Stephan Boes and Peter Pflaumer (2002)
Ascertaining Forecast Intervals Using Monte Carlo Methods – A Procedure to Evaluate the Uncertainty of Student Forecasts*
In: Zeitschrift für Bevölkerungswissenschaft, Vol. 27, 4/2002, p. 465-491, Opladen: Verlag Leske + Budrich, ISSN: 0340-2398
The official projection of numbers of students total and new enrolments state ranges for future trends, a method also frequently being used in the context of demographic forecasts. One repeatedly finds when comparing the forecast with reality that it is difficult to make a reliable assessment of future trends. For instance, the numbers of new enrolments have recently increased much faster than projected, their value being far outside the range of the most recent official forecast. In stating ranges, it is not possible to make assertions on the probability of forecasts actually taking place. However, it is such statements in particular that are important to be able to judge current developments and to draw short- and medium-term conclusions. Demographic literature contains a large number of works referring to this problem and developing appropriate procedures.
One of these approaches is taken up in this work. On the basis of the transition models of the Conference of Ministers of Education and Cultural Affairs, a possibility is developed to go beyond stating ranges and make concrete assertions as to the probability of forecasts coming true for the field of higher education. Here, distribution assumptions are made using the parameters varied in the forecast, and Monte Carlo methods are used to calculate point forecasts and forecast intervals. This procedure offers the additional advantage of incorporating a large number of possibilities of various distribution assumptions and subjective assumptions in one single projection, without in doing so losing track of one's observations in the calculation of just as many variants. Rather, this permits all assumptions to be summarised in one result with statements on uncertainty.
* Original title: Ermittlung von Prognoseintervallen mit Hilfe von Monte-Carlo-Methoden – Ein Verfahren zur Beurteilung der Unsicherheit von Studierendenprognosen (full text in German only)