Brav Ge Czy Go Mpers 2000

Journal of Financial Economics 56 (2000) 209}249
Is the abnormal return following equity

issuances anomalous?
Alon Brav , Christopher Geczy, Paul A. Gompers*
Fuqua School of Business, Duke University, Durham, NC, 27708-0120, USA
The Wharton School, University of Pennsylvania, Philadelphia, PA 19104-6367, USA
Graduate School of Business Administration, Harvard University, Boston, MA 02163, USA
Received 19 October 1998; received in revised form 10 August 1999
Abstract
We examine whether a distinct equity issuer underperformance anomaly exists. In

a sample of initial public o!ering (IPO) and seasoned equity o!ering (SEO) "rms from
1975 to 1992, we "nd that underperformance is concentrated primarily in small issuing
"rms with low book-to-market ratios. SEO "rms, that underperform these standard
benchmarks have time series returns that covary with factor returns constructed from
nonissuing "rms. We conclude that the stock returns following equity issues re#ect
a more pervasive return pattern in the broader set of publicly traded companies. 2000
Elsevier Science S.A. All rights reserved.
JEL classixcation: G1; G2
Keywords: Seasoned equity o!erings; Initial public o!erings; Long-run performance;

Anomalies
1. Introduction
Tests of long-run stock returns have become increasingly common within the
academic "nance literature. Papers purporting to demonstrate long-run anom-
alous returns include Asquith (1983), Agarawal et al. (1992), and Loughran and
* Corresponding author. Tel.: #617-495-6297; fax: 617-496-8443.

E-mail address: [email protected] (P.A Gompers).
0304-405X/00/$ - see front matter 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 3 0 4 - 4 0 5 X ( 0 0 ) 0 0 0 4 0 - 4
210 A. Brav et al. / Journal of Financial Economics 56 (2000) 209}249
Vijh (1997), who examine underperformance in stock returns following mergers;

Michaely et al. (1995), who document poor returns following dividend
omissions; and Dharan and Ikenberry (1995), who explore underperformance
following exchange listing changes. Ikenberry et al. (1995), on the other hand,
"nd overperformance following share repurchases. In addition, Ritter (1991),
Loughran and Ritter (1995), and Spiess and A%eck-Graves (1995) show that
stock market performance subsequent to initial public o!erings (IPOs) and
seasoned equity o!erings (SEOs) is poor. Both sets of authors attribute under-
performance to the e!ects of investor sentiment on returns: investors who
purchase shares in IPO and SEO "rms systematically over-value the shares at
the time of an equity issue.
We show that the poor long-run stock returns following equity issues are not
unique. In event time performance tests we show that IPO issuer returns are
similar to benchmarks matched on "rm size and book-to-market ratios, al-
though SEO returns still show some underperformance relative to various
benchmarks. We calculate abnormal performance using both buy and hold
returns and cumulative abnormal returns, and show that using buy and hold
returns tends to magnify the underperformance of IPOs and SEOs.
We also utilize time-series factor models to test underperformance, "nding
that Fama and French's (1993) three-factor model can capture joint covariation
of IPO returns, while the addition of Carhart's (1997) momentum factor is
needed to capture the covariation of SEO returns. The covariation of equity
issuer returns with a wider class of stock returns is potentially consistent with
either a risk-based or a behavioral model. Fama and French (1993) interpret
covariation of portfolio returns with their three factors as measures of a "rm's
riskiness. Our results are also consistent with behavioral models as presented in
Black (1986); De Long et al. (1990); Lee et al. (1991); and Shleifer and Vishny
(1997), in which noise traders drive stock prices away from fundamental values.
Since these misvaluations tend to cluster in speci"c time periods, the joint
covariation that we document may be driven by such market-wide misvalu-
ations rather than by risk. We demonstrate that the low average return on
equity issuer stocks is not a distinct anomaly; rather it is a manifestation of
a broader pattern in returns.
In addition, evidence is provided showing that tests of long-run abnormal
returns su!er from model misspeci"cation, a concern raised by Fama (1998a).
We show that reasonable changes to Fama and French's (1993) three-factor
model signi"cantly improve its explanatory power. Much like Fama and
French's original model speci"cation, our modi"cations are based on empirical
return patterns and are designed to uncover underlying common e!ects.
Other papers that examine long-run stock returns subsequent to issuances of corporate secur-
ities include Jegadeesh (1998), Brous et al. (1998), Mitchell and Sta!ord (1998), and Eckbo et al.
(2000) for equity and Spiess and A%eck-Graves (1999) for debt.
A. Brav et al. / Journal of Financial Economics 56 (2000) 209}249 211
Recently, Loughran and Ritter (1999) have criticized the ability of certain
methodologies to detect long-run abnormal price reactions. They argue that
using benchmark and factor returns that themselves contain a large number of
issuing "rms reduces the power of performance tests. Furthermore, they contend
that time-series regressions, which are commonly used to detect departures from
market e$ciency, tend to have low power. They attribute this problem to
speci"c weighting schemes used in the implementation of these regressions.
Providing direct evidence to refute these claims, we show that purging
benchmarks and factor returns of issuing "rms does not substantially change the
time-series regression results or the characteristics-matched performance
mentioned above.
The rest of the paper is organized as follows. In Section 2 we begin with
a discussion regarding the implementation of performance tests. Section 3 pres-
ents the sample selection and the data. Next, we critically examine the long-run
performance of equity issuers and provide new evidence regarding the unique-
ness of equity issuer long-run performance. Section 6 contrasts our results with
recent research on the returns of equity issuers. Section 7 concludes.
2. Performance tests, time horizons, trading strategies, and methodology
This section presents the economic framework for our analysis. We do not
develop new methodologies for evaluating long-run abnormal performance.
Instead, we identify the explicit and implicit assumptions made in long-horizon
stock return tests.
One component underpinning long-run performance tests is the choice of
trading strategy implicit in the performance measurement methodology. Any
study examining relative performance of speci"c classes of securities must also
assume portfolio formation rules used to generate those returns. Similarly Fama
(1970) points out that, any performance test must have some notion of what
&normal' returns are. Financial theory typically suggests that if there are perva-
sive and measurable sources of economic risk, or if "rm characteristics help
determine expected returns, then the researcher must match the sample returns
to those on benchmarks composed of "rms whose fundamental riskiness or
return-determining characteristics are similar to the target group.
Thus, performance tests are joint tests of the null hypothesis of no abnormal
performance, the formal model, and the trading rules or portfolio formation
strategies invoked. While most authors recognize that abnormal performance
measurement is always conditional on an asset-pricing model, few authors
recognize that the method of measurement of abnormal performance also a!ects
inferences. Recently, Barber and Lyon (1997), Kothari and Warner (1997), Lyon
et al. (2000), Fama (1998a), Mitchell and Sta!ord (1998), Brav (1999), and
Loughran and Ritter (1999) have addressed this issue. These authors argue that
the method of performance measurement in#uences both the magnitude of the

measured abnormal performance as well as the size and power of the statistical
test. By varying models, trading rules, and benchmarks, we show here that these
choices can lead to di!erent conclusions and we demonstrate how fragile certain
results in the literature actually are.
For example, conditional on a method of measuring abnormal performance
and an asset-pricing model, should sample "rm returns be equal or value
weighted? Loughran and Ritter (1999) point out that the choice of weighting
scheme is important for power considerations. To illustrate this point, consider
a simple scenario in which a sample contains 1000 "rms, 999 of which have a $1
million market capitalization (the &small' "rms) and one "rm that has a $1,001
million market capitalization (the &large' "rm). Assume that the small "rms have
all underperformed by an equal percentage rate equal to 50% while the large
"rm has overperformed by 50%. It is easy to see that an equal weighted measure
of abnormal performance will indicate severe mispricing (!50%), while value
weighting will lead the researcher to conclude that the sample abnormal perfor-
mance is virtually zero. The "rst method provides strong evidence against
market e$ciency whereas the second does not.
We argue that if an alternative hypothesis suggests that small stocks are likely
to be mispriced more than large stocks, then the power consideration alone
implies the use of equal weighting. Similarly, if we are interested in the manageri-
al implications of potential stock market mispricing, equal weighting returns
might be more appropriate. If the researcher's goal, however, is to quantify
investors' average wealth change subsequent to an event, then it follows that
value weighting is the correct method. Therefore, we present results using both
value and equal weighting to highlight cross-sectional di!erences in abnormal
performance wherever they are present.
Several recent papers are related to our choice of tests of long horizon
abnormal performance. Barber and Lyon (1997), Lyon et al. (2000), and Kothari
and Warner (1997) provide thorough evidence on various methods of measuring
abnormal performance. These papers do not "nd that one method is always
preferred. Fama (1998b) and Mitchell and Sta!ord (1998), however, provide
additional considerations regarding the merits of such methodologies. They
argue that abnormal performance measures, such as cumulative abnormal
returns (CARs) and time-series regressions, are less likely to yield spurious
rejections of market e$ciency relative to methodologies that calculate buy-and-
hold returns by compounding single period returns at the monthly frequency.
First, the buy-and-hold method can magnify underperformance } even if it
occurs in only a single period } due to the nature of compounding single-period
returns. Second, distributional properties and test statistics for cumulative
abnormal returns are better understood.
Finally, and perhaps central to any test of long-run stock return performance,
is the model misspeci"cation problem as discussed by Fama (1998b). Fama
points out that all models of expected returns are incomplete descriptions of the
systematic variation of expected returns across "rms. As we extend the measured
horizon in an event study, it becomes more likely that the inadequacy of the
maintained asset-pricing model is detected due to our compounding of the
pricing errors induced by model misspeci"cation. The approach we take in this
paper is to examine the robustness of SEO and IPO underperformance with
respect to various model speci"cations. As a result, we provide direct evidence of
the importance of the model misspeci"cation problem for equity issuers.
A troubling aspect of the model misspeci"cation problem is that in absence of
a clear alternative hypothesis, this misspeci"cation cannot be rejected. In other
words, supporters of market e$ciency can always claim that the abnormal
performance of a given sample is due to incorrect measurement of &normal'
returns rather than systematic security mispricing. Thus, our ability to identify
the true underlying sources of pricing errors is in this sense limited.
Moreover, normative models of expected returns, such as the Capital Asset
Pricing Model (CAPM) developed by Sharpe (1964) and Lintner (1965), have
shown little ability to explain the cross-section of stock returns. In general,
asset-pricing theory has had limited success in determining the exact sources of
risk that investors care about when purchasing investments (Fama, 1998a).
Consequently, empirical asset pricing has relied on empirically motivated factor
models. Since the researcher has no guidance in the construction of these factors,
the model misspeci"cation problem and its critique are exacerbated. Our contri-
bution to this debate is in showing that tests of long-run performance of equity
issuers are severely a!ected by this problem. Alternative speci"cations of Fama
and French's (1993) three-factor model have greatly improved pricing ability
and eliminate issuer underperformance on the whole. These alternative model
speci"cations, which exclude issuers in their factor formation, show that the
returns of equity issuers strongly covary with those of non-issuing "rms.
3. Data
Our sample of SEOs was identi"ed in the Investment Dealers+ Digest of

