BMI’s industry forecasts are generated using the best-practice techniques of time-series modelling. The precise form of time-series model we use varies from industry to industry, in each case determined, as per standard practice, by the prevailing features of the industry data being examined. For example, data for some industries may be particularly prone to seasonality, meaning seasonal trends. In other industries, there may be pronounced non-linearity, whereby large recessions, for example, may occur more frequently than cyclical booms.
Our approach varies from industry to industry. Common to our analysis of every industry, however, is the use of vector autoregressions. Vector autoregressions allow us to forecast a variable using more than the variable’s own history as explanatory information. For example, when forecasting oil prices, we can include information about oil consumption, supply and capacity.
When forecasting for some of our industry sub-component variables, however, using a variable’s own history is often the most desirable method of analysis. Such single-variable analysis is called univariate modelling. We use the most common and versatile form of univariate models: the autoregressive moving average model (ARMA).
In some cases, ARMA techniques are inappropriate because there is insufficient historic data or data quality is poor. In such cases, we use either traditional decomposition methods or smoothing methods as a basis for analysis and forecasting.
It must be remembered that human intervention plays a necessary and desirable part of all our industry forecasting techniques. Intimate knowledge of the data and industry ensures we spot structural breaks, anomalous data, turning points and seasonal features where a purely mechanical forecasting process would not.
There are a number of principal criteria that drive our forecasts for each Power indicator.
Development of generating capacity is determined largely by investment levels, with generation levels dictated by available capacity, plant utilisation rates and national policy. We therefore examine:
A mixture of methods is used to generate demand forecasts, applied as appropriate to each individual country:
Whenever possible, we compare government and/or third-party agency projections with the declared spending and capacity expansion plans of the companies operating in each country. Where there are discrepancies, we use company-specific data as physical spending patterns ultimately to determine capacity and supply capability. Similarly, we compare capacity expansion plans and demand projections to check the energy balance of each country. Where the data suggest imports or exports, we check that necessary capacity exists or that the required investment in infrastructure is taking place.
Sources include those international bodies mentioned above, such as the IEA, and EIA, as well as local energy ministries, official company information, and international and national news agencies.