During the next fifty years, the growth of the population of Finland is expected to slow down, and turn into decline. The age-distribution is expected to become older because mortality declines. In particular, the share of the working age population will decline. This development is accentuated by the fact that the large post war birth cohorts reach retirement age during the next decade. Although there is general agreement about these broad features, it is difficult to say exactly when a decline might begin, or how high the age-dependency ratio will be like in the future. A study of past forecasts shows that demographic developments have repeatedly taken forecasters by surprise. We show that the forecast errors in Finland have not been related to other social phenomena, such as wars or economic crises, in a simple way. In fact, our understanding of the causes of the past errors is poor. Therefore, it is important to recognize that our current view of future population development may similarly be in error. We account for the uncertainty by statistical modeling. In this way it is possible to estimate, how large errors one should expect, if future demographic development is as volatile as in the past. Using stochastic simulation, we derive a predictive distribution for the future population vector that gives a realistic indication of the uncertainty to be expected.
