The authors have several objectives that motivated their research. First, they seek to determine possible explanations of why the credit default swap (CDS) market played a dominant role in price discovery during the financial crisis, whereas the stock market was dominant in price discovery before the financial crisis. Second, they investigate the effects of the financial crisis on price discovery among stock prices, CDS spreads, and bond prices within the financial sector because these companies experienced the greatest volatility. Third, the authors model price discovery in these three markets (stocks, CDS, and bonds) simultaneously using a vector error correction model (VECM) along with forward estimation and rolling-window techniques. Finally, they use behavioral finance models to help explain changes in price discovery dynamics before the financial crisis and after the financial crisis.

The authors build on prior studies to evaluate a small sample of financial firms during 3 January 2005 through 30 December 2008. They conclude that during the financial crisis, CDS spreads dominated price discovery as a result of trader behavior and government regulations. For example, the informed trader hypothesis indicates that informed traders in the CDS market disseminate information before other markets do, thus causing CDS markets to lead price discovery.

Of the model scenarios or specifications tested—stocks versus CDS spreads, stocks versus bonds, CDS spreads versus bonds, and stocks versus CDS spreads versus bonds—the stock and CDS markets lead the bond market in both periods (before and during the financial crisis), and the CDS market has a stronger more stable relationship with the bond market. The behavioral finance models based on overconfidence and disposition imply that overconfidence was not a main factor during the crisis. The disposition effect occurs more often during bull markets and causes noise trading from selling winners too early and holding losers too long. Because the market was pessimistic and volatile, the lack of noise trading in the CDS market improved the quality or information of the CDS price. The improved quality of the CDS price dominated price discovery. The results can also be explained by government regulation. On 19 September 2008, the US SEC placed a ban on short sales of financial company stocks and removed the ban 8 October 2008. The statistical results of the recursive cointegration and price discovery tests of stocks versus CDS indicate 53.5% price discovery for CDS and 46.5% for stocks during the financial crisis.

The target audience for this research is practitioners and academics. The research suggests that during distressed financial markets, new information is captured in the CDS market before the stock and bond markets.

The authors’ CDS sample data are obtained from CMA, an industry-accepted source used by Bloomberg. The CDS sample contains daily observations of 39 public financial firms and midquote CDS spreads as of the close of business in New York City, but there is enough bond data for only 10 reference entities for the sample period. Furthermore, for each reference entity, the research includes two bonds issued each day: one with a maturity of more than five years and one with a maturity less than five years in order to calculate an estimated yield to maturity of approximately five years using linear interpolation. All of the bonds in the data sample have a maturity of five years and the following attributes: senior unsecured seniority, modified restructuring, no embedded options, denominated in US dollars, readily available pricing, and a fixed coupon rate. The five-year swap rate is used as a proxy for the five-year risk-free rate.

The authors use the Phillips and Perron (Biometrica 1988) unit root test, which is generalized from the Dickey–Fuller test and an application of a VECM, and price discovery to decompose a nonstationary vector of a time series into its permanent and temporary components in order to capture the long-run and short-term equilibrium price, respectively. They then use the permanent factor to capture the efficient price or source of permanent changes in the cointegrated relationship and then normalize and test the coefficients for evidence in the price discovery calculations. A main assumption is that any temporary shocks do not alter the long-run equilibrium. The research is successful in attaining its objectives. Statistical evidence is presented in nine exhibits, which clearly provide support for the conclusions.

The authors effectively leverage prior research from notable publications and also delve deeper into an analysis of price discovery during the recent financial crisis. The research uses forward estimation and rolling-window time estimation techniques. In forward estimation, the beginning of the data starts with a subsample on the first day and adds observations to obtain a series of longer samples, whereas the rolling-window approach translates the subsample forward in time for a constant sample size. The latter approach is an effective way to further analyze the dataset in the exhibits and two different time estimation techniques. The research is limited because of the list of data requirements and the small sample of 10 financial firms.