COMPETITIVE EFFECTS OF IPOS: EVIDENCE
FROM CHINESE LISTING SUSPENSIONS
FRANK PACKER AND MARK M. SPIEGEL
ABSTRAct. Theory suggests that initial public offerings (IPOs) can adversely im- pact
listed firms, both directly by increasing intra-industry competition, and in- directly by
completing related asset market spaces. However, the endogeneity of individual IPO
activity hinders testing these channels. This paper examines listing suspensions in China in
a panel specification that accounts for macroeconomic and financial conditions, isolating
the firm-level IPO impact. We measure the competi- tive impact of listing suspensions
through the value share of postponed firms in the IPO queue in their industry, and asset-
space competition by firms’ historical covari- ance with a synthetic portfolio of listed firms
with the IPO queue industry mix at the time of suspension. Our results support the
predicted IPO effects through both channels. We also document heterogeneity in IPO
effects. Stronger firms–measured through a variety of proxies–benefit less from the
suspension news. These results are robust to a battery of sensitivity tests.
Date: September 16, 2020.
Key words and phrases. Initial public offerings, China, competition, asset space.
JEL classification: G14, G18, G32
Packer: Bank for International Settlements; Email: @; Spiegel: Federal Re- serve
Bank of San Francisco; Email: @. Remy Beauregard and Jimmy Shek provided
excellent research assistance. We are grateful to Jing Liu, Jay Ritter and seminar partici- pants at the Bank for
International Settlements and the Federal Reserve Bank of San Francisco for helpful comments. This paper
was written in part while Spiegel was a visiting scholar at the Bank for International Settlements, whose
members he thanks for their hospitality and helpful comments. Spiegel’s research is also supported by the
National Natural Science Foundation of China, Project Number 71633003. The views expressed in this paper
are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of San Francisco,
the Federal Reserve System, or the Bank for International Settlements.
1
mailto:@
I. INTRODUCTION
Do IPOs have competitive effects on other firms? At least two channels have been cited
in the literature. In one, firms going public threaten listed firms profits in the same industry,
either by posing a direct competitive threat to other firms operating in their industry (.
Akhigbe et al. (2003) and Hsu et al. (2010)), or by changing the strategic dynamics of the
industry more generally (Spiegel and Tookes (2020)). Alternatively, the assets generated by
initial public offerings may provide a valuable alternative to investors, and thus reduce
demand for listed firm assets with similar risk characteristics (Braun and Larrain (2009)).
One challenge encountered in testing the importance of these channels is that in- dividual
firm listing decisions are endogenous. For example, it has been shown that waves of IPO
activity may signal overvaluation in an industry or covariance group (Ritter (1991),
Henderson et al. (2006)).
In this paper, we make use of the market implications of blanket suspensions of IPO
activity by the China Securities Regulatory Commission (CSRC) to examine the importance
of IPO competitive effects. In particular, we evaluate the equity valuation impact of
announcements of blanket suspensions in IPO activity. These suspensions were largely
unanticipated, and have been shown to have measurable implications for overall market
valuations (. Shi et al. (2018)).
Our analysis utilizes a panel sample to examine the implications of CSRC suspen- sions
on individual listed firms. We examine the implications of the 3 blanket IPO suspension
episodes in China that have been imposed since 2008. These suspensions were unanticipated,
and initially of unknown duration. Our panel specification allows us to condition for
macroeconomic and market conditions at the time of the suspension announcements. Combined
with knowledge of the composition of the queue of firms about to go public ahead of each
suspension, our investigation allows us to identify disparities in listed firm exposure to the
direct or asset space competition posed by queue firms at the time of the IPO suspension
announcements. This knowledge allows us to measure the market’s assessment of the
importance of these channels based on the impact of the announcements on firm equity
valuations.
Our results indicate that anticipated competition from new IPOs is evident through both
channels. IPO suspensions in China benefit those listed firms in industries heavily
represented in the queue of firms approved to go public, consistent with the expectation that
the IPO suspension would mitigate competition in product markets. Asset supply effects also
matter. Listed firm shares with greater covariance in returns
with a synthetic portfolio that replicates the industry composition of suspended IPOs earn
higher returns on suspension announcement dates. This suggests that these firms are expected
to benefit from a restriction in the supply of assets by firms with similar asset return
characteristics.
Our analysis also documents significant heterogeneity by firm health in the expo- sure of
firms to these competitive effects: we find that more profitable and productive firms are
significantly less sensitive to the competitive challenges presented by IPOs. In particular, we
find that equity changes in response to the suspension announce- ments among more
profitable and productive firms – measured through a variety of alternative metrics – are
less sensitive to the representation of their industry in the IPO queue at the time of the
announcement than less profitable and productive firms. Our findings therefore complement
those of Hsu, Reed and Rocholl (2010), who find increased sensitivity to the competitive
effects of IPOs among more leveraged and less research intensive firms. We also find a
similar pattern through the asset space channel, as healthier firms benefit less from the
reduced supply of correlated assets resulting the IPO suspension. However, this latter
channel is not as strong or robust in terms of statistical significance as the direct competition
channel.
The roadmap to the rest of the paper is as follows. In part 2, we review the
literature; and in part 3, we discuss institutional details with regard to the practice of
suspensions in China. In part 4, we provide an overview of the data and variable
construction. We present the methodology of the empirical tests in more detail in part 5.
After reporting the empirical results and a battery of robustness checks in parts 6 and 7,
respectively, we summarise our conclusions and policy implications in part 8.
II. LITERATURE REVIEW
A large literature exists examining how the IPO event—characterised by a discrete change
in scale of operations, capital structure and public visibility—affects firm per- formance,
valuation, and innovation. Of more recent vintage is the research that has examined how the
firm IPO affects other firms.
There are at least two channels through which IPOs have been shown to influ- ence
already public companies. First, there is the increased competition that newly listed firms
bring to their industry, which adversely affects direct competitors. While Akhigbe et al.
(2003) had tested for such an effect for more than 2000 IPOs between 1989-2000 and not
uncovered evidence for it, Hsu et al. (2010)—by focusing on large
IPOs and allowing for anticipation of events before the event date—found that sea- soned
industry competitors experience a negative share price reaction around the time of the IPO,
and post-IPO their operating performance declines as well. Hsu et al also show the negative
impact to be greater the more leveraged and less R&D intensive the listed firm; they
further show the impact goes in reverse for withdrawn More recently, Spiegel and
Tookes (2020) have demonstrated that rival firms incur losses upon IPOs, but more because
IPOs are an indicator of increased competition in the industry, rather than the newly listed
firms becoming stronger after the IPO.
The second channel through which IPOs have been documented to affect seasoned firms is
through a financial asset supply effect. Hong et al. (2008) show that local jurisdictions’
relative asset supply can affect the relative valuations of listed firms; Baschieri et al.
(2015) also document that local IPOs diminish the value of neighbor- ing listed firms.
Increases in traded assets generated by an IPO may also adversely affect the valuation of
firms with similar financial characteristics. In their examination of more than 250 IPOs in 22
emerging markets, Braun and Larrain (2009) document that highly covarying listed securities
experience a price decline upon IPOs in emerg- ing market economies (EMEs), and the
effect is larger the bigger the IPO and the less integrated the EME market is globally. Li
et al. (2018) examine IPO approval announcements in China. They find a negative (though
transitory) impact on listed share prices which is more prominent the higher the correlation
with the industry of the IPO firms, consistent with the asset supply hypothesis.
