Dimensional stability is one of the most important properties of wood used as timber. However, it is very time-consuming to measure this trait. Near-infrared spectroscopy is a tool adapted for prediction of many properties of wood, as chemical contents, basic density, and the time needed for the analysis is highly reduced. The issue here is to check the effectiveness of NIRS tool to build models and to predict the radial and tangential shrinkages of Teak (Togo) and Liquidambar (Madagascar) wood both from plantations. It was possible to predict the dimensional stability by means of NIR spectroscopy: For Teak, the transversal cross-section spectra yielded better NIRS-based models while for Liquidambar, the tangential surface gave the best statistics model. For the two species, the prediction models for tangential shrinkage showed similar results in term of efficiency. After independent validation or cross-validation, respectively for Teak, and for Liquidambar wood, we assume that NIRS can be used to predict radial and tangential shrinkages for screening.