Regional Determinants of Tourism Seasonality in Mediterranean EU NUTS-2 Regions: A Panel Analysis Using the Gini Coefficient (2020–2024)

1. Introduction

Tourism seasonality — the temporal concentration of tourist demand within specific periods of the year — represents one of the most structurally persistent challenges facing Mediterranean coastal destinations. While seasonal peaks generate substantial short-term economic activity, they simultaneously impose disproportionate pressures on natural environments, local infrastructure, and labour markets, while leaving productive capacity severely underutilized during off-peak periods (Zvaigzne et al., 2022; Turrión-Prats & Duro, 2018). From an economic standpoint, high seasonality reduces destination profitability, constrains long-term investment, and generates precarious employment conditions, while from an environmental perspective it intensifies resource use and degradation within compressed time windows (Duro & Turrión-Prats, 2019; Grossi & Mussini, 2021). Understanding the spatial distribution and structural drivers of seasonality has therefore become an increasingly urgent policy and research priority, particularly in the context of growing sustainability concerns and the ongoing debate about overtourism.

The Mediterranean EU constitutes the most tourism-intensive region in the world, and its destinations consistently record the highest levels of seasonal concentration globally. Duro & Turrión-Prats (2019) demonstrate that, contrary to a declining global trend in tourism seasonality observed since 2011, Mediterranean countries continue to exhibit the highest and increasingly concentrated seasonal demand patterns among all world regions, suggesting that region-specific structural factors resistant to general market forces are at work. At the sub-national level, this concentration is particularly pronounced in coastal and island regions of Croatia, Greece, and Portugal, where the overwhelming majority of overnight stays are recorded within a narrow summer window (Eurostat, 2024).

These patterns underscore that seasonality is not merely a national-level phenomenon but a deeply regional one, shaped by specific geographic, economic, and structural characteristics that vary substantially even within individual countries. Despite the regional nature of seasonality, the majority of existing comparative studies have been conducted at the national level, limiting the granularity of insights available to destination managers and policymakers. Krabokoukis & Polyzos (2024), in their analysis of Mediterranean countries, identify distinct seasonal profiles across national tourism systems but acknowledge that sub-national heterogeneity remains largely unexplored.

Studies that do address the regional dimension tend to focus on single-country contexts: Duro (2016) analyses Spanish provinces, Cuccia & Rizzo (2011) examine Sicilian destinations, and Grossi & Mussini (2021) focus on Veneto regions. Methodologies furthermore vary considerably across studies, limiting cross-regional comparability. Fernández-Morales et al. (2016) propose the Gini coefficient as a standardized and decomposable measure of seasonal concentration, demonstrating its applicability across diverse regional contexts, yet its use as a dependent variable in a multi-country panel regression framework remains among the less explored methodological approaches in the literature. A further gap concerns the post-pandemic period. The COVID-19 pandemic caused an unprecedented collapse of international tourism demand in 2020 and 2021, raising important questions about whether the resulting disruption might have catalysed a lasting structural reconfiguration of seasonal patterns — for instance, through the redistribution of demand across the calendar, shifts in accommodation preferences, or changes in source market composition. Ruggieri & Platania (2024) highlight that the imbalance of tourism demand, specifically in fragile territories such as islands, requires specific policies based on scientific evidence, yet the post-pandemic regional evidence base remains limited.

This study addresses these gaps by examining the level, spatial distribution, and structural determinants of tourism seasonality across Mediterranean EU NUTS-2 regions from Croatia, Spain, Greece, Portugal, and Italy over the 2020–2024 period. Building on the Gini coefficient as a standardized and methodologically robust measure of seasonal concentration (Fernández-Morales et al., 2016; Rosselló & Sansó, 2017), this research employs a balanced panel regression framework to empirically test three hypotheses. Prior empirical evidence consistently demonstrates that destinations with higher shares of international visitors exhibit more pronounced seasonal concentration, as foreign tourists are more constrained by institutional factors such as school holidays and travel costs in their home countries (Rosselló et al., 2004; Turrión-Prats & Duro, 2019). Based on this evidence, the following hypothesis is proposed:

H1: Regions with a higher share of international overnight stays exhibit higher levels of tourism seasonality.

The structure of accommodation supply has been identified as a significant moderator of seasonal concentration. Hotel establishments, owing to their fixed cost structures and professional revenue management capabilities, have stronger incentives to stimulate demand during shoulder and off-peak seasons compared to private rental providers (Cuccia & Rizzo, 2011; Duro, 2016). This leads to the following hypothesis:

H2: Regions with a higher share of hotel accommodation exhibit lower levels of tourism seasonality.

Island destinations are structurally predisposed to higher seasonality due to geographic isolation, limited transport accessibility during off-peak periods, and a stronger dependence on sun-and-sea tourism products that are inherently climate-sensitive (Ruggieri & Platania, 2024; Duro & Turrión-Prats, 2019). This leads to the following hypothesis:

H3: Island regions exhibit higher levels of tourism seasonality than mainland coastal regions.

Two research questions further structure the descriptive component of the empirical analysis:

RQ1: What is the level and spatial distribution of tourism seasonality across Mediterranean EU NUTS-2 regions in the 2020–2024 period?

RQ2: Has the COVID-19 pandemic caused a lasting structural shift in regional tourism seasonality patterns across Mediterranean EU destinations?

The study makes two primary contributions to the existing literature. Methodologically, it operationalizes the Gini coefficient as a dependent variable within a regional panel regression framework — an approach that remains underutilized in tourism seasonality research and provides a replicable template for future comparative studies. Empirically, it contributes a systematic NUTS-2 level comparative analysis of Mediterranean EU seasonality covering the post-pandemic recovery period, generating evidence directly relevant to regional destination management and EU cohesion policy. The remainder of the paper is structured as follows. Section 2 reviews the relevant literature on tourism seasonality, its measurement, and its determinants. Section 3 presents the data, variables, and methodological approach. Section 4 reports the empirical results. Section 5 discusses the findings in relation to the existing literature and their policy implications. Section 6 concludes with a summary of contributions, limitations, and directions for future research.

