In the rapidly growing literature on endogenous specialization, of which the modern pioneering studies are Rosen (1978), Becker (1981) and Yang (1988) among others, a technically challenging problem is that many interesting models with intra-firm and inter-firm division of labour among ex ante heterogeneous agents, are not tractable by the existing toolkit, and therefore an obstacle to further progress needs to be overcome. This chapter develops a new apparatus to fill the void. We review some of the existing analytical tools before introducing the new method. To solve for the equilibrium of economies with ex ante identical agents (largely for addressing the Smithian specialization, rather than the more discussed Ricardian specialization based on ex ante comparative advantages among people/countries), Yang (1988) proposes what may be referred to as a two-step approach to identify the general equilibrium structure of the division of labour (as well as the associated equilibrium prices) among all possible structures. In the first step, for each structure, the ‘equilibrium’prices and utilities are calculated by equalizing utilities across all specialization patterns in the said structure. Secondly, the structure with the highest utility is selected as the general equilibrium structure of specialization, in which no agent has an incentive to deviate from her specialization pattern. Years later, in their ambitious monograph, Yang and Ng (1993) cover and apply the two-step algorithm to many specific models. Sun, Yang and Yao (1999) further argue that the two-step approach applies to a much larger class of models. It is clear that the two-step approach has proved quite powerful in solving many endogenous specialization models (for a recent survey, see Yang, 2003). The assumption of ex ante identical