Corporate Financing for the period 1975}1992. For inclusion in our sample,
a "rm performing a seasoned equity o!ering must have returns included in the
"les of the Center for Research on Security Prices (CRSP) sometime within "ve
years of the o!ering date. Our sample includes 4526 o!erings made by 2772
"rms. These include all seasoned primary equity o!erings, both pure primary
share o!erings and combination primary and secondary share o!erings.
Figs. 1 and 2 shows the clustering of SEOs in our sample. Unlike an IPO, in
which a "rm by de"nition can only go public once, many "rms in our sample are
multiple issuers of seasoned equity. 1783 (64.3%) "rms perform only one
seasoned equity issue; 614 (22.2%) issue twice; 211 (7.6%) issue three times; 81
Fig. 1. Time series of seasoned equity o!erings listed by exchange at time of o!ering. The sample of
seasoned equity o!erings was identi"ed in the Investment Dealers' Digest of Corporate Financing for
the period 1975}1992. For inclusion in our sample, a "rm performing a seasoned equity o!ering
must be followed by CRSP sometime after the o!ering date. The sample includes 4526 o!erings
made by 2772 "rms. O!erings are listed by the exchange on which the "rm was trading at the time of
the o!ering.
(2.9%) issue four times; 35 (1.3%) issue "ve times; 17 (0.6%) issue six times; and
31 (1.1%) issue seven times or more. Fig. 1 also shows the distribution of issues
by exchange. In the 1970s, most SEOs were NYSE "rms. In later years, Nasdaq
"rms account for a larger fraction of issuances.
Our sample of IPOs is collected from various issues of the Investment Dealers+
Digest of Corporate Financing for the period 1975}1992 as well as from the
Securities Data Corporation database. Jay Ritter also provided data on initial
public o!erings for the period 1975}1984. For inclusion in our sample, a "rm
performing an IPO must have stock returns in the CRSP database within "ve
years of the o!ering date. Our "nal sample includes 4622 IPOs for the period
1975}1992.
Our data set is di!erent from those of Loughran and Ritter (1995) and Spiess
and A%eck-Graves (1995) in several ways. First, unlike Loughran and Ritter, we
exclude all unit o!erings from our sample because separating the value of the
o!erings' components (usually common stock with warrants) is di$cult. Utili-
ties are also included in our sample even though Loughran and Ritter exclude
them. The impact of including utilities is small: they comprise only 5.5% of
either sample. We reestimated our results excluding utilities and, although not
report here, these results were similar to those which we have reported.
In addition, unlike Spiess and A%eck-Graves, who look only at o!erings that
are entirely primary shares, we include o!erings that also have a secondary
Fig. 2. Time series of initial public o!erings. The sample of initial public o!erings is collected from
various issues of the Investment Dealers' Digest of Corporate Financing for the period 1975}1995 as
well as from the Securities Data Corporation database. For inclusion in the sample, a "rm
performing an IPO must be followed by the Center for Research in Securities Prices (CRSP) at some
point after the o!ering date. The "nal sample includes 4622 IPOs for the period 1975}1992 of which
539 list on NYSE, 215 list on ASE, and 3869 list on Nasdaq.
component in which insiders sells some of their shares in the o!ering. If, as
Myers and Majluf (1984) suggest, insiders attempt to sell their shares when they
perceive them to be overvalued, inclusion of these issues may reinforce the
long-run underperformance e!ect we "nd in the data as the market revises its
assessment of these "rms' prospects. Including secondary issues along with
primary issues is conservative in this sense. Because we can replicate the
magnitude of underperformance documented in Loughran and Ritter (1995) and
Spiess and A%eck-Graves (1995) using our sample, we do not feel our sample
selection imparts any biases in our conclusions.
We examine the size and book-to-market characteristics of our sample.
Starting in January of 1964, we use all New York Stock Exchange (NYSE)
stocks to create size quintile breakpoints. This approach is based on Fama and
French (1992), who use only NYSE stocks to create the breakpoints to ensure
dispersion in the characteristics across portfolios. Size is measured as the
number of shares outstanding times the stock price at the end of the preceding
month. We obtain our accounting measures from the COMPUSTAT quarterly
and annual "les and de"ne book value as book common equity plus balance
sheet deferred taxes and investment tax credits for the "scal quarter ending two
quarters before the sorting date. If the book value is missing from the quarterly
Table 1
The distribution of IPOs by book-to-market and size.
The sample is all IPOs from 1975 through 1992. Size breakpoints are formed quarterly by dividing
all NYSE stocks into size quintiles. Book-to-market breakpoints are formed quarterly with equal
numbers of NYSE "rms allocated to each of "ve book-to-market portfolios. The intersection of
these quintile breakpoints results in 25 possible allocations.
Book-to-market Size quintile

quintiles
Smallest Quintile 2 Quintile 3 Quintile 4 Largest Total
(%) (%) (%) (%) (%) (%)
Low 51.80 18.20 5.50 1.70 0.40 77.60

2 8.20 1.90 0.90 0.20 0.10 11.50
3 3.00 1.20 0.60 0.10 0.10 4.90
4 1.60 0.60 0.30 0.10 0.00 2.60
High 2.00 0.80 0.40 0.20 0.00 3.40
Total 66.60 22.70 7.70 2.30 0.60 100.00
statements or if "rms are missing altogether from the quarterly "les, we search
for the book value in the annual "les. We create book-to-market quintile
breakpoints using only NYSE "rms and then form 25 (5;5) size and book-to-
market portfolios by intersecting the portfolio breakpoints and allocating all
NYSE, Amex, and Nasdaq "rms into these portfolios. Our benchmark port-
folios are reformed each quarter.
These breakpoints also create 25 portfolios to which each SEO or IPO stock
can be assigned. For the sample of SEO "rms, we calculate the market value of
the "rm one month prior to the issuance. For the IPOs, the market value of
equity is calculated at the end of the "rst month after the o!ering. We calculate
book values using COMPUSTAT data. Book values for SEO "rms are taken six
months prior to the o!ering to allow for reporting lags. For the book value of
equity of IPO "rms, we use COMPUSTAT and record the "rst book value after
the IPO as long as it is within eighteen months of the o!ering month. Shortening
this interval does not a!ect any of the results that we report herein, although it
does lead to a smaller sample size. The bias in book value should not be too
large because the increment in book value due to retained earnings in the "rst
year is likely to be very small for newly public companies; it is only the small
recently incorporated "rms that are missing data.
Table 1 presents the size and book-to-market portfolio distribution for the
IPO sample by quintiles. IPOs are heavily weighted in the small, low book-to-
market portfolios. Over half (51.8%) of the IPO "rms are in the smallest, lowest
book to market portfolio, the (1,1) portfolio. Two thirds of the IPO "rms are in
&The (X, Y) portfolio' refers to the Xth size row and the Yth book-to-market column in the 5;5
display of the 25 size and book-to-market portfolios.
Table 2
The distribution of SEOs by book-to-market and size.
The sample is all SEOs from 1975 through 1992. Size breakpoints are formed quarterly by dividing
all NYSE stocks into size quintiles. Book-to-market breakpoints are formed quarterly with equal
numbers of NYSE "rms allocated to each of "ve book-to-market portfolios. The intersection of
these quintile breakpoints results in 25 possible allocations
Book-to-market Size quintile

quintiles
Smallest Quintile 2 Quintile 3 Quintile 4 Largest Total
(%) (%) (%) (%) (%) (%)
Low 24.10 12.60 7.30 4.20 2.40 50.70

2 7.40 5.10 3.30 1.70 1.30 18.70
3 4.10 3.20 2.20 2.30 1.50 13.20
4 2.70 2.00 1.90 1.80 1.20 9.70
High 2.20 1.50 1.30 1.40 1.40 7.80
Total 40.50 24.50 16.00 11.30 7.80 100.00
the smallest quintile of stocks. Similarly, less than 1% of the IPOs are in the
largest quintile of stocks when breakpoints are de"ned by NYSE "rms. In terms
of book-to-market ratios, over 77% of the IPO "rms have book-to-market
ratios in the lowest quintile when breakpoints are de"ned by NYSE stocks.
It is not surprising that the sample tends to have this pattern. Smaller "rms
may generate less internal cash and thus issue equity to "nance expansion.
Similarly, managers of smaller "rms may have substantially larger equity stakes
in the "rms and want to reduce their holdings via secondary o!erings. The low
book-to-market grouping may be explained in one of two ways. First, these may
represent future growth options. Smaller "rms are growth "rms and may have
many good investment opportunities for which they need to raise cash. On the
other hand, the low book-to-market ratio for these "rms may just be
an indicator of overpricing. Loughran and Ritter (1995) for example argue
that issuers time the market for new shares when their "rms are relatively
overvalued.
We present summary statistics for the SEO "rms in Table 2, where the
distribution is shown in the 5;5 size and book-to-market portfolios. The SEO
sample is more evenly distributed across size and book-to-market quintiles than
the IPO sample. Only 24% of the SEO sample is in the (1,1) portfolio, less than
half the fraction of IPO "rms that fall in the (1,1) portfolio. Almost 8% of the
seasoned equity-issuing "rms are in the largest quintile of NYSE stocks and
almost 50% of SEO "rms are outside of the lowest book-to-market quintile.
These di!erences lead us to examine the performance of IPO and SEO "rms
separately.
4. Underperformance of equity issuers
4.1. Event time returns
Various benchmarks are utilized to measure IPO and SEO long-run perfor-
mance throughout this paper. First, the performance of issuing "rms is matched
to the following broad market indexes: the S&P 500, Nasdaq composite index,
CRSP value weight index, and CRSP equal weight index. CRSP value and equal
weight indexes include all stocks listed on the NYSE, Amex, and Nasdaq. We
also create benchmark portfolios which will be assumed to represent normal
returns by matching issuing "rms with groups of "rms drawn from the universe
of stocks (NYSE, AMEX, and Nasdaq) based on "rm characteristics. These
characteristics are either "rm size and book-to-market or size, book-to-market,
and price momentum portfolios.
We form size and book-to-market portfolios as described in the previous
section. In order to eliminate the benchmark bias discussed in Loughran and
Ritter (1999), we eliminate all IPO and SEO "rms from the benchmark port-
folios for a period of "ve years after their issuance. We further reform our
benchmark portfolios each quarter and calculate equal weighted returns for the
next three months, repeating the above procedure in January, April, July, and
October of each year.
We form the size, book-to-market, and momentum portfolios as follows.
Starting in January 1964, we form size quartile breakpoints using NYSE "rms
only. Within each size quartile, we then form quartile book-to-market break-
points. Each of the 16 (4;4) size and book-to-market portfolios is in turn split
further into price momentum quartiles. Size and book-to-market are calculated
as described in the previous paragraph. Price momentum is calculated as the
buy-and-hold return over the previous year excluding the month before the
sorting date, i.e., the same procedure used in Carhart (1997). This sorting
procedure results in 64 (4;4;4) possible portfolios into which we allocate the
entire universe of CRSP-listed "rms (i.e., all NYSE, Amex, and Nasdaq stocks).
We use only quartile breakpoints (based on NYSE "rms) when we form size,
book-to-market, and price momentum benchmarks because "ner divisions of
"rms (i.e., quintiles) leave certain portfolios with few stocks. Unlike portfolio
formation based on accounting measures, we repeat the momentum-based
formation process monthly. For each of the 64 benchmark portfolios, we
calculate equal weighted monthly returns.
Table 3 presents long-run performance for the sample of SEO stocks from an
event time strategy in which each SEO constitutes an event. Panel A calculates
buy-and-hold returns. To construct our "gures, we use the monthly tapes and
follow each o!ering beginning in the month after the event for the earlier of 60
months or the delisting month. Our "ve-year buy-and-hold intervals are set to
match those chosen by Loughran and Ritter (1995).
On the left-hand side of Panel A, we equal weight the returns of the SEOs and
their benchmarks. Loughran and Ritter (1995) and Spiess and A%eck-Graves
(1995) calculate wealth relatives for the "ve year period by taking the ratio of
one plus the equal weighted return on the SEO portfolio over one plus the equal
weighted return on the chosen benchmark. We also report the ratio of the
average return on the SEOs relative to the chosen benchmark.
The wealth relatives in Panel A are very similar to those in Loughran and
Ritter and Spiess and Graves reassuring us that any di!erences in sample
selection has negligible impact on our results. The "ve-year wealth relative is less
than 1.0 for all benchmarks, ranging from 0.84 to 0.89. The "ve-year excess
returns are all negative, anywhere from !19.5% versus the Nasdaq Composite
index to !30.1% versus the S and P 500 index.
Unlike Ritter and Loughran, no attempt is made to match SEOs to individual
"rms. Loughran and Ritter (1995) match "rms on size and require the matching
"rms to be nonissuing "rms for at least "ve years. Therefore, because Nasdaq
returns are picked up by CRSP in 1972, 1978 is the "rst year that any Nasdaq
"rm can be used as a matching "rm in their study. As Table 2 shows, most
issuing "rms tend to be small. By matching the early sample only to small
NYSE/ASE "rms, Loughran and Ritter may be matching SEO "rms to the De
Bondt and Thaler (1985, 1987) long-term loser stocks that we know outperform
similarly sized stocks. Their benchmarks therefore are likely to be biased
towards "nding underperformance. This potential problem is avoided by our
use of all CRSP-listed "rms, including Nasdaq.
The right half of panel A presents value weighted results for our sample. In
Panel A value weighting reduces, but does not eliminate, underperformance.
Wealth relatives are now between 0.87 and 0.93 and the excess return is between
!14.2% and !25%. This indicates that small issuers underperform more
than large issuers.
In Panel B of Table 3, we calculate the cumulative abnormal returns (CARs)
for the "ve years following the issuance. The CARs are all closer to zero than the
excess returns calculated in Panel A. In fact, average CARs are between !8%
and !17% in the value weighted SEO sample. While value weighting increases
CARs relative to the broad market benchmarks, there is little change in perfor-
mance relative to the size and book-to-market or size, book-to-market, and
price momentum portfolios. The results in Table 3 also show that the underper-
formance of SEO stocks is magni"ed by the calculation of buy-and-hold returns.
The returns of the IPO sample are similarly examined in Table 4 where we
again present buy-and-hold returns in Panel A and cumulative abnormal
returns in Panel B. The results are similar in both panels. On an equal-weighted
basis, IPO "rms underperform broad market benchmarks by a wide margin,
underperforming the S&P by 44% and Nasdaq by 31%. Value weighting the
IPO stock returns, however, cuts this underperformance in half. Once IPO
"rm returns are matched to size and book-to-market portfolios, which are
220
Table 3
The long-run performance of seasoned equity o!erings. The sample is 4526 seasoned equity o!erings from 1975 through 1992.
Five year equal weighted and value weighted buy and hold returns on SEOs are compared with alternative benchmarks. In the "rst four rows we employ
the S&P 500, Nasdaq Composite, and NYSE-ASE-Nasdaq equal and value weighted indices. Value weighted results for these indices employ 4512
observations because market capitalization in the month after the SEO for 14 "rms could not be found. Value weighted results are adjusted for in#ation.
Rows "ve through seven present abnormal performance measured relative to attribute-based portfolios. Size and book-to-market portfolios are generated
by "rst forming size quintile breakpoints using NYSE "rms only. These quintiles are split further into book-to-market quintiles using NYSE breakpoints
only. The universe of NYSE, ASE and Nasdaq "rms is allocated into the resulting 25 portfolios and equal weighted monthly returns are calculated. The
breakpoints for these portfolios are recalculated quarterly. Size, book-to-market and momentum portfolios are formed as follows. We "rst form size
quartile breakpoints using NYSE "rms only and then, within each size quartile, form book-to-market quartile breakpoints using NYSE "rms only. We
allocate the universe of NYSE, ASE, and Nasdaq "rms into these 16 portfolios and then within each portfolio form additional quartile breakpoints based
on prior year returns excluding the previous month. We calculate the resulting portfolio average returns for these 64 portfolios. This portfolio formation is
repeated monthly. In the "rst six rows we report buy-and-hold results while in the seventh row we report event-time rebalanced results. Buy-and-hold
returns are generated by compounding 60 monthly returns starting in the month following the equity issue. If the SEO delists before the 60th month we
compound the return up until the delisting month. Event-time rebalancing results are generated by compounding event-time monthly averages for the
SEOs and their benchmark. Since some "rms lack the accounting information necessary for the attribute matching, we report results for 3775 "rms in rows
"ve and six and for 4091 in row seven. Abnormal return is the simple di!erence between SEO "ve-year average return and the corresponding benchmark.
Wealth relatives are calculated as R(1#R )/R(1#R ), where R is the holding period return on the SEO i for period ¹ and R is the return
G2 2 G2 2
on the benchmark portfolio over the same period. All returns on the SEOs and benchmark portfolios are taken from the CRSP "les.
Panel A: Buy-and-hold returns