2. Literature Review

2.1. Tourism Seasonality and Sustainabillity

Tourism seasonality is broadly defined as the systematic temporal imbalance in the distribution of tourist demand across the calendar year, manifested through fluctuations in visitor numbers, overnight stays, tourism receipts, and related economic activity (Zvaigzne et al., 2022). Although the phenomenon is universally recognized across the tourism literature, its causes are multidimensional and interrelated, and no single theoretical framework has achieved universal acceptance (Duro & Turrión-Prats, 2019). The literature conventionally distinguishes between two broad categories of causal factors. Natural seasonality arises from climatic and environmental conditions — most prominently temperature, sunshine hours, and precipitation — which render certain destinations inherently more attractive during specific periods of the year (Zvaigzne et al., 2022). This is particularly relevant for Mediterranean coastal destinations, where the summer concentration of demand is primarily driven by the appeal of warm temperatures and sea conditions that are unavailable during other seasons. Institutional seasonality, by contrast, reflects socially and economically constructed temporal constraints, including the scheduling of school holidays, public holidays, and industrial vacation norms that channel the bulk of tourist flows within a narrow seasonal window (Hartmann, 1986; Zvaigzne et al., 2022; Turrión-Prats & Duro, 2018). Ridderstaat & Croes (2020) propose a systematic framework for classifying the causal factors of tourism demand seasonality, distinguishing between interseason and intraseason effects, and identify climatic perception and price sensitivity as key drivers of peak-season concentration. Institutional factors are widely regarded as both more resistant to policy intervention and more variable across different source markets, which has important implications for destination management.

A third category of causal factors — inertia and social imitation — has received growing attention in more recent literature. Turrión-Prats & Duro (2019) demonstrate that historical patterns of seasonal demand tend to be self-reinforcing, as tourists, businesses, and infrastructure planning decisions adapt to established seasonal rhythms in ways that make structural change difficult. This inertia effect helps explain why Mediterranean destinations have maintained high levels of seasonality despite sustained policy efforts toward diversification, and why external shocks such as the COVID-19 pandemic may not necessarily produce lasting structural changes in seasonal distribution patterns.

The relationship between tourism seasonality and destination sustainability is multifaceted and operates across economic, environmental, and social dimensions. From an economic perspective, high seasonality creates structural inefficiencies: accommodation, transport, and food service operators must invest in fixed capacity sufficient to meet peak demand, while this capacity remains severely underutilized during off-peak periods, reducing annual profitability and constraining access to investment capital (Zvaigzne et al., 2022). Labour market consequences are particularly severe, as the concentration of economic activity in a narrow seasonal window generates widespread temporary and insecure employment, undermines the capacity of firms to retain skilled workers year-round, and limits human capital accumulation in tourism-dependent regional economies (Duro & Turrión-Prats, 2019).

From an environmental perspective, peak-season concentration imposes disproportionate pressures on natural ecosystems, water resources, waste management systems, and coastal areas within compressed temporal windows (Ruggieri & Platania, 2024). Mediterranean coastal and island destinations are particularly exposed to these pressures given their often limited territorial extent and carrying capacity, the sensitivity of their natural assets, and the extreme intensity of summer tourism flows relative to resident population. Zhang & Xie (2023) find empirically that higher tourism seasonality significantly elevates the probability of hotel exit, providing direct financial evidence of the sustainability threat that seasonal concentration poses to accommodation businesses. Ruggieri & Platania (2024) demonstrate that Mediterranean island destinations consistently record among the highest tourism intensity indices in Europe, with summer visitor-to-resident ratios that exceed sustainable thresholds in several cases.

The social dimension encompasses the disruption of local community life during peak season, including congestion, price inflation, noise, and loss of public space access, which accumulate into what the literature increasingly frames as overtourism phenomena (Krabokoukis & Polyzos, 2024). The concentration of environmental and social costs within the peak season thus constitutes a structural sustainability challenge that cannot be adequately addressed through demand-side measures alone. The literature increasingly emphasizes that effective deseasonalization requires structural interventions targeting supply-side composition, market diversification, and regional product development, rather than merely promotional incentives for off-peak travel (Duro & Turrión-Prats, 2019; Zvaigzne et al., 2022). This policy perspective motivates the empirical focus of the present study on the structural determinants of regional seasonality.

2.2. Measuring Tourism Seasonality: Methodological Approaches

A substantial body of methodological literature has developed around the measurement of tourism seasonality, with various indices proposed to capture different aspects of seasonal concentration. The earliest approaches relied on seasonal indices derived from time series decomposition, which identify the systematic seasonal component of a demand series but do not readily lend themselves to cross-destination comparison (Rosselló & Sansó, 2017). The coefficient of variation (CV) has been widely used as a simple and intuitive measure of the relative dispersion of monthly demand, but it is sensitive to distributional shape and does not distinguish between, for instance, a bimodal pattern with two seasonal peaks and a unimodal pattern with a single summer concentration (Duro, 2016). De Cantis, Ferrante, and Vaccina (2011) provide an early systematic application of the Gini coefficient to tourism seasonality measurement in their analysis of Sicilian hotel bed occupancy, demonstrating its analytical value for capturing both pattern and amplitude dimensions of seasonal variation.

The Gini coefficient, originally developed as a measure of income inequality (Lerman & Yitzhaki, 1985, as cited in Fernández-Morales et al., 2016), has emerged as the most methodologically robust synthetic measure of seasonal concentration in the tourism literature. It captures the overall degree of inequality in the distribution of overnight stays across the twelve months of the year, with a value of zero indicating perfect distributional equality and a value approaching unity indicating maximal concentration in a single month. Crucially, the Gini coefficient is insensitive to the particular months in which demand concentrates, making it suitable for cross-destination comparison across locations with different peak seasons, and it admits Lorenz curve visualization that facilitates intuitive interpretation (Fernández-Morales et al., 2016). Fernández-Morales et al. (2016) further demonstrate that the Gini coefficient can be additively decomposed by source market, enabling identification of the markets that contribute most to overall seasonal concentration — a feature directly relevant to destination management.

Building on this methodological foundation, Ferrante et al. (2018) conduct a systematic cross-country analysis of tourism seasonality across European countries using Eurostat data, identifying distinct clusters of national seasonal patterns. Their study demonstrates that Mediterranean EU member states form a distinct cluster characterized by high and summer-concentrated Gini values, in contrast to Northern and Central European destinations that exhibit more balanced seasonal distributions. Ferrante et al. (2018) additionally propose a new composite index that integrates both the pattern and amplitude dimensions of seasonality, arguing that the Gini coefficient alone does not fully capture the shape of the seasonal distribution. Martín Martín et al. (2019) propose the DP2 composite indicator as an alternative multidimensional measure of seasonality, grouping information from multiple seasonal variables, though this approach has seen limited application in cross-country comparative contexts.

