David X. Li

David X. Li
Born	1960s Nanjing, China
Occupation(s)	Actuary, quantitative analyst, professor
Years active	1997–present
Known for	Application of Gaussian copula models to valuation of collateralized debt obligations (CDOs)
Notable work	"On Default Correlation: A Copula Function Approach" (2000)

David X. Li (), originally from Nanjing, China and born in the 1960s, is a Canadian actuary and quantitative analyst of Chinese origin, recognized for his groundbreaking application of Gaussian copula models to the valuation of collateralized debt obligations (CDOs) during the early 2000s. The Financial Times once referred to him as "the world's most influential actuary." Nonetheless, in the aftermath of the 2008 financial crisis, his methodology was partly held responsible for exacerbating the market collapse, being described as a "recipe for disaster" when improperly implemented by financial practitioners. The extensive adoption of simplified Gaussian copula frameworks in financial products such as securities has been associated with heightened systemic risk during that era. Presently, Li holds an adjunct professorship in Statistics and Actuarial Sciences at the University of Waterloo.

Early life and education

Born Li Xianglin and raised in rural China amid the 1960s, Li's family experienced forced relocation during the Cultural Revolution for "re-education." He completed a master's degree in economics at Nankai University. In 1987, responding to a Chinese government initiative to study Western capitalism, he earned an MBA from Laval University in Quebec. Subsequently, he attained both a Master of Mathematics in Actuarial Science and a PhD in statistics from the University of Waterloo in 1995, with his dissertation titled An estimating function approach to credibility theory, supervised by Harry Panjer.

Career

Li embarked on his finance profession in 1997 within Canadian Imperial Bank of Commerce's World Markets division. By 2000, he had joined J.P. Morgan’s RiskMetrics group in New York City as a partner. In 2003, he advanced to director and global head of credit derivatives research at Citigroup. The following year, he led Barclays Capital’s credit quantitative analytics team. In 2008, Li relocated to Beijing to assume leadership of risk management at China International Capital Corporation.

Currently, he serves as an adjunct professor at the University of Waterloo’s Department of Statistics and Actuarial Sciences and holds a professorship at the Shanghai Advanced Institute of Finance (SAIF).

CDOs and Gaussian copula

In his landmark 2000 publication "On Default Correlation: A Copula Function Approach," Li introduced the Gaussian copula model tailored for CDO applications, allowing financial firms to price intricate securities such as mortgage-backed assets by modeling correlations among multiple components. This approach swiftly became a cornerstone technique for valuing CDOs.

Despite its initial acclaim, after the 2008 financial meltdown, the model faced criticism as a "recipe for disaster." Detractors like Nassim Nicholas Taleb argued that depending on correlation metrics was inherently inadequate for capturing interdependencies between securities. Li himself admitted certain limitations, acknowledging in 2005 that many users misunderstood the model’s fundamental principles and attributing its widespread adoption more to its ease of use than robust theoretical foundations. Industry analysts maintain that while Li developed the model, its improper application by banking institutions played a major role in its shortcomings.

Li's paper

In his 2000 publication titled On Default Correlation: A Copula Function Approach, featured in the Journal of Fixed Income, Volume 9, Issue 4, pages 43–54, Li proposes a technique for measuring default correlations through the Gaussian copula function:

<math>C_\rho(u,v) = \Phi \left(\Phi^{-1}(u), \Phi^{-1}(v); \rho \right)</math>

This method captures the interdependence between default occurrences of different entities—such as "House A" and "House B"—by examining their respective survival durations. Although this model demonstrates utility in sectors like real estate, the study cautions that depending solely on a single correlation parameter oversimplifies past data and may reduce forecasting reliability. Starting from section 6, Li provides empirical analyses illustrating the application of the Gaussian copula within credit risk frameworks.