Journal of Finance and Accountancy

An empirical analysis of initial public offering (IPO) price

performance in the United States
Zachary A. Smith
Saint Leo University

Abstract:

For decades, researchers have disagreed about the magnitude and predictability of
abnormal securities’ price performance generated by initial public offerings (IPOs). The purpose
of this study was to identify the best specified and most powerful method of abnormal
performance detection and to apply ply this method to examine the price performance of IPOs.
Matched by size, industry, and book
book-to-market
market ratios this study explored which ofo the resulting
seven portfolio and matched-firm
firm methods of abnormal performance detection produced the best
specified and most powerful test statistics. The paper additionally analyzes IPO price
performance to determine if IPOs generate abnormal performance. The researcher used the event
study approach for the research design along with the buy and hold abnormal return (BHAR)(
method of calculating abnormal returns to conduct this analysis. The findings were that (a) all of
the matched-firm
firm methods of abnormal performance detection were well specified and powerful
(matching by industry affiliation generated the best power and specification result)
result and (b) that
the IPOs generated statistically significant abnormal priprice
ce performances occurring in: (1) short-
short
term analyses, (2) longer-term
term analyses, and (3(3)) analyses of the lockup and quiet periods.

Key words: Event study, IPO performance, Quiet period, Lockup period, Specification and
power analysis, Short- and long-term
term abnormal perfor
performance,
mance, Initial public offering

An Empirical Analysis, Page 1

 Journal of Finance and Accountancy

INTRODUCTION

This research project will provide the reader with a thorough understanding of the
anomalies relateded to IPO price performance, by canvassing the population of IPOs that went
public on U.S. financial
nancial exchanges from 19851985-2008. There were four tests of abnormal IPO
performance carried out in this research project, they are tests for: (a) abnormally positive
positi pre-
market and initial day of trade performance, (b) abnormally negative longer longer-term
term performance,
(c) abnormally negative performance occurring around the expiration of the lockup period, and
(d) abnormally positive performance occurring during the expi expiration
ration of the quiet period. In
addition to the preceding tests, this study seeks to determine which of seven portfolio-matching
portfolio
(PM) and matched firm (MF) strategies are the best-specified
specified and most powerful estimators of
normal performance. The matching st strategies
rategies evaluated in this analysis were PM techniques by
market capitalization, industry affiliation, and market capitalization and book book-to--market ratios
and MF techniques by market capitalization, industry affiliation, industry affiliation and market
capitalization,
italization, and market capitalization and book
book-to-market
market ratios. This study relied on the use
of the event study methodology throughout the analysis and the calculation of abnormal returns
by the buy and hold abnormal return (BHAR) method.
The main results
lts and conclusions reached in this analysis were as follows. First, this
paper illustrates how poor, in regards to specification and power, PM techniques performed in
detecting abnormal performance;; on a positive note, all of the MF strategies used to estimate
est
abnormal performance
mance performed remarkably well. IInterestingly
nterestingly enough, the MF approach by
industry affiliation outperformed the more popular approach
approach—matching
matching by market capitalization
and book-to-market
market ratios. Second, it is apparent that the initial abnormal performance, between
the offer and initial trading of shares, is substantial
substantial—the
the current study estimates this abnormal
performance at 11.74%—however,
however, abnormal performance is not constrained to pre-market
pre
trading.. During the initial trading day, IPOs in this sample generated abnormally positive
performance of 3.44%. Third, this analysis illustrates that IPOs experience substantial long-term
long
underperformance up to three years after their initial unseasoned equity offering, when compared
against
st firms matched based upon industry affiliation. Finally, IPOs experience significant
abnormally positive performance in the five five-day
day period surrounding the expiration of the quiet
period of 1.64% and a significantly negative abnormal performance of 1.00% around the
expiration of the lockup period.
This paper continues as follows. Section I introduces the theory, empirical work, and
conceptual framework of the hypotheses related to IPO performance. Section II presents the
proposed methodology. Section II IIII presents the results of the current analysis. Section IV
provides a summary of the work and concludes.

LITERARY REVIEW

Studies of IPO performance have been concentrated in two general veins of inquiry: (a)
why do IPOs generate abnormal performance and (b) to what extent is this performance
abnormal. This project focus on addressing the second of the two preceding questions, namely
the significance of this abnormal
ormal performance
performance.. Many researchers have attempted to answer this
question (e.g. Affleck-Graves,
Graves, Hedge, & Miller, 1996; Ibbotson, 1975; Loughran & Ritter, 2004;
Reilly & Hatfield, 1969), but questions regarding their methods used to identify abnormal
performance
rmance have arisen (e.g. Brav, Geczy, & Gompers, 2000; Brown & Weinstein, 1985;
Cheng, Chueng, & Po, 2004; Schultz, 2003). The question that this project seeks to illuminate is

An Empirical Analysis, Page 2

 Journal of Finance and Accountancy

as follows: If IPOs generate abnormal performance, when is this abnormal IPO performance
pe
significant and how should academics measure this performance?