Benchmarks Equal weighted buy-and-hold returns (%) Value weighted buy-and-hold returns (%)
SEO Benchmark Abnormal Wealth SEO Benchmark Abnormal Wealth

return relative return relative
S&P 500 index 57.5 87.6 !30.1 0.84 75.3 94.3 !19.0 0.90
A. Brav et al. / Journal of Financial Economics 56 (2000) 209}249
Nasdaq Comp. 57.5 76.9 !19.5 0.89 75.3 89.5 !14.2 0.93
CRSP VW 57.5 81.5 !24.0 0.87 75.3 89.5 !14.1 0.93
CRSP EW 57.5 81.6 !24.2 0.87 75.3 102.2 !26.9 0.87
Size and Book-to-Market (3,775 obs) 57.6 83.9 !26.3 0.86 72.5 97.5 !25.0 0.87
Size and Book-to-Market and PR12 57.0 84.3 !27.3 0.85 75.7 99.5 !23.8 0.88
(3775 obs)
The long-run performance of seasoned equity o!erings.
The sample is 4526 seasoned equity o!erings from 1975 through 1992. Five year equal weighted and value weighted cumulative abnormal returns on SEOs
are compared with alternative benchmarks. In the "rst four rows we employ the S&P 500, Nasdaq Composite, and NYSE-ASE-Nasdaq equal and value
weighted indices. Value weighted results for these indices employ 4512 observations because market capitalization in the month after the SEO for 14 "rms
could not be found. Value weighted results are adjusted for in#ation. Rows "ve through seven present abnormal performance measured relative to
attribute-based portfolios. Size and book-to market portfolios are generated by "rst forming size quintile breakpoints using NYSE "rms only. These
quintiles are split further into book-to-market quintiles using NYSE breakpoints only. The universe of NYSE, ASE and Nasdaq "rms are allocated into the
resulting 25 portfolios and equal weighted monthly returns are calculated. The breakpoints for these portfolios are recalculated quarterly. Size,
book-to-market, and momentum portfolios are formed as follows. We "rst form size quartile breakpoints using NYSE "rms only and then, within each size
quartile, form book-to-market quartile breakpoints using NYSE "rms only. We allocate the universe of NYSE, ASE, and Nasdaq "rms into these 16
portfolios and then within each portfolio form additional quartile breakpoints based on prior year returns excluding the previous month. We calculate the
resulting portfolio average returns for these 64 portfolios. This portfolio formation is repeated monthly. Cumulative abnormal returns are generated by
summing 60 monthly abnormal returns starting in the month following the equity issue. If the SEO delists before the 60th month, we sum the return up to
the delisting month. Event-time rebalancing results are generated by summing event-time monthly averages for the SEOs and their benchmark. Since some
"rms lack the accounting information necessary for the attribute matching, we report results for 3775 "rms in rows "ve and six. All returns on the SEOs and
benchmark portfolios are taken from the CRSP "les.
Panel B: Cumulative abnormal returns
Benchmarks Equal weighted CARs (%) Value weighted CARs (%)
SEO Benchmark CAR SEO Benchmark CAR
S&P 500 index 51.5 75.5 !24.0 60.4 70.2 !9.8

Nasdaq composite 51.5 70.7 !19.2 60.4 69.2 !8.8
CRSP value-weighted 51.5 72.0 !20.5 60.4 68.1 !7.8
CRSP equal-weighted 51.5 68.7 !17.15 60.4 70.4 !10.1
Size and book-to-market (3775 obs) 53.2 68.6 !15.4 56.0 71.9 !15.9
Size and book-to-market and PR12 52.4 67.7 !15.3 56.6 73.8 !17.1
(3775 obs)
221
Table 4
222
The long-run performance of initial public o!erings.
The sample is 4622 initial public o!erings from 1975 through 1992. Five year equal weighted and value weighted buy-and-hold returns on IPOs are
compared with alternative benchmarks. In the "rst four rows we employ the S&P 500, Nasdaq Composite, and NYSE-ASE-Nasdaq equal and value
weighted indices. Throughout, value weighted results are adjusted for in#ation. In row "ve we present abnormal performance measured relative to size and
book-to market portfolios. These portfolios are generated by "rst forming size quintile breakpoints using NYSE "rms only. These quintiles are split
further into book-to-market quintiles using NYSE breakpoints only. The universe of NYSE, ASE and Nasdaq "rms is allocated into the resulting 25
portfolios and we calculate equal weighted monthly returns. The breakpoints for these portfolios are recalculated quarterly. Buy-and-hold returns are
generated by compounding 60 monthly returns starting in the month following the equity issue. Since some "rms lack the accounting information
necessary for the attribute matching, we report results for 3501 in row "ve. If the IPO delists before the 60th month, we compound the return up to the
delisting month. Abnormal return is the simple di!erence between IPO "ve-year average return and the corresponding benchmark. Wealth relatives are
calculated as R(1#R )/R(1#R ), where R is the holding period return on the IPO i for period ¹ and R is the return on the benchmark
G2 2 G2 2
portfolio over the same period. All returns on the IPOs and benchmark portfolios are taken from the CRSP "les.
Panel A: Buy-and-hold returns
Benchmarks Equal weighted buy-and-hold returns (%) Value weighted buy-and-hold returns (%)
IPO Benchmark Abnormal Wealth IPO Benchmark Abnormal Wealth

return relative return relative
S&P 500 index 33.1 77.3 !44.2 0.75 52.6 78.3 !25.7 0.86
Nasdaq composite 33.1 64.2 !31.1 0.81 52.6 68.2 !15.6 0.91
CRSP VW 33.1 71.7 !38.6 0.78 52.6 72.4 !19.8 0.89
CRSP EW 33.1 61.5 !28.4 0.82 52.6 61.4 !8.8 0.95
Size and book-to-market (3501 obs) 35.8 29.2 6.6 1.05 56.6 55.2 1.4 1.01
The long-run performance of initial public o!erings.

