IPO Valuation in the New and Old economies

1
IPO Valuation in the New and Old economies

• Sanjai Bhagat
• Srinivasan Rangan
• University of Colorado at Boulder

2
Outline

• Motivation
• Research Questions
• Prior Literature
• Model Specification
• Sample and Data
• Results
• Conclusions

3
Motivation

• The second half of the 1990s, which witnessed
  several significant innovations in the technology
  sector and the rise of the internet sector, has
  been labeled as the new economy period.
• In the new economy period (or boom period),
  equity values, especially those of initial public
  offering (IPO) firms, reached unprecedented
  heights and outpaced fundamentals
• This has prompted several commentators to raise
  questions about whether traditional valuation
  methods remain valid for IPOs.

4
Motivation

• Early profitability is not the key to value in a
  company like this (Inktomi).
• Jerry Kennelly, Chief Financial Officer of
  Inktomi Inc (1999)
• But valuations are just as often based on gut
  feel. As one entrepreneur told me, Its as if
  everybody just settles on a number that they are
  comfortable with.
• Gove (2000)

5
Motivation

• Were traditional value-relevant variables such as
  income and book value of equity valued
  differently in the boom period relative to an
  earlier time period for IPO firms?
• Also beginning from March 2000, the stock market
  in general took a dive (crash period). So the
  other question is How did these variables fare
  in the crash period?
• In this paper, we seek to provide descriptive
  evidence on shifts in the IPO valuation function
  in the boom period and crash period relative to a
  more stable period the late 1980s.

6
Motivation

• Another question that we consider is whether
  investment bankers and first-day investors agree
  on their valuations of different variables.
• We conduct this analysis by regresssing first-day
  market values on offer values and other
  variables.
• If the two sets of individuals agree with each
  other, the coefficient on offer value should
  equal one and the coefficient on other variables
  should equal zero.

7
Motivation

• Why is this important/interesting?
• Separate fact from fiction/anecdotes
• Fills gaps in the IPO valuation literature
• Stimulate further research into what factors
  drove the shifts in the valuation function

8
Discounted Cashflow Valuation

• where,
• n Life of the asset
• CFt Cashflow in period t
• r Discount rate reflecting the riskiness of the
  estimated cashflows

9
Research Questions

• Were the following variables valued differently
  by investment bankers and first-day investors in
  the boom and crash periods relative the second
  half of the 1980s?
• Income
• Book value of equity
• Sales
• RD
• Industry price-to-sales ratios
• Insider retention
• Were the valuation of these variables different
  for tech firms, internet firms, and loss firms?

10
Priors / Expectations

• Based on anecdotes, we expected that income would
  be valued less in the boom period relative to the
  1980s
• Based on anecdotes, we expected that sales would
  be valued more in the boom period relative to the
  1980s
• We had no priors on how things would change in
  the crash period and so we let the data speak.
• We also expected insider retention to be valued
  more in the boom period relative to the 1980s
  (substitution)
• For technology and internet firms, we expected
  income and sales to be less valuable and insider
  retention to be more valuable (substitution)
• For loss firms, we expected income to be valued
  less (Hayn (1995) and Basu (1997)) and insider
  retention to be valued more (substitution).

11
Prior Literature

• Valuation of ownership
• Leland and Pyle (1977)
• Downes and Heinkel (1982)
• Ritter (1984)
• Feltham, Hughes, and Simunic (1991)
• Valuation of financial information
• Klein (1996)
• Kim and Ritter (1999)
• Beatty, Riffe, and Thompson (2000)

12
Prior Literature

• Valuation of internet IPOs
• Hand (2000)
• Bartov, Mohanram, and Seethamraju (2002)
• Inter-temporal changes in the valuation function
• Core, Guay, and Buskirk (2003)
• Demers and Lev (2001)
• Keating, Lys, and Magee (2003)

13
Prior Literature

• How do we extend the literature?
• First, sample periods of most prior studies of
  IPO valuation do not cover the late nineties,
  2000, and 2001.
• Second, none of the prior studies share our
  research focus, which is to examine
  inter-temporal shifts in the IPO valuation
  function.
• Third, with the exception of BRT, sample sizes in
  prior studies are small or based on one industry,
  the internet, and hence limit the
  generalizability of their conclusions.

14
Prior Literature

• Fourth, while prior studies have examined the
  determinants of the first-day closing value, they
  have not modeled this value conditional on the
  offer value.
• Fifth, prior research has examined ownership
  retention by pre-IPO shareholders only as an
  aggregate signal we extend this research by
  studying the value implications of the ownership
  of four classes of shareholders CEOs, other
  officers and directors, venture capitalists, and
  other five percent blockholders.

15
Model Specification

• Dependent variable choices
• Price-to-earnings ratios or Price-to-sales ratios
• Price per share
• Offer value in millions of dollars
• Logarithm of Offer value
• Independent variables (expected signs)
• Year -1 Income before extraordinary items and RD
  ()
• Year -1 Book value of equity ()
• Year -1 Sales ()
• Year -1 RD ()
• Pre-IPO Industry median price-to-sales ratio ()
• Post-IPO insider retention ()

16
Model Specification

• Basic Model

OV Offer value INCBRD Income before
extraordinary items and RD in year 1 BV
Book value equity at the end of year 1 SALES
Sales for year 1 RD Research and development
costs in year 1 INDPS Median industry
price-to-sales comparable of recent
IPOs INSRET Percentage of the post-IPO firm
owned by pre-offering shareholders.
17
Model Specification

• Measuring Median Industry price-to-sales ratio
• Based on five most recent IPOs within the last
  two years from the same four digit SIC code
• Multiplied by Sales of Firm
• RD add back.
• Using RD stock instead annual RD in year -1
  does not change results.