METHOD

Traditionally, researchers attempt

attempted to measure abnormal performance in research
conducted on financial data sets using the event study methodology. This method seems to have ha
been pioneered by Ball and Brown (1968) and Fama, Fisher, Jensen, and Roll (1969), in which
the researchers analyze the impact of information on the performance of a publicly traded
security; however, according to Campbell, Lo, and MacKinlay (1997) the first published event
study was conducted 1933. When en the researcher refers to traditional event study designs, like
those mentioned in Campbell et al. (1997), normally the researcher has the luxury of (a) an
estimation window, (b) event window, and (c) post post-event
event window. However, when dealing with
IPOs, the data lacks this estimation
ation period, which is suppose to provide the researcher with
normalized expectations of return behavior. Therefore, researchers have to find other means of
estimating normal return behavior.
Researchers have used various methods to accomplish this task. In recent research, these
methods normally fall into one of two categories: matched firm (MF) approaches or portfolio
matching strategies (PM). PM matching strategies have been carried out by Brav and Gompers
(1997), Carter, Dark, and Singh (1998), Gompers and Lerner (2005); contrarily, Bhabra and
Pettway (2003) and Perfect and Peterson (1997) focused solely on the MF approach; Finally,
Ritter, J. (1991) used both MF and PM strategies. Attempts have been made by Barber and Lyon
(1997), Mitchell and Stafford (2000), and others to generate tests of different methods analyzing
the ability of these different methods to detect abnormal performance
performance—thethe two general
approaches are categorized as event and calendar time analyses.
There are good arguments for and ag against
ainst using various methods; researchers typically
prefer one method to the other. This project is interested in evaluating the investor’s buy and
hold investment experience; therefore, because the buy buy-and-hold
hold investor’s returns are most
appropriately modeled
eled in event time
time, the method used to analyze the investment performance in
the current analysis was the buy and hold abnormal return (BHAR) method. The question that is
inherently difficult to answer is that if the researcher is interested in analyzing the he buy and hold
investor’s experience, in terms of abnormal performance, what would the best best-suited
suited method
look like? Mitchell
chell and Stafford (2000) found serious flaws in using the BHAR method to
conduct their analysis,, but they focus
focused their critiques on PM techniques, using bootstrapping
procedures or bootstrapped t-statistics
statistics, and stated that “BHARS have poor statistical properties,
producing biased test statistics in random samples” (p.302). Portfolio
Portfolio-matching
matching techniques
produce misspecified test statisticsics in random samples; however, the method ignored in the
Mitchell and Stafford (2000) 00) analysis is the MF approach to benchmarking, which produced
well-specified
specified test statistics in the present analysis. Barber and Lyon (1997) concluded that the
MF approach,, relying on a firm’s market capitalization and book book-to-market
market ratios to evaluate
firm performance, generated well well-specified
specified test statistics throughout their analysis. This project
focuses on measuring the buy-and and-hold investor’s investment experience; therefore,
refore, the
researcher applied the BHAR method throughout this analysis.

An Empirical Analysis, Page 3

 Journal of Finance and Accountancy

Research Questions

This paper is partitioned into four different research questions; this section
tion develops these
questions.. The first question that this research project endeavors to answer is which general
method, matched-firmfirm or portfolio
portfolio-matching technique, paired with firm-specific
specific information
(i.e. market capitalization, industry affiliation, and book
book-to-market ratios) provides
rovides the best
proxy for expected return. The next topic that this project seeks to address is whether short-term
abnormal performance occurs, in the process of issuing unseasoned equity shares; the researcher
will analyze the specific time horizons in ssegments, studied by pre-trade
trade and initial trading day
results. Lengthening this analysis the project will then seek to determine whether IPOs
underperform the market in longer
longer-term analyses. Finally, the researcher will evaluate whether
IPOs generate significant
ficant abnormal performance in the five
five-day
day period surrounding the
expiration of the quiet and lockup periods.

Specification and power analyses

nalyses.

The first portion of the hypothesis testing section will evaluate the performance of
potential methods used to identify abnormal performance in similar studies. There has been
significant debate regarding whether researchers should use the CAR or BHAR method of
calculating abnormal returnss when conducting event studies. IIn n the previous subsection, this
debate was articulated.. In this section, the discussion centers around which method of estimating
expected return should be use to conduct event studies, given that the BHAR method is the
appropriate method to use to estimate the extent of abnormal performance.
The
he majority of research projects that attempt to determine which method of abnormal
performance detection, PM or MF MF,, to use when conducting event studies conclude that the MF
approach works quite well (Ang and Zhang, 2004; Barber and Lyon, 1997). However,
researchers seem to continuously revert back to attempting to identify a method of abnormal
performance detection that relies on the construction of portfolio benchmarks. In Lyon, Barber,
and Tsai (1999) the researchers used skewness adjusted t statistics and empirically generated
distributions of mean long-term
term stock returns generated from pseudo portfolios, to compensate
for the following biases: (a) the new listing bias, (b) the rebalancing bias, (c) skewness bias, (d)
cross-sectional
sectional dependence, and/or (e) a bad model problem (p. 197). However, after these
efforts are undertaken, the MF approach used to detect abnormal performance, using market
capitalization and book-to-market
market ratio data to match, generated better
better-specified
specified test statistic
than either adjusted
justed portfolio technique.
This analysis will illustrate tha
that the entire set of MF approaches, used to detect abnormal
performance, generated well-specified
specified and relat
relatively powerful test statistics—prior
prior to making
additional changes to the models of abno
abnormal
rmal performance detection. This study assumes that the
sample sizes used to conduct the analyses and the 18 18+ year period that the study was runr will
minimize the impact of this bias on the results obtain
obtained.. Finally, this research project provides
evidence that whenn researchers use the MF approach to detect abnormal performance, combined
with an independent sampling technique, the model performs very well regardless of the
techniques researchers use to match the event firms. In summary, first, there are many man biases that
researchers can fall prey to when attempting to conduct event studies, second, when researchers
re
use PM techniques
chniques instead of MF approaches these biases are magnified, and,, third, the biases

An Empirical Analysis, Page 4

 Journal of Finance and Accountancy

affect the results of the PM techniques more than th

the MF approaches used to detect abnormal
performance.

Short-term
term abnormally positive p
performance.

The most visible abnormality that currently exists in studies of IPO performance is that
IPOs tend to produce extremely abnormally positive performance results a short duration after
going public. This excess abnormal return occurs either in the pre pre-issuance
issuance period or in the one-
one
day performance of the post-offering
offering period (see Krigman, Shaw, & Womack, 1999; Loughran
& Ritter, 2004; McDonald & Fisher, 1972 1972;; Reily & Hatfield, 1969). Miller and Reilly (1987)
found that the extent of this underperformance was approximately 9.87% (p. 34) and Ibbotson,
Sindelar, and Ritter (1994) reiterated this sentiment by concluding that “first
“first–day
day returns average
10-15%” (p. 66). Cheng, Cheung, and Po (2004) found, while studying IPO price performance
on the Hong Kong financial market, that no trading profits were obtainable once IPOs began
trading publicly (p. 853), this finding contrasts those reached in Miller and Reilly (1987),
( an
analysis of IPOs listed in the U.S. markets. Historically, researchers seem to have assumed that
IPOs obtained profits in the first trading day. Perhaps, they have ignored the negative social and
process implications attached to an empirical find
finding
ing that the positive IPO performance is
constrained to the pre-trading
trading period. If the abnormal performance is constrained between the
offer and issuance, then the distributions of shares, and whom the shares are distributed, become
a more fundamental question,
tion, in regards to affording investors with equal opportunities to profit.
This question is relevant because the underwriting syndicate holds an unfair informational
advantage over the majority of the investing public.

Long-term underperformance..

