Methodology
The strength of our forecasting methodology, other than its accuracy, is its transparency. We have deliberately eliminated the need to speculate, and replaced it with clear, rigorous analysis. This methodology comprises three separate evaluation models:
Time Series model
Our Time Series model uses patterns identified by historical data, such as long-term trends, business cycles and seasonal influences to create forecasts. Using the ARIMA technique (Auto Regressive Integrated Moving Average), our model combines a regressive analysis of historical data with a moving average of more recent data to produce highly accurate short-term forecasts.
Whenever necessary, and particularly for those products whose markets are highly reactive, we create a more complex Time Series model that incorporates the effect of critical factors such as crude oil prices – ARIMA X. For these more volatile products this method increases accuracy by as much as 7%.
Multivariate model
Our second model identifies relationships between petrochemical prices and other economic and market variables. We have mined our extensive product data to compare price fluctuations between specific products and drivers such as consumer indicators, industrial indicators, trade data and other petrochemical price data. By determining the correlation between such drivers and specific product prices over time, we are not only able to forecast price, but to explain price movement.
We combine the results of the Time Series and Multivariate models to produce a Statistical Forecast.
ICIS Market Sentiment Index™
The third model in our methodology is utterly unique. We have all felt the effects of news and current events on price. This information stimulates a reaction in our industry, and indeed, in all drivers affecting the price of chemical products. We can think of this reaction as ‘market sentiment’, and while its influence has long been recognized, it has never been measured or harnessed.
Our model, the ICIS Market Sentiment Index™ (IMSI), quantifies the inflationary or deflationary value of specific types of news events on product price, giving them a number. This number is positive in months where more inflationary than deflationary events have occurred, and negative when the combination of events is likely to have a deflationary impact.
To create the index, we have capitalized on our position as the world’s largest recorder and provider of chemical and oil industry news, drawing on our vast historical data of events and comparing them with price fluctuations. In doing so, we can treat news as a valid statistic to create a predictor of market behaviour.