The sample is 4622 initial public o!erings from 1975 through 1992. Five year equal weighted and value weighted cumulative abnormal returns on IPOs are
compared with alternative benchmarks. In the "rst four rows we employ the S&P 500, Nasdaq Composite and NYSE-ASE-Nasdaq equal and value
weighted indices. Throughout, value weighted results are adjusted for in#ation. In row "ve we present abnormal performance measured relative to size and
book-to market portfolios. These portfolios are generated by "rst forming size quintile breakpoints using NYSE "rms only. These quintiles are split
further into book-to-market quintiles using NYSE breakpoints only. The universe of NYSE, ASE and Nasdaq "rms is allocated into the resulting
25 portfolios and we calculate equal weighted monthly returns. The breakpoints for these portfolios are recalculated quarterly. Since some "rms lack the
accounting information necessary for the attribute matching, we report results for 3501 in row "ve. Cumulative abnormal returns are generated by
summing 60 monthly abnormal returns starting in the month following the equity issue. If the IPO delists before the 60th month, we sum up to the
delisting month. All returns on the IPOs and benchmark portfolios are taken from the CRSP "les.
Panel B: Cumulative abnormal returns
Equal weighted CARs (%) Value weighted CARs (%)
Benchmarks IPO Benchmark CAR IPO Benchmark CAR
S&P 500 index 35.9 74.2 !38.3 40.7 61.5 !20.8

Nasdaq composite 35.9 67.9 !32.0 40.7 57.5 !16.8
CRSP value-weighted 35.9 70.2 !34.3 40.7 58.1 !17.4
CRSP equal-weighted 35.9 62.4 !26.5 40.7 51.8 !11.1
Size and Book-to-market (3501 obs) 42.3 32.6 9.7 53.0 49.7 3.3
223
themselves free of issuing "rms, there is no underperformance. Average excess

returns are actually positive. As Brav and Gompers (1997) and Brav (1999)
show, initial public o!ering "rms have returns that are comparable to similar,
non-issuing size and book-to-market "rms.
As we will demonstrate in a later section, the reason that matching by size and
book-to-market eliminates the underperformance of the IPO stocks, but not the
SEO stocks, is that the majority of IPO "rms (51.8%) are in the (1, 1) portfolio;
that is, the smallest, lowest book to market portfolio. Brav and Gompers (1997)
show that small, growth "rms do particularly poorly following heavy periods of
equity issuances. SEO "rms are more evenly distributed across size and book-
to-market quintiles (with only 24.1% of SEO "rms in the (1, 1) portfolio), and
hence the size and book-to-market benchmark for 75% of SEO stocks is not the
(1, 1) portfolio. However, we show in section IV that the time series of returns on
SEO stocks are mimicked by the time series of returns of non-issuing "rms
controlling for size and book-to-market. This covariation indicates that the
factors a!ecting the returns on small growth companies, whether they may be
risk or sentiment, a!ect the returns on SEO "rms at the same time.
4.2. Factor regression analysis
The characteristics-based approach of the previous section assumes that

equity risk is captured by an observable set of "rm-speci"c characteristics such
as size, book-to-market, or stock price momentum. If these observable charac-
teristics are only imperfect proxies for risk, then the characteristics-based
approach might misclassify "rms' riskiness. Fama (1998a) argues that factor-
based approaches to performance evaluation as documented in Fama and
French (1993) are potentially useful in capturing systematic patterns in average
returns, although he acknowledges that both factor-based approaches and
corresponding characteristic-based approaches likely su!er from model mis-
speci"cation as discussed earlier.
In the Fama}French (1993) three-factor model, the "rst factor is the excess
return on the value weighted market portfolio (RMRF). The second factor,
called SMB by Fama and French, is the return on a zero investment portfolio
formed by subtracting the return on a large "rm portfolio from the return on
a small "rm portfolio. Similarly, the third factor is the return of another
mimicking portfolio, named HML by Fama and French, de"ned as the return
on a portfolio of high book-to-market stocks (This portfolio represents the top
30% of all "rms on COMPUSTAT, less the return on a portfolio of low
book-to-market stocks (This portfolio contains "rms in the lowest 30% of the
The breakpoints for small and large are determined by NYSE "rms alone, but the portfolios
contain all "rms traded on NYSE, ASE, and Nasdaq exchanges.
COMPUSTAT universe of "rms.)

r !r "a#bRMRF #sSMB #hHM¸ #e . (1)
NR DR R R R R
Carhart (1997) explores extensions of the Fama}French model. In particular,
he argues that a fourth factor, that is nearly orthogonal to the Fama}French
factors, can increase the explanatory power of the Fama}French model for the
returns of mutual funds. Carhart's fourth factor is based on ranking "rms by
their return over the previous year, or price momentum. This notion follows the
empirical observation of Jegadeesh and Titman (1993) that "rms having high
returns in the previous year tend to continue to enjoy high returns in the next
year. His factor, PR12, is formed by taking the return on high momentum stocks
minus the return on low momentum stocks.
r !r "a#bRMRF #sSMB #hHM¸ #pPR12 #e . (2)
NR DR R R R R R
As discussed in Section 1, however, Loughran and Ritter (1999) argue that
factor models have low statistical power for uncovering underperformance if
factor returns are calculated using, in part, the returns of the sample being tested
for underperformance. To deal with this, we run our factor regressions using
returns created two ways: by using all CRSP-listed securities and also by using
only those "rms that have not issued equity in an IPO or SEO within the
previous "ve years.
We interpret the intercept in the three and four factor time series regressions
as a measure of average abnormal performance. This measure has a role
analogous to Jensen's alpha in the CAPM framework. To benchmark the full
sample and issuer-purged factor regressions, Table 5 presents four-factor model
regression results for the 25 size and book-to-market portfolios discussed
previously. These 25 portfolios contain all CRSP-listed "rms. The results in
Table 5 help establish how the issuer-purged factors perform in standard tests of
model performance "rst benchmarked in Fama and French (1993). In Panel A,
the full sample factors price nearly all of the test portfolios. Much like in the
earlier Fama and French (1993) results, the four-factor model has di$culty
pricing the (1, 1) portfolio of the smallest, lowest book-to-market "rms, produ-
cing a large negative intercept of !0.0053, 53 basis points per month. In
addition, the (2, 1) portfolio has a signi"cantly negative intercept and the (1, 4)
and the (2, 4) portfolios have signi"cantly positive intercepts, indicating that the
ability of these pricing models to explain all portfolio returns is somewhat
limited.
The inability of the models to price, for example, the (1, 1) portfolio, implies
that any group of stocks which strongly covaries with the return on the (1, 1)
Momentum here is de"ned as the previous 11-month nominal stock return lagged one month.
We create the factor breakpoints by taking the average return on the top 50% of all "rms (winners)
minus the return on the bottom 50% of all "rms (losers).
Table 5
226
Four factor regressions for twenty-"ve test portfolios formed on the basis of size and book-to-market.
Panel A presents estimated coe$cients from factor regressions for the 25 benchmark portfolios using factors formed from all CRSP-listed "rms. Panel
B forms the factors excluding all issuing "rms, i.e., excluding all "rms that issued equity in an initial public o!ering or seasoned equity o!ering during the
previous "ve years. RMRF is the return on a portfolio formed by subtracting the value weighted market return on all NYSE/ASE/Nasdaq "rms (RM)
minus the risk free rate (RF) which is the one-month Treasury bill rate. SMB (small minus big) is the di!erence each month between the return on small
"rms and big "rms. HML (high minus low) is the di!erence each month between the return on a portfolio of high book-to-market stocks and the return on
a portfolio of low book-to-market stocks. PR12 is the excess return on winners versus losers based on prior year returns excluding the previous month.
[t-statistics are in brackets.]
Panel A: Full sample factors
Intercept Size quintiles
Book-to-market Small 2 3 4 Large

quintiles
Low !0.0053 [4.16] !0.0016 [!1.90] !0.0004 [!0.45] 0.0013 [1.57] 0.0013 [1.52]
2 0.0012 [1.33] 0.0002 [0.21] 0.0007 [0.90] !0.0004 [*0.54] 0.0009 [1.01]
3 0.0001 [0.09] 0.0012 [1.64] 0.0000 [0.02] !0.0005 [!0.59] !0.0005 [!0.45]
4 0.0016 [2.17] 0.0016 [2.07] 0.0013 [1.62] 0.0000 [0.02] !0.0003 [!0.31]
High 0.0003 [0.41] !0.001 [!1.20] !0.0006 [!0.57] 0.0004 [0.33] !0.0007 [!0.51]
RMFR Size quintiles

quintiles
Low 1.0498 [34.87] 1.1060 [57.12] 1.0785 [54.22] 1.0464 [52.99] 0.9518 [46.36]
2 1.0171 [48.70] 1.0506 [55.73] 1.0286 [52.57] 1.0909 [55.84] 1.0501 [49.48]
3 0.9599 [58.57] 0.9981 [57.94]] 0.9657 51.82] 1.0948 [49.82] 0.9790 [40.79]
4 0.9056 [52.80] 0.9779 [54.41] 0.9777 [50.35] 1.0528 [44.48] 1.0172 [47.02]
High 0.9925 [52.60] 1.1008 [57.18] 1.1037 [45.03] 1.1686 [41.20] 1.349
SMB Size quintiles

quintiles
Low 1.3100 [27.10] 1.0104 [32.50] 0.7222 [22.62] 0.3533 [11.15] !0.1966 [!5.96]
2 1.1893 [35.47] 0.9297 [30.71] 0.6557 [20.87] 0.3191 [10.17] !0.1690 [!4.96]
3 1.0774 [40.95] 0.8378 [30.29] 0.6076 [20.30] 0.3343 [9.48] !0.2357 [!6.12]
4 1.0693 [38.83] 0.7561 26.20] 0.4651 [14.91] 0.1508 [3.97] !0.1657 [!4.77]
High 1.1794 [38.93] 0.9220 [29.83] 0.6224 [15.82] 0.2816 [6.18] !0.0995 [!1.88]
HML Size quintiles

quintiles
Low !0.2800 [!5.46] !0.5064 [!15.34] !0.4221 [!12.45] !0.4393 [!13.05] !0.4317 [!12.33]
2 0.0324 [0.91] !0.018 [!0.56] 0.0418 [1.25] !0.0075 [!0.23] !0.0793 [!2.19]
3 0.2237 [8.00] 0.2179 [7.42] 0.3069 [9.66] 0.2895 [7.73] 0.1444 [3.53]
4 0.3607 [12.33] 0.4337 [14.15] 0.4214 [12.73] 0.5133 [12.72] 0.482 [13.07]
High 0.6500 [20.21] 0.6813 [20.76] 0.7622 [18.24] 0.7129 [14.74] 0.8193 [14.57]
PR12 Size quintiles

quintiles
Low !0.0814 [!1.97] !0.0033 [!0.13] 0.0382 [1.40] !0.0045 [!0.17] !0.0749 [!2.66]
2 !0.1465 [!5.12] 0.0124 [0.48] 0.0249 [0.93] !0.0384 [!1.43] 0.0078 [0.27]
3 !0.046 [!2.05] 0.0213 [0.90] 0.0420 [1.65] !0.0350 [!1.16] 0.0774 [2.35]
4 !0.0673 [!2.86] 0.0494 [2.00] !0.0337 [!1.27] !0.0667 [!2.06] !0.075 !2.53]
High !0.0574 [!2.22] 0.0273 [1.04] 0.1213 [3.61] 0.0205 [0.53] !0.0426 [!0.94]
227
Table 5 (continued) 228
Adjusted R Size quintiles

quintiles
Low 92.70 96.83 95.96 95.16 92.78

2 95.61 96.04 94.82 94.58 92.39
3 96.68 95.97 94.25 92.72 88.22
4 96.00 95.07 93.24 89.98 90.03
High 95.92 95.63 91.92 88.87 80.36
Panel B: Purged factor regressions

quintiles
Low !0.0060 [!4.26] !0.0015 [!1.39] 0.0001 [0.06] 0.0013 [1.45] 0.0015 [1.78]
2 0.0004 [0.44] 0.0005 [0.60] 0.0011 [1.30] !0.0007 [!0.86] 0.0007 [0.83]
3 !0.0001 [!0.06] 0.0016 [2.08] 0.0003 [0.42] !0.0004 [!0.47] !0.0008 [!0.78]
4 0.0015 [1.80] 0.0018 [2.28] 0.0013 [1.53] !0.0002 [!0.23] !0.0009 [!0.95]
High 0.0000 [0.00] !0.0005 [!0.53] !0.0001 [!0.06] 0.0005 [0.39] !0.0009 [!0.58]
RMRF Size quintiles

quintiles
Low 1.1018 [33.34] 1.1685 [47.35] 1.1248 [50.48] 1.0804 [50.56] 0.9489 [47.33]
2 1.0483 [43.84] 1.0819 [53.19] 1.0524 [52.01] 1.1042 [58.16] 1.0497 [50.10]
3 0.9798 [51.58] 1.0192 [56.59] 0.9773 [50.72] 1.0978 [51.14] 0.9744 [41.93]
4 0.9097 [45.70] 0.9909 [53.37] 0.9793 [50.87] 1.0441 [47.00] 0.9989 [47.17]
High 0.9903 [42.58] 1.0928 [45.79] 1.0883 [39.47] 1.1468 [37.65] 0.9834 [27.12]
SMB Size quintiles