Researchers
rchers have also provided evidence in support of the theory that IPOs suffer from
long-term
term price underperformance when measured against standard benchmarks (see Affleck-Affleck
Graves, Hedge, & Miller, 1996; Ibbotson, 1975; Loughran, & Ritter, 1995; Ritter, 1991).
1991) Ritter
(1989) found that, in his sample of IPOs issued from 1975
1975-84, IPO’s 3-year
year holding period
returns (HPR) underperformed portfolios matched based upon market capitalization and industry
characteristics by 27.39% (p. 4); Ibbotson, Sindelar, and Ritte
Ritterr (1994) found similar results
analyzing IPO data from 1970-1990.
1990. Ritter (1989) and Ibbotson (1994) suggested that on
average IPOs underperform standard benchmarks from the end of the initial trading day to at
least the firm’s five-year
year publicly traded ann
anniversary.

Event-specific
specific Abnormal Performance
Performance.

Two events that occur systematically after a company issues unseasoned equity to the
public are thee expiration of the quiet and lockup periods.. Researchers have illustrated that these
two events produce abnormal performances in empirical analyses of event studies. However, the
directions of the abnormal performances that the two events generate are divergent, and
researchers have questioned the magnitude and causes of these abnormal performances. The
following two sections will define and review the literature related to the abnormal performance
that purportedly occurs during the expiration of the quiet and lockup periods.

An Empirical Analysis, Page 5

 Journal of Finance and Accountancy

At the conclusion of the

he quiet period, the SEC allows investment firms to initiate
coverage of a security. The reason why this period is so interesting is that Bradley, Jordan, and
Ritter (2003) have found that from 1996
1996-2000,
2000, for all IPOs issued, analysts initiated coverage on
76% of the newly issued IPOs, and of these 76%, analysts initiated coverage on 96% of these
issues as a strong buy or a buy (p. 33). This is not what the resea
researcher
rcher expected; structurally, the
researcher would prefer to see a distribution that, from a proprobabilistic
babilistic standpoint, firms rated
would just as likely receive a positive rating as a negative rating. According to Bradley et al.
(2003), when analysts initiate coverage immediately after the quiet period, the IPOs affected by
this event experienced a significantly
gnificantly positive abnormal return of 4.1% in a five five--day window
surrounding the quiet period (p. 33). If analysts left the newly issued IPOs uncovered at the
conclusion of their quiet period, firms experienced an insignificant abnormal return of 0.1% (see (se
Bradley et al., 2003, p. 33). In 2004, Bradley, Jordan, Ritter, and Wolf (2004) attempted to
expand this study to include IPOs that went public from January 2001 through July 2002; the
impact of the expiration of the quiet period during this time horizon was insignificant (p. 11). In
this study, the researcher endeavored to answer why the two research projects differed in regards
to their results and analyze whether abnormal performance is significant during the expiration of
the quiet period.
Researchers,, in the past, have not built a solid case to declare that abnormal performance
occurs as the lockup period expires. However, Field, and Hanka (2001) found that from 1988 to
1997, during the expiration of the lockup period, investors experienced a three three-da
day abnormally
negative performance of 1.5% (p. 471). The results from Garfinkle, Malkiel, and Bontas (2002)
were in agreement with Field et al. (2001), although the Garfinkle et al. (2002) found that
negative performance experienced during the expiration of the lockup period was to 4.47%. The
two different percentages vary remarkably and the methods that the researchers used to calculate
abnormal returns are quite different. A goal of this research project is to add clarity and
specificity to this potential anomaly.

METHODOLOGY

A discussion of the rationale behind the decisions to use the BHAR method to calculate
abnormal returns over the choice of the CAR method was address earlier in this document; this
section will describe how the researcher will implement the method and run the power and
specification analyses. Furthermore, the sample sizes are different in many of the analyses, even
though the study canvassed the entire time horizon, fr
from
om January 1985 to December 2008;
2008
because of incomplete records,
ds, the researcher conducted some of the analyses with much smaller
sample sizes than originally anticipated. This section will review the methodological procedures
applied to conduct the power and specification analyses as well as the test of abnormal
performance
formance and the samples sizes of each test.

Power and Specification Analyses

In this analysis, the researcher conducted the power and specification analyses based
partially on the methodology described in Ang and Zhang (2004). The following paragraphs
describe the adjustments made to the Ang et al. (2004) methodology. It is appropriate to note
here that two different power and specification analyses were run to determine how increases in
sample sizes would influence the metrics ability two identify abno
abnormal
rmal performance. In the first

An Empirical Analysis, Page 6

 Journal of Finance and Accountancy

analysis, the researcher took 10 nonnon-repeating

repeating samples of 50 companies taken from the list of
Russell 3000 constituents each year of this analysis. Next, these 10 yearly samples of 50
companies were combined to produce larger samples (n = 500).
To evaluate which of the seven different benchmarking techniques generated the best
specified test statistics, the project needs a pool of random companies to evaluate the metric used
as a proxy for normal performance, a method used tto o calculate abnormal performance, and a
method to aggregate
gregate abnormal performances aacross
cross the sample. This analysis uses the BHAR
methodology to calculate abnormal performance and seven different methods based upon either
portfolio matching or matched-firm firm methods used to proxy for expected returns.
This project used two different procedures to obtain proxies for expected returns: (a) PM
and (b) MF. For the MF approach, the researcher extracted the sample firms used in this analysis
from a list of the components
onents of the Russell 3000 Index each year. If a company was included in
the Russell 3000 list of companies, the company w was eligible to be a MF in this analysis—each
analysis
year, the list was updated, from 1985
1985-2002,
2002, due to the addition and deletion of firms from
fro the list
of constituents each year. If a firm is matched based upon any singular firm characteristic (i.e.
market capitalizationn or industry affiliation), a pool of potential matched firms are identified
based upon this criteria, then these firms are then sorted and the closet match is selected;
furthermore, if there are multiple firms that meet the matching requirements, a number is
assigned to each potential match and a firm is randomly selected from the potential matches. If
two factors are included in the matching procedure (i.e. industry and market capitalization and
market capitalization and book-to to-market
market ratios), the firms are sorted by the most appropriate
factor (i.e. industry affiliation for the industry/market capitalization sort and market
capitalization
italization for the market capitalization and book
book-to-market
market sort) first, and then they
the are
sorted based upon the second factor.
When this research project
ject used the PM technique, it relied on external portfolios to
match the firm to a portfolio with a similar likeness. The procedure for matching was simple: the
researcher paired the event firm with a portfolio grouping compiled and maintained on Dr.
Kenneth French’s website (URL:
http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/d
http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html).
ata_library.html). The researcher then
carried out this matching procedure for PM approaches based upon industry affiliation, market
capitalization, and market capitalization and bookbook-to-market ratios.
After the pairings were made, the performance of the simulated event firm (randomly
drawn from the Russell 3000 list each year) and the matched firm or portfolio match were
compared, in terms of specification, for 1, 2, 3, and 44-year
year time horizons. The researcher then ran
the specification analysis to deter
determine
mine if a metric would conclude that abnormal performance
had occurred when in actuality it had not not—throughout
throughout this analysis the level of significance
significa was
set at 5%. Therefore, for each pairing the following Buy and Hold Abnormal Return (BHAR)
was calculated:

,  ∏
1  ,  ∏
1  ,  (1)

The researcher then pooled the result of this formula for each sample taken for this analysis and
then compiled the following summary statistics: (a) sample size, (b) sample average, (c) sample
standard deviation.
After the researcher compiled these statistics for each sample taken, he calculated the
following statistic, Barber and Lyon (1997):

An Empirical Analysis, Page 7

 Journal of Finance and Accountancy


,
  (2)
!", #/√&

The researcher then took the t statistics and grouped them based upon their respective sample to
generate the empirical
irical size (ES) statistic. The researcher calculated this statistic by taking each
sample of 50 or 500 observations contained in the yearly cohorts, summing the number of times
that the given metric identified abnormal performance, and dividing this sum bby y total number of
observations contained in the cohort.
The power analysis uses the results of the specification analysis as a base to continue the
evaluation of the given method of abnormal performance det detection.
ection. All models of normal
performance that have properly constructed should identify no abnormal performance,
performance given that
enough random draws were taken from a randomly selected population of return data. From a
base of zero abnormal performance, abnormal performance is simulated across the entire sample samp
by taking the average performance add adding either positive or negative percentage movements
of 1, 5, 10, 15, 20, 30, 50, and 75%.
Using the outcome from these simulations, this project then calculates the tBHAR for each
level of simulated abnormal performance.
rformance. The researcher expects abnormal performance to be
negligible where zero abnormal performance is simulated and increase as both positive and
negative abnormal performances are simulated; therefore, when charting the results of this
analysis the researcher is looking to obtain a v or uu-shaped
shaped power curve, centered on zero
abnormal performance and increasing substantially as abnormal performance is simulated. simulated After
simulating abnormal performance, for each simulation and each metric, the Empirical Power
(EP) statistic was calculated. The EP statistic is similar to the ES statistic, but when the EP
statistic is calculated the researcher is analyzing how well a metric identifies the researcher’s
simulation of abnormal performance and when it does not. Thinking back to the u or v-shaped
curve, at zero percent simulated abnormal performance the researcher wants to see that the
metric does not identify abnormal per performance, therefore, the ES statistic would be zero, but at
each increment (positive and negative) from 1% to ∞, the researcher wants to see that the
metric’s ability to detect abnormal performance increases.

erformance.
Short-term abnormal performance

The time horizon

orizon used to evaluate short
short-term
term abnormal performance shrank significantly
when compared to the other studies in this analysis. Using the sources that were available
(Hoovers IPO Central, Edgar IPO), this study was able to obtain premarket offering prices price for
the IPOs included in this analysis. The researcher used the following time horizon, January 1,
1997 to December 22, 2005, for tests conducted on the performance of the initial day of public
trading and April 12, 1996 to January 28, 2008 for tests con conducted on pre-trade
trade performance.
Although, this is a substantial reduction in the intended sample, there were still a significant
number of observations in each samplesample—thethe researcher identified 1,876 observations for
premarket performance and 2,143 observa observations
tions for the initial day of trading. Even if it was
possible to obtain performance data prior to January 1, 1997 for the initial day of trading in IPOs,
the CRSP database, which was used to obtain daily pricing data in this analysis, did not have
initial day trading data for IPOs listed prior to January 1997.
Itt is important to note, in the analysis of pre-trade
trade performance, the public does not
openly share their return expectations and the research project lacks a specific time horizons (e.g.

An Empirical Analysis, Page 8

 Journal of Finance and Accountancy

does the offering

fering to issue period last 12 hours, 24 hours, 36 hours, or more) to match the return
of the event firm against. Therefore, the researcher compared the aggregate returns obtained in
the pre-public
public trading period with the returns obtained by investor’s invinvesting
esting in a market
proxy—the the researcher uses standard market indices to obtain this performance (e.g. Russell
3000, S & P 500, NASDAQ Index, etc.). Therefore, when the researcher evaluated abnormal
performances occurring in pre-public
public trading he started wi
with
th the assumption that the aggregate
IPO will produce a return of 0%, and compared this return against the return of the DJIA, Russell
3000, and NASDAQ indices, to gain some insight on how substantial premarket IPO
performance is. The researcher will rever
revert to using the best-specified
specified and most powerful method
used to detect abnormal performance in the remainder of the analyses, because the researcher has
obtained public trading data that can be compared to the event firm’s performance. Therefore, in
the initial
tial day of trading, the researcher will take the returns obtained in from the sample of IPOs
and, using the BHAR method to detect abnormal performance, match these firm’s to the best- best
specified and most powerful metric identified in the preceding po power and simulation analysis to
determine whether abnormal performance has occurredoccurred.

erformance.
Long-term abnormal performance

To obtain a general sample to run tests for longer

longer-term
term abnormal IPO performance, the
researcher used the Field-Ritter
Ritter dataset of founding dates, identified in Loughran and Ritter
(2004; as noted in http://bear.cba.ufl.edu/ritter/foundingdates.htm) for com
companies
panies that went
public from 1985 to 1996. Additionally, the researcher obtained information pertaining to IPO
issuance from 1996 to 2002 from on on-line
line IPO databases (e.g. Hoovers IPO Central, Edgar IPO).
The total sample of IPOs used in this analysis was 55,883.
,883. Any company that had an offer price of
less than $5 or were foreign offerings were removed from this listlist—this
this is the base IPO list used
for both the analysis of longer-term
term abnormal IPO performance and event specific abnormal IPO
performance. Using the best-specified
specified and most power method of abnormal performance
detection, the researcher paired the IPOs in this sample with the benchmark to determine whether
abnormal performance is evident using time horizons ranging from day 2 to trading day 750.

