A method is presented for unravelling the demographic equation for str
uctured populations. A solution to the McKendrick-von Foerster equatio
n is constructed using spline functions and this is fitted to stage-st
ructured population data in such a way that the solution is smooth, po
sitive, and does not imply negative death rates. The smoothness of the
surface, and hence the complexity of the population model, is determi
ned in a statistically optimum manner using cross validation. Time- an
d age-dependent death rates can be obtained as well as time-dependent
birth rates. Confidence intervals are obtained for population size and
death rates that give a 95% probability that the true population dyna
mics are within the intervals. Practical application of the method is
demonstrated, and comparison made with three alternative methods. The
method requires estimates of stage populations at a series of times an
d estimates of the durations of those stages, or age ranges covered by
the stages. The stage durations may vary with time.