Classical forecasting tools

Advantages of these approaches are that they are simple to perform and easy to understand. The time trend deals with the data sequentiality. Heterogeneity can be captured by the extended approaches that allow for different intercepts and slopes. Volatility and non-linearities can only be dealt with in a limited way.

Auditability is very high as modelers explicitly choose the model structure and can use statistical tests to support their choices.

• Allowing for different intercepts β_0i for various countries (linear panel data regression)

• Allowing for different slopes β_1i for various countries (random coefficient model or model separately analysed for each country)

• Including drivers, such as policy variables and prices in the above models
• Allowing for quadratic and interaction terms of variables to capture non-linear patterns

Classical forecasting tools can be summarized as linear trend line fitting. Simple examples are drawing or fitting simple straight lines through existing datapoints. Forecasts are based on linear extrapolation of such lines.

Statistically this could be done using linear regression in which an intercept and slope parameter of a trend variable are estimated:

acreage_t=β₀+β₁⋅t+ε_t