Event-Specific
Specific Abnormal IPO Performance

For tests of abnormal performance occurring during the expiration of the quiet and
lockup
up periods, the sample size shra
shrank
nk to 5,529 due to firm attrition. In this analysis, the event
horizon was the five-day
day period su
surrounding the day of the specific event—either
either the expiration
of the lockup period or the conclusion of the quiet period. The researcher compared the BHAR
obtained from the IPO experiencing the event against the benchmark and the results of these
individuall analyses were aggregate
aggregated to give an average BHAR for the entire sample of IPOs
issuing shares over this period.

RESULTS

This section provides the results of the tests that the researcher conducted to indentify
abnormal price performance related to th thee issuance of unseasoned IPO issuance. The first section
provides the results of the specification and power tests the researcher conducted on seven
metrics used to identify abnormal performance. Sections 2 through 4 will display the results of

An Empirical Analysis, Page 9

 Journal of Finance and Accountancy

tests that the researcher conducted to identify abnormal performance, using the best-specified
best
and most power testing procedure.

Specification and Power

The purpose of this section was to determine, which method of benchmarking was most
effective in testing for abnormal performance during the sample time horizon.. Based upon the
review of literature, the researcher employed two broad methodological strategies to conduct the
specification and power analyses
analyses--the portfolio matching and the matched-firm
firm approaches.
approache The
first subsection will present the specification results and the second subsection will present the
results of the power analysis.

Specification Analysis.

The first question that this analysis answered is as follows: in samples of 50 and 500
companies, how often did the randomly drawn event firm (i.e. drawn from the list of Russell
3000 constituents each year) generate statistically significant abnormal performance. After the
researcher conducted the hypothesis
othesis test for each sample, the number of rejections were added
together and divided by the number of observations, thus resulting in the ES statistics. The
researcher displayed the results of the specification analysis Table 1.
The researcher found that all of the approaches using the MF technique (i.e. matching
based upon market capitalization, industry affiliation, industry affiliation and market
capitalization, market capitalization and book
book-to-market
market ratios) were generally well specified,
using a level of significance of 5%. The MF approach based upon market capitalization, alone,
did incorrectly identify abnormal performance in 5.56% of its samples, using sample sizes of 50
and 11.11% with sample size of 500.
To determine if the metrics were misspecified, the researcher conducted an additional test
to determine whether the ES was significantly different from the theoretical 5% level of
significance—where α was the level of significance and n was the sample size. The ES interval
ranges from 1.82% to 8.18% for the 180 samples of 50 comp companies
anies and from negative 5.07% to
15.07% in the 18 samples of 500 companies
companies;; therefore, the percentage of the observations in the
samples that were previously rejected were within our error boundaries
boundaries. MF approaches,
matched based upon Market Capitalization
Capitalization,, Industry Affiliation and Market Capitalization, and
Market Capitalization and Book--to-Market
Market Ratios Companies, generated spurious rejections;
however, these rejections where not statistically different than the theoretical level of
significance. Even though
ugh they were not statistically different from the theoretical level of
significance used in this analysis, they were different. The best
best-specified MF approach used in
this analysis and the approach that did not identify abnormal performance greater than the t
theoretical level of significance
gnificance in any of the analyse
analyses was the MF technique based solely upon
industry affiliation.
Each of the PM strategies (i.e. matched by market capitalization, industry affiliation, and
market capitalization and book-to to-market ratios)
tios) rejected the null hypothesis; this indicates an
identification of abnormal
bnormal performance even though the researcher had not simulated abnormal
performance. Every specification test, using the PM techniques, regardless of how it was
matched to the event firm, generated misspecified test statistics and in all cases was significantly
different from the theoretical level of significance.

An Empirical Analysis, Page 10

 Journal of Finance and Accountancy

As illustrated in Table
le 1, as the sample size increases from 50 to 500, the observed
percentage of spurious rejections decreased using the MF approach. This occurs because, without
simulating abnormal performance, researchers would expect to detect no abnormal performance.
Given the preceding results, the research
researcher found and believes that it is evident that the MF
approach is a better-specified
specified method of abnormal performance detection than the PM strategies.
The researcher concluded that the best
best-specified MF approach iss implemented using MF
strategy, matching by industry affiliation.

Power Analysis.

The purpose of powerr analysis was to determine which method had the least type II error,
and which methodology had the highest power. This research project relied on running the power
analysis by simulating abnormal performance in +/ +/- .01, .05, .10, .15, .20, .30, .50, and .75
intervals to the individual BHARs derived from the results of the specification analysis. In
essence, this analysis forced the average abnormal performance away from zero and imposed
abnormal performance on the BHAR. The researcher calculated the EP statistic atistic by adding each
of the sample average BHARs, for each level of simulated abnormal performance, and dividing
this sample average by the size of each sample. Again, the researcher obtained 180 samples of 50
observations in the first round of the analy
analysis
sis and 18 samples of 500 companies in the second
round of the analysis.
All of the MF approaches had defined power curves—the the traditional U or V shaped—the
shaped
power curve are centered approximately centered on zero, the point where no abnormal
performance is simulated. In comparison,
omparison, the PM benchmarks had no defined structure or at least
not the structure needed to make credible inferences pertaining to the power of the benchmark.
Again, the PM benchmarks failed to approach acceptable standards that are necessary to judge
the benchmarks ability to detect abnormal performance; in the remaining analyses, the PM
techniques were not included because the researcher did not considered them to be meaningful meani
alternatives to the MF approach.
Iff the researcher simulated abnormal returns of 15%, the competing matched-firm
matched
approaches only rejected the null hypothesis (identifying abnormal performance) in
approximately 30% of samples using sample sizes of 50 observations
observations. When the sample is
expanded to 500 from 50, the MF approach oach identified abnormal performance in 80% of the
samples.. Therefore, as the sample size increases, the power curve narrows making the employed
methodology appropriate.
There is still no statistically significant difference between the various MF approaches to
benchmarking. When conducting the remainder of the tests the research project was concerned
with the speed at which the metric deteriorates. As the event horizon increased the method’s
ability to detect abnormal performance decreases. Comparing the event horizons of one-, one two-,
three-, and four-years
years using sample sizes of 50, this study found that a simulated abnormal
performance of +/- 10% will be detected in 17%, 10%, 6%, and 6% of the samples; in samples of
500 observations the percentage of detection are 80%, 55%, and 25%, and 11%, respectively. To
analyze the general ability lity of each of the MF approaches to detect abnormal performance, this
project now will ill identify when the metrics identify abnormal performance in 95% of the
analyses. The EP reached 95% at 15%, 15%, 30%, and ~40% of simulated abnormal
performance using an event horizon of one one-, two-, three-, and four-years,
years, respectively, and
sample sizes of 500 observations. Therefore, if researchers intend on using the matched firm

An Empirical Analysis, Page 11

 Journal of Finance and Accountancy

approaches identified in this analysis their sample sizes and predicted level of abnormal
performance should be significantly large.