quintiles
Low 1.2635 [23.80] 0.9235 [23.29] 0.6543 [18.27] 0.3097 [9.02] !0.2019 [!6.27]
2 1.1869 [30.89] 0.8928 [27.31] 0.6241 [19.19] 0.3252 [10.66] !0.1590 [!4.72]
3 1.0759 [35.25] 0.8252 [28.51] 0.6020 [19.44] 0.3421 [9.92] !0.2152 [!5.76]
4 1.0948 [34.23] 0.7648 [25.63] 0.4911 [15.88] 0.2020 [5.66] !0.1223 [!3.59]
High 1.2296 [38.90] 0.9500 [24.78] 0.6599 [14.88] 0.3183 [6.50] !0.0097 [!0.17]
HML Size quintiles

quintiles
Low !0.3701 [[!5.62] !0.5854 [!11.90] !0.5037 [!11.34] !0.4644 [!10.90] !0.4791 [!11.99]
2 !0.0462[!0.97] !0.1227 [!3.03] !0.0247 [!0.61] !0.0066 [!0.17] !0.0468 [!1.12]
3 0.1196 [3.16] 0.1351 [3.76] 0.2545 [6.63] 0.2769 [6.47] 0.2166 [4.67]
4 0.2235 [5.63] 0.3847 [10.39] 0.4118 [10.73] 0.5625 [12.70] 0.5686 [13.47]
High 0.5331 [11.50] 0.5517 [11.60] 0.6734 [12.25] 0.6878 [11.33] 0.8093 [11.20]
PR12 Size quintiles

quintiles
Low !0.0044 [!0.06] !0.0894 [!1.55] !0.0643 [!1.23] !0.0798 [!1.60] !0.2185 [!4.66]
2 !0.1232 [!2.20] !0.0329 [!0.69] !0.0240 [!0.51] !0.0552 [!1.24] 0.0230 [0.47]
3 !0.0073 [!0.16] !0.0254 [!0.60] 0.0458 [1.01] !0.0948 [!1.89] 0.2126 [3.91]
4 !0.0240 [!0.52] 0.0792 [1.82] !0.0410 [!0.91] !0.1063 [!2.04] !0.0286 [!0.58]
High 0.0843 [1.55] 0.1024 [1.83] 0.2756 [4.27] 0.1244 [1.75] 0.0911 [1.07]
229
230
Table 5 (continued)
Adjusted R Size quintiles

quintiles
Low 90.73 94.58 94.65 94.03 92.74

2 93.94 95.14 94.16 94.60 92.19
3 95.30 95.36 93.51 92.68 88.36
4 94.33 94.46 93.00 90.70 89.93
High 93.46 92.92 89.20 86.47 74.97
may also display underperformance even if the negative intercept is due to the
model misspeci"cation problem. In Section 5, we demonstrate quantitatively
that the returns on underperforming SEO "rms covary with small growth
companies in general. For instance, the simple correlation between the monthly
return over the market of the (1, 1) portfolio and the smallest SEO tercile is
0.867 (p"0.000), between the (1, 1) and the middle tercile is 0.625 ( p"0.000),
and between the (1, 1) and the largest tercile of SEOs is 0.101 ( p"0.129).
Section 5 also develops a new HML factor that helps the models to price the (1,
1) portfolio. As the results there show, the new HML factor can price both SEO
and IPO stock returns.
In Panel B, the purged factor returns appear to price the test portfolios almost
as well as the full sample returns. While the coe$cients do change slightly, there
is no dramatic shift in the average pricing errors, or intercepts, from Panel A.
For example, the issuer-purged factor returns still have di$culty pricing small
growth companies. The largest e!ect of purging the factor returns is on the PR12
coe$cients. It seems that issuing "rms contribute signi"cantly to PR12 returns.
Similarly, the adjusted RYs are slightly smaller in all the regressions with the
purged factor returns, although the models continue to explain the vast majority
of the time series of returns.
Table 6 presents the three and four factor regressions for various SEO
portfolios using both the full sample factor returns (Panel A) and the issuer
purged factor returns (Panel B). The results in Panel A show that both the three
and four factor regressions have di$culty pricing the equal-weighted SEO
portfolio returns. Once returns are value weighted, however, intercepts in three
and four factor regressions become small and statistically insigni"cant. If issuers
are broken down by size tercile, only the small and medium size terciles display
signi"cant underperformance with intercepts of !0.0069 and !0.0026 per
month. Interestingly, all SEO portfolios load negatively on PR12, meaning they
covary positively with low momentum stocks. This is one reason that matching
the buy-and-hold returns on the basis of momentum in Table 3 did not lower the
underperformance. SEO "rms actually have high past returns prior to issuing.
The average stock return in the year prior to issuance is 76%. But in the period
immediately following the SEO, their returns look like the returns of low past
return stocks. A risk-based interpretation of the negative factor loading on
PR12 would imply that SEO stocks are less risky after they issue. A sentiment-
based interpretation would argue that PR12 is just picking up SEO mispricing.
The results in Panel B with the purged factors are qualitatively similar to the full
sample results except that the R's of the regressions decline slightly. It therefore
does not appear that a large portion of the explanatory power of the
Fama}French model comes from factor contamination; as Loughran and Ritter
(1999) claim.
We present similar results for IPO "rms in Table 7. In the three factor
regressions for IPO returns, only the equal weighted and smallest tercile returns
232
Table 6
Fama}French (1993) time-series regressions on SEO rolling portfolios for the whole sample and sorted on the basis of size.
The sample of SEOs is all "rms that issued equity between 1975 and 1992. Portfolios of SEOs are formed by including all issues that were done within the
previous "ve years. RMRF is the return on a portfolio formed by subtracting the value weighted market return on all NYSE/ASE/Nasdaq "rms (RM)
minus the risk free rate (RF) which is the one!month Treasury bill rate. SMB (small minus big) is the di!erence each month between the return on small
"rms and big "rms. HML (high minus low) is the di!erence each month between the return on a portfolio of high book-to-market stocks and the return on
a portfolio of low book-to-market stocks. PR12 is the excess return on winners versus losers based on prior year returns excluding the previous month.
The "rst "ve columns presents results using the Fama and French three factor model. The next "ve columns add Carhart's momentum factor to the Fama
and French three factors. Within each set of regressions we report equal and value weighted results for the entire sample as well as equal weighted results
for three size sorts. [t-statistics are in brackets.]
Panel A: SEOs } Fama}French and Carhart's factors/size terciles
Full sample Size terciles Full sample Size terciles
Equal Value Small Medium Large Equal Value Small Medium Large
weighted weighted weighted weighted
Intercepts !0.0037 !0.0014 !0.0069 !0.0026 !0.0016 !0.0019 !0.0019 !0.0006 !0.0021 !0.0008
[!4.81] [!1.36] [!4.32] [!3.16] [!1.77] [!1.77] [!2.56] [!0.53] [!2.42] [!0.85]
RMRF 1.0491 0.9238 1.0693 1.0697 1.0082 1.0722 0.9340 1.1219 1.0762 1.0184
[53.72] [36.10] [26.60] [50.76] [42.72] [60.87] [36.42] [31.90] [50.73] [43.16]
SMB 0.7668 0.0328 1.2771 0.8292 0.1950 0.7364 0.0194 1.2078 0.8207 0.1816
[25.09] [0.82] [20.30] [25.14] [5.28] [26.81] [0.49] [22.02] [24.81] [4.94]
HML !0.1021 !0.0264 0.0887 !0.2547 !0.1408 !0.1507 !0.0478 !0.0218 !0.2683 !0.1623
[!3.17] [!0.63] [1.34] [!7.34] [!3.62] [!5.15] [!1.12] [!0.37] [!7.61] [!4.14]
PR12 !0.1984 !0.0874 !0.4518 !0.0557 !0.0876
[!7.94] [!2.40] [!9.05] [!1.85] [!2.61]
Adjusted R 95.69 87.65 86.95 95.58 91.76 96.57 87.89 90.24 95.62 91.95
Panel B: SEOs } issuer purged factors/size terciles
Intercepts !0.0040 !0.0017 !0.0069 !0.0030 !0.0020 !0.0028 !0.0007 !0.0050 !0.0023 !0.0010
[!4.65] [!1.63] [!4.17] [!3.01] [!2.14] [!3.34] [!0.65] [!3.04] [!2.30] [!1.04]
RMRF 1.0900 0.9345 1.1157 1.1205 1.0336 1.0971 0.9402 1.1269 1.1244 1.0398
[50.39] [36.74] [26.68] [44.61] [42.95] [54.03] [38.01] [28.07] [45.30] [44.83]
SMB 0.7355 0.0514 1.2557 0.7667 0.1848 0.7355 0.0515 1.2558 0.7667 0.1849
[20.42] [1.27] [18.92] [19.23] [4.84] [22.86] [1.31] [19.74] [19.50] [5.03]
HML !0.1183 0.0171 !0.0064 !0.2693 !0.0925 !0.1742 !0.0283 !0.0820 !0.3000 !0.1411
[!2.83] [0.35] [!0.08] [!5.54] [!1.99] [!4.31] [!0.58] [!1.03] [!6.07] [!3.06]
PR12 !0.2745 !0.2227 !0.4339 !0.1506 !0.2386
[!5.76] [!3.84] [!4.61] [!2.59] [!4.39]
Adjusted R 94.66 87.68 85.74 93.66 91.35 95.29 88.35 86.85 93.81 91.96
233
234
Table 7
Fama}French (1993) time-series regressions on IPO rolling portfolios for the whole sample and sorted on the basis of size.
The sample is all IPOs that issued equity between 1975 and 1992. Portfolios of IPOs are formed by including all issues that were done within the previous
"ve years. RMRF is the return on a portfolio formed by subtracting the value weighted market return on all NYSE/ASE/Nasdaq "rms (RM) minus the
risk free rate (RF) which is the one-month Treasury bill rate. SMB (small minus big) is the di!erence each month between the return on small "rms and big
"rms. HML (high minus low) is the di!erence each month between the return on a portfolio of high book-to-market stocks and the return on a portfolio of
low book-to-market stocks. PR12 is the excess return on winners versus losers based on prior year returns excluding the previous month. The "rst "ve
columns presents results using the Fama and French three factor model. The next "ve columns add Carhart's momentum factor to the Fama and French
three factors. Within each set of regressions we report equal and value weighted results for the entire sample as well as equal weighted results for three size
sorts. [t-statistics are in brackets.]
Panel A: IPOs } Fama}French and Carhart's factors/size terciles
Intercepts !0.0037 !0.0011 !0.0065 !0.0021 !0.0006 !0.0019 !0.0015 !0.0011 !0.0006 !0.0010
[!1.94] [!0.66] [!2.29] [!1.15] [!0.39] [!0.57] [!0.87] [!0.39] [!0.33] [!0.60]
RMRF 0.9798 1.0684 0.8385 1.0005 1.1013 1.0067 1.0633 0.9055 1.0189 1.0966
[24.39] [25.62] [11.73] [21.31] [28.06] [25.58] [25.17] [13.40] [21.65] [27.58]
SMB 1.2214 0.8459 1.4267 1.2729 0.9640 1.1860 0.8527 1.3384 1.2486 0.9702
[19.42] [12.96] [12.75] [17.32] [15.69] [19.33] [12.95] [12.71] [17.02] [15.65]
!0.2568 !0.5808 !0.2795 !0.5637 !0.3134 !0.570 !0.0624 !0.3183 !0.5539
HML 0.0784
[!3.88] [!8.45] [0.67] [!3.61] [!8.72] [!4.79] [!8.12] [!0.56] [!4.07] [!8.38]
PR12 !0.2310 !0.0443 !0.5752 !0.1584 !0.0403
[!4.13] [0.74] [!6.00] [!2.37] [0.71]
Adjusted R 86.68 86.57 63.59 83.53 88.79 87.52 86.54 68.22 83.84 88.76
Panel B: IPOs } issuer purged factors/size terciles
Intercepts !0.0030 !0.0017 !0.0059 !0.0019 !0.0011 !0.0023 !0.0019 !0.0041 !0.0016 !0.0012
[!1.79] [!0.91] [!2.05] [!1.01] [!0.61] [!1.35] [!1.03] [!1.41] [!0.79] [!0.69]
RMRF 1.0174 1.1295 0.8530 1.0360 1.1633 1.0212 1.1279 0.8631 1.0381 1.1624
[24.13] [24.75] [11.86] [21.33] [26.59] [24.29] [24.69] [12.11] [21.36] [26.53]
SMB 1.1548 0.7498 1.4044 1.2023 0.8579 1.1548 0.7498 1.4045 1.2023 0.8578
[17.25] [10.35] [12.30] [15.59] [12.35] [17.33] [10.36] [12.44] [15.61] [12.35]
HML !0.4175 0.6010 !0.1767 !0.4648 !0.6082 !0.4472 !0.5879 !0.2564 !0.4818 !0.6006
[!5.12] [6.81] [!1.27] [!4.95] [!7.18] [!5.35] [!6.47] [!1.81] [!4.98] [!6.89]
PR12 !0.1456 !0.0642 !0.3911 !0.0836 !0.0373
[!1.48] [!0.60] [!2.34] [!0.73] [!0.36]
Adjusted R 85.19 83.77 62.76 82.20 85.94 85.26 83.72 63.44 82.16 85.88
235
Fig. 3. Calendar}time cumulative abnormal returns by size of SEO issuer. The sample is 4526
seasoned equity o!ering (SEO) "rms that issued equity between 1976 and 1992. For each "rm in the
sample we calculate a time}series regression of its excess return on the Fama and French three
factors including a momentum factor where all factors have been purged of issuing "rms. From each
such regression we keep the intercept and residuals. We calculate size breakpoints each month using
market capitalization at the beginning of the month and plot the equal weighted cumulative sum of
these estimates for three size sorts.
show any underperformance. The four factor results, however, show that even
the equal weight portfolio and the smallest IPO "rm returns can be priced.
Intercepts are insigni"cant in all of the four-factor regressions.
The results from both the IPO and SEO sample indicate that the explanatory
power of the four-factor model declines modestly when the factors are purged of
issuers. This explanatory power is particularly low for the smallest tercile of IPO
and SEO stocks. Similarly, IPO and SEO stocks tend to load negatively on
PR12, the price momentum factor.
4.3. Calendar time returns
Loughran and Ritter (1999) also express concern that the time-series regres-
sion approach weights each calendar month equally rather than taking into
account the number of observations in each month. They argue that managers
will respond to temporary misvaluations by issuing overpriced equity that tends
to underperform subsequently. As a result, the time-series approach, which