While this critique has merit, the Gini coefficient remains the most widely cited and comparable single-measure indicator in the comparative literature, and its use as a dependent variable in panel regression frameworks — as proposed in the present study — has been adopted by Duro (2016) and Turrión-Prats & Duro (2018) in national-level analyses, yet has not been systematically applied at the NUTS-2 regional level across multiple countries. Grossi & Mussini (2021) extend the Gini-based approach by developing a decomposition methodology that enables the attribution of changes in seasonal concentration over time to specific sub-periods, thereby identifying whether concentration is intensifying or moderating in particular parts of the seasonal distribution. Their analysis of Italian NUTS-3 regions reveals substantial within-country heterogeneity in seasonal concentration and its dynamics, reinforcing the argument that national-level analyses mask important regional variation. In the present study, the Gini coefficient is complemented by the Q3 seasonal ratio — the share of annual overnight stays occurring in the third quarter — as a straightforward cross-validation measure that facilitates communication with non-specialist policy audiences.

2.3. Structural Determinants of Tourism Seasonality

The empirical literature on the determinants of tourism seasonality has progressively shifted from descriptive documentation toward explanatory modelling, with panel data approaches playing an increasingly prominent role. Cisneros-Martínez & Fernández-Morales (2015) demonstrate that cultural tourism segments can contribute to desesonalisation in coastal areas by attracting visitors with less climate-sensitive travel motivations, providing early evidence for the role of tourism product composition in moderating seasonal concentration. Rosselló et al. (2004) provide one of the earliest systematic econometric analyses of seasonality determinants, identifying the share of international tourists in total demand, relative prices, and income levels in source countries as significant explanatory variables. Their findings establish that international tourism flows are more strongly subject to institutional constraints — particularly the alignment of holiday calendars across multiple source countries — than domestic flows, making higher international dependency a robust positive predictor of seasonal concentration. Subsequent work has consistently replicated and extended this finding across different destinations and methodological approaches (Turrión-Prats & Duro, 2018; Duro, 2016).

The role of accommodation structure has been examined in several studies, with convergent findings pointing to hotel dominance as a moderating factor on seasonality. Cuccia & Rizzo (2011), in their analysis of Sicilian tourist destinations, demonstrate that areas with higher concentrations of hotel accommodation tend to exhibit lower Gini coefficients relative to areas dominated by private rental and second-home accommodation. The mechanism proposed is primarily supply-side: hotels, as profit-maximizing operators with high fixed costs and professional revenue management capabilities, have strong incentives to attract shoulder-season demand through pricing strategies, targeted marketing, and product diversification. Private rental providers, by contrast, are more likely to be available only during peak periods, amplifying seasonal concentration. Duro (2016) confirms this relationship in the context of Spanish provinces, finding that higher hotel market share is negatively associated with seasonal concentration even after controlling for other demand-side factors.

The geographic dimension of seasonality determinants has received particular attention in the context of island destinations. Ruggieri & Platania (2024) conduct a comprehensive analysis of Mediterranean island tourism from 2008 to 2018, demonstrating that island regions consistently record higher seasonality levels than mainland coastal destinations with comparable climatic profiles. The authors attribute this pattern to a combination of factors including limited transport accessibility outside the summer season, the dominance of sun-and-sea product offerings that are inherently climate-sensitive, relatively undiversified tourism product portfolios, and the high proportion of international visitors whose institutional constraints are particularly binding. These findings provide the theoretical basis for the island geography hypothesis (H3) tested in the present study.

Regional economic development, proxied by GDP per capita, has been hypothesized as a negative determinant of seasonality on the grounds that higher levels of economic development are associated with greater diversification of economic activities, stronger service sector capabilities outside tourism, and a more developed infrastructure for year-round visitor attraction. Turrión-Prats & Duro (2019), in their global analysis, find partial support for this relationship, though the effect is not uniformly significant across specifications. In the present study, regional GDP per capita is included as a control variable rather than as the basis for an explicit hypothesis, given the mixed evidence in the literature. Comparative analyses of tourism seasonality across Mediterranean EU destinations have primarily been conducted at the national level.

Krabokoukis & Polyzos (2024) analyse monthly overnight stay data for Mediterranean countries from 2000 to 2019, identifying three distinct seasonal profile clusters and demonstrating that Croatia and Greece consistently record the most extreme levels of seasonal concentration among Mediterranean EU members, while Spain — primarily owing to the Canary Islands — and Portugal exhibit comparatively more balanced distributions. Their study confirms that Mediterranean seasonality is not a homogeneous phenomenon but varies systematically with structural destination characteristics, and calls for more granular regional analyses to inform effective policy.

Duro & Turrión-Prats (2019), in their global seasonality analysis, similarly highlight the Mediterranean as the most seasonally concentrated tourism region in the world and note a concerning trend of increasing rather than decreasing concentration in Mediterranean countries since 2011, despite sustained policy attention to deseasonalisation at both national and EU level. This trend stands in marked contrast to the global pattern, which shows a modest decline in average seasonality across all world regions over the same period, suggesting that Mediterranean-specific structural factors are reinforcing concentration in ways that general market forces and policy interventions have been unable to overcome. At the sub-national level, existing evidence is more limited.

Duro (2016) analyses Spanish provinces and finds substantial within-country heterogeneity, with island provinces (Balearic Islands, Canary Islands) recording notably higher Gini values than mainland coastal provinces. Vergori & Arima (2022) show that transport mode availability shapes Italian regional seasonality, with air-dependent destinations exhibiting more extreme seasonal peaks than road-accessible alternatives. Cuccia & Rizzo (2011) demonstrate similar patterns within Sicily, with coastal and island municipalities exhibiting more extreme concentration than inland cultural destinations. Grossi & Mussini (2021) find comparable heterogeneity within Italian NUTS-3 regions. However, to the best of the authors’ knowledge, limited research has systematically compared NUTS-2 regional seasonality across multiple Mediterranean EU countries using a harmonized dataset and a unified panel regression framework covering the post-pandemic period — the gap that the present study addresses.