The empirical literature on the market timing of equity issuance poses a challenge to
research on the competitive effects of IPOs. Ritter (1991) documents long-term under-
performance of IPOs accounted for by poorly performing companies going pub- lic in high-
volume years, which is consistent with firms taking advantage of windows of opportunity to
issue when investors are “irrationally overoptimistic about the fu- ture potential of certain
industries.” Pagano et al. (1998) finds that private firms are more likely to go public when
industry market-to-book are unusually high. Boeh and
1Chemmanur and He (2011) also document that IPOs in an industry are associated with market
share decline of competing firms in the same industry. Similarly, in their study of Australian firms,
McGuilvery et al. (2012) show listed companies to be negatively affected by the completion of an IPO in
their industry. Chod and Lyandres (2011) present a model of the decision to go public in the presence of
product market competition whereby firms going public not only increase their market share, but adversely
affect the market value of industry rivals. Nguyen et al. (2014) document that in reaction to perceived
overreaction of the market to the competitive challenges posed by competing firm IPOs, listed firms in the
same industry increase their share repurchases.
Dunbar (2014) document higher IPO issuance when average priori IPO returns have been
high. Baker and Wurgler (2000) document downward adjustment in the prices of shares
related to firms issuing new shares, as firms issue just before periods of low market
Market timing of IPOs by opportunistic issuers means that a negative relationship between
the announcement of the IPO and the valuation of similar, already listed firms, cannot
necessarily be taken as evidence of competitive effects. Since firm man- agement are more
informed about the firm’s industry more generally, announced IPOs can reflect overly generous
valuations of similar firms. IPO announcements could then trigger a price reversal among
existing firms in the industry even without any com- petitive effects. Problems of the same
nature can emerge when IPOs are withdrawn in response to perceived excess doldrums of
market conditions.
It follows that a shock to activity is needed for identification of competitive IPO
effects. However, individual IPOs (or withdrawn IPOs) can generally not be observed on that
basis. Nevertheless, over the past few decades the regulators of mainland China, one of
the world’s largest IPO markets, have implemented blanket IPO sus- pensions of IPO
activity. Though these events are are often launched in response to macroeconomic and/or
market conditions, their blanket nature creates an identifiable exogenous and unanticipated
shock to IPO activity in the cross section for a large number of identifiable firms.
We are not the first to recognize the research value of IPO suspensions in China. In related
empirical work, Cong and Howell (2019) examine the impact of IPO suspen- sions in China
on the firms whose IPOs were suspended. They document significant declines in growth and
innovative activity as a result. However, they do not examine the impact on the currently
listed competitors of suspended firms. In a related paper,
2In addition to IPOs, opportunistic issuance is evident in the case of seasoned equity offerings
as well. Loughran and Ritter (1997) document systematic long-term underperformance of seasoned equity
offerings, which they interpret as being due in large part to firms taking advantage of windows of opportunity to
issue equity when they are overvalued. The above cited Baker and Wurgler (2000) paper considers seasoned
as well as new equity issuance, and concludes that the evidence suggests that “. . . firms time the market
component of their returns when issuing securities.” Bradley and Yuan (2013) also document that rival firms
can be affected by seasoned equity offerings – negatively so in the case of seasoned secondary share
offerings. The theoretical model of Chemmanur and He (2011) predicts lower post-IPO profitability and
productivity for firms going public during an IPO “wave”.
Shi et al. (2018) examine whether large IPOs in China affect overall market condi- tions
and document a negative impact which is larger the larger the IPO. By contrast, they do not
find that IPO suspensions have been viewed as good news by the overall market. But Shi et
al look at aggregate effects of IPOs, while our examination of the impact of the suspensions
on individual shares allows us to condition for differences in firm characteristics at the time
of suspension announcements.
III. IPO SUSPENSIONS IN CHINA: THE INSTITUTIONAL BACKDROP
. Chinese IPO suspensions. As described above, China’s securities regulator, the
CSRC, is authorised to impose suspensions on new IPOs of indeterminate length, and has
done so nine times since 1994 and five times since 2005. The motivation for the suspensions
appears to have been concern over market stability, in particular that liquidity might be
reduced, market prices depressed, or demand from existing stocks depressed by new IPOs
(Tian (2011), Shi et al. (2018), Cong and Howell (2019)).3
The length of the suspension is indeterminate at its launch; the lifting date is only
announced later on. In fact, the length is quite variable: in the three most recent
suspensions of 2008-09, 2012-14, and 2015 that are the focus of this study, the length of the
suspensions were 214, 438 and 156 days, Cong and Howell (2019) demonstrate
that the suspensions were costly to firms that had been approved and were in the queue
for listing, likely due to increased market uncertainty and lost strategic opportunities.
Despite the stated objective of the regulatory authorities to achieve stabilization of the
broader market, the evidence of the impact on the market of the IPO suspensions has been
mixed. Figure 1 shows the Shanghai stock exchange index (SSE) from 2008 through 2016, a
period which includes the three suspensions at the core of this study. While the suspensions
followed sustained declines in 2008 and 2012, and a sharper decline in 2015, Graph 1
shows no consistent direction in the longer-term movement of the SSE during the
suspension periods. Though in 2008-2009 the stock market rebounded a significant amount
during the suspension period, over the 2012-2014
3The last IPO suspension was lifted in November 2015, and the head of the CSRC who oversaw the last two
of the suspensions, Xiao Gang, was removed from his post in February 2016. The Chinese securities
regulator has since not imposed another despite episodes of volatility in the overall market.
4Our choice of dates follow Shi et al. (2018), who in turn refer to Hexun, a Chinese financial
news web-site ( Our analysis covers only back to the
2008-09 suspension and the two that follow, since the availability of IPO approval dates does not extend
reliably further back.
suspension, it remained fairly constant, and in the immediate aftermath of the 2015
suspension announcement share prices continued an ongoing decline that was only partially
reversed late in the
The above results are consistent with Packer and Spiegel (2016), who showed that there is
little correlation between the size of market issuance of IPOs and overall market
movement. They are also consistent with Shi et al. (2018)’s examination of the Shanghai
Stock Exchange Index response to nine IPO suspensions from 1995, which concludes that
the market does not respond positively to the announcement of the suspension, and is
“inconsistent with the view that IPO moratoriums can mitigate the downward price pressure
during bad times.” It is possible that the mixed results on the broader market could be due to
the endogeneity of suspensions to macroeconomic conditions. The suspensions may have
been viewed as a signal of adverse aggregate news, which weighed on market sentiment.
. Cross-sectional implications for individual listed ftrms. Regardless of the factors
that might be affecting the overall market at the time of an IPO sus- pension, there is
reason, as mentioned in the above literature review, to expect that differences in the cross
− section of listed firms should be reflected in the impact of IPO suspensions.
Particularly firms with a similar profile—in terms of industry affiliation or asset returns
characteristics—as the prospective firms whose IPOs were
5Both the CSI 300 and the Shenzhen stock exchange index show very similar patterns.
suspended should be affected more greatly by an IPO suspension. And the related
hypotheses can be tested because, at the time of each of the three suspensions, there was a
readily identifiable list of firms about to go public.
There are a series of steps required for Chinese firms to go public in mainland China.
