Creating an Optimal Subset of Financial Statement Variables to Value Private Companies

Abstract

This thesis analyzes the best method for producing the optimal subset of financial statement variables in order to predict private company valuation. Since private company data is hard to obtain, companies at their initial public offering (IPO) are used as a proxy. The data used was a combination of the valuation of companies at their IPO, as well as all of the financial statement variables found on their S-1. Three optimal subsets are produced using the methods of forward stepwise selection, backward stepwise selection and sequential replacement. The resulting models were compared based on their RMSE and R2 values. The resulting optimal subset was determined to be the one of subset size 5 produced from the backward stepwise selection model. It includes the financial statement variables: Common Shares Used to Calculate Earnings Per Share Basic, Deferred Revenue Current, Earnings Before Interest and Taxes (EBIT), Current Liabilities Other Sundry, and Current Liabilities Other Excluding Deferred Revenue. The coefficients for EBIT and Common Shares Used to Calculate Earnings Per Share Basic were compared with the different subset sizes they are found in and it was discovered that their value remains steady as new variables are added into the model. This thesis lays the groundwork to potentially apply the resulting optimal model to data surrounding private companies partaking in M&A transactions to assist in their valuation.