TEXT

Abstract

"Wisdom of Crowds" is the theory that an aggregation of distinct estimates will be more accurate than each individual estimate. Dating from its original application in the 19th century, when it was employed to estimate the weight of an ox at a county fair, this theory now has applications across science and financial markets. An extension of the theory claims that subsets of wise crowds can sometimes have greater predictive power than the crowd as a whole. This thesis aims to take advantage of the "wisdom of crowds" to distill predictive information from large sets of survey data representing the spending intentions of major commercial businesses on technology products. This work proposes two models to assess the survey data's ability to predict the financial performance of the vendors of these products, one focused on earnings surprise and another on stock return. After assessing the information from the entire survey using these models, we then seek to identify subsets within the survey respondent crowd whose intentions correlate more precisely with the performance metrics. Thus, we are able to demonstrate a predictive relationship between survey respondents' spending intentions and vendor performance, as well as identify more predictive "expert" sub-crowds.