EXAMPLE: POPULATION FORECAST
The exercise objective is to make a simple forecast of population for Australia for 1994, based on actual data available for 1987 to 1993. Two alternative approaches are taken:
1. Forecast the population increase in 1994 from the net increase in the years 1988 to 1993;
2. Forecast the population increase in 1994 from defined change factors 1988 to 1993 (the MAF approach).
The
results are then compared to the actual population increase for 1994.
The exercise is aimed at comparing very simple forecasting methods: deliberately
avoiding any regression analysis. One simple approach (method 1A) takes
the average net increases in the years 1998 to 1993 and suggests the same
average increase for 1994. This estimate is found to be 24% over the actual
increase for 1994.
Another simple approach ( 1B) identifies that the population is increasing,
but in decreasing increments. Assuming a linear extrapolation, 1994 is
estimated to increase 138,631. This estimate is 26% under the actual increase
for 1994.
The MAF approach (method 2) identifies a data framework system, identifies the change factors in the system, forecasts the change factors separately (still using simple methods), and then derives the total net change. 1994 is here estimated as 197,096 or 5% over the actual increase.
The point is that improved forecasts may be obtained by disaggregating a system, identifying change factors, and forecasting the change factors separately but subject to the system discipline. Forecasting efficiency may be improved by a good data control framework as well as more sophisticated forecasting procedures. Statistical differences are forced to be disclosed and considered, using a simple bookkeeping discipline.
The forecasting method has been deliberately kept simple, to illustrate the importance of a data framework. The advantages of the this approach are:
(1) Accuracy: the absolute percentage error in forecasting can be reduced by data disaggregation and control measures.
(2) Clarity: data arrays show the flow of actual into forecast; the forecasts can be seen to "grow" from actuals.
(3) Understanding: change factors or components are highlighted, clarifying an understanding of the reasons for changes.
(4) Exceptions: differences, such as the statistical differences noted above, can be forced out for separate consideration.