2.4. Post-Pandemic Tourism Seasonality: An Emerging Research Agenda

The COVID-19 pandemic constitutes the most severe disruption to global tourism demand since reliable measurement began, with international arrivals declining by approximately 74% in 2020 relative to 2019 (UNWTO, 2021). Beyond the aggregate collapse in demand, the pandemic raised substantive questions about whether the structural reconfiguration of tourism behaviour during 2020–2021 — including the shift toward domestic tourism, shorter booking windows, and greater preference for outdoor and less crowded destinations — might persist as lasting changes in seasonal distribution patterns following the restoration of unrestricted mobility.

Evidence on post-pandemic seasonality changes at the destination level remains limited, but the available findings are largely consistent with a pattern of temporary disruption followed by structural reversion. Krabokoukis & Polyzos (2024) note that while 2020 and 2021 recorded anomalous seasonal distributions in all Mediterranean countries, primarily driven by the near-complete absence of international arrivals during non-summer months when travel restrictions were most severe, the recovery in 2022 and 2023 was accompanied by a rapid return to pre-pandemic seasonal profiles. This pattern is consistent with the broader literature on the inertia of seasonal demand (Turrión-Prats & Duro, 2019), which suggests that deeply embedded institutional and social factors driving seasonal concentration are resistant to even substantial external shocks. The present study tests this proposition empirically through the inclusion of a COVID-19 period dummy variable in the panel regression framework (RQ2), contributing new evidence at the NUTS-2 regional level where post-pandemic dynamics have not previously been systematically examined

3. Materials and Methods

3.1. Research Design

This study employs a quantitative comparative research design combining two complementary analytical approaches. The first component consists of a descriptive cross-sectional and longitudinal analysis of tourism seasonality across Mediterranean EU NUTS-2 regions for the period 2020–2024, addressing RQ1 and RQ2. The second component consists of a balanced panel regression analysis in which the Gini coefficient of seasonality serves as the dependent variable, enabling the empirical testing of hypotheses H1, H2, and H3 regarding the structural determinants of regional seasonal concentration. The combination of descriptive and inferential approaches is designed to first document the spatial and temporal distribution of seasonality before proceeding to explain its variation through structural factors (Fernández-Morales et al., 2016; Turrión-Prats & Duro, 2018).

The panel structure of the data — with multiple NUTS-2 regions observed over five consecutive years — allows for the control of unobserved region-specific heterogeneity through fixed or random effects estimators, thereby reducing the risk of omitted variable bias that would affect cross-sectional analyses conducted at a single point in time (Wooldridge, 2010). The choice between fixed and random effects specifications is determined empirically through the Hausman (1978) specification test, as detailed in Section 3.4.

3.2. Data Sources and Sample

All data used in this study are drawn exclusively from Eurostat, the statistical office of the European Union, ensuring cross-national comparability through the application of harmonized data collection standards mandated by Regulation (EU) No 692/2011 of the European Parliament and of the Council concerning European statistics on tourism (Regulation (EU) No 692/2011). The use of a single institutional data source eliminates potential inconsistencies arising from differences in national measurement methodologies, which would otherwise compromise the validity of cross-regional comparisons (Ferrante et al., 2018). The primary dataset for the dependent variable is Eurostat’s monthly overnight stays at tourist accommodation establishments at the NUTS-2 regional level (dataset code: tour_occ_nin2m), which provides harmonized monthly data for nights spent by domestic and international guests in registered accommodation establishments. Additional Eurostat datasets are used to construct explanatory and control variables, as specified in Table 1. The study covers 2020–2024, thereby capturing both the pandemic disruption phase (2020–2021) and the subsequent recovery period (2022–2024). The analytical sample consists of NUTS-2 regions in Croatia, Spain, Greece, Portugal and Italy for which complete annual observations can be constructed for the variables used in the balanced panel. Regions with more than one missing year for any model variable are excluded. Because the empirical objective is to compare Mediterranean tourism systems rather than only administrative coastline, the sample also retains selected metropolitan and special-status regions where Eurostat tourism flows provide relevant year-round reference cases. This inclusion is explicitly considered when interpreting the descriptive results.

3.3. Calculation of the Gini Coefficient of Seasonality

The Gini coefficient of seasonality is calculated for each NUTS-2 region i and year t from the twelve-monthly values of overnight stays extracted from dataset tour_occ_nin2m. Following the methodology established by Fernández-Morales et al. (2016) and subsequently applied in comparative tourism studies (Duro, 2016; Rosselló & Sansó, 2017; Turrión-Prats & Duro, 2018), the Gini coefficient is computed from the sorted monthly distribution using the following formula:

Gᵢₜ = 1 − (2/n) · Σᵢ (n − i + 0.5) · (xᵢ / Σx)

(1)

Equation (1): Gini coefficient formula for seasonal concentration. n = 12 (months), xᵢ = overnight stays in month i, sorted in ascending order.

In this formulation, monthly overnight stay values are first sorted in ascending order, and the weighted sum reflects the Lorenz curve area below the line of perfect equality. A value of G = 0 indicates perfectly equal distribution of overnight stays across all twelve months, while a value approaching G = 1 indicates the maximum possible concentration of all overnight stays within a single month. Mediterranean coastal destinations typically record Gini values in the range of 0.40–0.70, with Croatian and Greek island destinations consistently at the higher end of this range (Krabokoukis & Polyzos, 2024).

As a robustness check and to facilitate communication with policy audiences, the Gini coefficient is complemented by the Q3 seasonal ratio, defined as the share of annual overnight stays occurring in the third calendar quarter (July–September). While the Gini coefficient captures the overall degree of distributional inequality across all twelve months, the Q3 ratio provides a direct and intuitive measure of summer concentration that is particularly relevant for the Mediterranean context (Ferrante et al., 2018). Pearson correlation between the two measures is expected to be high, and any systematic divergence between the two indicators will be reported and discussed in the results section.

3.4. Panel Regression Model

The panel regression model specification follows the approach established in the tourism seasonality determinants literature (Rosselló et al., 2004; Turrión-Prats & Duro, 2019), adapted to the NUTS-2 regional level and the specific variables available from Eurostat. The baseline model is specified as follows:

Gᵢₜ = α + β₁ INTᵢₜ + β₂ HOTᵢₜ + β₃ ISLANDᵢ + β₄ GDPᵢₜ + β₅ TIᵢₜ + β₆ COVIDₜ + μᵢ + εᵢₜ

(2)

Equation (2): Baseline panel regression model. i = NUTS-2 region; t = year; μᵢ = region-specific unobserved effect; εᵢₜ = idiosyncratic error term.

