This paper documents the construction of a 1992 social accounting matrix (SAM) for
Tanzania. On the basis of recently generated national accounts data, a 56-sector SAM is built
focusing on the disaggregation of agriculture — which comprises 21 of the 56 sectors. First,
a highly aggregated SAM (macrosam) is designed to set the macroeconomic framework that
provides the control totals for the disaggregation procedure. Then, the sector disaggregation
of the microeconomic SAM (microsam) is done. Data sources and the data adjustments made
are presented. The microsam differentiates 4 household types and 5 labor categories. Special
features of the microsam include non-monetary, own-household consumption and separate
marketing margins on domestic products, exports, and imports — which play a crucial role
in the low income economies of sub-Saharan Africa. Since the data base is to be used for
economic policy modeling, consideration of these features will have a significant influence on
the results of the analysis. Due to data insufficiencies the first microsam obtained from
adjusted raw data (protosam) is highly unbalanced. A cross-entropy estimation method is
applied to balance the protosam and generate the final estimated 1992 microsam for Tanzania
that uses all available information in a consistent framework.