A Comprehensive Survey of Steel Demand Forecasting Methodologies and their Practical Application for the Steel Industry
AuthorJi-Mi Chu,Moon-Kee Kong,Center for Economic Research and Information Analysis
View : 34718
Like : 0
This article classifies and compiles the methodologies through a comprehensive review of the literature, and then finds clues to enhance the accuracy of steel demand forecasting.
The approaches for forecasting steel demand can broadly be classified into the econometric and intensity of use (IU) approaches.
Econometric approaches are divided into the econometric demand model and vector autoregression (VAR). The econometric approach widely uses a simple single equation or a simultaneous equation to forecast steel consumption, considering that steel demand is affected by macroeconomic variables including GDP, industrial production, trade structure, and economic volatility. The VAR methodology has the merit of avoiding the weakness of econometric demand model that requires forecasts of exogenous variables since VAR assumes all variables in a model are endogenous.
The intensity of use (IU) approaches rose to prominence in the early 1970s when some OECD member countries observed their steel demand fall while macroeconomic indicators grew. The IU approach is a useful concept that attempts to link steel consumption to the technological and structural changes in an economy.
Mathematical methodologies and computational approaches
Hybrid mathematical methodologies seek to enhance predictability based on the grey model, algorithm, and fuzzy ARIMA model. The steel weighted industrial production (SWIP) index is broadly used by worldsteel and other steel associations.
To complement the weakness of top-down macro methodologies which directly predict total steel demand, POSRI is concurrently applying a bottom-up micro methodology to predict demand for 16 steel products and summing them to forecast total demand.