Carbon allocation to sapwood in tropical canopy trees is a key process determining forest carbon sequestration, and is at the heart of tree growth and dynamic global vegetation models (DGVM). Several allocation hypotheses exist including those applying assumptions on fixed allocation, pipe model, and hierarchical allocation between plant organs. We use a tree growth model (IBTREE) to evaluate these hypotheses by comparing simulated sapwood growth with 30 year tree ring records of the tropical long-lived tree Toona ciliata M. Roem. in Thailand. Simulated annual variation in wood production varied among hypotheses. Observed and simulated growth patterns matched most closely (r2 = 0.70) when hierarchical allocation was implemented, with low priority for sapwood. This allocation method showed realistic results with respect to reserve dynamics, partitioning and productivity and was the only one able to capture the large annual variation in tree ring width. Consequently, this method might also explain the large temporal variation in diameter growth and the occurrence of missing rings often encountered in other tropical tree species. Overall, our results show that sapwood growth is highly sensitive to allocation principles, and that allocation assumptions may greatly influence estimated carbon sequestration of tropical forests under climatic change.