Statistical Model to Predict IPO Prices for Semiconductor
1. Statistical Model to Predict IPO Prices for Semiconductor & Semiconductor
Equipment Industry
Manuel G. Russon, Ph.D.
St. John’s University
Xuanhua “Peter” Yin
St. John’s University
Abstract
In the process of IPO, multiple valuation methods are utilized. Conventionally, issuer valuate
equities using intrinsic and relative valuation methods to determine the value of common equity.
However, now more and more statistical modeling is used in the valuation process. The purpose
of this research is to test multiple variables regarding the performance and book value of firms in
semiconductor & semiconductor equipment industry and create a model to predict IPO price in
the industry. Addition to that, this research will utilize mean absolute percentage error(MAPE) to
determine the degree of fitness rather than R-squared conventionally and test the effect of
dividend in corporation operation in semiconductor & semiconductor equipment industry.
I. Introduction
IPO is a crucial process in a corporation’s life cycle. Through IPO, corporation will be able to
gain the capital it needs for its future growth, and in this process equity pricing is the most
delicate but important procedure. A fair IPO valuation will help the corporation’s success in the
IPO and deliver the capital in need.
Parties, who are interest in the results, are investment bankers, institutional investors and
semiconductor & semiconductor equipment companies.
II. Methodology
+ + ?
P = f(Income Statement, Asset Value, Dividend Rate)
1. Data
The research using cross section data from FactSet in the year 2013. The dependent variable will
be stock price and the independent variables are in three categories, representing the three source
of equity value: book value, profitability and dividend.
i. Book value per(bvps) share will be used, representing book value.
ii. Dividend per sahre(dvps) will be used, representing dividend.
2. iii. For profitability(isvar), earnings per share(eps), earnings per share forward(epsfd),
earnings per share without extraordinary item(epsxo), cash flow per share(cfps), sales
per share(sps), operating income per share(oincps), and net income per share(nincps),
will be used.
2. Techniques
This research will use histogram, scatter plots and sequence chart to rule out outliers and
descriptive statistics and correlation matrix for multicollinearity. And use regression analysis,
linear and gam model, to come up with the prediction model.
Importantly, this research will use R-squared and MAPE as criteria to select the optimum model.
Conventionally, R-squared, the coefficient of determination is used to evaluation the quality of a
regression model. But R-squared is highly sensitive to outliers, which in a lot of situations affects
the quality of R-squared. Both R-squared and MAPE will be used in the research and conclusion
will be drawn based on both.
The mean absolute percentage error(MAPE), displayed in Eqn. 1 below, also known as mean
absolute percentage deviation(MAPD), is a measure of prediction accuracy of forecasting
method in statistics.
𝑀 =
100
𝑛
∑ |
𝐴 𝑡−𝐹𝑡
𝐴 𝑡
|𝑛
𝑡=1 (1)
Where: At is the actual value
Ft is the predicted value
n is the sample size
3. Model
Eqn 2, 3 and 4 display the functional specification, population regression line and sample
regression line in the estimation of price.
+ + +
Price = f(bvps, dvps, isvar) (2)
Price = α+β1*bvps+ β2*dvps+ β3*isvar (3)
Price = α+b1*bvps+ b2*dvps+ b3*isvar (4)
III. Results
Table 1 displays the descriptive statistics of variables tested.
Table 1 Descriptive Statistics
vars n mean sd
media
n skew
kurtosi
s