The model is estimated under both fixed effects (FE) and random effects (RE) specifications. The FE estimator controls for all time-invariant unobserved regional heterogeneity by demeaning variables within each region, and is consistent regardless of whether the unobserved regional effect μᵢ is correlated with the regressors. However, it cannot identify the coefficient on the time-invariant ISLAND dummy, which is absorbed into the regional fixed effect.

The RE estimator treats μᵢ as a random variable uncorrelated with the regressors, which allows identification of all coefficients including ISLAND, but produces inconsistent estimates if this orthogonality assumption is violated (Wooldridge, 2010). The choice between FE and RE specifications is determined by the Hausman (1978) specification test, which evaluates whether the unobserved regional effects are systematically correlated with the included regressors. Under the null hypothesis of no correlation, the RE estimator is both consistent and efficient; rejection of the null hypothesis indicates that the FE estimator is preferred on consistency grounds (Hausman, 1978). Both sets of estimates are reported in the results section to allow assessment of robustness across specifications. In cases where the Hausman test favours FE estimation but the ISLAND coefficient is of substantive interest, a correlated random effects (Mundlak) approach is applied as an additional robustness check (Wooldridge, 2010). To address potential heteroskedasticity and within-region serial correlation of the error term εᵢₜ, cluster-robust standard errors are applied throughout all specifications, with clustering at the NUTS-2 region level. This approach is consistent with standard practice in panel regression analyses of tourism data and ensures that inference remains valid even when the residuals are not independently and identically distributed within regions across time (Wooldridge, 2010). Variance inflation factors (VIF) are calculated to assess potential multicollinearity among the independent variables prior to estimation.

3.5. Construction of the ISLAND Variable

The ISLAND binary variable is manually coded based on the official Eurostat NUTS-2 classification and geographic definitions. A NUTS-2 region is coded as ISLAND (= 1) if it corresponds to a territory that is geographically fully or predominantly island-based, such that the majority of its tourism activity occurs on island territory without a continuous land connection to the mainland of the respective country. Based on this criterion, the following regions in the sample receive an ISLAND coding of 1: Notio Aigaio and Voreio Aigaio in Greece, Illes Balears and Canarias in Spain, Sardegna and Sicilia in Italy, and Regió Autonoma dos Açores in Portugal. All other Mediterranean coastal NUTS-2 regions in the sample, including Jadranska Hrvatska, receive an ISLAND coding of 0, notwithstanding the presence of offshore islands within their territory, as their predominant land mass and tourism infrastructure are mainland-based. This coding decision is conservative: NUTS-2 regions that contain islands but are not predominantly island-based, such as Jadranska Hrvatska, are treated as mainland regions. The implication is that any estimated island effect should be interpreted as the effect of predominantly island-based NUTS-2 territories rather than the presence of island tourism within a broader coastal region.

4. Results

4.1. Descriptive Statistics

The final balanced panel comprises 305 region-year observations across 61 NUTS-2 regions from five Mediterranean EU member states: Croatia (4 regions), Spain (19), Greece (13), Portugal (4) and Italy (21). The dependent variable, the Gini coefficient of monthly overnight stay distribution (G), ranges from 0.039 to 0.742 across the full sample, with a mean of 0.397 and a standard deviation of 0.162, indicating substantial cross-regional heterogeneity in seasonal concentration. The mean share of international overnight stays (INT) is 61.3%, with variation from 4.3% in strongly domestic regions to 95.4% in highly internationalised destinations. The hotel accommodation share (HOT) averages 53.6%, but ranges from 15.8% to 100%, reflecting highly heterogeneous accommodation structures across the five countries. Regional GDP per capita (PPS) ranges from 39% to 166% of the EU average. Tourism intensity (TI) varies from 434 to 127,157 overnight stays per 1,000 inhabitants, illustrating the sharp differences in tourism pressure between metropolitan, coastal and island regions.

4.2. Gini Coefficient of Seasonality: Cross-Regional Distribution

The Gini coefficients calculated from monthly overnight stay data reveal substantial spatial heterogeneity in seasonal concentration across the 61 Mediterranean EU NUTS-2 regions. Country-level averages for the 2020–2024 period are presented in Table 2. The full region-year dataset should be retained as an online appendix or supplementary file when submitting the article.

Greece records the highest mean Gini coefficient (0.484), followed by Italy (0.446) and Croatia (0.358), while Spain (0.312) and Portugal (0.301) exhibit comparatively lower levels of seasonal concentration. The higher mean for Spain and Portugal compared to regional minima reflects within-country heterogeneity, with year-round island destinations (Canarias, Madeira, Açores) substantially reducing national averages. The Q3 ratio confirms these patterns: Greek regions concentrate on average 56.8% of annual overnight stays within the three summer months, compared to 34.8% in Portugal.

At the individual region level, the five most seasonally concentrated regions across the full observation period are: HR03 — Jadranska Hrvatska (mean Gini = 0.664, mean Q3 = 76.6%), ITF6 — Calabria (0.652, 77.8%), EL62 — Ionia Nisia (0.632, 72.1%), ITG2 — Sardegna (0.605, 71.4%), and EL42 — Notio Aigaio (0.578, 64.1%). Of these five regions, three are island or predominantly island territories (ITG2, EL42, and the Ionian Islands in EL62). Jadranska Hrvatska (HR03) is the most extreme non-island case, recording the highest Gini coefficient in the entire sample despite not meeting the island geography criterion defined for hypothesis testing. The five least seasonal regions are dominated by Spanish metropolitan and special-status territories with year-round cultural, business, and administrative tourism: ES64 (Ceuta, mean Gini = 0.160), ES30 (Comunidad de Madrid, 0.169), HR05 (Sjeverna Hrvatska including Zagreb, 0.190), ES63 (Ciudad Autónoma de Melilla, 0.201), and ES70 (Canarias, 0.221), the latter benefiting from its year-round climatic appeal.

4.3. Temporal Trends in Seasonality (2020–2024)

The temporal evolution of seasonality across the five-year observation window is dominated by the structural disruption of the COVID-19 pandemic in 2020 and 2021, followed by a rapid return to pre-pandemic patterns. In 2020 and 2021, Gini values across all countries increased substantially relative to 2019 benchmarks, reflecting the near-complete collapse of international tourism during these years and the disproportionate absence of demand during shoulder and off-peak months. This concentration effect is captured by the positive and statistically significant COVID dummy coefficient in the regression model (see Section 4.5).

