The renewal of the forestry codes of the countries of the Congo Basin in the 90's has legalized management plans for forest concessions. These plans intend, among other objectives, to ensure the sustainable exploitation of commercial species. Sustainability is assessed using the stock recovery rate, which is defined in national directives as the ratio of the exploitable timber stock at the end of a felling cycle over the exploitable timber stock at the beginning of this cycle. Computing this rate requires forecasting the temporal development of each species during a felling cycle, which is usually achieved using the so-called "stock recovery formula". This paper shows that this formula corresponds to a Leslie model, and then proposes a generalization as a Usher matrix model. Using the data from the M'Baïki experimental plots in the Central African Republic, the stock recovery rate for sapelli (Entandrophragma cylindricum, Meliaceæ), a major timber species in Central Africa, was estimated. The estimate was completed by its confidence interval using bootstrap methods. Although 225 observations were available for sapelli, the stock recovery rate was estimated with no more than an accuracy of about 45% at confidence level 95%. This did not permit to conclude whether the asymptotic stock recovery rate was greater or less than one. This suggested that much more observations than usually acknowledged are required to estimate the stock recovery rate with an acceptable accuracy. Different logging scenarios were finally tested to assess the impact of management parameters on the stock recovery rate of sapelli.