Selecting a suitable level of accuracy in stand growth estimations is a question of practical importance. A detailed model is more complicated and demands more initial information from the user to produce reliable results. Despite being correct and more precise, it may well be that the more detailed results give no practical bonus value. The essential part of the estimations could also have been received with coarser means. We have used three levels of stand description accuracy. The first is a mean tree model (MTM), where all trees are assumed identical. The second is a distribution-based model (DBM), where size differences are allowed. Both versions are distance-independent and the spatial configuration of the stand is considered random. The third is a spatially explicit model (SEM), where all trees are described individually and their spatial locations are taken into account. The core dynamics in all models are kept the same, so a previous model is a straightforward aggregation of the latter one. The possibility of a perfect aggregation was examined both theoretically and by practical simulations. A process-based stand growth model ACROBAS was used to simulate the different levels of aggregation. The results demonstrate that under the Poisson assumption of stand spatial configuration there are hardly any differences among model aggregations. As mortality is not totally random in the SEM, the Poisson assumption does not hold through the whole simulation period. As the spatial configuration eventually tends toward regularity, the total growth in the stand is somewhat increased, which is evident in variables such as total stem volume and foliage mass. However, mean height and stocking density developments are almost identical in all models. This suggests that the mean height is “an almost perfect” aggregation variable.