Yanwei (Wayne) Zhang

Currently, I am a senior statistician at the CNA insurance company, located at downtown Chicago. Hired by the statistical research department in 2009, I have been working with a group of statisticians and data specialists to provide predictive analytics service to internal customers. Much of the work involves the construction of statistical models to enable better quantification and identification of insurance loss risks, and more accurate underwriting and marketing efforts for a variety of property and casualty insurance products. In this process, I designed a number of analytics tools that provide innovative solutions to many challenging problems that current insurance companies are struggling with. An incomplete list of these includes:

• A SAS tool kit that visualizes the goodness-of-fit, statisticial significance and the effectiveness of a predictive model;
• A SAS macro function that implements a customized variable selection procedure within the generalized linear models, allowing automation and user intervention;
• A mapping tool that creates a visual representation of demographic information or insurance exposures.

Aside from insurance pricing, I have also been actively working in the area of insurance loss reserving, a practice to forecast ultimate losses for a cohort of insurance polices in-force or written in the past. The work Zhang (2010a) builds the connection between many existing loss reserving models in the actuarial literature and the seemingly unrelated regressions from the statistics field, and further proposes a general framework to analyze loss reserving problem to account for dependencies among multiple triangles. A more recent work Zhang et al. (2010b) employs a nonlinear Bayesian heirarchical model to address many existing problems in the area of loss reserves: e.g., estimating predictive distribution of loss reserves; extrapolation of the development pattern beyond the range of observed data and appropriately accounts for the corresponding uncertainty; blending data from multiple companies to make robust estimation and carry out industry-level inference; input of expert opinion to reflect information not available in the data.

Moreover, I have great interest in statistical programming. I co-authored the R package ChainLadder that is broadly used in insurance loss reserving analysis. I also created several SAS macro functions (Zhang 2010c) that facilitate easy use of Perl regular expression in the SAS macro environment.