MSc

The present study entitled “PREDICTION OF FUTURES PRICES OF RUBBER” was conducted with the following objectives

i) to examine the price movements of rubber futures through NMCE;
ii) to predict the rubber futures prices and
iii) to compare the forecasting performance of univariate and multivariate models.

Futures trading perform two important functions - price discovery and hedging of price risk, hence an effort to predict the futures prices of rubber, a predominant crop of Kerala, is of contemporary significance to the rubber growers and traders.

The study was based on secondary data. Futures prices of daily open, low, high, close and spot and volume traded of rubber were collected for a period from April 2003 to August 2008 from National Multi Commodity Exchange. The daily data were converted into monthly averages for the analysis. The price movements of rubber futures have been examined using ordinary line graph, correlation, candlestick chart, Compound Annual Growth Rate (CAGR) and ANOVA. Correlation has been found inorder to measure the relation between the domestic rubber prices and the crude oil prices. ANOVA was used to find the significance in the growth rate of rubber prices over different time periods. Prediction of futures prices of rubber has been done using the Multiple Linear Regression, Principal Component Analysis and ARIMA and the results of these models were compared to measure the forecasting performance of these models.

The price movements of rubber futures using the line graph showed that both the spot and futures prices were highly related and hence prediction of one with the other is possible. Rubber is an internationally traded commodity and the hike in the rubber prices globally influence the domestic the rubber pries. Moreover the rise in the crude oil prices influenced the natural rubber prices, since the movements of domestic rubber prices and the crude oil prices showed a positive correlation. The volumes traded were also fluctuating over the years. The ban on futures trading in rubber drastically reduced the volume traded due to loss of investors’ confidence. Candlestick chart showed that the prices were fluctuating with bullish, bearish and neutral trend. Even though the rubber prices increased, the growth rate of rubber prices and volume traded over the years revealed a lower annualized gain making it clear that there was no abnormal hike in the rubber prices. Hence the rise in rubber prices cannot be attributed to futures trading.

The prediction of futures prices of rubber were done by different forecasting models, viz., ARIMA, MLR and PCA. MLR got R square and adjusted R square of 92.1 per cent and 91.5 per cent both values showing the significance of the model for predicting the futures prices. Even though the value of R square is very high none of the regression coefficients were significant in the multiple linear regression model. This might be due to the multicollinearity of the independent variables viz; open, high, low and volume traded which are highly correlated. Hence the principal component analysis was done. The R square and Adjusted R square for the regression equation fitted using the Principal components as regressor are 91.7 per cent and 91.6 per cent respectively. So with P1 ie., the first component generated using open price, it was able to predict 91.7 per cent of the variation in the close price of rubber futures. The ARIMA results got R square of 99 per cent wth MAPE 1.97 per cent indicating that the forecasting inaccuracy was very low and the Normalized Bayesian Criteria (BIC) of 10.478 indicated goodness of fit of the model and the accuracy of the prediction.

While comparing the results of MLR, PCA and ARIMA, it was found ARIMA performed better in prediction. Also the forecasting errors of ARIMA were negligible indicating the forecasting efficiency of the model. Hence the study concluded that the univariate model outperforms the multivariate model with better accuracy in prediction.