This New Experimental Method of Pricing Securities Appears to Improve Prediction
The blended, multifactor model co-developed at Stevens and Cornell proved superior in tests when building international portfolios
A Stevens-Cornell team has proposed a novel multi-factor model for researchers pricing companies and assets.
Stevens School of Business researcher Ying Wu, working with Cornell University economist Andrew Karolyi, detailed the model in the Journal of Financial and Quantitative Analysis in April. The model promises greater power and accuracy — particularly when assessing smaller, lesser-known companies in a known industry space — by accounting for both local and global information and effects.
Mining big financial data for insights
Previous studies chiefly had utilized two distinct methodologies to price assets: one a purely integrated scenario, considering global trade to be completely unhindered by national boundaries and economies, and another essentially treating nations as isolated economic entities.
"We don't agree with either model," says Wu. "A blend seems intuitive, but in fact the methods being used have not attempted to bridge this gap. Our work is aimed at finding a middle way more closely aligned with contemporary markets."
To build their new method, Wu and Karolyi began by evaluating returns for more than 37,000 international stocks in 46 developed and emerging nations, dating back over a period of 20 years.
After dividing companies into two categories — approximately 5,700 traded on multiple major exchanges, and thus considered to be globally traded, the rest considered locally traded — Wu and Karolyi simulated and tested a host of portfolios. Information about industry trends and leaders, they found, colored perceptions of unknown, smaller companies' valuations and operations in given spaces.
The duo then used those insights to build an improved pricing model integrating traditional factors, such as company size, market valuation and momentum, as well as variables quantifying the influence of information about global firms on investment activity in local companies.
Why a hybrid model fits a global economy
Theoretical models constructed this way can help better guide researchers and investors, Wu believes, because they better reflect modern economics.
"Today we have a real hybrid between a truly global economy and, in some cases, still very localized industry sectors," explains Wu. "A foreign investor may not know much at all about a small internet company in China or be able to get information on the ground like a local investor would.
"So that investor will naturally look to a company like Baidu or Alibaba as a guide to pricing, and to news about those companies as a guide to its future potential when trying to understand it.
"This model, in a sense, quantifies that intuitive process."
Wu and Karolyi also performed additional research on the influence of currency risk in the pricing of public stocks, analyzing the same 37,000-stock dataset for relationships to approximately 40 currencies. They learned that currency risks are sometimes (but not always) integrated in stock-pricing decisions. The resulting publication was awarded a Best Paper award at the Multinational Finance Society's annual conference in Budapest in June.
