The COVID-19 dummy is coded as 1 for 2020 only and 0 otherwise; 2021 is retained in the panel as a post-shock recovery year, not as a second COVID-dummy year and not with any weighting. With N × T = 27 × 11 = 297 observations, the reported mean of 0.091 corresponds to 27 country–year observations (0.091 × 297 ≈ 27), i.e., one full year across all EU-27 Member States.

The extended controls reveal substantial heterogeneity in EU-27 economic structure: trade openness ranges from 0.56 (more closed Continental economies) to 3.81 (small open economies such as Luxembourg, Malta, and Ireland), tertiary educational attainment ranges from 18% to 53%, and institutional quality differs by more than two standard deviations across Member States. This heterogeneity is precisely what motivates the moderation and sub-sample analyses.

Table 2 presents the pairwise correlation matrix. Off-diagonal correlations remain moderate, with no value approaching conventional multicollinearity thresholds. Because pairwise correlations alone are not a sufficient diagnostic, we additionally report formal Variance Inflation Factor (VIF) and condition-index statistics in Table 2a. The bivariate correlation between real GDP growth and tourism activity is relatively weak (0.367), reinforcing the need for a conditional panel specification rather than reliance on unconditional comovements.

Table 2. Pairwise correlation matrix (selected variables).

Variable	GDP	Tour	GovC	Inv	COVID	Open	HC	Inst
Real GDP growth	1.000
Tourism nights gr.	0.367	1.000
Gov consumption	−0.365	−0.117	1.000
Investment (t − 1)	0.084	−0.073	−0.002	1.000
COVID dummy	−0.485	−0.466	0.133	0.088	1.000
Trade openness	0.142	0.061	−0.243	0.087	−0.038	1.000
Tertiary education	0.058	0.029	0.114	0.062	−0.011	0.187	1.000
Institutional qual.	0.119	0.034	0.198	0.052	0.027	0.342	0.502	1.000
(a). Formal multicollinearity diagnostics for the full FE specification
Regressor	VIF				Tolerance
Tourism nights growth	1.31				0.76
Gov. consumption (% GDP)	1.24				0.81
Investment (% GDP, t − 1)	1.06				0.94
COVID-19 dummy	1.36				0.74
Trade openness	1.32				0.76
Tertiary education	1.47				0.68
Institutional quality	1.64				0.61
Inflation	1.18				0.85

Notes: VIF values below 5 and condition-index values below 10 are generally interpreted as indicating that harmful multicollinearity is unlikely. The maximum VIF in the full specification is 1.64, and the maximum condition index is 2.26, indicating that coefficient instability due to multicollinearity is not a material concern.

5.2. Baseline Fixed-Effects Results

Table 3 reports the baseline fixed-effects estimates. Columns (1)–(3) progressively add the extended controls, allowing us to assess the stability of the tourism coefficient. The coefficient on tourism activity is positive and statistically significant across all specifications, with point estimates between 0.038 and 0.044, indicating that a one percentage point increase in tourism nights growth is associated with an increase of approximately 0.04 percentage points in real GDP growth, ceteris paribus. The sign and magnitude are consistent with the TLGH and confirm Proposition P1.

Table 3. Baseline and extended fixed-effects estimates (dependent variable: real GDP growth).

Variable	(1) Baseline FE	(2) + Openness	(3) Full FE
Tourism nights growth	0.0384 *** (0.0076)	0.0398 *** (0.0078)	0.0407 *** (0.0081)
Gov. consumption (% GDP)	−0.4125 ** (0.1862)	−0.3942 ** (0.1797)	−0.3654 ** (0.1721)
Investment (% GDP, t − 1)	0.1401 ** (0.0666)	0.1356 ** (0.0651)	0.1289 * (0.0648)
COVID-19 dummy	−0.0458 *** (0.0080)	−0.0461 *** (0.0079)	−0.0463 *** (0.0078)
Trade openness	—	0.0083 * (0.0046)	0.0071 (0.0049)
Tertiary education	—	—	0.0612 * (0.0334)
Institutional quality	—	—	0.0094 (0.0086)
Inflation	—	—	−0.1452 ** (0.0667)
Constant	0.0806 ** (0.0398)	0.0671 * (0.0392)	0.0411 (0.0431)
Observations	270	270	270
Groups (countries)	27	27	27
R2 within	0.488	0.502	0.521
F-statistic	57.08	48.92	37.81
Country FE	Yes	Yes	Yes
Cluster-robust SE	Yes	Yes	Yes

Notes: Standard errors are clustered at the country level in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10. The F-test of joint significance is rejected at p < 0.001 in all columns. The F-test for the joint significance of country fixed effects is also rejected, supporting the within-country specification. The (—) indicates that the variable was not included in that specific model specification.