Initial Performance

The following section on focuses on detecting abnormal performance during the initial
trading period. The main questions posit
posited in the following section was whether unseasoned
IPOs produced abnormalmal performances in the time pre preceding
ceding public trading and if this abnormal
performance continued into the first day of public trading. The results of the analysis conducted
prior to public trading are reported first and then an analysis of whether IPOs produced
produce abnormal
performances on their
heir first day of trading is reported.

IPO Performance (Pre-issuance)

issuance)

This project uses the average returns in this round of the analysis; there is no way to pair
event firms with another firm base upon firm specific criteria, because this performance occurs oc
prior to public trading. The average return that IPOs generated prior to public trading or from
their offering to their issuance to the public was 11.74%, with a sample standard deviation of
31.16%, and 1876 observations taken from April 12, 1996 to JJanuary anuary 29, 2008. The researcher
conducted a t test to determine if the 11.74% performance was statistically different from zero.
The resulting t statistic was 16.32, which was outside the critical value of 1.645 for a one-tailed
one
statistical test, given a 5%% level of significance.
The preceding analysis illustrated the difference between the performance obtained by
IPOs pre-public
public trading and an expectati
expectation
on of zero abnormal performance. Since this is the pre-
public trading period, there is no specific way to pair the individual IPO performance with a
benchmark. Therefore, the researcher aggregated the returns into monthly IPO cohorts, these
performance cohorts assumes that the investor obtains shares of the IPO O in the offering and sells
the shares at the initial
ial trade on the first day of public trading. In Table 22,, the researcher has
illustrated how abnormal IPOs perform
performance is in pre-public trading.
To make this analysis comparable to the results ob obtained
tained in the remainder of the analyses
contained in this project
oject the researcher paired these returns with the performances of standard
benchmarks over our time horizon. The researcher displayed the results of these comparisons in
Table 3. Table 3 shows the average monthly performance of IPO cohort versus those of DJIA,
Russell 3000, and the NASDAQ Composite IIndices over the period analyzed. As the numbers in i
Table 3 indicate, at 5% level of significance for a oneone-tail t test (t critical of 1.66), the researcher
rejected the null hypothesis for only the IPO sample, implying that the IPO group experienced
significant abnormally positive returns. None of the be benchmark
nchmark indices produced abnormal
returns.
The DJIA was the best performing benchmark out of the three potential benchmarks
chosen for this analysis; the project continues to analyze whether the IPO cohort significantly
outperformed the best performing in index, which was the DJIA in this period. The average
difference between the IPO cohort and the DJIA’s yearly average return was 8.41%, with a
sample standard deviation of 13.86%, and observations’ occurring over 139 months--the months
computed t statistics was 7.15. 5. Again, with a 95% level of significance for a one-tailed
one test the
critical value of t is 1.66; therefore, this research project rejects the nul
nulll hypothesis and identifies

An Empirical Analysis, Page 12

 Journal of Finance and Accountancy

statistically significant evidence that abnormal performance occurred during the pre-public
p
trading period when compared against standard indices.

Initial Day of Public Trading

The next analysis determine

determines whether IPOs generate abnormal performance on the first
day of public trading. To answer the question, the analysis evaluates the returns of IPOs issued to
the public from January 1, 1997 to December 22, 2005, the sample contains 2,143 observations.
observati
Using a standard t test, this analysis uncovered the following: the average rreturn
eturn across the IPOs
was 3.44% and the average performance of the matchedmatched-firm
firm benchmark was 0.13%. The sample
standard deviation was 16.27%; resulting in a t value of 9.423, which when en compared to a
critical value of 1.645, at a 95% level of significance, indicated that the IPOs abnormal returns
on the first day of trade are statistically significant. The returns of IPOs on the first day of trade
are significantly different from the rreturns obtained for the MF benchmark.

Long-term
term Abnormal Performance

This round of the analysis turns to evaluating whether significant abnormal performances
occur after the short-term
term abnormal performances. This project accomplished its longer-term
longer
analysis by canvassing the population of IPOs issued in the U.S. from Janua
January
ry 1, 1985 to
December 31, 2002. The study identified 5,583 IPOs to use in this analysis; the researcher
matched these IPO based upon industry affiliation to a benchmark firm. The BHAR was
calculated and the researcher identified the sample average and stastandard
ndard deviation given the
individual BHARs. The output, which encompasses trading day 2 through 750, is the averaged
BHAR across the entire sample over the specified time horizon. The researcher evaluated the
data and generated a two-tailed t test for all 749 time-horizons.
The analysis of the data illustrates that, from trading days 5 through 12 IPOs significantly
underperformed the MF benchmark, at day 17 the trend changed positive, and it was statistically
significantly positive until trading day number 120 (with one insignificant reading on o day 33)--at
day 120 the BHAR wass 1.934%. The averaged BHAR continued along insignificantly, but
positive, until reaching trading day 161. However, the BHAR did not generate a significantly
negative BHAR until it reached 201 trading day. The BHAR remained significantly negative
through the
he remainder of the analysis. Moreover, at the end of year three the highest abnormally
negative performance occurred, which was -22.41%.

Quiet and Lockup Expiration

To construct a test for abnormal performance at the expiration of the lockup and quiet
periods this project canvasses the same population of IPOs used in the longer
longer-term
term analysis. The
number of observations for the quiet and lockup period analyses was 5529. To carry out these
analyses this section calculates the 55-day BHAR surrounding the date
te in which the quiet period
ended and the lockup period expired.

An Empirical Analysis, Page 13

 Journal of Finance and Accountancy

Quiet Period.

For the analysis of performance surrounding the expiration of the quiet period, the sample
average BHAR was 1.64%, for the five five-day
day period surrounding the event and the sample
standard deviation was 13.9%. The resulting t statistic was 8.75, using a 95% level
vel of
significance the critical value was 1.645; the null hypothesis is rejected
rejected—at
at the conclusion of the
quiet period IPOs produce a significantl
significantly positive abnormal performance.