After applying for approval by the CSRC, or Chinese Securities Regulatory Commission,
there is a preliminary review of the application that may take years. This process usually
involves multiple meetings and repeated requests to the applicant for more information. This
is followed by a formal assessment by the Stock Issuance Examination and Verification
Committee of the CSRC as to whether listing criteria are met, and a decision by this
Committee whether or not approve the listing. Around 70-80% of firms gain approval, and the
results of the IPO approval process are publicly announced. The firm receiving formal
approval may then apply to go public listing at one of the domestic exchanges within six
months, though the exchange approval is merely a formality due to exchange rules being
identical to CSRC requirements. The firm with the help of underwriters builds a book,
conducts a road show and decides on a share subscription day. After subscription it takes
around 4 weeks for the shares to list; in total the time between approval and listing averages
around 3 months, though the interval has varied between two and five months (when there
has not been an intervening suspension).
This process allows us to identify the group of firms that have been approved for IPO that
have yet to issue at the time of the suspension announcement. Namely, ahead of the 2008-
09, 2012-14 and 2015 suspensions that are examined in this paper, 30, 66 and 62 firms had
been approved for an IPO and were waiting to list, respectively. The average size of the total
158 postponed IPOs, when they later occurred, was around one billion RMB, though IPO
size shows considerable variation across suspensions, averaging more than 3 billion RMB in
2008, compared to and billion RMB in 2012 and 2015, respectively (Table 1). The
difference in average size in turn is what drives a much larger sum of suspended IPOs in 2008
(91 billion RMB) relative to those in the 2012 and 2015 suspensions (29 and 35 billion RMB).
Importantly, we know the industry affiliation of each of these firms, and thus the percent
of each industry’s market capitalisation the IPOs would have occupied at the time of the
suspension: in Table 1, we see that this averages % for the entire queue sample, though
again considerably more for the 2008 sub-sample (%).
Not only is there a variety of industries represented in the prospective IPO queue, there is
also considerable heterogeneity in the length of time the postponed IPOs
TABLE 1. Postponed IPO Firm Characteristics
Pooled suspensions (158 ftrms) Mean Median Std. Dev. Min Max
Size of postponed IPO RMB bn
Postponed IPO/Industry Market Cap %
Length of Suspension Days 214 156 438
Days from Approval to IPO Days 175 1500
SUM of postponed IPOs RMB bn
2008 Suspension (30 ftrms)
Size of postponed IPO RMB bn
Postponed IPO/Industry Market Cap %
Length of Suspension Days 214
Days from Approval to IPO Days 386 367 726
SUM of postponed IPOs RMB bn
2012 Suspension (66 ftrms)
Size of postponed IPO RMB bn
Postponed IPO/Industry Market Cap %
Length of Suspension Days 438
Days from Approval to IPO Days 546 1500
SUM of postponed IPOs RMB bn
2015 Suspension (62 ftrms)
Size of postponed IPO RMB bn
Postponed IPO/Industry Market Cap %
Length of Suspension Days 156
Days from Approval to IPO Days 236 175 417
SUM of postponed IPOs RMB bn
Note: The size of the postponed IPO is taken from the actual IPOs of queue firms that listed subsequent to the end of the IPO
suspension. Industry market capitalisation measures are taken at the time of the suspension. Days from approval to IPO is measured for
each queue firm going public from its approval date for IPO prior to the suspension to its IPO date subsequent to the end of the suspension.
Sum of postponed IPOs adds up the sizes of each IPO that was postponed across queue firms, both for the pooled sample and for each
suspension separately.
Source: .
are delayed. The actual lengths of the suspensions themselves vary from 214 days in 2008-
2009 to more than one year at 438 days in 2012-2014 and to a relatively modest 156 days in
2015. However, variation in the length of time IPOs have been in queue ahead of the
suspension and the time it subsequently takes to go public imply that the overall time from
approval to the eventual IPO vary within each suspension as well,
. ranging from 367 to 726 days for IPOs postponed by the 2008-09 suspension, and
from 546 to 1500 days for the 2012-14 We later test whether industry and
suspension-based differences in the listing delays, as proxied by the time between approval and
listing, matter for the competitive impact of the IPO suspensions.
IV. DATA
. Sample and variable deftnitions. Our panel sample consists of pooled data for listed
firms on the Shenzhen and Shanghai stock exchanges at the time of the an- nouncements of
the 2008, 2012 and 2015 suspensions. Our base specification includes 6,045 observations, with
1,484 firms from the 2008 suspension, 2,390 firms from the 2012 suspension, and 2,171 firms
from the 2015 suspension. Our dependent variable is ri,t, the one-day return on equity values.
The returns are taken over the one-day win- dow corresponding to the closing price of the first
trading day after the announcement of the suspension over the end of the previous trading
day’s close.
Our variables of interest are proxies for the intensity of our two channels of potential
competition raised by the firms at the time of the suspensions. We first consider an
industry-level measure of potential delay in direct competition based on the firms from each
listed firm’s industry at the time of the suspension. First, we identify the pre- IPO firms in
the queue at the time of the IPO suspension, and sort by industry, using the CSRC industry
definitions from WIND. For each of the firms in the queue, we then take the realized public
offering amounts at the time of their later IPO (for queue firms that ended up never going
public, the number is zero) and sum them up across all firms within each industry i at time
t. This yields a proxy for the expected total potential market cap of queue firms in industry
i at time t, or MCQi,t. Our measure implictly assumes that investors have unbiased
expectations of both the ultimate size of the IPO, and those IPOs that will never take
place. Because the impact on a particular industry should be greater, the higher the
proportion of suspended IPOs relative to existing industry market capitalization, we divide
this sum by the total market capitalization of all listed firms within industry i at the time of
the suspension t, or MCLi,t.
6Among the queue firms in our sample, there is a negative correlation between the time from
approval to suspension and the time from the end of suspension to the eventual IPO, in the case of both the
2008-2009 and 2015 episodes, which suggests that the queue was respected, at least to some extent. However,
the same correlation is negative for prospective IPO firms ahead of the 2012 event, where perhaps the length
of the suspension—438 days—implied that the order of approval prior to the suspension was no longer
viewed as relevant when the IPO market opened up again.
Σ
Σ
Σ
Our primary proxy for potential direct competition from firms in the IPO queue at the time
of suspension, which we term IP Oi,t, then satisfies
IP Oi,t =
MCQi,t . (1)
MCL
i,t
As an alternative proxy, we also adjust for the length of time that the suspension delayed
the IPO process. We again assume unbiased foresight and base these expec- tations on
realized IPO delays. In particular, for each pre-IPO firm f in industry i in the IPO queue
we evaluate delay as the days between the eventual IPO date and the approval date by the
CSRC, which we term Delayf,i,t. We then multiply the days of delay for each queue firm
by its public offering amount MCQf,i,t, and take the sum across all f firms in the queue in
industry i at time t, f (Delayf,i,t · MCQf,i,t). We subsequently divide this total by total
market capitalization of the same industry
used earlier, or
Σ
n MCLi,t. Our alternative proxy, which we term DIP Oi,t satisfies
DIP Oi,t =
(Delayf,i,t · MCQf,i,t ). (2)MCL
f i,t
DIP Oi,t therefore corresponds to the average delay in days for IPOs in an industry,
weighted by the public offering amount of the firms in the queue relative to industry market
capitalization. For example, if a single IPO corresponding to 10% of the industry
market capitalization was delayed 100 days, the number of average days delay impacting
already listed firms in the industry, or DIP Oi,t, would be estimated at 10.