Among controls, the coefficient on general government final consumption is negative and statistically significant, indicating that excessive public consumption acts as a fiscal drag on growth—a finding particularly relevant for EU economies operating under Stability and Growth Pact constraints. Lagged investment is positive and significant, consistent with standard capital-accumulation channels. The COVID-19 dummy enters with a large negative coefficient (−0.046), confirming the magnitude of the pandemic shock. In the extended specification, inflation exerts a significant negative effect on growth, consistent with the post-2022 inflation episode, while tertiary educational attainment enters with a positive sign, providing initial evidence in favour of the human-capital channel.

Specification diagnostics confirm the validity of the fixed-effects approach. The cluster-robust Hausman (Sargan–Hansen) test rejects the null of orthogonality between the country effect and the regressors with a statistic of 127.92 (4 d.f., p < 0.001). The Mundlak–Chamberlain correlated random-effects cross-check yields a joint Wald statistic of χ²(3) = 10.54 (p = 0.0145), also rejecting the random-effects orthogonality restriction. Both tests indicate that unobserved, time-invariant country characteristics (institutional, structural, geographic) are systematically related to the covariates, and that the fixed-effects specification is the econometrically valid basis for inference.

5.3. Dynamic Panel Estimates (System GMM)

Table 4 reports the two-step System GMM estimates with Windmeijer-corrected standard errors. The lagged dependent variable is positive and highly significant, confirming the persistence of growth. The tourism coefficient remains positive and significant under the dynamic specification. In column (1), the short-run tourism coefficient is β = 0.0211 and the autoregressive coefficient is ρ = 0.265; therefore, the implied long-run effect is β/(1 − ρ) = 0.0211/(1 − 0.265) = 0.0287, approximately 0.029. Relative to the full FE coefficient of 0.0407, the short-run GMM estimates of 0.0211–0.0239 imply that the dynamic adjustment reduces the immediate tourism coefficient by about 40–50%, while leaving the qualitative conclusion unchanged.

Table 4. System GMM dynamic panel estimates.

Variable	(1) Two-Step	(2) Two-Step + IQ	(3) Two-Step Full
GDP growth (t − 1)	0.265 *** (0.078)	0.249 *** (0.081)	0.231 *** (0.083)
Tourism nights growth	0.0211 *** (0.0072)	0.0224 *** (0.0074)	0.0239 *** (0.0078)
Gov. consumption	−0.3214 ** (0.1452)	−0.3098 ** (0.1418)	−0.2876 * (0.1531)
Investment (t − 1)	0.1187 ** (0.0521)	0.1156 ** (0.0518)	0.1098 ** (0.0529)
COVID-19 dummy	−0.0421 *** (0.0072)	−0.0418 *** (0.0071)	−0.0413 *** (0.0070)
Trade openness	—	0.0061 (0.0042)	0.0058 (0.0044)
Institutional quality	—	0.0084 (0.0072)	0.0079 (0.0074)
Tertiary education	—	—	0.0521 (0.0298)
Observations	243	243	243
Groups	27	27	27
Instruments	21	23	25
AR(1) p-value	0.004	0.005	0.006
AR(2) p-value	0.418	0.421	0.443
Hansen J (p)	0.327	0.341	0.352
Diff-Hansen (p)	0.412	0.396	0.421

Notes: Two-step System GMM with Windmeijer-corrected standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10. The instrument matrix is collapsed; tourism activity, government consumption, and investment are treated as predetermined. AR(2) is not rejected in any specification, ruling out second-order serial correlation in the differenced residuals. The Hansen J test does not reject the validity of the over-identifying restrictions. The number of instruments is deliberately restricted below the number of groups. However, column (3) uses 25 instruments for 27 groups, which is close to the conventional upper bound; it is therefore reported for transparency as a full-specification sensitivity check, while the main GMM interpretation relies more heavily on the parsimonious columns (1) and (2) and on consistency with the FE, interaction and threshold results. The (—) indicates that the variable was not included in that specific model specification.

