Forecasting Tourist Arrivals in Tanzania Using INGARCH Models: Accounting for Seasonality and COVID-19 Impacts

Tourism plays a great role in social economic development in many developing countries. It significantly increases foreign exchange profits, accelerate creation of jobs, and promotes infrastructure development. The aim of this study is to predict monthly tourist arrivals in Tanzanian by using Integer-valued GARCH (INGARCH) which is a framework within Time Series Generalized Linear Models (TSGLM). A negative binomial distribution was used in line with INGARCH (1,1) process to account for overdispersion while seasonality was captured through Fourier terms, and the structural shock of COVID-19 was represented via an intervention dummy. Results shown a strong seasonality while the COVID-19 dummy is statistically insignificant suggesting a minimal long-term effects on tourist arrivals in Tanzania. Diagnostic tests such as the Ljung-Box (p-values > 0.25) and residual ACF/PACF plots validate the suitability of the proposed model and confirm the absence of serial correlation. The predictions have indicated a seasonal peak around mid-year and a trough in the early months of the year which are consistent with established tourism patterns. The INGARCH-TSGLM model has shown its ability to effectively capture seasonal dynamics and the overdispersed count structure of tourist arrivals thus providing reliable forecasts for planning.