Naslov
Functional time series approaches to forecast tourist arrivals: The case from Croatia
(Functional time series approaches to forecast tourist arrivals: The case from Croatia)

Autori
Bošnjak, Mile ; Novak, Ivan ; Bašić, Maja

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Proceedings of FEB Zagreb 14th International Odyssey Conference on Economics and Business / Sever Mališ, Sanja ; Načinović Braje, Ivana ; Galetić, Fran - : Ekonomski fakultet Sveučilišta u Zagrebu, 2023, 125-133

Skup
FEB Zagreb 14 th International Odyssey Conference on Economics and Business

Mjesto i datum
Poreč, Hrvatska, 10.05.2023. - 13.05.2023

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
forecasting ; tourism arrivals ; functional time series ; principal components

Sažetak
Modelling and forecasting tourist demand is a contemporary issue in tourism and with it connected international economics. Its importance stems from prediction of exchange rate changes, volatilities in terms on national currency, but also a demand for domestic goods and services, and consequently gross domestic product. As tourist arrivals increase on an annual basis, forecasts of inbound tourist arrivals are a prerequisite for forecasting labour market trends, demand for domestic goods and services, exchange rate volatilities, investment and gross domestic product. Both parametric and nonparametric approaches were used in previous studies, whereby non-linear forecasts and multivariate parametric approaches, methods such as spatial dependence and spatial heterogeneity, or fuzzy time series approach, which predicted tourist arrivals with higher degree of accuracy compared to traditional methods. Contemporary methods include machine learning and internet search approaches. This paper takes on a nonparametric approach and univariate specification previously tested in finances and demographic studies, and not evaluated in cases of tourist arrivals series. This paper aims to evaluate forecasting performance of functional time series approaches to forecast tourist arrivals, i.e., inbound tourism. Monthly data sample of tourist arrivals in Croatia from January 2005 up to December 2017 was considered as a training sample while the data from January 2018 up to December 2019 were considered as a testing sample. Tourist arrivals time series in Croatia exhibits unit root and seasonal unit root, which is confirmed with standard unit root tests and seasonal unit root tests. Mean squared error and root mean squared error was used to evaluate forecast accuracy of the considered specifications. Based on empirical evaluation from this research, functional time series approach outperforms seasonal ARIMA as a benchmark. Considering dynamic updates, which were evaluated for different methods, forecast from penalized least squares were more accurate comparing to block moving method and ridge regression. Empirical results from this paper suggested functional time series approach as a promising alternative to forecast tourist arrivals.

Izvorni jezik
Engleski

POVEZANOST RADA