The author examines "the problem of measuring projection accuracy. Experience with applying the accuracy measure 'quality of prediction', as proposed by Keyfitz, leads the author to a critical evaluation and elaboration of this accuracy measure. Time series of prediction quality values may show a remarkable temporal instability, partly depending on the chosen bench mark, which seriously hinders interpretation. This interpretation problem may be solved by an easily applicable graphical solution, a convenient short-circuit device to assess a projection's accuracy without restrictions as to size of population or length of projection period." The concepts are illustrated with data used for population forecasts for the Netherlands during the 1970s. (SUMMARY IN FRE)