weights each period equally, understates the severity of the underperformance
following equity issues.
To address this concern, we undertake the following analysis. For each SEO
"rm we regress each "rm's post-issuance excess return on the Fama and French
factor returns including Carhart's momentum factor from Eq. (2), where all
factor returns have been purged of equity issuers. From each such regression, the
intercept and the residual are summed to measure "rm-level time-series abnor-
mal returns (ARs). Using the individual ARs, we form three equal weighted size
portfolios } small, medium and large } monthly whose breakpoints are based on
market capitalizations at the beginning of the month. In Fig. 3 we plot the equal
weighted cumulative sum of these ARs is plotted for the three size sorts
beginning on the "rst date of our sample, January 1976.
As the "gure shows, the smallest tercile of issuers has consistently negative
abnormal performance. The medium and large issuers, however, do not exhibit
substantial underperformance over the entire sample period, although they do
substantially decline in value relative to the benchmarks between 1983 and 1984
and again in 1995. Fama and French (1995) argue that a potentially unexpected
earnings shock a!ected small, growth companies in the early and mid-1980s.
This is also the same period in which Brav and Gompers (1997) show that IPO
"rms underperformed.
Since the cumulative abnormal returns in Fig. 3 are averages of individual
SEO "rms' abnormal returns, the Loughran and Ritter (1999) critique that
weighting time periods equally reduces the power of tests of no underperfor-
mance does not apply here. It can be seen that for the large and medium size
terciles, the exact number of observations in each month is virtually irrelevant.
Apart from a two-year period of underperformance from 1983 to 1984 and
a small decline in 1995, both graphs are nearly #at, indicating that weighting the
calendar months equally does not lead to reduced power to reject the null
hypothesis of no underperformance.
5. Model misspeci5cation
In this section, we extend the analysis conducted in Section 4. In particular,

we demonstrate that the measured underperformance of equity issuers is likely
driven by the problem that the benchmark models available misprice certain
Fama (1998a) discusses how this time series portfolio approach can be modi"ed to allow for
alternative weighting schemes and suggests the approach of Ja!ee (1974) and Mandelker (1975),
which corrects for heteroskedasticity in portfolios' abnormal returns arising from changing com-
positions over time. While we do not explicitly correct for possible heteroskedasticity, the evidence
in Fig. 3 makes the relevant point notwithstanding.
238
Table 8
Fama}French (1993) time-series regressions on SEO and IPO rolling portfolios for the whole sample and sorted on the basis of size.
The sample is all "rms that issued equity between 1975 and 1992. Portfolios of issuers are formed by including all issues that were done within the previous
"ve years. RMRF is the return on a portfolio formed by subtracting the value weighted market return on all NYSE/ASE/Nasdaq "rms (RM) minus the
risk free rate (RF) which is the one-month Treasury bill rate. SMB (small minus big) is the di!erence each month between the return on small "rms and big
"rms. HMLH (high minus low) is the di!erence each month between the return on a portfolio of high book-to-market stocks and the return on a portfolio
of low book-to-market stocks excluding xrms listed on the NYSE. PR12 is the excess return on winners versus losers based on prior year returns excluding
the previous month. The "rst "ve columns presents results using the Fama and French three factor model. The next "ve columns add Carhart's
momentum factor to the Fama and French three factors. Within each set of regressions we report equal and value weighted results for the entire sample as
well as equal weighted results for three size sorts. [t-statistics are in brackets.]
Panel A: SEOs } Fama}French and Carhart's issuers purged factors/size terciles
Intercepts !0.0023 !0.0017 !0.0036 !0.0015 !0.0019 !0.0007 !0.0006 !0.0007 !0.0007 !0.0007
[!2.79] [!1.60] [!2.17] [!1.51] [!1.89] [!0.91] [!0.52] [!0.46] [!0.65] [!0.72]
RMRF 1.0560 0.9341 1.0213 1.1072 1.0395 1.0647 0.9402 1.0365 1.1117 1.0458
[52.70] [36.78] [25.89] [45.65] [43.10] [58.90] [38.06] [28.44] [46.58] [44.84]
SMB 0.7058 0.0529 1.2058 0.7334 0.1790 0.6999 0.0487 1.1954 0.7303 0.1747
[21.93] [1.30] [19.03] [18.82] [4.62] [24.14] [1.22] [20.45] [19.08] [4.67]
HMLH !0.1911 0.0110 !0.2911 !0.2348 !0.0478 !0.3170 !0.2226 !0.5545 !0.1662 !0.2300
[!6.92] [0.32] [!5.36] [!7.03] [!1.44] [!7.48] [!3.84] [!6.49] [!2.97] [!4.20]
PR12 !0.2359 !0.0205 !0.3695 !0.2583 !0.0804
[!9.23] [!0.59] [!7.17] [!7.65] [!4.20]
Adjusted R 95.40 87.68 87.27 94.07 91.28 96.26 88.35 89.13 94.26 91.85
Panel B: IPOs } Fama}French and Carhart's Issuer purged factors/size terciles
Intercepts 0.0009 0.0012 !0.0012 0.0027 0.0014 0.0021 0.0010 0.0015 0.0036 0.0012
[0.59] [0.64] [!0.43] [1.48] [0.77] [1.30] [0.49] [0.52] [1.94] [0.66]
RMRF 0.9516 1.1125 0.7415 0.9554 1.1574 0.9578 1.1112 0.7561 0.9604 1.1565
[25.41] [25.12] [10.82] [22.29] [26.80] [25.89] [25.05] [11.28] [22.53] [26.73]
SMB 1.0790 0.6824 1.3276 1.1141 0.7957 1.0748 0.6832 1.3176 1.1106 0.7963
[17.93] [9.59] [12.06] [16.18] [11.47] [18.11] [9.60] [12.25] [16.24] [11.48]
HMLH !0.5054 !0.4841 !0.4758 !0.5855 !0.4544 !0.2283 0.0447 !0.5332 !0.1853 0.0322
[!9.80] [!7.94] [!5.04] [!9.92] [!7.64] [!2.63] [0.43] [!3.39] [!1.85] [0.32]
PR12 !0.5377 !0.4777 !0.5512 !0.6117 !0.4499
[!10.28] [!7.62] [!5.81] [!10.15] [!7.35]
Adjusted R 88.28 84.66 66.12 86.10 86.26 88.56 84.61 67.53 86.24 86.20
239
240
Table 9
Four factor regressions for twenty-"ve test portfolios formed on the basis of size and book-to-market.
Coe$cients from four factor regressions for the 25 benchmark portfolios using purged factors, i.e., excluding all "rms that issued equity in an initial public
o!ering or seasoned equity o!ering during the previous "ve years. RMRF is the return on a portfolio formed by subtracting the value weighted market
return on all NYSE/ASE/Nasdaq "rms (RM) minus the risk free rate (RF) which is the one-month Treasury bill rate. SMB (small minus big) is the
di!erence each month between the return on small "rms and big "rms. HMLH (high minus low) is the di!erence each month between the return on
a portfolio of high book-to-market stocks and the return on a portfolio of low book-to-market stocks excluding xrms listed on the NYSE. PR12 is the excess
return on winners versus losers based on prior year returns excluding the previous month. [t-statistics are in brackets.]

