Economic Catastrophe Bonds: Inefficient Market or Inadequate Model?

Haitao Li, Feng ZhaoSSRN 2012 · DOI

TL;DR

This paper challenges the claim that senior CDX tranches are overpriced, arguing that previous analyses suffered from flawed calibration and overly restrictive models. By correcting the calibration and employing a simpler, well-established model, the authors demonstrate that CDX tranches can be priced accurately, suggesting the market is not necessarily inefficient.

What Problem Does This Paper Address?

The paper tackles a significant debate in financial economics: whether the market for credit derivatives, specifically senior CDX tranches, is inefficient. A prominent study by Coval, Jurek, and Stafford (CJS) claimed that these tranches, which are akin to economic catastrophe bonds, are overpriced relative to index options. CJS argued that investors are not adequately compensated for the systematic risks they bear, leading to a persistent inefficiency. This paper aims to re-evaluate this claim by scrutinizing the methodology used by CJS and proposing alternative modeling approaches. The core question is whether the observed pricing discrepancies are evidence of market inefficiency or simply a consequence of inadequate modeling techniques.

Methods

The authors' approach involves a critical re-examination of the CJS model and its calibration. They identify two main issues with the CJS methodology:

  • Problematic Calibration Procedure: The paper highlights that CJS's method for estimating key parameters, particularly the variance ratio ($\rho$), leads to significant biases in CDX tranche pricing. They propose a simple correction to this calibration process.
  • Restrictive Model Assumptions: The CJS model relies on specific distributional assumptions for systematic and idiosyncratic factors. The authors argue these assumptions are too restrictive to accurately capture the observed market prices.

To demonstrate their points, the authors:

  • Correct CJS Calibration: They adjust the calibration of the variance ratio ($\rho$) and show how this significantly reduces the overpricing observed by CJS.
  • Employ an Alternative Model: They utilize the "double-t" model proposed by Hull and White (2004), which is a parsimonious model with minimal free parameters. This model uses Student-t distributions for both systematic and idiosyncratic factors, allowing for fatter tails than the Gaussian assumptions in CJS.
  • In-Sample Fitting: They also conduct an in-sample analysis where they allow the CJS model maximum flexibility to fit tranche prices, demonstrating that even with this flexibility, the model struggles.

Datasets

The paper uses data on CDX index and tranches, as well as S&P 500 index options. The specific data period appears to be from 2003 to 2007, as indicated by references to this timeframe in the text. No new datasets are introduced; rather, the authors analyze existing market data.

Key Insights and Ideas

  • Calibration is Crucial: The paper strongly emphasizes that the way parameters are calibrated can have a profound impact on pricing results. A seemingly small issue in calibration can lead to dramatically different conclusions about market efficiency.
  • Model Choice Matters: The choice of distributional assumptions for asset returns is critical, especially for instruments like CDX tranches that are sensitive to extreme events. Gaussian assumptions, as used in CJS, may not be sufficient.
  • Parsimony Can Be Powerful: The success of the Hull and White (2004) "double-t" model, which is simple and has few free parameters, suggests that complex models are not always superior and can sometimes be a source of error.
  • Challenging Established Findings: The paper directly challenges a widely cited conclusion about market inefficiency, highlighting the importance of rigorous methodological scrutiny in financial research.

Key Results

  • A simple correction to the CJS calibration procedure reduces the average overpricing of the senior 15-30 tranche by 80% (from 16.2 bps to 3.3 bps).
  • The Hull and White (2004) "double-t" model, with fixed parameters, can price CDX tranches and index options well, exhibiting significantly smaller pricing errors compared to the CJS model.
  • Even when the CJS model is given maximum flexibility (in-sample fit), it still exhibits substantial pricing errors for several tranches.
  • Table 1 shows that the CJS model with $p=0.2$ (a corrected variance ratio) significantly improves pricing compared to the original CJS model, and the double-t model provides even better fits. For the 15-30 tranche, the CJS model with $p=0.2$ predicts 11.9 bps (vs. 8.6 actual), while the double-t model predicts 10.1 bps.

Limitations and Future Work

  • Model Specificity: While the double-t model performs well, the authors acknowledge they are not claiming it's the "perfect" model for CDO pricing. They suggest that other models might also achieve good fits.
  • Focus on CDX.NA.IG: The empirical analysis primarily focuses on the CDX.NA.IG index. Future work could explore other indices or types of credit derivatives.
  • Further Model Refinements: The paper mentions introducing negative skewness into the double-t model (creating a skewed-t model) to better capture volatility skew in index options, which also shows excellent fit. This suggests that further exploration of distributional assumptions could be fruitful.
  • Robustness Checks: While the double-t model's performance is robust even when fitted to only tranche prices, further robustness checks across different market conditions or data periods could be valuable.

Summary updated

BibTeX
@misc{li2012economic,
  author = {Haitao Li and Feng Zhao},
  title = {{Economic Catastrophe Bonds: Inefficient Market or Inadequate Model?}},
  year = {2012},
  doi = {10.2139/ssrn.2023717},
  howpublished = {SSRN},
}
Figure