The Market Cost of Capital approach is emerging as a standard for estimating risk margins for non-hedgeable risk on an insurer’s fair value balance sheet. This paper develops a conceptually rigorous formulation of the cost of capital method for estimating margins for mortality, lapse, expense and other forms of underwriting risk. For any risk situation we develop a three-step modelling approach that starts with i) a best estimate model and then adds ii) a static margin for contagion risk (the risk that current experience differs from the best estimate) and iii) a dynamic margin for parameter risk (the risk that the best estimate is wrong and must be revised). We show that the solution to the parameter risk problem is fundamentally a regime-switching model that can be solved by Monte Carlo simulation. The paper then goes on to develop a number of more pragmatic methods that can be thought of as shortcut approximations to the first-principles model. One of these shortcuts is the prospective method currently used in Europe. None of these methods require stochastic-on-stochastic projections to get useful results.