18
Model Specification

• We substitute aggregate insider retention with
  ownership levels of four categories of owners
• CEO
• Officers Directors
• Venture Capitalists
• Other 5 Blockholders
• We also examine the impact of changes in
  percentage ownership around the IPO.
• We include sales growth as an additional proxy
  for growth prospects and our results are
  unchanged.

19
Model Specification

• Our main research goal is to test for
  inter-temporal shifts in the valuation function.
• Therefore, we expand the basic model by adding
  binary dummy variables for the boom period and
  the crash period.
• We also construct dummies for loss firms,
  technology firms, internet firms and for
  interactions of these dummies with the six basic
  variables.

20
Model Specification

• Expanded Model

21
Model Specification

• Boom 1 if the offer date is during
  1/1997-3/2000, and 0 otherwise.
• Crash 1 if the offer date is during
  4/2000-12/2001, and 0 otherwise.
• Loss 1 if income before extraordinary items is
  negative, and 0 otherwise.
• Tech 1 if a firm belongs a technology
  industry, and 0 otherwise.
• Internet 1 if a firm belongs to an internet
  industry, and 0 otherwise.

22
Model Specification

• Logarithmic specification
• Hand (2000) and Ye and Finn (2000), Beatty,
  Riffe, and Thompson (2001).
• Reduces heteroscedasticity and influence of
  outliers
• L(W) loge(1W) when W gt 0 in millions
• L(W) -loge(1-W) when W lt 0 in millions.
• The transformation is monotone and one-to-one and
  ensures that L(W) is defined when W is zero or
  close to zero.
• Is the best model based on Box-Cox analysis.

23
Model Estimation

• Robust regression
• OLS is justified by the fact that it is best
  linear unbiased estimate of linear model
  coefficients, and the overall best estimate when
  regression residuals are normally distributed.
• Additionally, if residuals are normally
  distributed we have convenient access to a
  distribution theory for inference.
• However, when residuals are non-normal, OLS is no
  longer the most efficient estimator.
• In contrast to OLS which estimates the
  conditional mean, quantile regression estimates
  the conditional median.
• It is less sensitive to outliers in the dependent
  variable.
• It may be more efficient than OLS.

24
Sample and Data

• Eighties Nineties
• Initial Sample non-financial companies,
• firm-commitment offerings, US companies,
• not units, not spinoffs, not LPs,
• proceeds gt 5 million 718
  1,381
• Delete misclassifications 51
  64
• Delete not listed on Compustat
  26 0
• Delete dual-class IPOs 0 87
• No prospectuses 2 10
• Final sample 633
  1,222
• Final sample with data for all variables
  (1855-230) 1,625 IPOs

25
Sample and Data

• Data sources
• Financial Data (years -1,-2,-3) prospectuses
• Ownership data prospectuses
• Offer date, offer price, shares issued SDC
• Shares Outstanding Ljungqvist and Wilhelm
  (2003), prospectuses
• Stock price and return data CRSP
• Industry comparables COMPUSTAT
• Industry classification Loughran and Ritter
  (2003)
• In process of collecting data on cash flows, age,
  underwriter reputation, long-term debt.

26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
Results Offer Values

• For profitable non-tech firms in the 1980s
• income, sales, RD, industry price-to-sales
  ratios, and insider retention are positively
  related to offer values
• book value of equity is unrelated to offer value
  of equity.
• Income is valued more in the boom period, for
  tech firms, and internet firms (contrary to
  expectation). Valuation of income has remained
  stable across the boom and crash periods.
• Income of loss firms is valued negatively.

31
Results Offer Values

• In the boom period, sales were valued less in the
  boom period relative to the 1980s (contrary to
  expectation). Coefficients on sales are lower
  for tech and internet firms Sales became more
  valuable in the crash period.
• Consistent with expectation, insider retention is
  valued higher in the boom period, for tech firms,
  for internet firms, and for loss firms.

32
Results Market Values

• In regressions of first-day market values
  conditional on offer values we find that
• Except for sales and RD in the crash period,
  investment bankers and first-day investors tend
  to value financial and growth variables similarly
• In general, first-day investors and investment
  bankers tend to disagree about the value
  relevance of insider retention. For example, for
  profitable non-tech firms first-day investors
  value insider retention less by 1.65 million
• Value differential became smaller in the boom
  period relative to the 1980s
• First-day investors assign a lower valuation than
  do investment bankers for tech firms, for
  internet firms, and for loss firms.

33
Conclusions

• Shifts in firm characteristics contributed to
  shifts in offer values in recent times.
• We also find evidence of parameter variation for
  several financial variables across time-periods
  and industries.
• Detailed information on ownership structure is
  incrementally useful in explaining IPO values
  (compared to one aggregate number).
• We believe the analysis of first-day market
  values conditional on offer values is a fruitful
  way to understanding differences between
  investment banker and market assessments of IPO
  prospects.