Modeling the link between tourism and economic development: evidence from homogeneous panels of countries

Tourism has grown rapidly since the mid-20th century and is a major global sector contributing significantly to GDP and employment. While many studies have examined tourism’s link to economic growth, fewer have explored its role in broader economic development, which encompasses not only income but also health, education, and inequality. International organizations posit that tourism can promote socio-economic development, but realizing these benefits requires appropriate policies and initial conditions (e.g., regulation, taxation, infrastructure). This paper investigates whether tourism contributes to economic development in a multidimensional sense, using panel data for 123 countries (1995–2019). It also critically evaluates causal analyses that generalize results from heterogeneous panels by comparing results for the full panel with those for homogeneous groups formed by tourism specialization and development levels. The study aims to clarify whether investments in tourism—often funded by international organizations to improve socio-economic conditions, especially in less developed countries—translate into measurable development gains, and whether conclusions differ across country groups with distinct dynamics.

The literature on tourism and economic development is comparatively sparse relative to work on tourism and economic growth. Theoretically, tourism can foster development by driving growth and enabling policies that improve living conditions. The Human Development Index (HDI) is widely used to operationalize development as a multidimensional construct (health, education, income). Empirical evidence is mixed and often based on heterogeneous panels. Some studies find that tourism promotes development or a bidirectional link (e.g., Nicaragua and Costa Rica; Central America and the Caribbean; various cross-country panels), while others find no relationship (e.g., Poland; a panel of 133 countries). A key critique is that large, heterogeneous panels can yield generalized conclusions driven by a subset of countries. A few studies address this by grouping countries by development or tourism dynamics and report that effects differ by group (e.g., tourism contributing to development mainly in more developed countries or that development conditions tourism growth in high-development contexts). This paper responds to these gaps by analyzing both the whole panel and homogeneous subgroups, highlighting the risk of overgeneralization from heterogeneous samples.

Data: Two indicators are used: (1) Tourism Specialization Rate (TIR), measured as international tourist arrivals per capita (UNWTO for arrivals; World Bank for population), and (2) Human Development Index (HDI) from UNDP. The sample covers 123 countries worldwide from 1995 to 2019 (pre-COVID endpoint due to pandemic disruptions). Variables are used in natural logs (l.TIR, l.HDI) and first differences (d.l.TIR, d.l.HDI) to represent growth. Grouping: Countries are classified into homogeneous groups based on shared dynamics of tourism specialization and development using Brida et al. (2023), with primary focus on three main groups: Group A (high tourism specialization, high development; N=36), Group B (low specialization, high development; N=29), and Group C (low specialization, low development; N=43). Smaller groups and outliers are identified but not central to the analysis. Econometric approach: The study first tests for cross-sectional dependence using Pesaran’s CD test (2004) for the full panel and each group. Given significant cross-sectional dependence, second-generation panel unit root tests (Pesaran CIPS, 2007) are applied, showing variables are I(1) and stationary in first differences. Granger causality in panels is then tested using the Dumitrescu and Hurlin (2012) approach, which allows for heterogeneity across countries. Models are estimated with 1–3 lags, using first-differenced variables to ensure stationarity. To address cross-sectional dependence, p-values are obtained via bootstrap with 500 replications. Causality is examined in both directions: tourism → development and development → tourism, for the full sample and for each homogeneous group.

- Cross-sectional dependence: Significant dependence exists across countries for both tourism and development variables (CD test significant for logs and first differences in full panel and groups). - Stationarity: Pesaran CIPS tests indicate unit roots in levels but stationarity in first differences for both variables across the full panel and all groups (1% significance), so analysis proceeds in differences. - Tourism → Development: For the full panel, the null of no causality is rejected at 3 lags (e.g., Z-bar significant with p≈0.044; Z-bar tilde p≈0.056), indicating one-way causality from tourism specialization to development. By homogeneous groups, this causality holds only for Group C (low tourism specialization, low development) at 3 lags (Z-bar and Z-bar tilde significant at p≈0.01), and is not supported for Groups A or B. - Development → Tourism: For the full panel, no causality is found in any lag specification. By groups, causality from development to tourism is present only in Group A (high specialization, high development) at 2–3 lags when considering the Z-bar tilde statistic (p≈0.014 at 2 lags; p≈0.036 at 3 lags), indicating that higher development levels foster increased tourism specialization in these countries. - Overall: Results demonstrate that conclusions drawn from heterogeneous panels can be misleading; causal relations are group-specific: tourism drives development in low-specialization, low-development countries (Group C), whereas development drives tourism in high-specialization, high-development countries (Group A).

The findings address the central question by showing that tourism can contribute to economic development but not uniformly across countries. In less developed, low-specialization contexts (Group C), increased tourism precedes improvements in development, supporting the rationale for international and national investments in tourism to enhance socio-economic conditions. However, limited existing specialization may constrain the speed and magnitude of development gains, suggesting a need for foundational investments in tourism infrastructure and market positioning. Conversely, in highly developed, highly specialized destinations (Group A), better socio-economic conditions—such as robust infrastructure, health, education, security, and trained human resources—enhance destination attractiveness and increase tourism specialization. The absence of these relationships in other groups underscores the risk of inferring universal effects from heterogeneous panels. The results reinforce the importance of tailoring policy interventions to country characteristics and of conducting causal analyses within homogeneous strata to avoid overgeneralization.

This study contributes by: (1) examining the tourism–development nexus using a multidimensional development measure (HDI) across 123 countries (1995–2019) and (2) contrasting heterogeneous panel results with those from homogeneous country groups. For the full panel, tourism Granger-causes development at longer lags, but subgroup analysis reveals this result is driven by countries with low tourism specialization and low development (Group C). Development Granger-causes tourism only in countries with high development and high tourism specialization (Group A). Policy implications include prioritizing medium- to long-term investments in tourism infrastructure, supply capacity, and international market positioning in low-specialization, low-development countries, and strengthening socio-economic foundations (security, health, transport, digital infrastructure, human capital) to sustain tourism in advanced destinations. Future research should extend analyses to regional/local levels and incorporate additional country-specific factors that may condition these causal links.

- Scope is national; tourism impacts are often regional or local, suggesting the value of subnational analyses using similar homogeneous grouping. - Causality is detected only for specific groups; other unobserved country-specific factors may mediate the tourism–development relationship. - Data limitations necessitated use of TIR and HDI; alternative measures (e.g., tourism receipts, inequality, poverty) are less consistently available across countries and time. - Dumitrescu–Hurlin test indicates presence of causality in at least some units but does not identify which specific countries drive the result.