Four points deserve emphasis. First, the dynamic specification confirms the baseline TLGH result (Proposition P1) but adjusts the magnitude downward, indicating that part of what the static model attributes to tourism reflects the autoregressive persistence of growth and possible reverse causality. The decline from approximately 0.041 in the full FE specification to 0.021–0.024 in the System GMM estimates is economically meaningful: expressed consistently, the dynamic adjustment reduces the immediate tourism coefficient by about 40–50% relative to the FE estimate. Second, the implied long-run GMM effect remains positive (approximately 0.029 in column 1), which means that the endogeneity-adjusted estimate is smaller but not negligible. A 10 percentage point acceleration in tourism nights growth is associated with roughly 0.21–0.24 percentage points higher real GDP growth in the short run and about 0.29 percentage points in the long run, compared with about 0.41 percentage points in the static FE model. Third, the GMM diagnostics support the reported specification: AR(2) is not rejected, Hansen J does not reject instrument validity, and the Diff-in-Hansen test supports the additional level instruments. However, the instrument count requires caution: column (3) contains 25 instruments for 27 groups, so it should be read as a sensitivity check rather than as a stand-alone basis for causal inference. Fourth, given the small-N structure of the EU-27 panel, these diagnostics should be interpreted cautiously and jointly with the instrument-count restriction and the robustness checks, rather than as definitive proof of causal identification.

5.4. Moderation: Interactions with Institutional Quality and Human Capital

Table 5 reports the interaction analysis that tests Propositions P2 and P3. We augment the baseline FE model with interaction terms between tourism activity and (i) institutional quality and (ii) tertiary educational attainment.

Table 5. Moderation effects: interactions with institutions and human capital.

Variable	(1) IQ Interaction	(2) HC Interaction	(3) Both
Tourism nights growth	0.0312 *** (0.0089)	0.0286 *** (0.0094)	0.0241 *** (0.0098)
Tourism × Inst. quality	0.0094 ** (0.0044)	—	0.0081 * (0.0046)
Tourism × Tertiary educ.	—	0.0247 ** (0.0114)	0.0213 * (0.0119)
Institutional quality	0.0072 (0.0085)	—	0.0066 (0.0089)
Tertiary education	—	0.0584 * (0.0331)	0.0541 (0.0348)
Gov. consumption	−0.3712 ** (0.1734)	−0.3621 ** (0.1722)	−0.3534 ** (0.1718)
Investment (t − 1)	0.1318 ** (0.0651)	0.1289 ** (0.0648)	0.1267 * (0.0651)
COVID-19 dummy	−0.0459 *** (0.0079)	−0.0461 *** (0.0079)	−0.0457 *** (0.0079)
Observations	270	270	270
R2 within	0.514	0.518	0.531
Joint F (tour + interactions)	29.4 [0.000]	27.1 [0.000]	21.6 [0.000]

Notes: Standard errors are clustered at the country level in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10. All regressors are mean-centred before forming the interaction terms to facilitate interpretation. The F-statistic in the last row tests the joint significance of tourism and its interactions. The (—) indicates that the variable was not included in that specific model specification.

Both interaction terms are positive and statistically significant, providing direct empirical support for Propositions P2 and P3. Quantitatively, a one-standard-deviation increase in institutional quality amplifies the tourism elasticity by approximately 15%, while a comparable increase in tertiary educational attainment raises it by roughly 18%. These magnitudes are economically meaningful: they imply that two otherwise identical Member States differing only in institutional quality (e.g., a Northern versus a Southeastern European economy) would derive substantially different growth dividends from the same tourism boom. This is a result that the standard, non-moderated TLGH framework cannot deliver.

5.5. Sub-Sample Heterogeneity

Table 6 reports the tourism coefficient (and its standard error) from baseline FE regressions estimated on the three sub-samples described in Section 4.4. The results reveal pronounced heterogeneity. The tourism elasticity is highest in emerging tourism destinations, low- to moderate-intensity destinations, and Continental economies; it is substantially smaller, and in some specifications, loses significance, in mature, high-intensity, Mediterranean destinations. These findings are consistent with diminishing returns and overtourism mechanisms.

Table 6. Tourism coefficient across sub-samples (FE estimates).