Our proxy for potential competition in asset space from the firms in the queue is based
on the covariance of each individual security with a synthetic portfolio of listed firms whose
industry composition matches that of the delayed IPO queue portfolio. This is calculated as
the weighted sum (across industries) of covariances between the monthly return (of the three
years prior to suspension t) of each listed firm’s stock price, Rf , and the monthly return in
each industry, Ri. This proxy, which we term COVf,t, satisfies
COVf,t =
i(Cov(Rf , Ri) · MCQi,t) , (3)MCQ
t
where the weights are the eventual market capitalisation of all the suspended IPOs in the
relevant industry (MCQi,t)) relative to the total of suspended IPOs at the time of
suspension t, MCQt. The industry index returns are calculated as the averages
of equity returns within an industry, weighted by ratio of company market cap to
industry market cap.
As discussed below, we allow for heterogeneity by firm profitability, under the hy-
pothesis that more profitable firms will be less vulnerable, and hence their equity values
less sensitive to the suspension announcement. We therefore interact our mea- sures of
potential IPO competition from firms in the queue with indicators of firm profitability.
We consider five different gauges of profitability: a) net profit margin, NP M , defined as
income before extraordinary items divided by sales/revenue; b) re- turn on assets, ROA,
defined as income before extraordinary items divided by average of the beginning balance and
ending balance of total assets; c) return on equity, ROE, defined as net income available for
common shareholders divided by the average of the beginning balance and ending balance
of the total common equity, measured in book values; d) return on invested capital, ROI ,
defined as profits divided by sum of debt and equity; and e) Operating profitability,
OROC, defined as operating rev- enues divided by operating Data is obtained from
WIND. For further details on variable construction, see Appendix Table 2.
Finally, in addition to our inclusion of firm fixed effects we also include a number of
conditioning variables to control for individual firm heterogeneity. As with the prof-
itability measures above, data are from and defined in more detail in Appendix Table 2.
We include a measure of market capitalization, MKT CAP , measured in billions of
RMB before the suspension date. It is possible that, holding all else equal, larger firms
would be less vulnerable to the impact of suspended IPOs of a given size in the same
industry. Another motivation for using size is that large firms are likely to have more
information flows available to investors, and thus reduced uncertainty (Ismail et al. (2015)).
We also include a measure of leverage, LEV , measured as the three-year average, taken
from the year prior to the suspension, of the ratio of total assets to total equity in
book values. The higher the leverage, the more burdensome are a firm’s debt payments,
and the greater the vulnerability to creditors unwilling to roll over obligations. A higher
ratio indicates relatively little equity to cover losses in the firm’s value to pay back debt
holders, and should be positively related to vulnerability to the disruption posed by new
IPOs.
7Profits defined in this latter way is independent of interest expenses and thus should be indepen-
dent of the firm’s capital structure.
The price to book ratio, P BOOK, is often viewed as measure of value, and enters
significantly in some asset pricing models. We take the average of the price to book ratio
for the three years before the suspension. Firms with high price-to-book ratios might be
construed as more vulnerable to the disruption caused by IPOs in the same industry.
Firms with more volatile earnings could be more sensitive to a suspension of IPOs. We
measure earnings variability, SDEBIT , as the standard deviation of earnings before
interest and taxes/totas, over the three years before the suspension.
We also include a 0-1 dummy to indicate state ownership, SOE. A large number of
firms in China are state-owned enterprises. These firms have been shown to enjoy
preferential access to borrowing from state-owned banks (. Chang et al. (2019)).
State-owned firms may therefore differ from private firms in creditworthiness and
profitability. However, as preferential treatment to state-owned firms has waned in
the wake of reforms (Liu et al. (2020)), SOEs might also be viewed as more vulner-
able to competition from an IPO in their industry, and thus more likely to benefit
from a suspension. We include a dummy variable, denoted by SOE, that equals 1 if the
company is state owned, meaning the firm’s biggest shareholder (or controlling
shareholder) is the central government or its agencies, or the local government or its
agencies, according to ’s definition. In addition to acting as a control variable on
its own, the state-ownership variable is used to segment samples to test the main
hypotheses separately.
Finally, we include SHANGHAI , a 0 − 1 dummy that takes value 1 if the firm is
listed on the Shanghai exchange, and value 0 if it is instead listed on the Shenzhen ex- change.
As one exchange tends to include larger firms that are more established, while the other
exchange more representing of technology and the new economy, it is con- ceivable that
suspensions could affect firms differentially depending on the exchange on which they are
listed.
In addition to the above explanatory variables, we also include separate dummies for
each suspension to take into account possible particularities of the individual suspension
episodes, including any differences in macroeconomic and overall financial conditions.
. Summary statistics. Summary statistics for the full sample of over 6000 ob- servations
of listed firms, pooled across the three periods covering the suspensions, are shown in
Table 2. The key dependent variable R1—share price return on the
TABLE 2. Summary Statistics
Variable Mean Median Std. Dev. Min Max
R1
IPO 0
DIPO 0
COV
NPM
ROA
ROE
ROI
OROC
MKTCAP
LEV
PBOOK
SDEBIT
SOE 0 0 1
SHANGHAI 0 0 1
Observations 6045
Note: Variables defined as in Appendix Table 2.
Source: .
first trading day after the announcement of the suspension —averages a % de- cline,
and the median is only %. However, the standard deviation of returns of
percentage points, and maximum and minimum values of plus and minus 10%,
respectively, are indicative of considerable cross-sectional variation in share price re- turns on
suspension days across the full listed firm sample. The percentage of market cap that
postponed IPOs represent in any industry, the IP O variable, averages only %, though
this can range as high as %. While the mean (weighted by percent of suspended IPOs
of industry market cap) length of delay of IPOs (DIP O) is days, the standard
deviation is 5 days, and the measure ranges as high as 66 days. The listed firm return
covariance with delayed IPO firm assets (COV ) also shows considerable cross-sectional
variation, with a standard deviation more than twice the mean and a maximum observation
of 40 times the mean.
Market capitalization of the listed firms averages 14 billion RMB, compared to a median
of 4 billion, and ranging as high as 2165 billion. Price-to-book ratios average around 10
TABLE 3. Summary Statistics
Pooled Mean Median Std. Dev. Min Max
R1
IPO 0
DIPO 0
Delay 0 959
COV
Observations 6045
2008 Mean Median Std. Dev. Min Max
R1
IPO 0
DIPO 0
Delay 368 0
COV
Observations 1484
2012 Mean Median Std. Dev. Min Max
R1
IPO 0
DIPO 0
Delay 0 959
COV
Observations 2390
2015 Mean Median Std. Dev. Min Max
R1
IPO 0
DIPO 0
Delay 219 0 398
COV
Observations 2171
Note: R1, IPO, DIPO, COV defined as in Appendix Table 2. Delay is defined as the
average number of days between approval and IPO among all queue firms in the listed firm’s
industry, and thus has the same value across all firms in the same industry.
Source: .
with a median value of 5. The mean of the profitability measures – , , , and
% for NP M , ROA, ROE, ROI and OROC, respectively—are all quite close to the median
observations. The mean and medians of the measure for leverage (Assets over book equity) are
both around 2, with a standard deviation of 1. Roughly % percent of the entire listed firm
sample (which includes double-counting across suspensions) are state- owned, while % are listed
on the Shanghai exchange (as opposed to Shenzhen).
