Bayesian Stochastic Betas for Stocks

The attached paper, “Bayesian Analysis of Stochastic Betas??, provides a method for estimating the beta of a stock to the market. This method, SBETA, uses a Gibbs Sampler, and the Appendix of the paper provides the required conditional posterior distributions to employ with the Gibbs Sampler. The model being estimated is described on pages 3-5. The estimation methodology is described on pages 6-8 and in the Appendix. We want code in C++ or C# (preferred) that can read in a time series of stock returns and market excess returns, (i.e., market return minus risk-free rate) and output estimates of the four parameter values of the SBETA, {a, d, sb, s}, along with the numerical standard error of each parameter estimate using Geweke’s method (see page 15). Sample stock and market data will be provided. ## Deliverables Complete and fully-functional code in C# or C++ that can read our sample data and output the four SBETA parameters and associated standard error. Code must be fully commented in clear English. We also require that the simulation study outlined on page 9 of the paper be implemented as a test and confirmation of the implementation. I.e., it must create simulation series like those described in the paper and be able to recover the simulation parameters using the same methods that it applies to our sample data. An example of what we expect is given in Table 1 of the paper. Performance must be on the order of that described in the paper at the bottom of page 11. * * *This broadcast message was sent to all bidders on Friday Apr 15, 2005 10:34:27 AM: As a point of clarification: This project does not require a GUI or anything other than file I/O as an interface. We will provide sample data as text CSV files (and we can even provide simple C#/C++ functions to parse CSV fields). Likewise, the test function can output to a text file. What we will be focusing on is the code itself, where we expect the comments to clearly show that it implements the substance of this paper and produces the same results. ## Platform We will compile and test the code on Windows XP with Microsoft Visual C++ or C#.