From 2022 onwards, seasonality patterns reverted rapidly toward pre-pandemic profiles. By 2023, mean Gini values across all five countries had returned to levels broadly comparable to pre-pandemic observations reported in the literature for these destinations (Krabokoukis & Polyzos, 2024). This reversion supports a preliminary answer to RQ2: the COVID-19 pandemic did not catalyse a lasting structural reconfiguration of seasonal demand patterns, as regional tourism systems reverted to their structural equilibria within approximately two years of the restoration of unrestricted mobility. The 2024 data confirm the stability of this post-pandemic seasonal equilibrium.

4.4. Island vs. Mainland Regional Seasonality

A descriptive comparison between island and mainland NUTS-2 regions in the sample provides preliminary support for H3. The nine island regions identified in the sample — Notio Aigaio (EL42), Voreio Aigaio (EL41), Kriti (EL43), Illes Balears (ES53), Canarias (ES70), Sicilia (ITG1), Sardegna (ITG2), Açores (PT20), and Madeira (PT30) — record a mean Gini coefficient of 0.453 compared to 0.387 for the 52 mainland coastal regions, a difference of 0.066 Gini units. However, island regions exhibit considerable internal heterogeneity: Canarias (ES70) records among the lowest Gini values among island regions in the entire sample (mean 0.221) owing to its year-round climate appeal, while Notio Aigaio and Voreio Aigaio record among the highest. This within-group variation underscores the importance of controlling for additional structural factors in the regression framework.

4.5. Panel Regression Results

Prior to model estimation, variance inflation factors (VIF) were calculated for all independent variables. All VIF values fall below the conventional threshold of 10 (INT: 1.81; HOT: 1.11; GDP_log: 1.11; TI: 1.66; COVID: 1.11; ISLAND: 1.46), indicating the absence of problematic multicollinearity. The Hausman specification test yields a test statistic of χ²(5) = 27.967 (p < 0.001), rejecting the null hypothesis of no systematic correlation between regional effects and the regressors. Accordingly, the fixed effects (FE) estimator is preferred on consistency grounds (Hausman, 1978; Wooldridge, 2010), and FE estimates are treated as the primary specification. Random effects (RE) estimates are reported alongside for comparison. As the time-invariant ISLAND variable is absorbed into regional fixed effects under the FE specification, its coefficient is estimated exclusively in the RE model and interpreted with appropriate caution given the rejected null of the Hausman test.

Table 3 reports the regression results for both specifications. Under the FE estimator, the hotel accommodation share (HOT) is negative and statistically significant (β = −0.0038, p = 0.029), supporting H2: regions in which hotels account for a larger share of overnight stays exhibit systematically lower Gini coefficients, net of within-region changes in other covariates. Regional GDP per capita (log-transformed) is also negative and highly significant (β = −0.5245, p = 0.004), suggesting that increases in relative economic development are associated with lower seasonal concentration. The COVID-19 dummy is positive and highly significant (β = 0.1395, p < 0.001), confirming that the pandemic years 2020–2021 were associated with elevated Gini values. The share of international overnight stays (INT) is positive but statistically insignificant under the FE specification (β = 0.0006, p = 0.553), meaning that H1 is not supported in the within-region model. Tourism intensity (TI) is also insignificant under FE (p = 0.222); because TI is measured on a large numerical scale, its coefficient rounds to approximately zero in the table and should be interpreted through its statistical significance rather than its rounded magnitude.

4.6. Hypothesis Testing Summary

Table 4 Summarises the outcomes of the three hypotheses tested in this study.

H2 is supported: the hotel accommodation share emerges as a robust negative determinant of seasonal concentration under both FE and RE specifications, consistent with findings of Cuccia & Rizzo (2011) and Duro (2016). H1 is not supported under the FE estimator: within-region variation in the international tourist share does not predict changes in Gini, suggesting that the cross-sectional association between internationalisation and seasonality documented in the literature (Rosselló et al., 2004) may reflect stable between-region differences rather than a causal within-region mechanism. H3 cannot be tested under the FE model due to the time-invariant nature of the island dummy; under RE, the ISLAND coefficient is positive but falls short of conventional significance thresholds (β = 0.044, p = 0.112), providing only weak and inconclusive evidence for the island geography hypothesis.

4.7. Robustness Check

As a robustness check, the FE model was re-estimated using the Q3 seasonal ratio (share of annual overnight stays in July–September) as an alternative dependent variable in place of the Gini coefficient. The direction and significance of all coefficients remain consistent with the primary specification: HOT is negative and significant (β = −0.074, p < 0.05), GDP_log is negative and significant, and COVID is positive and highly significant. The consistency of findings across both seasonality measures reinforces the validity of the primary results and confirms that the identified relationships are not an artefact of the specific measurement approach adopted.

5. Discussion

Addressing RQ1, the analysis reveals that tourism seasonality across Mediterranean EU NUTS-2 regions in the 2020–2024 period is both high in absolute terms and markedly heterogeneous in spatial distribution. Mean Gini coefficients across all region-year observations range from 0.039 (ES70 — Canarias, 2023) to 0.742 (HR03 — Jadranska Hrvatska, 2020), with regional mean values ranging from 0.160 (ES64 — Ceuta) to 0.664 (HR03), and an overall sample mean of 0.397, placing the Mediterranean EU firmly within the high-seasonality cluster identified at national level by Duro & Turrión-Prats (2019) and Krabokoukis & Polyzos (2024). The spatial distribution is not random: coastal and island destinations in Greece and southern Italy systematically record higher Gini values than their mainland counterparts in Spain, Portugal, and northern Italy, while metropolitan regions with diversified tourism bases consistently record the lowest seasonal concentration in the sample. Within-country heterogeneity is at least as large as between-country heterogeneity, confirming that national-level analyses are insufficient for understanding the spatial dynamics of Mediterranean seasonality and validating the NUTS-2 regional approach adopted in this study.