Table 3 reports sample values both pooled and separated by suspension dates, as well as mean
IPO delays. The average single day return was much higher for the first suspension in 2008 (%)
than the last in 2015 (%), while the middle 2012 suspension was in between at (%). The
mean (unweighted) length of delay averaged 217 days in 2008, vs. 446 days in 2012, and 164 days in
2015. DIPO results in much smaller numbers across the board than the unweighted delay length,
given the weighting scheme by the proportion of queue firms to industry market cap.
V. METHODOLOGY
We estimate a conventional panel specification. Our base specifications include an in- dicator
of firm performance on its own, and interacted with the two channels of potential exposure to
competition from IPO activity. Our first is IP Oj,t, which represents the value of IPOs in the queue
in industry j at time t, and our second is COVi,j,t, which is measured as the average covariance of
returns over the previous three years of firm i in industry j at time t with a weighted portfolio of
industries in queue.
Our performance indicators consist of four alternative indicators of firm profitability and one
indicator of firm productivity. Our profitability indicators include net profit margin, NP Mi,j,t,
return on assets, ROAi,j,t, return on equity, ROEi,j,t, and return on investment, ROIi,j,t. Our proxy
for productivity is the ratio of operating revenue to operating costs, OROCi,j,t. For example, our
specification with the NP Mi,j,t performance indicator satisfies
ri,j,t = c + β1NP Mi,j,t + β2IP Oj,t + β3IP Oj,t · NP Mi,j,t + β4COVi,j,t (4)
+β5COVi,j,t · NP Mi,j,t + γXi,j,t + D12 + D15 + si,j,t
where in addition to the variables defined above, ri,j,t represents the one-day return on firm i in
industry j at time t; Xi,j,t is a vector of firm characteristics, including market capitalization
(MKT CAP ), leverage (LEV ), the price-to-book ratio (P BOOK), earnings volatility (SDEBIT
), and a 0-1 indicator that takes value 1 if the firm is listed on the Shanghai stock exchange
and 0 if listed on the Shenzhen (SHANttHAI); D12 and D15 are time dummies indicating
observations from the 2012 or 2015 suspensions respectively; and si,j,t is the residual, assumed to be
well-behaved.
We estimate this specification using ordinary least squares with standard errors clustered by
industry. Our time dummies account for differences in prevailing overall macroeconomic and financial
conditions prevailing on different suspension dates.
There are four primary variable coefficients of interest. The importance of direct competi- tion
effects of IPO activity are measured by the sensitivity of one-day returns to the eventual value of the
IPOs in industry j that were in the queue at the time of the suspension, normal- ized by the market
capitalization of industry j (β2), and the adjustment of that sensitivity for firm performance, here
measured by interacting IP Oj,t with firm net profit margins (β3). Similarly, the importance of the
asset-space competition effects of IPO activity are measured by the sensitivity of one-day returns to
the covariance of the returns to firm i in industry j over the previous three years with the returns
on a synthetic portfolio of firms with the industry mix of the IPO queue (β4), and the estimated
adjustment of that sensitivity for firm performance, here measured by interacting IP Oj,t with firm
net profit margins (β5).
As discussed in the previous section, as a robustness check we also consider an alternative indicator
of potential competitors in the IPO queue, DIP Oj,t. This specification weights eventual firm
IPO values in the queue by the length of time from the event date to their eventual listing. This
alternative measure accounts for the fact that the CSRC appears to respect firms’ places in the
queue. As such, the announcement of a blanket moratorium on IPOs represented a larger expected
delay on firms that had been approved for IPO listing for a longer period prior to the suspension
announcement. To keep the specification simple, we use the realized IPO delay as a proxy for the
expected delay at the time of suspension.
VI. RESULts
. Base speciftcation. Our base specification results are shown in Table 4. Each column interacts
the competition channel proxies IP O and COV with a different firm performance indicator.
Model 1 runs our base specification with the performance variable firm net profit margin (NP M
). It can be seen that firms with higher net profit margins fared better on suspension dates, as NP M
enters positively at statistically significant levels. This is not surprising, as suspensions all
occurred during relatively tumultuous periods where stronger firms were likely outperforming their
weaker counterparts. Given the standard errors in Table 1, our point estimate indicates that a one
standard deviation in NP M results in a 59 basis point increase in returns.
Both the proxy for potential direct IPO competition (IP O) and that variable interacted with
NP M enter at statistically significant levels with their expected positive and negative point
estimates respectively. Our point estimates indicate that these channels are econom- ically
significant as well. A one standard deviation increase in IP O is estimated to raise
TABLE 4. Base specification results
Performance indicator (PI)
(1)
NPM
(2)
ROA
(3)
ROE
(4)
ROI
(5)
OROC
PI
IPO
IPOxPI
COV
***
()
***
()
***
()
**
()
***
()
**
()
***
()
**
()
***
()
**
()
***
()
***
()
***
()
*
()
**
()
***
()
***
()
***
()
***
()
**
()
COVxPI ** *** *** **
MKTCAP
LEV
PBOOK
()
***
()
***
()
**
()
()
***
()
***
()
***
()
()
***
()
**
()
***
()
()
*
()
**
()
***
()
()
*
()
***
()
**
()
SDEBIT **
() () () () ()
SOE
SHANGHAI
()
***
()
()
***
()
()
***
()
()
***
()
()
***
()
Constant *** *** *** ***
() () () () ()
Observations 6,045 6,048 5,984 5,916 5,937
R-squared
Note: Dependent variable is one-day return on equity. Ordinary least squares estimation, with standard errors
clustered by industry in parentheses. Models (1) through (5) alternate performance indicators, as indicated at tops
of column. See text for variable definitions. *** p<, ** p<, * p<.
one-day returns by basis points, while our point estimate on the interactive term in- dicates
that a one standard deviation increase in NP M for firms with average IP O values reduces those
returns by basis points.
Our proxy for potential asset-space IPO competition (COV ) also enters at statistically
significant levels with its expected positive sign. Our coefficient estimate indicates that a
one-standard deviation increase in COV results in a basis point increase in one-day returns.
Our interactive term (COV xNP M ) misses statistical significance at conventional cutoffs, but the
point estimate indicates economically meaningful heterogeneity across firms in their sensitivity to this
channel as well, with a one standard deviation increase in NP M for firms with average COV values
reducing returns by basis points.
The four other models substitute the performance variables, ROA, ROE, ROI, and the
productivity measure, OROC, one at a time for NP M in our base regression. All of the
performance variables enter positively and significantly on their own, again demonstrating that
better-performing firms tended to have higher returns on the suspension dates.
Turning to the competition channel proxies, our qualitative results are largely robust to the use
of the alternative firm performance variables. Both the IP O and COV variables consistently
enter positively at statistically significant levels, although the COV variable only enters at a 10%
confidence level using the ROI performance variable. The interactive terms also continue to enter
negatively, with both interactive terms exhibiting statistical significance at least at a 5%
confidence level.
Overall, our results for our primary hypotheses are quite robust to the use of all of our firm
performance variables and send the same message: The suspensions of IPO activity was taken as
better news for firms with greater exposure through either the direct or asset-space competition
channels. Moreover, the sensitivity to these channels of competition through IPO activity was
measurably related to firm performance, with poorer-performing firms exhibiting more sensitivity.
Turning to the other covariates, their importance and robustness varies. Our MKT CAP proxy
enters positively and significantly throughout, although sometimes only at a 10% con- fidence level.