quintiles
Low !0.0011[!1.01] !0.0003 [!0.24] !0.0002 [!0.13] 0.0015 [1.38] !0.0002 [!0.17]
2 0.0026 [2.52] 0.0001 [0.14] 0.0004 [0.40] !0.0007 [!0.87] 0.0004 [0.44]
3 0.0004 [0.42] 0.0010 [1.27] 0.0001 [0.12] !0.0008 [!0.82] !0.0004 [!0.41]
4 0.0015 [1.63] 0.0017 [1.84] 0.0014 [1.39] 0.0002 [0.18] 0.0006 [0.52]
High !0.0007 [!0.59] !0.0013 [!1.12] !0.0002 [!0.17] 0.0010 [0.63] 0.0004 [0.19]
RMRF Size quintiles

quintiles
Low 1.0024 [38.83] 1.1650 [43.08] 1.1507 [43.66] 1.0959 [44.65] 1.0092 [39.57]
2 1.0006 [43.31] 1.0962 [51.63] 1.0713 [51.81] 1.1056 [56.70] 1.0594 [49.12]
3 0.9650 [48.59] 1.0274 [55.96] 0.9723 [47.36] 1.0953 [48.25] 0.9574 [38.84]
4 0.9010 [42.59] 0.9777 [46.21] 0.9599 [42.90] 1.0107 [36.95] 0.9402 [33.88]
High 0.9842 [37.61] 1.0882 [40.67] 1.0646 [32.66] 1.1068 [30.51] 0.9201 [20.99]
SMB Size quintiles

quintiles
Low 1.1708 [28.75] 0.8529 [19.99] 0.6078 [14.62] 0.2628 [6.79] !0.2283 [!5.67]
2 1.1579 [31.77] 0.8855 [26.43] 0.6306 [19.32] 0.3251 [10.57] !0.1598 [!4.70]
3 1.0827 [34.55] 0.8451 [29.18] 0.6295 [19.43] 0.3734 [10.42] !0.1981 [!5.09]
4 1.1168 [33.47] 0.8034 [24.07] 0.5299 [15.01] 0.2516 [5.83] !0.0845 [!1.93]
High 1.2892 [31.23] 1.0126 [23.99] 0.7273 [14.14] 0.3793 [6.63] 0.0541 [0.78]
HML Size quintiles

quintiles
Low !0.5759 [!15.41] !0.3792 [!9.68] !0.2262 [!5.93] !0.2375 [!6.68] !0.0917 [!2.48]
2 !0.1937 [!5.79] !0.0275 [!0.89] 0.0499 [1.66] 0.0007 [0.03] 0.0041 [0.13]
3 0.0240 [0.83] 0.1130 [4.25] 0.1423 [4.79] 0.1649 [5.02] 0.0771 [2.16]
4 0.1102 [3.60] 0.1952 [6.37] 0.1911 [5.89] 0.2371 [5.98] 0.1535 [3.82]
High 0.3130 [8.26] 0.3298 [8.51] 0.3400 [7.20] 0.2927 [5.57] 0.2883 [4.54]
PR12 Size quintiles

quintiles
Low !0.0893 [!1.57] 0.0023 [0.04] 0.0571 [0.98] 0.0198 [0.37] !0.0506 [!0.90]
2 !0.1856 [!3.64] 0.0084 [0.18] 0.0079 [0.17] !0.0521 [!1.21] 0.0449 [0.94]
3 !0.0487 [!1.11] !0.0357 [!0.88] !0.0037 [!0.08] !0.1443 [!2.88] 0.1518 [2.79]
4 !0.0737 [!1.58] !0.0040 [!0.08] !0.1377 [!2.79] !0.2484 [!4.12] !0.2089 [!3.41]
High !0.0129 [!0.22] 0.0043 [0.07] 0.1293 [1.80] !0.0482 [!0.60] !0.1359 [!1.40]
241
242
Table 9 (continued)
Adusted R Size quintiles

quintiles
Low 94.64 93.82 92.90 92.53 88.86

2 94.63 94.98 94.22 94.60 92.15
3 95.13 95.43 93.01 92.23 87.57
4 93.92 93.17 91.03 86.62 83.60
High 92.15 91.56 85.70 81.80 65.31
classes of stocks of which issuing "rms are a part. To motivate the analysis in
this section, it is useful to review recent research that bears directly on the
alternative speci"cations we suggest in response.
Loughran (1993) examines the return di!erentials between "rms listed on
Nasdaq and the NYSE. He documents that during the period July 1981 to
December 1988, Nasdaq-listed stocks underperform NYSE-listed stocks by
about 5.8% per year, controlling for di!erences in book-to-market and the fact
that Nasdaq is populated by young IPO "rms. He concludes that the underper-
formance of Nasdaq-listed "rms is partially explained by the low returns on
recent IPOs. More recently, Hansen (1998) examines NYSE, Amex and Nasdaq
industries around initial public o!ering announcements. He "nds that the long
run abnormal performance of Nasdaq industries is similar to that of IPO "rms,
whereas NYSE and Amex industries do not underperform subsequent to public
o!erings.
In addition, Loughran (1997) examines the robustness of the book-to-market
in explaining di!erences in average returns. He "nds that for NYSE-listed "rms,
the book-to-market e!ect exists only in the month of January. Furthermore, this
January seasonal is con"ned to small "rms on the NYSE. By contrast, book-to-
market does explain di!erences in average returns for Nasdaq-listed stocks
outside January. Loughran attributes the latter fact to the mispricing of young,
growth "rms which tend to list mainly on Nasdaq.
Taken together, these studies imply that if we are interested in capturing
common covariation due to book-to-market as in Fama and French (1993), then
using NYSE-listed "rms may bias our performance tests and exacerbate the
e!ects of model misspeci"cation.
To examine this possibility, we replace the book-to-market factor-mimicking
portfolio, HML, used in the previous time-series regressions with an alternative
zero investment portfolio HMLH. The latter portfolio is formed as follows.
Using the 5;5 size and book-to-market cuto!s introduced earlier, we equal
weight the returns of Amex and Nasdaq listed stocks which have not issued
equity in the previous "ve years and belong to the smallest size and lowest
book-to-market portfolio, the (1, 1) portfolio. Second, we equal weight the
return of Amex and Nasdaq listed stocks which have not issued equity and
belong to the smallest size and highest book-to-market allocation, the (1, 5)
portfolio. HMLH is formed monthly by subtracting the average return on the
latter portfolio from the former. As in Fama and French (1993), the construction
of HMLH guarantees a focus on the di!erences in returns due to book-to-market
e!ects since any other exposures are netted out by forming the spread between
tails of the book-to-market distribution of stocks at portfolio formation. At the
same time, we are focusing on small, non-issuing "rms that are listed either on
See Loughran (1993, p. 257}259). These results pertain to "rms in the smallest two deciles based
on NYSE cuto!s.
the Amex or Nasdaq which, according to Loughran (1997), drive much of the
book-to-market e!ect.
Table 8 presents the time-series results which employ HMLH in place
of HML both for SEO and IPO portfolios. For SEOs, the underperformance
of the full sample relative to the three-factor model is halved. Small size
SEO underperformance has declined from 69 basis points per month in
Table 6 panel A to 36 basis points per month, although it remains statistically
signi"cant. Medium size SEO underperformance has declined from 26 basis
points to 15 basis points and is not statistically signi"cant at traditional
levels. For IPOs, it can be seen that relative to the three-factor model there
is no discernible underperformance for the full sample or any of the three
size sorts. The four factor regressions which include HMLH provide the
strongest support for the model misspeci"cation explanation for measured
underperformance. The intercepts for both the SEO and IPO portfolios
are economically and statistically negligible.
The above analysis, and our contention that issuer underperformance is likely
driven by a model misspeci"cation problem, could be criticized as an outcome of
data mining. We may have by chance found a formulation that works in sample.
We believe, however, that the motivation for the formulation of HMLH given in
the beginning of this section as well as in the analysis in the remainder of this
section are su$cient to alleviate such concerns. Below we provide an out of
sample test that addresses this issue.
We use the 25 size and book-to-market portfolios presented in Section 3
as test portfolios and price them using the four factor model which has
been purged from recent issuers including the new de"nition of HML. If
book-to-market e!ects are better captured by HMLH, our alternative speci"ca-
tion should price these portfolios better than the original four factor model
(see Table 5). Table 9 presents the regression results. All but one of the
25 intercepts are insigni"cantly di!erent from zero, and most have economically
negligible intercepts. The ability of our model with HMLH to price these test
portfolios gives us con"dence in their ability to capture general covariation in
returns.
These results can be interpreted in one or both of two ways. First, we have
shown that slight variation in the formation of one factor mimicking portfolio
leads to the disappearance of underperformance. Second, these results support
the view that underperformance is not unique to issuers; issuer returns covary
with returns of nonissuing stocks.
The original Fama and French (1993) model is an empirically motivated speci"cation. Likewise,
we use additional information to create factor-mimicking portfolios that maximize the possibility of
matching return di!erentials.
6. Recent research on the returns of equity issuers
The abnormal price performance of SEOs has recently attracted further

scrutiny from several researchers. Mitchell and Sta!ord (1998), Eckbo et al.
(2000), Brous et al. (1998), and Jegadeesh (1998) examine di!erent samples of
seasoned equity o!erings to test the abnormal performance hypothesis. In this
section we contrast their results with those reported here.
Mitchell and Sta!ord (1998) examine a large sample of SEOs over the period
1958}1993. Much like Brav et al. (1996) and this paper, they also measure
abnormal performance by employing both factor regressions as well as
attribute-based control portfolios. Consistent with the "nding in this paper, they
show that value-weighting the abnormal returns of issuers reduces the measured
abnormal performance. Mitchell and Sta!ord's paper di!ers from ours in two
main aspects. First, since they examine several corporate events (acquisitions,
SEOs, and repurchases), they do not focus on the general equity issuance
underperformance and the relationship between IPO and SEO abnormal per-
formance. Second, we explicitly examine the relation between equity issuer
underperformance and common co-movement of equity issuer and nonissuer
"rms' stock returns.
More recently, Eckbo et al. (2000) have questioned the existence of underperfor-
mance for a sample of SEOs from the period 1963}1995. Following Brav et al.
(1996) they also employ factor regressions as well as attribute matching to
evaluate abnormal performance. Their main result is that a six-factor asset-pricing
model can price portfolios of issuing "rms. Eckbo et al. argue that SEO "rms are
less risky than similar size and book-to-market matched "rms since their exposure
to unexpected in#ation and default risks are lower. Since their factor model is
based on macroeconomic risk factors it is hard to compare the results to those
presented in this paper. We leave such comparison for future research.
Brous et al. (1998) have recently examined the performance of seasoned equity
o!ering "rms. Brous et al. (1998) however, do not focus on abnormal perfor-
mance measurement. Instead, they argue that if SEO "rms are indeed mispriced
due to excessive optimism by the market as suggested by Loughran and Ritter
(1995), these erroneous expectations should disproportionately correct themsel-
ves when SEO "rms make subsequent earning announcements. Brous et al.
(1998) "nd little evidence that investors are systematically disappointed by
earnings announcements that follow equity o!erings, implying that the negative
abnormal performance of SEO "rms is not systematically concentrated at times
of earnings news.
Jegadeesh (1998) explores equity issuer underperformance in a sample of
seasoned equity o!erings as well. He examines returns of issuers using a charac-
teristics-based approach and "nds results that are similar to our characteristics-
based results. Jegadeesh does not utilize factor regressions and in fact critiques
their use based on Daniel and Titman (1997). Recent evidence by Berk (2000)
and Davis et al. (2000) call the Daniel and Titman results into question.
Similarly, an objective researcher should not disregard factor models altogether.
The researcher should see what the factor models say to potentially gain deeper
insights into the phenomenon. Jegadeesh also presents evidence that a large
portion of the negative SEO return occurs on or around earnings announce-
ments. Jegadeesh's earnings announcement results, however, are contradictory
to those of the Brous, et al. paper mentioned above. It is di$cult to assess which
paper has the correct result.
7. Conclusions
Our paper sheds light on the distinctiveness of equity issuer returns and the
robustness of various long-horizon stock performance tests. We explore the
e!ect that various long horizon test methodologies have on measured perfor-
mance. The results demonstrate that IPO returns are similar to nonissuing "rm
returns matched on the basis of size and book-to-market ratios. While SEO
returns underperform various characteristic-based benchmarks in event time
performance tests, time series factor models, which can price SEO portfolio
returns, show that SEO returns covary with nonissuing "rm returns. Finally, we
show that model misspeci"cation is an important consideration in long horizon
performance tests. Small changes to the factor speci"cation in Fama and
French's model improve its ability to price equity issuer returns as well as
commonly used test portfolios.
The evidence suggests that many of the long-run stock return anomalies
found in the "nance literature are manifestations of the same return pattern in
the data. Various types of corporate decisions may be more heavily weighted in
the types of "rms that underperform without the decision or event being studied
causing the underperformance. This co-movement is shown in our multi-factor
time-series models and our event time return calculations. In other words, our
evidence shows that equity issuers' return patterns are not distinct. They are part
of more systematic price movements observed for Amex and Nasdaq companies
that have not issued equity.
Recent papers by Teoh et al. (1998, 2000) and Lee (1997) indicate that various
attributes of the IPO or SEO "rms can predict which issuers will underperform.
Teoh et al. (1998, 2000) "nd that a higher level of discretionary accruals prior to
issuance, potentially to boost earnings, is associated with poorer post-issue
performance. Lee (1997) shows that SEOs in which sell equity holdings, perform
better than SEOs in which managers do not sell. In addition, "rms with higher
In a previous version of this paper we have shown that the post-listing negative drift documented
by Dharan and Ikenberry (1995) is largely caused by equity issuing "rms in their sample. These
results are available from the authors upon request.
levels of discretionary accruals and that issue pure primary shares tend to be
smaller IPO and SEO stocks, so it is not surprising that their measures are
correlated with relative underperformance. Thus, our results are consistent with
their "ndings.
Our results are consistent with both behavioral "nance and equilibrium
rational asset pricing interpretations. As Fama (1998b) argues, our results may
imply that we do not yet have the correct model of security returns. In fact, the
ability to narrow the underperformance or explain it with various factor models
is evidence in that direction. A behavioral explanation of our evidence would
suggest that investor sentiment a!ects a large number of "rms simultaneously.
While we do not o!er such a uni"ed theory of behavioral "nance, our results
point to the need for one.
Acknowledgements
We thank Eugene Fama, Campbell Harvey, Steve Kaplan, Tim Loughran,

