Browsing by Author "Lunguli, Amani"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Forecasting inflation in Tanzania.
    (University of Dar es Salaam, 2018) Lunguli, Amani
    The research is based on forecasting inflation in Tanzania using time series techniques. The inflation series across January 2000 to December 2017 was obtained from the Tanzania National Bureau of Statistics (NBS). Two families of time series models autoregressive integrated moving average (ARIMA) models and Exponential smoothing models were fitted to the data. Time series model building strategies as postulated by Box-Jenkins (1976) are explored in detail from a theoretical and practical stand point. The stages in the model building namely, identification, estimation and diagnostic checking are explained explicitly and applied to the data. A best fitting model for each family of models was selected based on how well the model captures the variations in the data (goodness of fit).The goodness of fit was assessed by the use of the Schwartz Bayesian Information Criterion (SBIC) and Akaike Information Criteria (AIC). A seasonal model, SARIMA (1,1,1)×〖(0,0,1)〗_12 was chosen to be the best fitting from the ARIMA family of models, while the Winter’s exponential smoothing with alpha = 0.9999, beta = 0.0711, gamma = 0.0107 was chosen to be the best fitting from the Exponential smoothing models. The selected models were used to compute 24 months forecasts for inflation. Comparisons of the two selected models were carried out based on the goodness of fit and the forecasting power of the two models using RMSE, MAE and MAPE methods selection criteria. It was established that the SARIMA (1,1,1)×〖(0,0,1)〗_(12 )model was superior to the Winter’s exponential smoothing with alpha = 0.9999, beta = 0.0711, gamma = 0.0107 model according to both criteria. Analysis and write-up were done using E-VIEWS, STATGRAPHICS and R software. Presentation and explanations of results were aided by the use of graphs and tables. The model is recommended for use by policy planners since it has minimum error discrepancy of ±1 which follows closely with the actual data.