Our base specification point estimate indicates that a one standard deviation movement in market
capitalization results in a 50 basis point increase in expected return on IPO suspension dates. This
finding would be in keeping with the possibility that the sus- pensions occurred on turbulent dates,
which larger firms with greater market capitalization were more capable of weathering.
Our LEV ERAttE proxy also enters positively and significantly throughout. Our base
specification point estimate indicates that a one standard deviation increase in firm leverage results in
a basis point increase in expected return on IPO suspension dates. This finding may reflect
the expectation that the IPO suspension would be more helpful to more leveraged existing firms who
may face greater equity market needs going forward.
Our price-to-book measure, P BOOK, enters negatively at statistically significant levels
throughout as well. Our base specification point estimate indicates that a one standard deviation
increase in a firm’s price-to-book ratio results in a basis point decrease in expected return
on IPO suspension dates. This result appears in keeping with those for our
performance variables, which suggest that better-performing firms, who are also likely to have
higher price-to-book ratios holding all else equal, did better on the suspension dates.
Not all of the conditioning variables enter significantly. Our proxy for revenue risk,
SDEBIT , is insignificant throughout, with the exception of our specification with the OROC
firm-productivity variable. Our SOE dummy is also insignificant throughout, sug- gesting that after
controlling for difference in firm performance there are no systematic differences in exposure to
competition through IPO activity between SOE and non-SOE firms.
In contrast, our dummy indicating firms that trade on the Shanghai, rather than the
Shenzhen, exchange enter positively and significantly throughout. This may also reflect a
systematic superiority in more established firms, which are more likely to be listed on the Shanghai
exchange. Our base specification indicates that the average return on the Shanghai exchange on
suspension dates was basis points higher than firms listed on the Shenzhen exchange.
Lastly, we don’t report our dummies for the event dates, but our event dummies for the 2012 and
2015 suspensions both enter significantly negative. This contrasts with the constant term, which can
be interpreted as the impact of the 2008 suspension and enters significantly
. Adjusting for expected issuance delay. As the CSRC typically respected the IPO queue when
scheduling listing dates, firms who had been in the queue longer experienced a greater disruption to
their IPO timing than those who had not. To adjust for this potential heterogeneity in the news
content of the suspension announcement, we weight the IP O variable for each firm in the queue
by the realized period of delay. We term this alternative variable DIP O.
Our results for this alternative variable are shown in Table 5. As in our base specification, each
column interacts the competition channel proxies DIP O and COV with a different firm
performance indicator, as explained in the previous section.
Our overall results for our variables of interest are quite similar to those for our base IP O variable,
with similar coefficient estimates for the variables of interest. Model 1 runs our base specification
with firm net profit margin (NP M ). Firms with higher net profit margins again fared better on
suspension dates, as NP M enters positively at statistically significant levels, with an almost-
identical point estimate. Similarly, both our alternative the proxy for potential direct IPO
competition (DIP O) and that variable interacted with NP M also
8The full regression results are available in an online appendix. GIVE ADDRESS In addition, while we
are concerned about potential endogeneity issues, we report the results using our two potential measures
of direct competition for individual suspension dates in Appendix Table A1. The results are quite poor for all
of the individual suspension dates, illustrating the importance of our pursued panel approach.
TABLE 5. IPO queue adjusted for delay
Performance indicator (PI)
(1)
NPM
(2)
ROA
(3)
ROE
(4)
ROI
(5)
OROC
PI
DIPO
DIPOxPI
COV
***
()
***
()
***
()
**
()
***
()
**
()
***
()
**
()
***
()
**
()
**
()
***
()
***
()
*
()
**
()
***
()
***
()
***
()
***
()
**
()
COVxPI ** *** *** **
MKTCAP
LEV
PBOOK
()
***
()
***
()
**
()
()
***
()
***
()
***
()
()
***
()
**
()
***
()
()
*
()
**
()
***
()
()
*
()
***
()
**
()
SDEBIT **
() () () () ()
SOE
SHANGHAI
()
***
()
()
***
()
()
***
()
()
***
()
()
***
()
Constant *** *** *** ***
() () () () ()
Observations 6,045 6,048 5,984 5,916 5,937
R-squared
Note: Dependent variable is one-day return on equity. Ordinary least squares estimation, with standard errors
clustered by industry in parentheses. Models (1) through (5) alternate performance indicators, as indicated at tops
of column. See text for variable definitions. *** p<, ** p<, * p<.
continue to enter at statistically significant levels with their expected respective positive and negative
point estimates.
The performances for our proxy for potential asset-space IPO competition (COV ), as well as
its interactive term (COV xNP M ), are also similar to those in our base specification. Both enter
with their expected signs, at statistically significant levels for the COV variable
on its own, but again just missing 10% statistical significance for that variable interacted with
NP M . As before, however, the two variables both enter with statistically significant coefficient
estimates in all of the other models in Table 5. Overall, our results for our variables of interest are
robust to the use of the alternative DIP O measure of the intensity of potential direct competition in
the queue.
The performance of the other covariates also remain quite similar. The MKT CAP proxy
continues to enters positively and significantly throughout, as does the LEV ERAttE vari- able. The
price-to-book ratio variable (P BOOK) also continues to enter negatively through- out. While the
proxy for revenue risk (SDEBIT ) remains insignificant for four of the five specifications, it enters
positively and significantly at a 5% confidence level for Model 5 with our OROC performance
variable. The SOE dummy remains insignificant throughout, and the Shanghai dummy and event
dummies remain significantly positive and negative respectively as well.
Overall, then, our results are generally robust to the use of our alternative proxy for potential
direct competition in the queue, confirming that the IPO suspensions were viewed as better news for
firms with greater potential competition from either the direct or asset channels in the IPO queue
at the time of the suspension announcement.
VII. ROBUSTNESS CHECKS
This section subjects our results to a battery of robustness tests, concentrating on the NP M
indicator of firm performance. We organize the robustness tests into three tables, and concentrate
our discussion on the performances of our variables of interest. The first table considers changes
in the specification of the base regression. The second investigates the robustness of our results to
a variety of changes to our sample. Lastly, we examine the robustness of our results to perturbations
in our investigation methodology.
. Speciftcation changes. Table 6 displays a variety of alternative changes in our base regression
specification.
Model 1 drops the conditioning variables, only retaining the variables of interest and the time
dummies. Our results for this alternative specification are quite similar. NP M continues to enter
positively on its with an almost identical coefficient estimate. Both the IP O and COV variables
also continue to enter positively and significantly as well, while the interactive terms also continues
to enter negatively with the IP OxNP M variable again entering with statistical significance and the
COV xNP M variable just missing.
Model 2 drops our interactive terms. NP M and COV continue to enter positively and
significantly on their own, with modestly lower coefficient point estimates. However, the IP O
variable now enters significantly with the opposite sign. This sensitivity illustrates the importance of
allowing for differences across firms in sensitivity to the composition of the queue on suspension
dates by firm performance.