The descriptive findings confirm and extend existing evidence on the spatial distribution of tourism seasonality across Mediterranean EU destinations. The country-level ranking — Greece, Italy, Croatia, Spain, Portugal in descending order of mean Gini — is broadly consistent with national-level analyses reported by Krabokoukis & Polyzos (2024) and Duro & Turrión-Prats (2019), providing cross-validation for the Gini-based measurement approach applied here at the NUTS-2 regional level. However, the sub-national analysis reveals important patterns that national averages obscure. The high mean Gini for Italy (0.446) is driven predominantly by southern coastal regions such as Calabria (ITF6, Gini = 0.652) and Sardegna (ITG2, 0.605), while northern Italian metropolitan regions such as ITC1 (Piemonte) and major urban centres record lower seasonality relative to southern coastal regions, though they do not reach the low values observed in year-round destinations such as the Canary Islands or Ceuta. Similarly, Spain’s comparatively low mean Gini (0.312) reflects the moderating influence of the Canary Islands (ES70, mean Gini = 0.221), whose year-round climate effectively distributes demand across all twelve months and substantially reduces the national average. The case of Jadranska Hrvatska (HR03) merits particular attention. With a mean Gini of 0.664 and a Q3 ratio of 76.6%, this region records the highest seasonal concentration among all non-island NUTS-2 regions in the sample and ranks first overall in four of the five observation years. This finding is consistent with Eurostat’s own assessments of Mediterranean regional seasonality (Eurostat, 2024) and underscores the structural nature of Croatia’s seasonality challenge. The region’s extreme concentration — nearly four-fifths of annual overnight stays occurring in a single quarter — is particularly notable given Croatia’s high international tourism dependency (INT ≈ 90%), its dominance of private and holiday accommodation over hotels, and its geographic configuration as a long, narrow coastal region with limited inland tourism infrastructure. These structural characteristics, identifiable through the panel regression framework, collectively explain why deseasonalisation efforts to date have produced limited results.

The confirmation of H2 — that regions with higher hotel accommodation shares exhibit lower seasonal concentration — constitutes the central policy-relevant finding of this study. The FE estimate (β = −0.0038, p = 0.029) implies that a ten percentage point increase in hotel market share is associated, on average, with a reduction in the Gini coefficient of approximately 0.038 units, holding all other within-region factors constant. This effect, while modest in absolute terms, is economically meaningful given the typical Gini range of 0.30–0.70 observed across Mediterranean coastal regions. The result replicates and extends findings from single-country analyses by Cuccia & Rizzo (2011) for Sicily and Duro (2016) for Spanish provinces to a broader multi-country regional context, enhancing the generalizability of the accommodation structure hypothesis.

This result is further supported by Zhang & Xie (2023), who demonstrate that seasonality-exposed hotels face significantly elevated exit probabilities, confirming the two-way relationship between accommodation structure and seasonal concentration. The mechanism underlying this relationship is twofold. On the supply side, hotel operators — as profit-maximizing enterprises with high fixed costs — have strong incentives to deploy revenue management strategies, targeted promotional campaigns, and diversified product offerings to stimulate demand during shoulder and off-peak seasons. Empirical evidence from the yield management literature supports the capacity of hotels to shift demand across the seasonal distribution through differential pricing and package design (Turrión-Prats & Duro, 2018). On the demand side, hotel accommodation is more likely to attract business travellers, conference participants, and cultural tourists whose travel timing is less constrained by climatic and institutional factors than that of leisure-oriented beach tourists who constitute the primary clientele of private rental accommodation.

The disproportionately high share of holiday home and private rental accommodation in Croatia, Greece, and southern Italy — precisely the destinations with the highest Gini values — is thus not merely a symptom of seasonality but an active structural driver of it. From a policy perspective, this finding suggests that structural interventions targeting the accommodation composition of highly seasonal regions may be more effective in reducing seasonal concentration than demand-side promotional campaigns for off-peak travel. Policies that incentivize hotel development, facilitate the conversion of private rental properties to hotel-standard accommodation, or support the development of year-round tourism products compatible with hotel distribution channels could contribute to measurable reductions in Gini-measured seasonality over medium-term horizons.

This conclusion is consistent with recommendations emerging from the broader Mediterranean tourism sustainability literature (Zvaigzne et al., 2022; Ruggieri & Platania, 2024). The failure to confirm H1 under the fixed effects specification — the hypothesis that higher international tourist share drives higher seasonal concentration — represents a substantively important finding that warrants careful interpretation. At first glance, this result appears to contradict the established literature, in which the share of international visitors has been consistently identified as a positive determinant of seasonality (Rosselló et al., 2004; Turrión-Prats & Duro, 2019). The key to reconciling this apparent contradiction lies in the distinction between within-region and between-region variation that characterises fixed effects estimation.

The FE estimator identifies the effect of changes in INT within a given region over the five-year observation period. Given the relative stability of international tourism dependency within regions over such a short temporal window, the within-region variation in INT is limited, reducing statistical power and potentially rendering the coefficient insignificant even if the underlying relationship is genuine. The cross-sectional association between high internationalisation and high seasonality is well-established and visible in the descriptive results — the most international destinations (HR03, EL42, EL41) are also among the most seasonal — but this between-region relationship is absorbed by the regional fixed effects and cannot be estimated within the FE framework. The insignificance of INT under FE therefore reflects the estimator’s mechanics rather than the absence of a substantive relationship, and should not be interpreted as evidence against the hypothesis in a broader conceptual sense. Future research employing longer time series or exploiting more pronounced within-region variation in internationalisation could provide a more definitive test of the H1 proposition.

The inability to confirm H3 within the fixed effects framework is an inherent methodological limitation rather than a substantive finding: since island geography is a time-invariant characteristic of each NUTS-2 region, its effect is completely absorbed by the regional fixed effects and cannot be separately identified. Under the random effects specification, the ISLAND coefficient is positive (0.044) but falls short of the conventional significance threshold (p = 0.112), providing only inconclusive evidence. However, this RE estimate must be interpreted with caution given that the Hausman test rejects the RE model’s consistency assumption, suggesting that the island effect may be confounded with other region-specific unobserved factors correlated with the regressors.

Agius & Briguglio (2021) demonstrate in the Maltese context that ecotourism development can contribute to reducing island seasonality by attracting nature-oriented visitors whose travel timing is less constrained by summer peak demand, while Ruggieri et al. (2022) show in a panel analysis of Mediterranean island destinations that community-based tourism models are associated with lower seasonal concentration than industrial tourism models. The descriptive evidence nonetheless provides substantive support for the island geography proposition. Island regions in the sample record a mean Gini of 0.453 compared to 0.387 for mainland regions, a difference of 0.066 units that is consistent in direction with H3.

