Abstract:
Recent records show that the price of gold has been rising at a higher rate than in the past. This has been shown to be true for Sri Lankan gold prices as well. In this study an attempt has been made to develop a forecasting model for gold price and to examine the relationship between selected factors, that is the inflation rate, exchange rate and gold price. The data was mined from the World Gold Council and the Central Bank of Sri Lanka. The sample data of gold price were gathered from 2007 January to 2016 March in the currency of US dollars per troy ounce. It was converted into Sri Lankan rupees per 22 carat. Data until December 2015 were used to build the ARIMA model and the VEC model remainder was used to forecast the gold price and to check the accuracy of the model. Box-Jenkins, Auto Regressive Integrated Moving Average methodology (ARIMA) has been used to developed the model 𝐷[𝐿𝑛[𝐺𝑂𝐿𝐷 𝑃𝑅𝐼𝐶𝐸]]; with terms AR (3) and MA(3) and to forecast the future gold price. The MAPE value of fitted data in the appropriate model is 9.4%. To identify the relationship with gold price, inflation rate and exchange rate, quarter value data of all three factors were used. Two models were developed by based on the minimum AIC and the minimum SIC values. Firstly, the stationarity of the data is checked through the Augmented Dickey Fuller test and then the Johansen co-integration test and the Vector error correction model (VECM) are employed for analysis. The results of the Johansen co-integration test revealed that exchange and inflation rates are co-integrated with the gold price that led to run VECM. The VEC model developed for minimum AIC value provides evidence for the existence of long run and short run relationships between the gold prices, the exchange rate and the inflation rate and the model developed for minimum SIC value as well. The model developed based on minimum SIC value is rejected since the existence of serial correlation. The speed of adjustment to equilibrium is 12.1%, the model explains the gold price of the current quarter as 69.3% of the gold price of the previous quarter, and the exchange and inflation rates in the VEC model developed based on minimum AIC value. The MAPE value of fitted data from appropriate VEC model is 6.36%. When forecasting time period is increasing the percentage error in ARIMA model is higher than the percentage error increasing in appropriate VEC model. According to the mean absolute percentage error as forecasting accuracy measure the study concluded that the VEC model is more appropriate fitted model to forecast the gold price in Sri Lanka than the fitted ARIMA model.
