A new approach to SAM updating with an application to Egypt

When social accounting matrices (SAMs) are updated with fragmentary data the new column and row totals are often not all known in advance. Also, the overall total of all matrix entries may not be known. This results in a situation in which the generally used RAS method cannot be applied to update the SAM, as the outcomes become unreliable with unknown totals. In this paper a generalised version of the RAS method is developed, based on information theory. This method, which incorporates the RAS method as a special case, is more flexible with respect to necessary information and improves the accuracy of the updated SAMs. In addition to developing a generalised RAS method, we present an easily applied procedure for building SAMs which reduces the time needed to update them. The generalised RAS method and the SAM building procedure are applied to a 1991-92 SAM for Egypt.