Geographical and environmental dependencies in multivariate count models: Recreation demand for Italian national parks

We propose an innovative spatial-based multivariate Generalized Poisson regression model to explain the number of outdoor recreation trips to protected areas. The innovation consists of capturing the impact of geographical and environmental similarities across destinations. As a proof of concept, the approach is applied to data obtained from an online survey with 794 respondents sampled from across Italy between October and November 2022. The attention is placed on assessing travellers' preferences for four popular Italian National Parks: the Parco Nazionale delle Dolomiti Bellunesi, the Parco Nazionale dell'Appennino Tosco-Emiliano, the Parco Nazionale delle Cinque Terre, and the Parco Nazionale dell'Abruzzo, Lazio e Molise. The results corroborate the existence of strong spatial and environmental correlation patterns across the nature sites under assessment. Further, the multivariate approach enhances the predictive accuracy of trip count frequencies relative to simpler models used for comparison.