Sub-Sample	Tourism Coef.	Std. Err.	N
Full sample	0.0407 ***	0.0081	270
Mature tourism (9 countries)	0.0246 **	0.0118	90
Emerging tourism (18 countries)	0.0512 ***	0.0103	180
Mediterranean (9 countries)	0.0241 **	0.0119	90
Continental (18 countries)	0.0498 ***	0.0098	180
High intensity (top tercile, 9)	0.0187	0.0142	90
Low/medium intensity (18)	0.0531 ***	0.0102	180

Notes: Each row reports the tourism coefficient from a separate FE regression on the indicated sub-sample, including all controls from Table 3, column (3). Standard errors are clustered at the country level. *** p < 0.01, ** p < 0.05.

The pattern in Table 6 is consistent with the conceptual framework: in destinations where tourism is still scaling up and where the demand-multiplier and investment-spillover channels have room to operate, the marginal effect on growth is substantial. In mature, high-intensity destinations, congestion effects, housing-price externalities, and labour-market saturation appear to weaken the relationship. This heterogeneity is a key empirical contribution of the paper and the foundation for the differentiated policy recommendations developed in Section 7. These differences can be interpreted through the degree of saturation, seasonality and spatial concentration of tourism. In mature Mediterranean destinations, additional nights tend to be concentrated in already well-known coastal, island and heritage locations. The marginal euro of additional tourism demand therefore faces higher infrastructure, housing and environmental costs than in countries where tourism is still diversifying spatially. Croatia, Malta, Cyprus and Greece illustrate this mechanism particularly clearly: a large share of tourism pressure is concentrated in coastal or island spaces with limited land, water and housing capacity. Austria is a different but equally relevant case, because high tourism intensity is linked to Alpine and urban–cultural destinations where transport, accommodation and environmental carrying capacity are already highly managed. By contrast, the stronger coefficient for emerging and Continental destinations is consistent with a catching-up mechanism. Countries such as Poland, Czechia, Romania, Bulgaria and the Baltic states still have room to transform tourism growth into complementary investment in transport, accommodation quality, digital promotion, creative industries and local supply chains. In these contexts, an increase in nights spent is more likely to activate new domestic value chains rather than merely intensify pressure on already saturated destinations. The Mediterranean/Continental split should therefore not be read as a simple geographic distinction. It captures differences in seasonality, product structure and spatial concentration. Mediterranean tourism is often more dependent on summer coastal demand, while Continental tourism tends to be more diversified across city breaks, business travel, cultural routes, spa and wellness tourism, mountain tourism and domestic weekend travel. This diversification helps explain why the marginal growth effect remains stronger in the Continental group.

5.6. Threshold Effects

Table 7 reports the Hansen-type threshold estimates with tourism intensity (nights per inhabitant) as the threshold variable. A single-threshold model is preferred over no-threshold and double-threshold alternatives based on bootstrap likelihood-ratio tests (LR₁ = 18.4, p = 0.003; LR₂ = 6.7, p = 0.182). The estimated threshold value is 7.42 nights per inhabitant, which separates the EU-27 into a low/moderate-intensity regime (in which tourism elasticity is 0.0524, highly significant) and a high-intensity regime (in which the elasticity drops to 0.0089 and loses significance). This value should be interpreted as an indicative macro-panel benchmark rather than a precise universal policy cut-off, because the threshold depends on the selected indicator, country-level aggregation and the 2014–2024 sample period.

Table 7. Hansen-type panel threshold estimates.

Threshold Variable: Nights per Inhabitant	Coefficient	Std. Err.
Estimated threshold (γ)	7.42	[5.84; 9.17] 95% CI
Tourism × I(q ≤ γ) (regime 1: low/moderate)	0.0524 ***	0.0098
Tourism × I(q > γ) (regime 2: high intensity)	0.0089	0.0142
Difference (regime 1–regime 2)	0.0435 ***	0.0166
Bootstrap LR1 (single threshold) [p-value]	18.4 [0.003]
Bootstrap LR2 (double threshold) [p-value]	6.7 [0.182]

Notes: Hansen (1999, 2017) panel threshold model with country fixed effects. Bootstrap p-values based on 1000 replications. *** p < 0.01.

To avoid over-interpreting the threshold estimate, the following analysis treats 7.4 nights per inhabitant as a warning range rather than a mechanical carrying-capacity rule. The policy relevance of the threshold lies in identifying where the marginal contribution of additional tourism nights becomes weaker, not in imposing a single numerical limit on all EU destinations.

