(PRWEB) September 26, 2012
Sentrana Inc., a Washington DC-based leader in scientific marketing solutions, reported today on the publication of research by Principal Scientist Liuxia Wang, Ph.D., which reveals an unexpected relationship between the stock price and certain financial performance metrics of a top foodservice company. Dr. Wang’s paper, entitled “Bayesian principal component regression with data-driven component selection,” describes the creation of a hierarchical probabilistic principal component regression model that outperforms existing methods and more accurately identifies the factors most responsible for stock price movement.
The paper’s case study, which investigates the relationship between stock price and financial performance metrics, somewhat unexpectedly reveals that revenue increases do not always lift a stock price. For a foodservice company with a long history, such as the one in the case study, revenue increases by themselves do not positively impact stock price if gross profit or other financial achievement is negatively impacted to spur revenue growth. The model developed from historical performance data clearly shows that shareholders value gross profit, net earnings, and profit margin more than revenue increases.
For this analysis, Dr. Wang used financial performance and share price data from 12 quarters between 2002 and 2006. The covariates, which are variables that are possibly predictive of outcomes, were the S&P500 index, quarterly revenue, quarterly gross profit, quarterly net earnings, quarterly profit margin and quarterly dividend. The actual results over time were compared to the stock price using Bayesian principal component regression (BPCR), and compared to the Hwang-Nettleton method and the classical principal component regression (PCR) method. BPCR was the best performer compared to the actual results, but the additional value of BPCR is that it is well-suited for real world research when the researcher wants to understand the impact of all the variables, especially the highly correlated variables, used in the model.
“BPCR is valuable in real-world situations that involve highly correlated variables when a standard regression model can’t determine which has more impact,” said Dr. Wang. “For example, customer retention is an area where the margins as well as the total revenue from a customer are both important in determining whether that customer will continue to purchase from a distributor. BPCR can uncover the subtle differences between these two variables and identify which is the most important.”
As in the case of customer retention or any analysis involving highly correlated variables, this new method is optimal for identifying the true relationships between key variables and their impact on outcomes.
To preview or purchase the article, please visit: http://www.tandfonline.com/doi/abs/10.1080/02664763.2011.644524