The challenge of accurately forecasting commodity prices in times of market volatility
In collaboration with Ummar Farook Nazeer; additional inputs provided by Thomas Cherian
We all love gazing at the crystal ball to predict the future. The reasons can range from amateur curiosity to genuine business need.
While crystal balls or tarot cards fulfills one's curiosity, forecasting techniques based on sound arithmetic goes towards solving many business problems.
Many forecasters use regression models (linear and non-linear) to predict the price of various commodities. In simple terms, regression models are used to predict one variable based on a given set of variables.
This model held fort until 2008 when the credit crisis took the world by storm. The crisis, which has its origins in the collapse of the U.S. housing market, had ended up producing dramatic volatility in the price of commodities.
Crude oil, for example, touched nearly $150 during the heights of credit crisis. And the commodities market continued to remain volatile in the ensuing years.
As one could observe from the above graph, the jagged lines show the intensity of volatility, which can simply throw basic forecasting models off gear.
Many forecasters, if not all, usually resort to linear and non-linear regression models to forecast price of commodities. However, in times of volatility, simple regression models cease to be of much use as they end up throwing unrelated variables. In other words, in times of extreme price fluctuations, regression models do not provide accurate price forecast and the analyst will have to depend on the gut feeling to make predictions - a not so desirable scenario.
Volatility was not confined to traded commodities alone. As is evident from the above graph, even major non-traded commodities such as Ethylene, Polyethylene and Butadiene have witnessed significant variations in the past 4-5 years.
Monthly variations have been in the range of 0 - 80% for highly volatile commodities like Butadiene and 0-20% for Ethylene and its derivative polyethylene.
Forecasting is a handy tool for category managers in charge of procurement. And any uncertainty induced by forecasting models will add to the list of anxieties.
It is imperative for category managers to have robust forecasting techniques at their disposal. This is because procurement budgets usually have plenty of numbers that are derived out of forecast. A wrong or an inaccurate number can lead to confusing outcomes.
Category managers also use forecasting models to validate the commodity prices quoted by their suppliers. A robust forecasting model, which produces near-accurate prices, will help them to better negotiate with suppliers.
Prices of certain traded commodities play an indirect role in determining the contract prices of some non-traded commodities.
For example, average crude oil prices will be looked at by category managers while formulating the contract price of Ethylene, which is a feedstock of Polyethylene. Crude oil and Polyethylene prices have an 85.1% correlation. In which case, category managers would need dependable forecast of crude oil prices that would enable them to make critical business decisions.
The need for robust forecasting techniques is certainly non-negotiable. In times of sharp fluctuations of commodity prices, simple regression models won't be of much use; and category managers will need to employ a variety of auto regression models to forecast the prices of both traded as well as non-traded commodities.
As the procurement world is moving towards continuous mode, it would be tough for category managers to build and run robust models all by themselves. Instead, they can rely on category specialists capable of continuously tracking fundamental factors across value chain, and who can also double up as quant analysts to effectively plot price trends.
Related Insights:View All
Get more stories like this
Subscirbe for more news,updates and insights from Beroe