The Canary Islands (ES70) constitute the principal exception: their year-round climate and high share of international charter tourists from Northern European markets whose holiday calendars are more evenly distributed produce a mean Gini of 0.221 — the lowest among island regions in the sample — despite their island geography. This exception highlights that island geography per se does not mechanically generate high seasonality; rather, it creates conditions of transport dependency and product specialisation that tend toward seasonal concentration in the absence of countervailing factors such as climate differentiation or market diversification. The findings of Ruggieri & Platania (2024), who document significantly higher seasonality in Mediterranean island destinations using a different methodological approach over an earlier observation period, are broadly consistent with this interpretation. The negative and highly significant coefficient on regional GDP per capita (log) under both FE (β = −0.525, p = 0.004) and RE (β = −0.161, p < 0.001) specifications indicates that more economically developed regions exhibit lower seasonal concentration, net of other covariates. This finding is consistent with the theoretical proposition that higher regional economic development is associated with greater diversification of tourism product supply, stronger institutional capacity for year-round destination management, more developed infrastructure for non-beach tourism (cultural, gastronomy, conference, wellness), and a higher proportion of domestic visitors whose travel patterns are less tightly constrained by summer calendars. The larger FE coefficient relative to RE suggests that within-region economic development trajectories — reflecting improvements in regional GDP over time — have a particularly strong association with reductions in seasonal concentration, which is an encouraging finding from a long-term policy perspective: regions that successfully develop their economies may also simultaneously reduce their seasonality as a secondary benefit of diversification.

The positive and highly significant COVID-19 dummy coefficient (FE: β = 0.140, p < 0.001; RE: β = 0.163, p < 0.001) confirms that the pandemic years 2020–2021 were associated with substantially elevated Gini values across all regions in the sample. This is consistent with Duro et al. (2021), who demonstrate that Mediterranean coastal and island destinations exhibited the highest vulnerability to the COVID-19 shock due to their extreme dependence on international tourism and concentrated seasonal demand structures, all else equal. This finding is mechanically consistent with the near-complete collapse of international tourism demand during these years: since international visitors are concentrated in summer, their absence disproportionately affected non-summer months, paradoxically increasing the relative concentration of the remaining (predominantly domestic) demand in the summer period. The result thus reflects a disruption-induced increase in apparent seasonality rather than a genuine structural shift toward more seasonal behaviour.

The rapid reversion of Gini values to pre-pandemic levels from 2022 onwards — visible in the temporal trend analysis and confirmed by the stability of 2023 and 2024 values — provides a clear empirical answer to RQ2: the COVID-19 pandemic did not produce a lasting structural shift in regional tourism seasonality patterns across Mediterranean EU destinations. This finding is consistent with the theoretical prediction derived from the inertia literature (Turrión-Prats & Duro, 2019) that deeply embedded institutional and structural factors driving seasonal concentration are resistant to even substantial external shocks. From a policy perspective, this result is sobering: it suggests that the pandemic — the most severe disruption to international tourism in the modern era — represented a missed opportunity for lasting deseasonalisation, as market forces and structural factors rapidly reasserted the pre-existing seasonal equilibrium once mobility restrictions were lifted.

6. Conclusions

This study examined the level, spatial distribution, and structural determinants of tourism seasonality across 61 Mediterranean EU NUTS-2 regions from Croatia, Spain, Greece, Portugal and Italy during the 2020–2024 period. Using harmonised Eurostat data and a balanced panel framework, the analysis operationalised the Gini coefficient as the dependent variable in a regional panel regression to identify the structural factors driving cross-regional variation in seasonal concentration.

The findings make two principal contributions to the tourism seasonality literature. Methodologically, the study demonstrates that the Gini coefficient — previously used mainly as a descriptive measure — can be operationalised as a dependent variable within a multi-country NUTS-2 panel framework, providing a replicable analytical template for future comparative destination research. Empirically, the analysis generates the first systematic regional comparison of Mediterranean EU tourism seasonality covering the post-pandemic recovery period, with three central findings: hotel accommodation share emerges as a robust negative determinant of seasonal concentration (supporting H2), international tourism dependency does not predict within-region changes in seasonality under fixed effects estimation (H1 not supported), and island geography effects remain inconclusive due to methodological constraints associated with time-invariant variables (H3 inconclusive). The COVID-19 pandemic temporarily elevated seasonal concentration during 2020–2021 but did not produce a lasting structural shift, with regional patterns reverting to pre-pandemic equilibria from 2022 onwards.

For destination management and EU regional policy, the implications are direct. The robust hotel share effect indicates that structural interventions targeting accommodation composition — incentives for hotel development, support for upgrading private rental capacity, and product diversification programmes — are likely to produce more durable seasonality reductions than demand-side promotional campaigns alone. The rapid post-pandemic reversion further suggests that external shocks, however severe, do not by themselves drive structural change: deliberate policy interventions remain necessary. The case of Jadranska Hrvatska, recording the highest seasonal concentration among all non-island regions in the sample, illustrates the magnitude of the challenge facing destinations dominated by private accommodation and international leisure demand.

Several limitations should be acknowledged. The five-year observation window is relatively short for panel analysis, and the inclusion of two pandemic years constrains within-region variation available for fixed effects estimation. Eurostat accommodation statistics exclude short-term rentals intermediated through platforms such as Airbnb and Booking.com, which may lead to underestimation of seasonal concentration in regions with high platform penetration, particularly Croatia and Greece. Furthermore, the geographic scope is limited to five Mediterranean EU member states; the generalisability of findings to non-EU Mediterranean destinations or to other high-seasonality contexts warrants further investigation.

Future research should extend the temporal scope as longer post-pandemic time series become available, allowing more definitive testing of whether 2020–2024 represents a return to the prior equilibrium or the establishment of a new structural baseline. Incorporating platform-based accommodation data — once harmonised statistics become available at the regional level — would substantially improve the precision of seasonality measurement in regions where short-term rentals dominate. Finally, complementing the panel approach with case-study evidence from specific high-seasonality destinations such as Jadranska Hrvatska could illuminate the institutional and behavioural mechanisms through which structural determinants operate, providing finer-grained insights to inform regionally tailored policy interventions.