The threshold result provides direct evidence for Proposition P4: the tourism–growth relationship in the EU-27 is non-linear, with a clear inflexion around 7–8 nights per inhabitant. Below this threshold, tourism is a strong growth engine; above it, the marginal contribution is statistically indistinguishable from zero. The country-level results indicate that high-intensity regimes are most likely to characterise Member States such as Croatia, Cyprus, Malta, Greece and Austria. Spain, Italy and Portugal also contain well-known high-pressure sub-national destinations discussed in the overtourism literature, but these regional examples should be treated only as contextual illustrations because the present empirical model is estimated at the country level and does not classify NUTS-2 regions directly. To make the threshold estimate more operational, it is useful to distinguish between countries that are structurally close to or above the high-intensity regime and countries where the same threshold mainly signals future monitoring needs. Table 8 should be interpreted as a country-level reading of the empirical regimes, not as a fixed classification of every destination within each country. The threshold is estimated from country–year panel variation; local carrying-capacity limits may be reached much earlier in specific cities, islands or protected areas.

Table 8. Country-level reading of the tourism-intensity threshold.

Threshold Position	Illustrative Country Examples	Interpretation	Policy Implication
Above or close to the 7.4 nights-per-inhabitant regime	Croatia, Malta, Cyprus, Greece, Austria	Tourism intensity is high relative to the resident population and/or concentrated in coastal, island, Alpine or heritage destinations.	Shift from volume expansion to carrying-capacity management, seasonality reduction, quality upgrading and resident-oriented regulation.
Large countries with regional overtourism hotspots	Spain, Italy, Portugal	National averages may remain below or around the threshold even though cities, islands and coastal regions experience intense pressure.	Use regional and city-level monitoring; avoid interpreting national averages as proof that destinations are not congested.
Low/moderate-intensity regime	Many Central, Eastern and Northern EU Member States	Tourism expansion still has room to generate demand, investment and supply-chain spillovers.	Support sustainable scaling, connectivity, destination branding and local value capture.

This qualification is important because overtourism is inherently spatial. A country-level model can identify macroeconomic non-linearity, but it cannot fully observe visitor congestion in Barcelona, Venice, Dubrovnik, Amsterdam, island destinations or protected natural areas. Conversely, a country classified as high intensity may still contain inland regions with considerable unused tourism potential. The threshold should therefore be used as an early-warning macro indicator and should be complemented by NUTS-2, municipal or destination-level indicators such as visitor density, accommodation pressure, housing-market effects, water stress and seasonal peak loads.

5.7. Robustness Checks

The robustness checks confirm the qualitative conclusions while also clarifying the limits of precision. (i) Replacing the tourism nights measure with the growth rate of international tourist arrivals per capita yields a tourism coefficient of 0.0331 (s.e. 0.0091, p < 0.01), close to the baseline. (ii) Using real GDP per capita growth as the dependent variable produces a tourism coefficient of 0.0386 (s.e. 0.0083), confirming that the result is not driven by population dynamics. (iii) Excluding 2020–2021 to remove the pandemic episode yields a coefficient of 0.0421 (s.e. 0.0089), virtually unchanged, indicating that the baseline relationship is not driven by the extreme observations of the COVID-19 period. (iv) Adding year fixed effects (two-way FE) yields 0.0359 (s.e. 0.0086), confirming that the result is robust to global cyclical components. (v) Formal multicollinearity diagnostics report a maximum VIF of 1.64 and a maximum condition index of 2.26, well below conventional warning levels. (vi) Threshold sensitivity checks in Table 9 show that the non-linearity is qualitatively stable, although the precise threshold value changes with the threshold variable and sample definition.

Table 9. Sensitivity of the threshold result.

Specification	Threshold Variable	Estimated Threshold	Low/Moderate-Regime Tourism Effect	High-Regime Tourism Effect
Baseline	Nights per inhabitant	7.42	0.0524 ***	0.0089
Alternative intensity measure	International arrivals per inhabitant	2.91	0.0476 ***	0.0118
Absorptive-capacity threshold	Institutional quality (WGI avg.)	0.94	0.0251 **	0.0473 ***
Excluding the COVID years	Nights per inhabitant, excluding 2020–2021	7.86	0.0498 ***	0.0104

Notes: Bootstrap inference is applied as in Table 7. The sensitivity checks are intended to assess whether the qualitative non-linearity is stable, not to establish a single invariant numerical cut-off. *** p < 0.01, ** p < 0.05.

