Household Electricity Consumption of Middle Class Family in Chiitagong - A Case Study

Abstract

Introduction: Electricity is a necessity in the modern world. Electricity has attained a very important place in every household on this planet. It is a major contributor towards improvement of the standard of living of any individual, family and society at large. The aims of this study is to find out the monthly average household electricity consumption and fit a suitable time series model to predict the electricity use. Data and Analysis: A time series monthly electricity uses data of a middle class family in Chittagong from January 2001 to November 2015 is considered in this analysis. To check variability the descriptive statistics and different types of graphs are used. The volatility model ARCH family regression with ARIMA disturbances model is used for forecasting. Chow test statistic is used for checking the structural breaking point of the dataset. Results & Conclusion: From the ACF and PACF function we get the cut off point for AR and MA part are 2 and 3 respectively. Further as ARCH effect is significant for this data set we use ARCH family regression with ARMA disturbances model, After comparing the different value of the parameters, ARCH(1) with ARIMA (3,0,2) disturbances is best fit for this data set. There have a structural break point for the month of December in 2010. Before this date data, ARCH (1) regression family with ARIMA (2,0,2) disturbance is the best fitted model for the analysis. And for the post data follows only ARIMA(1,0,1) disturbance is the best fitted model for this analysis. Recommendation: For forecasting of the monthly electricity uses of a middle class family in Chittagong, ARCH(1) with ARIMA (3,0,2) disturbances time series model can be used. For better prediction one can consider to select a representative size of sample families with at least 20 years data. Also some covariates like family size and electronic items used in the family can be considered and can try to fit a GARCH or TGARCH model.