Full enumeration agriculture censuses have become less feasible financially and operationally in many developing countries. They had not always lived up to their promise to serve as frame for the intercensal surveys and to provide reasonably accurate benchmark data for small areas. Observed large deviations between the survey and census results point to serious non-sampling errors especially in the latter. Some developing countries had resorted to replacing agriculture censuses with sample enumerations that differ from the intercensal surveys only in the choice of a smaller domain, hence a bigger sample. There is need for strategies that link the sample census and the intercensal surveys, in a manner appropriate to the limited technical and operational capabilities of the developing countries. Using the Philippines as example, the paper presents the design of the 1991 sample census of agriculture and outlines a strategy to draw second phase replicated samples directly from it.