Lockup Expiration.

In the analysis of the performance resulting from the expiration of the lockup period, the
researcher found significantly negative performance of 1.00%. In addition, the sample standard
deviation was 13.74%, therefore, the resulting t test produced a test statistic of –5.41,
5.41, and with a
5% level of significance the critical t value is -1.645.
1.645. Therefore, again, the researcher rejected the
null hypothesis and concluded that significant negative abnormal performance of 1.00% occurred
at the expiration of the lockup
ckup period.

SUMMARY & CONCLUSION

In the preceding section, the researchers has (a) presented a well specified and powerful
method used to identify abnormal performance when conducting event studies, (b) shown that
short-term
term abnormal IPO performance is positive, (c) illustrated that events occurring throughout
the IPO process instigate abnormal performances, and (d) provided a description of IPO
performance over the initial three years of seasoning. The results of the analyses related to event
specific performances--abnormal
abnormal performances occurring at the expiration of the quiet and
lockup periods--generated
generated significant, but not substantial abnormal performance. However, the
pre-public
public trade abnormal performance of 11% and 3% abnormal performance in the initial in
trading day, together with long-term
term underperformance of IPOs in excess of 30%, seem to
suggest that substantial performance abnormalities occur when companies issue unseasoned
equity shares to the public.
Researchers focus the majority of their expl
explanations
anations that attempt to explain why short-
short
term abnormal performance occurs on the asymmetric information hypothesis. To summarize,
according to Ritter and Welch (2002), either investors are more informed than the issuer about
the market demand for the company’s
pany’s shares or the investor believes that the issuer knows more
about the firm’s prospects and need protection against potential market lemons (IPOs
(IPO that
underperform). Purnanandam and Swaminathan (2004) questioned the conventional wisdom that
companies initially discount their shares when they offer them to the public, for whatever reason.
Purnanandam et al. (2004) found that, in a sample of over 2,000 IPOs issued from 1980 to 1997,
companies typically overpriced IPOs, when the researchers compared these IPOs to their non- non
IPO counterparts the over pricing ranged from 15% to 50%, depending on the matching criteria.
Puranandam et al. provided the first real critique of what has become general knowledge in the
academic community: Companies typically under pr price
ice their shares when they issue unseasoned
equity. If IPOs are initially overpriced and this overpricing increases
increases—not
not only in the period
prior to public trading, but IPOs continue to generate significantly positive abnormal
performance in their first day of trading
trading—does this signal market inefficiency?

An Empirical Analysis, Page 14

 Journal of Finance and Accountancy

It would be a mistake to conclude that empirical evidence supports the conjecture that
markets are inefficient. However, this initial over
over-pricing,
pricing, followed by substantial short-term
short
abnormally positive performance,
rformance, which is followed by by—over
over a period of three years—a
years
reversal to longer-term
term underperformance could at least hint at market inefficiency. Efficient
market theory concedes that short
short-term
term departures from fundamental or intrinsic will exist in the
marketplace; however, prices will rapidly adjust and the market will eliminate pricing
discrepancies. In the longer term analysis of IPO performance, the researcher found that when
IPOs are trading under their lockup provision, the returns are generally po positive.
sitive. However, as the
IPOs approach the expiration of the lockup period the performances generated by the IPOs
evaluated in this analysis were resoundingly negative.
The expiration of the lockup period occurs at approximately trading day number 128 (i.e.
180 calendar day lockup period is equivalent to ~26 weeks, subtracting the weekends equals 128
trading days). In this research project’s ex post analysis, after testin
testing
g all ex ante hypotheses, it
became apparent that the downward trend in IPO prices, following the expiration of the lockup
period, was remarkable. From rom trading day 128 to 350, which the researcher has approximated at
241 calendar days—one-year, there was a decline of .05% every trading day when compared
against a firm matched based upon industry affiliation. The regression summary is appealing, the
R Squared value was in excess of .98 and the relationship was very significant (p p = .001). The
trend is undeniable
able and significant. A
After
fter the IPO researches it’s lockup expiration, it is likley to
experience a downward trend rend of losing approximately .05 percentage points in value each day
that it trades for approximately one
one-year.
The general conclusion that the researcher has reached in this analysis is as follows,
“when it comes to participating in the IPO market, buyer beware.” Fi First,
rst, and foremost,
fore the
process of issuance is not fair, there are not fair opportunities for economic profit. A class of
sophisticatedd investors reap the benefits of the 11.74% of performance occuring prior to public
trading and in the initial trading day investors may be able to obtain approximately 3 percentage
points of positive performance, however, the investors have to buy at the market open and sell at
the closing price on the security’s initial trading day. If the average investor does not sell at the
market close, holding onto the newly issued security will generate a negative 3% price
movement from trading day 2 through trading day 7. This is then followed by a substantial
upswing in performance and, ofcourse, eventually if held long enough investors will feel the
sting of longer-term
term negative abnormal performance of 22.41% after approximately three years.
The researcher has provided
vided investors an overview of the patterns that IPOs seem to t have exhibit
from 1985 to 2008;; hopefully, the average investor finds a meaningful way to put this
infomration to use.