The institutional-quality threshold specification in Table 9 deserves separate interpretation. Unlike the tourism-intensity threshold, which captures congestion and saturation, the WGI threshold captures absorptive capacity. The stronger tourism coefficient in the high-institutional-quality regime (0.0473 ***) indicates that tourism growth translates into larger GDP gains once governance exceeds the estimated threshold of 0.94. This reinforces Proposition P2 and suggests that transparent regulation, effective public administration and rule-of-law conditions help convert tourism rents into productive investment and spillovers rather than allowing them to dissipate through leakages, congestion or rent-seeking. This result does not contradict the intensity-threshold finding; rather, it shows that institutional capacity can partly offset diminishing returns where tourism is managed effectively.

Taken together, the multi-method evidence supports a refined version of the TLGH for the EU-27: tourism contributes to growth in advanced European economies, but the magnitude, persistence, and direction of this contribution depend on absorptive capacity, intensity, and destination type.

6. Discussion: Linking Results to Mechanisms

The empirical results map closely onto the five-channel framework developed in Section 2. The positive baseline coefficient on tourism activity captures the joint operation of the demand-multiplier, investment-spillover, foreign-exchange, productivity, and labour-market channels. The downward adjustment observed under the System GMM specification is consistent with growth persistence and partial endogeneity: tourism responds positively to favourable growth episodes, so part of the static coefficient absorbs reverse causation. The difference between the static FE and dynamic GMM estimates is substantively important and is interpreted throughout the paper using the same range: the dynamic adjustment reduces the immediate tourism coefficient by about 40–50% relative to the FE estimate. The FE coefficient captures the contemporaneous within-country association between tourism nights and GDP growth after controlling for observed macroeconomic variables and country fixed effects. The smaller GMM coefficient is more conservative because it accounts for lagged growth and the possibility that better-performing economies also attract more tourism. The preferred interpretation is therefore not that tourism loses importance in the dynamic model, but that tourism is a complementary growth driver whose immediate contribution is positive but more moderate than suggested by static regressions alone.

The interaction analysis provides the clearest mapping to specific channels. The positive interaction with institutional quality (Table 5) confirms that the productivity, investment, and knowledge-spillover channels are conditional on the quality of governance: rents from tourism are more likely to be reinvested into productivity-enhancing infrastructure and training in well-governed economies. The positive interaction with human capital reflects the absorptive-capacity argument: tourism’s productivity-spillover effect is realised only when the workforce can absorb, adapt, and diffuse the practices brought in by international visitors and operators. These findings imply that improving institutions and human capital is not separate from tourism policy—it is its enabling condition.

The threshold and heterogeneity results illuminate the negative-externality side of the framework. Below the estimated threshold of 7.4 nights per inhabitant, the demand-multiplier and investment-spillover channels dominate. Above the threshold, congestion, housing-price inflation, and the diversion of resources away from tradable manufacturing—the tourism-specific Dutch disease—appear to flatten the relationship. This is a finding that aligns with the empirical work of Mihajlović et al. (2025) on environmental thresholds, but extends it by providing a quantitative threshold value that policymakers can use as a benchmark for destination management. The country examples in Section 5.6 help clarify the mechanism behind the threshold result. In Croatia, Malta and Cyprus, high nights-per-inhabitant ratios are amplified by small population size and strong spatial concentration of tourism; in Greece, the island/coastal structure creates local carrying-capacity pressures even when national averages smooth regional variation; in Austria, high tourism intensity is linked to mature Alpine and urban–cultural destinations with advanced but capacity-constrained infrastructure. These examples show why the same increase in tourist nights does not have the same growth meaning across countries.

The COVID-19 dummy enters with a large and stable negative coefficient across the reported specifications, while the sub-sample patterns suggest that the shock may have differed across destination types. However, this evidence remains descriptive and should not be interpreted as a formal test of Proposition P5. A direct test would require additional interaction terms such as COVID-19 × mature tourism destination, COVID-19 × high tourism intensity, or COVID-19 × Mediterranean destination. The present manuscript therefore treats pandemic resilience as an exploratory interpretation and identifies it as a priority for future research rather than as a confirmed contribution.

Taken together, the discussion supports a repositioning of the TLGH in the EU context. Tourism is a growth-enhancing sector under most conditions, but it is not a universal engine: its marginal contribution is moderated by institutions and human capital, bounded by congestion and externality limits, and differentiated across destination types. The standard, undifferentiated TLGH formulation misses these nuances—precisely the gap our framework and multi-method evidence aim to fill.

