This paper studies the efficient arrangement among several agents that are subject to idiosyncratic, privately observed taste shocks affecting their marginal utility of current consumption. Agents accumulate capital and have access to a technology to produce goods. The framework deviates from previous literature, which assumed that (i) there is a continuum of agents and (ii) agents are exogenously endowed with output every period with no investment opportunities. A new method is proposed to solve for the optimal allocation that takes advantage of the fact that the utility possibility set is convex. Pareto weights play the role of promised utility in Abreu, Pearce, and Stacchetti (1990). The method is applied to study efficiency in a partnership between the founder, who faces the taste shock, and the partner, who provides funds but do not face shocks. New insights are derived. Under private information the ownership structure determines to what extent private information matters. If the founder’s share of the partnership is too big his incentives to cheat vanish. Additionally, efficiency implies that, while incentive constraints bind, equity shares must fluctuate to alleviate information problems. In the long run, there are two possible extreme structures: (1) the founder’s equity share becomes sufficiently large to make the incentive problem irrelevant and (2) the founder’s equity share converges to zero. Two alternative economies are studied to understand the role of key assumptions: (i) an endowment economy and (ii) an economy in which both partners face taste shocks. It turned out that to obtain the main results allowing for a production economy is crucial but having only one agent with shock is not.