TABLE 6. Changes in specification
(1) (2) (3) (4) (5) (6)
NPM
IPO
***
()
***
***
()
**
***
***
()
***
***
()
***
***
()
**
IPOxNPM
COV
()
***
()
**
()
*
()
()
***
()
()
***
()
**
()
***
()
**
() () () () ()
COVxNPM **
MKTCAP
()
***
()
*** ***
()
***
()
***
LEV
()
***
()
()
*
()
()
***
()
()
***
()
()
***
()
PBOOK *** **
() () () () ()
SDEBIT
() () () () ()
SOE ***
SHANGHAI
()
***
()
()
***
()
()
***
()
()
***
()
()
***
()
Constant *** *** *** ***
() () () () () ()
Observations 6,058 6,045 6,045 6,106 6,060 6,045
R-squared
Note: Dependent variable is one-day return on equity, except Model 5, whose dependent variable is the 1-
day excess return on equity, and Model 6, whose dependent variable is the 2-day return on equity. Ordinary
least squares estimation, with standard errors clustered by industry in parentheses. See text for variable
definitions and details on sample perturbations. *** p<, ** p<, * p<.
We next drop the IP O and its interactive term IP OxNP M (Model 3). Our NP M
variable continues to enter positively and significantly with a very similar coefficent estimate. Both the
positive coefficient estimates on our COV variable, and the negative point estimate on that variable
interacted with NP M are also robust to dropping the IP O variable, as both enter with coefficient
estimates that are modestly larger in absolute value.
Model 4 drops the COV and COV xNP M variables. Our NP M variable enters positively and
significantly with a very similar coefficient estimate. Both the positive coefficient esti- mates on
our IP O variable, and the negative point estimate on that variable interacted with NP M are also
robust to dropping the COV variable, as both again enter with coefficient estimates that are
statistically significant and modestly larger in absolute value.
Model 5 retains all of the explanatory variables of our base regression, but examines 1 day
excess, rather than raw returns as the dependent variable. Our results for the variables of interest
continue to enter at statistically significant levels with their expected signs, with the exception of
the COV xNP M variable, which is negative as expected, but statistically insignificant.
Lastly, Model 6 also retains all of the variables in our base repression, but extends our
dependent variable to a longer 2-day event window. Our results for the variables of interest continue
to enter at statistically significant levels with their expected signs, again with modestly larger
coefficient point estimates in absolute value.
Overall, our base regression results are quite robust to modest perturbations in our spec- ification.
The lone exception is that of Model 2, where the IP O variable actually entered with the
incorrect negative sign when the interactive terms were dropped. As discussed above, this
sensitivity illustrates the importance of accounting for heterogeneity across firms by performance in
assessing sensitivity to IPO activity among listed firms.
. Sample changes. We next investigate the robustness of our results to changes in our
sample. Because we investigate omitting outliers for a wide variety of reasons, we only report our
coefficient estimates for our variables of Our results are shown in Table 7.
Model 1 reduces our sample to a sub-sample that only includes SOE firms, resulting in 2,890
observations. NP M continues to enter significantly positive, with a modestly larger coefficient
estimate than our base specification. Moreover, the two measures of potential channels for IPO
competitive effects both enter significantly with their expected positive signs. Similarly the
interactive terms remain negative and statistically significant.
Model 2 displays the results for the Non-SOE sub-sample (3,155 observations). The NP M
performance variable continues to enter positively and significantly, as does the IP O variable and the
term interacting these (IP OxNP M ). However, the COV variable and its interactive term are both
insignificant. We therefore conclude that both SOE and Non-SOE firms face exposure to IPO
activity through direct competition, of roughly the same magnitude, but while SOE firms appear to
also face potential competition through the asset-space channel, we find no evidence that this
channel is at work for our Non-SOE sub-sample.
9Full results are available from the authors upon request at GIVE ADDRESS.
TABLE 7. Changes in sample
(1) (2) (3) (4) (5) (6)
IPO IPOxNPM NPM COV COVxNPM Constant
(1) SOE sample *** *** *** * ** ***
() () () () () ()
(2) Non-SOE sample * *** *** ***
() () () () () ()
(3) Shanghai listed *** *** *** ***
() () () () () ()
(4) Shenzhen listed *** *** *** ** * ***
() () () () () ()
(5) Balanced panel ** *** *** ***
() () () () () ()
(6) Drop profitable *** *** *** ** ***
() () () () () ()
(7) Drop unprofitable *** *** *** ** ***
() () () () () ()
(8) Drop productive * ** *** ** ***
() () () () ()
(9) Drop unproductive *** *** *** ** ***
() () () () () ()
(10) Drop big *** *** *** ** ***
() () () () () ()
(11) Drop small *** *** *** ** ***
() () () () () ()
(12) Drop high IPO *** *** *** *** ** ***
() () () () () ()
(13) Drop large ImpactM *** *** *** ** ***
() () () () () ()
Note: Dependent variable is one-day return on equity. Ordinary least squares estimation, with standard errors clustered by
industry in parentheses. See text for variable definitions. *** p<, ** p<, * p<.
Models 3 and 4 divide our sample into the sub-samples of firms listed on the Shanghai stock
exchange and those listed on the Shenzhen stock exchange. The two sample have 2,617 and 3,428
observations respectively. While the SHANttHAI variable itself was positive and significant in
our base specification, indicating that variables listed on the Shanghai exchange enjoyed modestly
superior returns on suspension dates, our results for the sub- samples divided by exchange listings
are quite similar.
For both sub-samples, the NP M performance variable continues to enter positively and
significantly, as does the IP O variable and the term interacting these (IP OxNP M ). How- ever, we
do observe differences in the COV variable and its interactive term. For the obser- vations from
firms listed on the Shenzhen exchange, our results for the COV variable and its interactive term are
the same as those for our base specification, with the former entering positively and significantly,
while the interctive term is significantly negative. In contrast, both of the variables enter
insignificantly with the wrong sign for observations from firms listed on the Shanghai exchange. We
conclude that the relative lack of robustness we find for our asset-space competition channel is
attributable to firms listed on the Shanghai exchange, although it is not clear why this would be the
case.
Model 5 reduces our sample to a balanced panel of listed firms with observations from each of
the suspension dates. This leaves us with the majority of observations (5,021) in our base panel,
but this specification is obviously exposed to some extent to both potential survivorship bias and
potential differences associated with firm age. For that reason, we only examine this specification as a
robustness check.
The performance variable NP M continues to enter significantly positive, with a modestly larger
coefficient estimate. However, the two measures of the competition channels are both insignificant.
The IP O variable does better, continuing to enter positively with a smaller point estimate at close
to a 10% confidence level, while the interactive term remains negative and statistically significant.
However, the COV variable is very insignificant and enters with the wrong sign, although its
interactive term also remains significantly negative. Overall, our results do not appear to be robust
to estimation under a truncated balanced panel.
The remainder of specifications in Table 7 examine the robustness of our results to drop- ping a
variety of outliers from our sample, with outliers defined as realizations more than three standard
deviations from our sample mean. Models 6 and 7 drop extremely profitable and unprofitable firms,
defined by the NP M measure, respectively. Models 8 and 9 drop extremely productive and
unproductive firms, defined by the OROC measure, respectively. Models 10 and 11 drop extremely
large and small firms, defined by extreme values of firm asset holdings. Model 12 drops industries
that had extremely large IPOs from our queue sample. Lastly, Model 13 drops firms from industries
most closely correlated with the sample queue, identified as those with very high values of IP O.
Our results are generally robust to the omission of all of these outliers. The NP M variable enters
significantly with its expected sign throughout, as does the IP O variable and its interactive
term, as well as the COV variable. However, the interactive term COV xNP M is often
insignificant, with the exception of Model 12, which drops the high IPO industries. Overall, our
results are quite robust to our sample perturbations, with the exception of the interactive COV
xNP M term, which continues to display some fragility. However, even
here the variable consistently enters with its predicted negative sign and is usually close to the 10%
confidence level.
