Coherent risk measures and good-deal bounds (Q5936314): Difference between revisions

The mathematical structure of valuation bounds in incomplete markets was first established in the context of super-hedging and arbitrage pricing. More recently it has been shown that arbitrage bounds can be generalized to ``good-deal bounds'', and most mathematical results from no-arbitrage pricing just carry over. Independently, risk measures like ``Value at Risk'' (VaR) appeared. Deficiencies of (quantile-) VaR led to the question what the characteristics of economically sensible risk measures are. It turns out that the mathematical structure behind coherent risk measures and good-deal bounds is exactly the same. This leads to interesting connections between the two fields, which started from different viewpoints. This paper works out the relation between risk measures, valuation bounds, and certain classes of portfolio optimization problems. It is economically general in the sense that it provides a common framework for applications in any cash stream spaces, be it in dynamic trading setting, one-step models, or deterministic cash stream. It is mathematically general in the sense that the core results are established for general linear spaces (which may be infinite-dimensional). The valuation theory presented fills a gap between arbitrage valuation on the one hand and utility maximization (or equilibrium theory) on the other hand. Arbitrage valuation has the advantage that the derived price bounds are completely independent of (estimated) probabilities and personal preferences. The paper contains some conceptual insights that seem to be new: -first, coherent risk measures as functions on a space of random variables can be generalized to general spaces of economic objects like commodities, delivery contracts, stochastic payment streams, or consumption plans; -second, there is a one-to-one correspondence between the following economic objects: 1) ``coherent risk measures'' \(\rho,\) 2) cones \(A\) of ``acceptable risks'' or ``desirable claims'', 3) partial ordering ``\(x\succeq y\)'', meaning ``\(x\) is at least as good as \(y\)'', 4) valuation bounds, 5) sets of ``admissible'' price systems; -third, the two examples of generating a coherent risk measure ``from standard risks'' and the ``largest coherent risk measure below a given function'' are shown to be special cases of the same principle: from any set \(B,\) define a (possibly coherent) acceptance set \(A\) by taking the conic hull \(A=\text{cone}(B)\); -fourth, a coherent risk measure and a set of ``cash streams available in a market'' define a new risk measure, which is again coherent if this set is a cone; -fifth, a series of coherent risk measures and valuation bounds can be defined for deterministic cash streams; -sixth, using the classical \((\mu,\sigma)\)-portfolio optimization theory of Markovitz as a blueprint, \((\mu,\rho)\)-portfolio optimization, where \(\rho\) is a coherent risk measure, can be considered.