7. Policy Implications: Segmented Strategies for the EU-27

The standard TLGH literature typically concludes with general recommendations to support tourism investment, innovation, and sustainability. Our results show that such generic advice is unlikely to be useful: the marginal contribution of tourism to growth, and the marginal cost of overtourism, vary systematically across destination types. We therefore organise the policy implications into four segmented strategies, each tied to specific EU funding instruments and destination–management challenges.

7.1. Strategy A—Capacity Management for High-Intensity Mature Destinations

For Member States falling in or close to the high-intensity regime identified by the country-level threshold model (notably Croatia, Cyprus, Malta, Greece and Austria), and for sub-national destinations in Spain, Italy and Portugal that are widely discussed in the overtourism literature but are not directly classified by our model, additional tourism growth may contribute little to aggregate output while generating substantial congestion costs. The policy priority is therefore no longer undifferentiated expansion but carrying-capacity management. For these countries, the empirical result does not imply that tourism should be reduced mechanically. Rather, it suggests that additional public support should be redirected from quantitative expansion toward higher value added per visitor, dispersion of flows, climate adaptation, housing affordability, labour-market quality and environmental carrying capacity. Concrete actions include:

(a)

Visitor-flow regulation through dynamic pricing of access to congested heritage sites, city-centre tourist taxes calibrated to seasonal demand, and timed-entry systems modelled on the existing Acropolis, Alhambra, and Pompeii pilots.

(b)

Geographic dispersal strategies that channel demand toward lesser-known interior and rural destinations through targeted marketing, transport–infrastructure links, and EU Cohesion Policy/Just Transition Fund support.

(c)

Short-term-rental regulation to mitigate housing-price externalities, drawing on the models implemented in Barcelona, Lisbon, and Amsterdam, and aligned with the current EU framework under Regulation (EU) 2024/1028 on data collection and sharing relating to short-term accommodation rental services (European Parliament & Council of the European Union, 2024).

(d)

Investment redirection: In this regime, the marginal productive use of EU funds is no longer additional accommodation capacity but rather quality upgrading, digitalisation, and decarbonisation of existing facilities.

7.2. Strategy B—Acceleration for Emerging Tourism Destinations

For Member States in the low-/moderate-intensity regime (most Central, Eastern, and Northern European countries), the empirical results indicate substantial unrealised growth potential from tourism expansion. The policy priority is to scale up tourism in ways that maximise domestic value capture and minimise leakages. In this group, the main policy risk is the opposite: underinvestment and weak integration with local suppliers may prevent tourism growth from translating into broader regional development. The emphasis should therefore be on building local linkages before large-scale volume growth produces leakages or future congestion. Concrete actions include:

(a)

Supply chain deepening: Targeted support for local food-and-beverage producers, handicrafts, and creative industries integrated with the tourism value chain, using ERDF and ERASMUS+ instruments.

(b)

Destination branding and connectivity: Regional airport networks, high-speed rail extensions financed by CEF Transport, and joint cross-border destinations co-financed by Interreg programmes.

(c)

Quality-assurance frameworks: Adoption of EU eco-label and EMAS certifications across the accommodation sector to position emerging destinations in higher-margin segments from the outset, rather than competing on volume.

(d)

Workforce development: Targeted use of European Social Fund Plus and the Pact for Skills Tourism Partnership to address the structural shortage of mid- and high-skill tourism professionals identified in OECD (2025).

7.3. Strategy C—Institutional and Human-Capital Complementarities

Our interaction estimates show that improvements in institutional quality and human capital amplify the growth contribution of tourism. This implies that tourism policy cannot be designed in isolation from broader governance and education agendas. Three operational implications follow:

(a)

Tourism policy should be tightly coordinated with national rule of law and anti-corruption reforms, particularly in destinations where governance gaps risk diverting tourism rents (NextGenerationEU rule-of-law conditionality is the existing vehicle).

(b)

Vocational training and dual-education systems for hospitality, gastronomy, digital marketing, and destination management should be scaled up using ESF+ and the Pact for Skills, with a particular focus on lifelong-learning participation in southern and eastern Member States.

(c)

Investment in research and innovation infrastructure (Horizon Europe Cluster 4: Digital, Industry and Space; Cluster 6: Food, Bioeconomy) should explicitly include tourism-related themes such as smart destinations, predictive flow management, and circular hospitality.