Aharon Ofer, Jay Ritter, Bill Schwert (the editor), Rene Stulz, Robert Vishny,
Luigi Zingales, two anonymous referees, and seminar participants at the Uni-
versity of Chicago, the 1995 Western Finance Association meetings and the
second Red-Sea Finance Conference at Taba for their helpful comments and
suggestions. Jay Ritter graciously provided access to data on IPOs for part of his
sample, and Michael Bradley provided access to the SDC database. Victor
Hollender and Laura Miller provided excellent research assistance. Financial
support was provided by the Center for Research on Security Prices, University
of Chicago, and the Division of Research at the Harvard Business School. Any
errors or omissions are our own.
References
Agarawal, A., Ja!e, J., Mandelker, G., 1992. The post-merger performance of acquiring "rms:
A re-examination of an anomaly. Journal of Finance 47, 1605}1621.
Asquith, P., 1983. Merger bids, uncertainty, and stockholder returns. Journal of Financial Econ-
omics 11, 51}83.
Barber, B., Lyon, J., 1997. Detecting long-run abnormal stock returns: the empirical power and
speci"cation of test statistics. Journal of Financial Economics 43, 341}372.
Berk, J., 2000. Sorting the sorts. Journal of Finance, forthcoming.
Black, F., 1986. Noise. Journal of Finance 41, 529}543.
Brav, A., 1999. Inferences in long-horizon event studies: a Bayesian approach with applications to
initial public o!erings. Duke University working paper.
Brav, A., Geczy, C., Gompers, P.A., 1996. The long-run underperformance of seasoned equity
o!erings revisited. Working paper, University of Chicago.
Brav, A., Gompers, P.A., 1997. Myth or Reality? The long-run underperformance of initial public
o!erings: Evidence from venture capital and nonventure capital-backed companies. Journal of
Finance 52, 1791}1822.
Brous, P., Datar, V., Kini, O., 1998. Is the market optimistic about the future earnings of seasoned
equity o!ering "rms? Seattle University working paper.
Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52, 57}82.
Daniel, K., Titman, S., 1997. Evidence of the characteristics of cross-sectional variation in stock
returns. Journal of Finance 52, 1}34.
Davis, J., Fama, E., French, K., 2000. Covariances and average returns: 1929}1997. Journal of
Finance, forthcoming.
De Bondt, W., Thaler, R., 1985. Does the stock market overreact? Journal of Finance 40,
793}808.
De Bondt, W., Thaler, R., 1987. Further evidence on investor overreaction and stock market
seasonality. Journal of Finance 42, 557}581.
De Long, J.B., Shleifer, A., Summers, L.H., Waldmann, R.J., 1990. Noise trader risk in "nancial
markets. Journal of Political Economy 98 (4), 703}738.
Dharan, B., Ikenberry, D., 1995. The long-run negative drift of post-listing stock returns. Journal of
Finance 50, 1547}1574.
Eckbo, E., Masulis, R.W., Norli, "., 2000. Conditional long-run performance following security
o!erings: is there a `new issues puzzlea? Journal of Financial Economics, forthcoming.
Fama, E., 1970. E$cient capital markets: a review of theory and empirical work. Journal of Finance
25, 383}417.
Fama, E., 1998a. Market e$ciency, long-term returns, and behavioral "nance. Journal of Financial
Economics 49, 283}306.
Fama, E., 1998b. Determining the number of priced state variables in the ICAPM. Journal of
Financial and Quantitative Analysis 33 (2), 217}231.
Fama, E., French, K., 1992. The cross-section of expected stock returns. Journal of Finance 47,
427}466.
Fama, E., French, K., 1993. Common risk factors in the returns of stocks and bonds. Journal of
Financial Economics 33, 3}55.
Fama, E., French, K., 1995. Size and book-to-market factors in earnings and returns. Journal of
Finance 50, 131}156.
Hansen, R.S., 1998. Industry performance around initial public o!erings: a missing piece to the new
issues puzzle. Working paper, Virginia Tech.
Ikenberry, D., Lakonishok, J., Vermaelen, T., 1995. Market underreaction to open market share
repurchases. Journal of Financial Economics 39, 181}208.
Ja!ee, J., 1974. Special information and insider trading. Journal of Business 47, 881}898.
Jegadeesh, N., 1998. Long-run performance of seasoned equity o!erings: benchmark errors and
biases in expectations. Working paper, University of Illinois at Urbana Champaign.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: Implications for stock
market e$ciency. Journal of Finance 48, 881}898.
Kothari, S.P., Warner, J., 1997. Measuring long horizon security price performance. Journal of
Financial Economics 43, 301}339.
Lee, I., 1997. Do "rms knowingly sell overvalued equity? Journal of Finance 52, 1439}1466.
Lee, C., Shleifer, A., Thaler, R., 1991. Investor sentiment and the closed-end fund puzzle. Journal of
Finance 46, 75}109.
Lintner, J., 1965. The valuation of risk assets and the selection of risky investments in stock
portfolios and capital budgets. Review of Economics and Statistics 47, 13}37.
Loughran, T., 1993. NYSE vs NASDAQ returns. Journal of Financial Economics 33, 241}260.
Loughran, T., 1997. Book-to-market across "rm size, exchange, and seasonality: is there an e!ect?
Journal of Financial and Quantitative Analysis 32(3), 249}268.
Loughran, T., Ritter, J., 1995. The new issues puzzle. Journal of Finance 50, 23}51.
Loughran, T., Ritter, J., 1999. Uniformly least powerful tests of market e$ciency. Forthcoming in
the Journal of Financial Economics.
Loughran, T., Vijh, A., 1997. Do long-term shareholders bene"t from corporate acquisitions?.
Journal of Finance 52, 1765}1790.
Lyon, J., Barber, B., Tsai, C., 2000. Improved methods for tests for long run abnormal stock returns.
Journal of Finance, forthcoming.
Mandelker, G., 1975. Risk and return: the case of merging "rms. Journal of Financial Economics 1,
303}335.
Michaely, R., Thaler, R., Womack, K., 1995. Price reactions to dividend initiations and omissions.
Mitchell, M., Sta!ord, E., 1998. Managerial decisions and long-term stock price performance.
Working paper, University of Chicago.
Myers, S., Majluf, N.S., 1984. Corporate "nancing and investment decisions when "rms have
information that investors do not have. Journal of Financial Economics 13, 187}221.
Ritter, J., 1991. The long-run performance of initial public o!erings. Journal of Finance 42, 365}394.
Sharpe, W.F., 1964. Capital asset prices: a theory of market equilibrium under conditions of risk.
Shleifer, A., Vishny, R., 1997. The limits to arbitrage. Journal of Finance 52, 35}55.
Spiess, K., A%eck-Graves, J., 1995. The long-run performance following seasoned equity issues.
Journal of Financial Economics 38, 243}267.
Spiess, K., A%eck-Graves, J., 1999. The long-run performance following public bond issues. Journal
of Financial Economics 54, 45}73.
Teoh, S., Welch, I., Wong, T.J., 1998. Earnings management and the long-run underperformance of
seasoned equity o!erings. Journal of Financial Economics 50, 63}100.
Teoh, S., Welch, I., Wong, T.J., 2000. Earnings management and the long-run underperformance of
initial public o!erings. Journal of Finance, forthcoming.

Brav Ge Czy Go Mpers 2000

Uploaded by

Copyright:

Available Formats

Brav Ge Czy Go Mpers 2000

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Brav Ge Czy Go Mpers 2000

Uploaded by

Copyright:

Available Formats

Journal of Financial Economics 56 (2000) 209}249

Is the abnormal return following equity

Received 19 October 1998; received in revised form 10 August 1999

We examine whether a distinct equity issuer underperformance anomaly exists. In

JEL classixcation: G1; G2

Keywords: Seasoned equity o!erings; Initial public o!erings; Long-run performance;

* Corresponding author. Tel.: #617-495-6297; fax: 617-496-8443.

Vijh (1997), who examine underperformance in stock returns following mergers;

2. Performance tests, time horizons, trading strategies, and methodology

the method of performance measurement in#uences both the magnitude of the

Our sample of SEOs was identi"ed in the Investment Dealers+ Digest of

Book-to-market Size quintile

Low 51.80 18.20 5.50 1.70 0.40 77.60

Book-to-market Size quintile

Low 24.10 12.60 7.30 4.20 2.40 50.70

4. Underperformance of equity issuers

4.1. Event time returns

Panel A: Buy-and-hold returns

SEO Benchmark Abnormal Wealth SEO Benchmark Abnormal Wealth

Panel B: Cumulative abnormal returns

Benchmarks Equal weighted CARs (%) Value weighted CARs (%)

SEO Benchmark CAR SEO Benchmark CAR

S&P 500 index 51.5 75.5 !24.0 60.4 70.2 !9.8

Panel A: Buy-and-hold returns

IPO Benchmark Abnormal Wealth IPO Benchmark Abnormal Wealth

The long-run performance of initial public o!erings.

Panel B: Cumulative abnormal returns

Equal weighted CARs (%) Value weighted CARs (%)

Benchmarks IPO Benchmark CAR IPO Benchmark CAR

S&P 500 index 35.9 74.2 !38.3 40.7 61.5 !20.8

themselves free of issuing "rms, there is no underperformance. Average excess

4.2. Factor regression analysis

The characteristics-based approach of the previous section assumes that

COMPUSTAT universe of "rms.)

Panel A: Full sample factors

Intercept Size quintiles

Book-to-market Small 2 3 4 Large

RMFR Size quintiles

Book-to-market Small 2 3 4 Large

Book-to-market Small 2 3 4 Large

HML Size quintiles

Book-to-market Small 2 3 4 Large

PR12 Size quintiles

Book-to-market Small 2 3 4 Large

Book-to-market Small 2 3 4 Large

Low 92.70 96.83 95.96 95.16 92.78

Panel B: Purged factor regressions

Intercept Size quintiles

Book-to-market Small 2 3 4 Large

RMRF Size quintiles

Book-to-market Small 2 3 4 Large

Book-to-market Small 2 3 4 Large

HML Size quintiles

Book-to-market Small 2 3 4 Large

PR12 Size quintiles

Book-to-market Small 2 3 4 Large

Adjusted R Size quintiles

Book-to-market Small 2 3 4 Large

Low 90.73 94.58 94.65 94.03 92.74

Adjusted R Size quintiles

Adusted R Size quintiles