demografie.net
Understanding population dynamics –
Projections and forecasts compared
Forecast or Projection?
Two terms often confused – yet fundamentally different.
Relative Prediction
A forecast starts from a defined initial situation and an assumed development path and states – with a – sometimes included – formulated probability of occurrence between zero and one hundred percent – a resulting event.
It always delivers a formally and logically meaningful result.
Model-based extrapolation
A projection simply extrapolates a trend from a fully defined initial and developmental state. It is therefore a model-based calculation that can develop several competing scenarios.
The result does not even have to be meaningful – on the contrary: it can, for example, show that a certain development cannot continue beyond a certain point in time. As a pure extrapolation, it naturally cannot deal with probabilities.
A concrete example
One year ago: 84 kg → diet change → today: 72 kg
kg\n1 year ago
kg\ntoday
kg\nfuture
If you continue like this and nothing unexpected happens, your weight will most likely stabilise at 60 kg in two years.
If you continue to lose one kilogram per month, in six years you will weigh nothing at all.
The German Federal Statistical Office publishes its population projections with twelve different scenarios every year. However, it is never claimed that one of these scenarios (or even all of them simultaneously) will become reality; it therefore deliberately refrains from making a forecast!
Statistisches Bundesamt, 2009
The frequently expressed and unsubstantiated criticism that a population projection spanning five decades is practically impossible – just look at the last five decades of German population development – falls flat.
A projection only becomes a forecast (sometimes in the hands of journalists and politicians) when it is claimed that this or that case will in all probability occur. In the example of population projections, this would logically mean: one of the scenarios – the other eleven not. And the projection makes no statement about precisely this probability – that is what distinguishes it from a forecast.
Rather, the projection clarifies what consequences various trends could have if important parameters – e.g. fertility, life expectancy or migration balance – were to develop in one way or another.
Thought experiments
Show what consequences various trends would have – without claiming that any of them will occur.
Relative predictions
Claim with a certain probability which scenario will actually come about.