A mixed model approach was utilized to develop stratum variance estimates for use in determining the sample allocation for estimation of “Not on Mail List” (NML) farm counts in the 2002 U.S. Census of Agriculture. Sample allocation was carried out by adding the NML farm item to the set of eight items currently used in the determination of sample allocation for the June Area Frame Survey. Changes in the sample allocations are recommended in the area frame sample design to support both the current sample survey based estimates and the 2002 Census of Agriculture NML farm count estimates.