7.4. Strategy D—Exploratory Post-Pandemic Resilience and Climate Adaptation

The COVID-19 dummy in the regression and the heterogeneous sub-sample patterns suggest that resilience is a destination characteristic rather than an inherent property of the sector. However, in line with the empirical limitation noted above, this section should be read as a policy-oriented extension of the results rather than as a formally tested resilience model. A formal COVID-19 interaction model should be estimated before treating this as a confirmed test of Proposition P5. Building resilience requires:

(a)

Product diversification: Reducing over-reliance on a single source market or a single season through targeted marketing in emerging source countries and through year-round product design (cultural, gastronomic, nature-based, business).

(b)

Digital infrastructure: Smart-destination platforms with real-time flow monitoring, interoperable booking and payments, and EU-wide digital identity integration via the European Digital Identity Wallet.

(c)

Climate adaptation, particularly in Mediterranean destinations facing increasing summer heat stress, water scarcity, and wildfire risk. The Recovery and Resilience Facility (RRF) and LIFE programmes provide direct funding lines.

(d)

Crisis-preparedness frameworks at the destination level, including business-continuity templates, parametric insurance instruments for SMEs, and EU-level early-warning systems for demand shocks.

7.5. EU-Level Coordination

Our threshold and heterogeneity results suggest that an EU-level tourism strategy should explicitly distinguish between expansion-oriented and carrying-capacity-oriented destinations and allocate financial and regulatory instruments accordingly. The proposed Transition Pathway for Tourism developed by the European Commission could be strengthened by (i) introducing a carrying-capacity scoreboard with destination-level indicators (nights per inhabitant, visitor density at peak), (ii) tying access to specific funding lines to the appropriate strategy track, and (iii) creating an EU-wide overtourism early-warning system. These interventions translate the empirical findings of this paper into operational governance tools.

8. Conclusions

This paper revisited the tourism-led growth hypothesis for the EU-27 over 2014–2024 by combining an integrative conceptual synthesis of transmission mechanisms with a multi-method empirical strategy that includes static fixed-effects, dynamic System GMM, interaction, sub-sample, and Hansen-type threshold models. The contribution of the paper is threefold.

First, on the theoretical side, we provide an integrative synthesis of the five channels through which tourism affects growth in advanced European economies and identify the moderating role of institutional quality, human capital, and trade openness. The value of this framework lies in organising existing theoretical insights into an empirically operational structure, rather than in claiming a wholly new theory of tourism-led growth.

Second, on the empirical side, the results confirm a positive and statistically significant contribution of tourism activity to real GDP growth (Proposition P1), but they refine this result in two important ways: the relationship is moderated by institutional quality and human capital (P2, P3), and it is non-linear with an indicative threshold around 7.4 nights per inhabitant (P4). The evidence is also consistent with heterogeneous exposure to the COVID-19 shock across destination types, but because no formal COVID-19 interaction model is estimated, P5 is not reported as a confirmed empirical contribution. The use of System GMM addresses endogeneity and persistence concerns, although its small-N application requires cautious interpretation.

Third, on the policy side, the results yield segmented recommendations tailored to four destination archetypes—high-intensity mature, emerging, institutionally complementary, and resilience-focused—and linked directly to specific EU funding instruments. These recommendations move the policy discussion from generic prescriptions to actionable, destination-specific guidance for tourism ministries, regional authorities, and EU funding agencies.

The paper has several limitations that suggest avenues for further research. First, the threshold estimate is conditional on the selected indicator, the available country-level sample and the 2014–2024 period; it should therefore be interpreted as an indicative warning range rather than as a precise policy rule. Second, although the System GMM estimates reduce concerns about endogeneity and growth persistence, the small number of EU Member States means that finite-sample bias and instrument-validity concerns cannot be fully eliminated. Third, the resilience proposition is not directly tested through COVID-19 interaction terms and should be examined in future work. Fourth, the analysis uses country-level data; a sub-national replication using NUTS-2, municipal or destination-level data would allow a finer characterisation of overtourism dynamics. Finally, incorporating tourism-related carbon emissions, water stress and biodiversity indicators directly into the growth specification would enable a fuller assessment of the sustainability dimension that Section 7 highlights as an emerging priority. The country-level nature of the dataset is especially important when interpreting overtourism. The estimated threshold is a macroeconomic signal, not a destination-level carrying-capacity limit. Future research should therefore replicate the analysis with NUTS-2 or municipal data and, where possible, combine official accommodation statistics with short-term-rental data, mobility indicators and local environmental pressure indicators. Such a design would allow the national growth effects identified here to be linked more directly to the city, island and regional processes through which overtourism is experienced.