An Empirical Analysis, Page 15

 Journal of Finance and Accountancy

REFERENCES

Affleck-Graves,
Graves, J., Hegde, S., & Miller, R. E. (1996). Conditional price trends in the aftermarket
for initial public offerings. Financial Management, 25(4)
25(4), 25-40.
Ang, J. S., and Zhang, S. (2004). An evaluation of testing procedures for long horizon event
studies. Review of Quantitative Finance and Accounti
Accounting, 23: 251-274.
Ball, R., & Brown, P. (1968). An empirical evaluation of accounting income numbers. Journal of
Accounting Research, 159-178. 178.
Barber, B., and Lyon, J. (1996). Detecting abnormal operating performance: The empirical
em
power and specification off test statistics. Journal of Financial Economics, 41:
41 359-399.
Barber, B., and Lyon, J. (1997). Detecting long
long-run
run abnormal stock returns: The empirical power
and specification of test statistics. Journal of Financial Economics, 43: 341-372.
372.
Bhabra, H. S.,, & Pettway, R. H. (2003). IPO prospectus information and subsequent
performance. The Financial Review
Review, 38, 369-397.
Brav, A., Geczy, C., and Gompers, P. (2000). Is abnormal performance following equity
issuances anomalous? Journal of Financial Economics, 56(2): 209-249.
Brown, S., and Warner, J. (1985). Using daily stock returns: The case of event studies. Journal
of Financial Economics, 14: 3-31.
Brown, S. J., & Weinstein, M. (1985). Derived factors in event studies. Journal of Financial
Economics, 14(3), 491-5.
Campbell, J. Y., Lo, A. W., & MacKinlay, A. C. (1997). The econometrics of financial markets.
Princeton, NJ: Princeton University Press.
Carter, R. B., Dark, F. H., & Singh, A. K. (1998). Underwriter reputation, initial returns, and the
long-run performance of IPO stocks. The Journal of Finance, 53(1), 285-311. 311.
Cheng, W., Cheung, Y. L., & Po, K. (2004). A note on the intraday patterns of initial public
offerings: Evidence from Hong Kong. Journal of Business Finance & Accounting,
Accounting 31 (5-6),
837-860.
Fama, E. F, Fisher, L., Jensen, M. C., & Roll, R. (1969). The adjustment of stock prices to new
information. International Economic Review
Review, 10(1), 1-21.
Gompers, P. A., & Lerner, J. (2003). The really longlong-run
run performance of Initial Public Offerings:
The pre-NASDAQ
NASDAQ evidence. The Journal of Finance, 57(4), 1355-1392.
Ibbotson, R. G. (1975). Price performance of common stock new issues. Journal of o Financial
Economics (2), 235-272.
Kothari, S., and Warner, J. (1997). Measuring longlong-horizon security price performance.
ormance. Journal
of Financial Economics, 43: 301301-339.
Krigman, L., Shaw, W., and Womack, K. (1999). The persistence of IPO mispricing and the
predictive power of flipping. The Journal of Finance, 54(3), 1015-1044.
Loughran, T. and Ritter, J. (2004). Why hhas as IPO underpricing changed over time? Financial
Management, 33(3), 5-37.
Lyon, J., Barber, B., and Tsai, C. (1999). Improved methods for tests of long
long-run
run abnormal stock
returns. The Journal of Finance
Finance, 54(1): 165-201.
McDonald, J. and Fisher, A. (1972). New-issue stock price behavior. Journal of Finance, 27(1),
97-102.
Perfect, S. B., & Peterson, D. R. (1997). Day
Day-of-the-week effects in the long-run
run performance of
initial public offerings. The Financial Review, 32(1)
32(1), 49-70.

An Empirical Analysis, Page 16

 Journal of Finance and Accountancy

Purnanandum, A. K., & Swaminatham, B. (2004). Are IPOs really underpriced? Review of
Financial Studies, 17(3),, 811
811-848.
Reilly, F. and Hatfield, K. (1969). Investor experience with new stock issues. Financial Analysts
Journal, 25(5), 73-80.
Ritter, J. R. (1991). The long-run
run performance of initial public offerings. The Journal of Finance,
45(1), 3-27.
Ritter, J. R., and Welch, I. (2002). A Review of IPO Activity, Pricing, and Allocations. The
Journal of Finance, 57(4),, 1795
1795-1828.
Schultz, P. (2003). Pseudo market timing and ththe long-run
run underperformance of IPOs.
58(2), 483-517.
The Journal of Finance, 58(2)

An Empirical Analysis, Page 17

 Journal of Finance and Accountancy

TABLES & FIGURES

Table 1. The results of the specification analysis conducted on 180 samples of 50 companies and
18 samples of 500 companies.

Specification Analysis
180 Samples of 50 Simulated Event Firms
Matched Firm Portfolio Match
Years Mcap Ind Ind & MCap MCap & B toM Mcap Mcap & B to M Industry
1 5.00% 3.89% 4.44% 3.33% 43.89% 44.44% 46.67%
2 2.78% 1.67% 3.33% 2.78% 31.67% 25.56% 56.67%
3 2.20% 1.67% 2.22% 1.11% 33.33% 28.89% 65.56%
4 5.56% 3.89% 3.89% 2.78% 47.22% 36.67% 79.44%

18 Samples of 500 Simulated Event Firms

Matched Firm Portfolio Match
Years Mcap Ind Ind & MCap MCap & B toM MCap MCap & B to M Industry
1 0.00% 0.00% 0.00% 0.00% 83.33% 83.33% 66.67%
2 0.00% 0.00% 0.00% 5.56% 66.67% 66.67% 83.33%
3 11.11% 0.00% 0.00% 5.56% 66.67% 77.78% 94.44%
4 0.00% 0.00% 5.56% 0.00% 66.67% 61.11% 94.44%
Notes: Mcap - Market Capitalization, Ind - Industry Affiliation, B to M - Book to Market Ratio.

Public Issuance IPO Returns.

Table 2: Pre-Public

Pre-Public Trade - Average Monthly IPO Performance from 1997 to 2007

Average Standard Deviation of

Year Return Returns Number of Months t value tCritical - 1.796
2007 11.94% 6.34% 12 6.52 Reject
2006 9.72% 5.00% 12 6.73 Reject
2005 9.75% 5.02% 12 6.73 Reject
2004 9.25% 4.12% 12 7.77 Reject
2003 7.14% 6.68% 12 3.70 Reject
2002 5.52% 4.09% 12 4.68 Reject
2001 7.79% 6.30% 12 4.29 Reject
2000 34.60% 32.56% 12 3.68 Reject
1999 5.23% 7.33% 12 2.47 Reject
1998 -0.08% 4.41% 12 -0.06 Accept
1997 1.55% 2.16% 12 2.49 Reject

Notes: Table 2, provides the yearly returns of IPOs assuming that an investor was issued shares of
each IPO and subsequently sold those shares on the market when the IPO began trading publicly.

An Empirical Analysis, Page 18

 Journal of Finance and Accountancy

Table 3: Average Monthly IPO Performance Compared Against Standard Benchmarks

Sample Average Return Comparison from July 1996 to January 2008

Test Ho </= 0, H1 > 0, tCritical=~2.61 (t→.005)

IPO DJIA Russell 3000 Nasdaq

Sample Average Return 8.96% 0.55% 0.46% 0.23%
Standard Deviation 13.52% 4.04% 4.42% 8.19%
Count 139 139 139 139
T-Value 7.81 1.60 1.22 0.33
Accept/Reject Decision Reject Accept Accept Accept

Notes: Table 3 provides an analysis of average yearly returns for a strategy that
invests in every IPO that went public from July 1996 to January 2008 and
compares the IPO performance result against standard benchmarks

An Empirical Analysis, Page 19