Extreme Measures in Continuous Time Conic Finace

Yoshihiro Shirai

2023-10-20

Abstract:Dynamic spectral risk measures define a claim's valuation bounds as supremum and infimum of expectations of the claim's payoff over a dominated set of measures. The measures at which such extrema are attained are called extreme measures. We determine explicit expressions for their Radon-Nykodim derivatives with respect to the common dominating measure. Based on the formulas found, we estimate the extreme measures in two cases. First, the dominating measure is calibrated to mid prices of options and valuation bounds are given by options bid and ask prices. Second, the dominating measure is estimated from historical mid equity prices and valuation bounds are given by historical 5-day high and low prices. In both experiments, we find that the market determines upper bounds by testing scenarios in which losses are significantly lower than expected under the dominating measure, while lower bounds by ones in which gains are only slightly lower than in the base case.

Risk Management,Probability

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to determine the extreme measures of dynamic spectral risk measures in the continuous - time conic finance framework. Specifically, the author aims to find the explicit expressions of the Radon - Nikodym derivatives of these extreme measures and, based on these expressions, estimate the performance of the extreme measures in two cases:

1. **Market Calibration Case**:
- **Dominant Measure**: Calibrated by the mid - price of options.
- **Valuation Range**: Given by the bid - ask spread of options.

2. **Historical Data Case**:
- **Dominant Measure**: Estimated from the historical mid - stock prices.
- **Valuation Range**: Given by the historical 5 - day high - low prices.

Through these two experiments, the author finds that the market tends to test situations where losses are significantly lower than expected when determining the upper bound, while when determining the lower bound, it tests scenarios where returns are slightly lower than the benchmark situation.

### Main Contributions
1. **Mathematical Contributions**:
- Provide explicit expressions of the Radon - Nikodym derivatives of extreme measures in continuous - time conic finance (Theorem 3.3).
- Prove the existence and uniqueness of these extreme measures.

2. **Empirical Research**:
- **Option Data**: By calibrating the bilateral gamma process to the mid - price of options of the SPY ETF, verify whether the valuation range of continuous - time conic finance can match the relative bid - ask spread and infer the market's uncertainty about different events.
- **Historical Stock Price Data**: Assuming that the law distribution under the reference measure is a bilateral gamma process, estimate the upper and lower bounds through the historical closing mid - prices of SPY, compare the digital moment estimator (DM) with the generalized method of moments (GMM), and infer the market's uncertainty about different events.

### Research Significance
- **Theoretical Significance**: Provide the mathematical basis for extreme measures in continuous - time conic finance, filling the gap in the existing literature.
- **Empirical Significance**: Through empirical analysis, verify the effectiveness of the theoretical model and provide new insights into the market's uncertainty about different events, which helps market participants better understand and respond to market fluctuations.

### Key Formulas
- **Radon - Nikodym Derivative**:
\[
\frac{dQ}{dQ}\bigg|_{\mathcal{F}_t} = E(M)_t, \quad \frac{dQ^*}{dQ}\bigg|_{\mathcal{F}_t} = E(M^*)_t
\]
where
\[
M_t = \int_{[0,t] \times \mathbb{R}^D \setminus \{0\}} \psi_s(y) \tilde{N}(ds, dy), \quad M^*_t = \int_{[0,t] \times \mathbb{R}^D \setminus \{0\}} \psi^*_s(y) \tilde{N}(ds, dy)
\]
and \(\psi\) and \(\psi^*\) are defined as follows respectively:
\[
\psi_t(y) = \Gamma'_+(\nu(\{Z_t > Z_t(y)\})) 1_{\{Z_t(y) > 0\}} - \Gamma'_-(\nu(\{Z_t < Z_t(y)\})) 1_{\{Z_t(y) < 0\}}
\]
\[
\psi^*_t(y) = -\Gamma'_-(\nu(\{Z_t > Z_t(y)\})) 1_{\{Z_t(y) > 0\}} + \Gamma'_+(\nu(\{Z_t < Z_t(y)\})) 1_{\{Z_t(y) < 0\}}
\]

Through these formulas, the author is able to be precise.