. Alternative estimation methodologies. Lastly, we consider the robustness of our base
specification results to a variety of estimation methodologies. Our results are displayed in Table 8.
Models 1 and 2 re-estimate the base specification with White’s heteroscedasticity robust and
conventional standard errors respectively (Table 8’s model 2 is thus the same specifi- cation as
reported in Table 4, model 1). It can be seen that all of our variables of interest continue to enter
at statistically significant levels in model 1, except for the interactive COV xNP M variable,
which fails to enter significantly under robust standard error estima- tion, despite entering
significantly with conventional standard errors.
Model 3 runs our base specification using weighted least squares, with firm size, measured by firm
asset holdings. Our performance variable continues to enter positively with statisti- cal significance
and a coefficient estimate similar to that in our base specification. Our IP O variable is also robust
to estimation under weighted least squares, as is that variable’s inter- active term, IP OxNP M .
Indeed, both coefficient point estimates are larger than we obtain in our base specification. However,
our COV variable and its interactive term COV xNP M are insignificant, suggesting that the
significance of the asset substitution channel in our earlier results may have been driven by the
smaller firms in our sample.
Model 4 winsorizes variables at a 1% level, rather than the 5% level in our base specifica- tion,
while Model 5 trims instead of winsorizing them at the 5% level. For both methods, our performance
variable continues to enter positively with statistical significance, as does our IP O variable, as well
as that variable interacted with our performance variable, IP OxNP M . Our COV also continues to
enter significantly positive, with comparable coefficient point estimates, but that variable interacted
with our performance variable just misses 10% signif- icance under 1% winsorizing, and enters with
only 10% statistical significance when we trim rather than winsorize.
Overall, our results for our variables of interest are robust to the perturbations in es- timation
methods, particularly for our measure of potential direct competition, the IP O variable and that
variable interacted with our performance variable. Our COV proxy for the asset-space competition
channel is also quite robust. However, we again observe sensitiv- ity and less robustness for that
variable interacted with our performance variables, although it consistently continues to enter with its
expected negative sign.
VIII. CONCLUSION
This paper uses data from Chinese IPO suspensions to evaluate the efficacy of two pro- posed
channels in the literature for competition from IPO activity to adversely affect listed firms. The
Chinese suspensions, which eliminated all IPO activity for uncertain periods of
TABLE 8. Changes in estimation method
(1)
Robust SE
(2)
Regular SE
(3)
Weighted LS
(4)
1% Winsor
(5)
5% Trim
NPM
IPO
IPOxNPM
COV
**
*
()
**
*
()
-
***
()
*
()
**
*
()
**
()
-
***
()
***
()
*
*
()
**
*
()
-
**
*
()
**
()
**
*
()
**
*
()
-
**
*
()
**
()
**
*
()
**
*
()
-
**
*
()
**
()
COVxNPM * * *
() () () () ()
MKTCAP *** *** ***
() () () () ()
LEV *** *** ** ***
() () () () ()
PBOOK *** ** ***
() () () () ()
SDEBIT
() () () () ()
SOE
() () () () ()
SHANGHAI *** *** *** ***
Constant
()
**
*
()
()
**
*
()
()
**
*
()
()
**
*
()
()
**
*
()
Observations 6,045 6,045 3,803 6,045 4,899
R-squared
Note: Dependent variable is one-day return on equity. Ordinary least squares estimation, with standard errors
clustered by industry in parentheses. See text for variable definitions. *** p<, ** p<, * p<.
time, provide an opportunity to evaluate the proposed competitive effects of IPOs without
contamination from the endogeneity of individual firm listing decisions. Because we have
multiple suspensions, we pursue a panel approach, which allows us to condition for disparities in
aggregate conditions at the time of the suspension announcements.
We evaluate the first channel, which we term “direct competition," through the share of firms in
the IPO queue in their industry at the time of the suspension, weighted by the size of their eventual
IPOs. The second channel is the “asset space" channel, where firms that enjoyed additional
demand due to desirable risk characteristics might find IPOs providing new competition for this
asset attribute. We measure this second channel by the covariance of a listed firm with the returns on
a synthetic portfolio of listed firms with the same industry mix as those firms in the IPO queue at the
time of the suspension announcement. Our results provide evidence of anticipated competition from
new IPO firms through both channels. These results are robust to a wide variety of sensitivity
tests.
We also examine the possibility of heterogeneity in exposure to new competition by firm
performance. We evaluate this heterogeneity through a term which interacts our proxies for
competition with a variety of performance measures. Our results demonstrate a meaningful and
robust degree of heterogeneity of firm exposure through the direct competition channel, with better-
performing firms exhibiting less sensitivity to the suspension announcements through this channel
than their weaker counterparts. We also find evidence of heterogeneity through the asset-space
channel, although measured heterogeneity by firm performance is weaker and less robust to
sensitivity tests. As this term is particularly weak under weighted least squares by firm size and for
the subset of firms listed on the Shanghai exchange, which tend to be larger than those on the
Shenzhen exchange, it seems likely that the heterogeneity in exposure to asset-space competition
through IPOs is more pronounced among smaller listed firms.
IX. APPENDIX
TABLE A1. Individual event dates
(1) (2) (3) (4) (5) (6)
1st event 1st event 2nd event 2nd event 3rd event 3rd event
NPM *** ***
() () () () () ()
IPO **
() () ()
IPOxNPM ***
DIPO
()
**
()
()
() () ()
DIPOxNPM ***
() () ()
COV **
() () () () () ()
COVxNPM
() () () () () ()
MKTCAP *** *** *** ***
() () () () () ()
LEV *** ***
() () () () () ()
PBOOK *** ***
() () () () () ()
SDEBIT ** ** ** **
() () () () () ()
SOE * *
() () () () () ()
SHANGHAI *** ***
() () () () () ()
Constant *** *** ***
() () () () () ()
Observations 1,484 1,484 2,390 2,390 2,171 2,171
R-squared
Note: Robust standard errors in parentheses
*** p<, ** p<, * p<
Table A2: Variable Definitions
Variable Deftnition Formula Unit Sources
Dependent variable
r1 One-day return
∆st × 100st−1 % WIND
r2 Two-day return
st+1−st−1 × 100st−1 % WIND
Independent variable
IPO Market capitalization of all % calculated
queue firms in industry i,
MCQi,t based on IPOs after
suspension at time t, divided by
market capitalization of all
listed firms in industry at time
of suspension t, MLCi,t .
DIPO Weighted sum of “delays of IPO Days Calculated
process" in each industry;
where the days of delay for
each delayed IPO firm f in
industry i (Delayf,i) are
multiplied by size of the
delayed IPO (MCQf,i), and then
summed across firms in queue
at time of suspension t. The
resulting sum is divided by total
market cap of the industry, MCLi,t,
at suspension t. Delay in IPO process
= IPO date - approval date by CSRC
COV Weighted sum of covariances Calculated
between monthly firm and
industry returns Rf and Ri,
estimated three years prior to
suspension at time t, where the
sum is over the industries and
weights are the market cap
of all suspended IPOs in the
industry i, MCQi,t to all the
suspended IPOs during
suspension t, MCQt.
Industry index return =
weighted average of equity
returns within an industry;
weighted by ratio of company
market cap to industry market cap
NPM1,2 Net profit margin
MCQi,t
MCLi,t
Σ
f (Delayf,i,t×MCQf,i,t)
MCLi,t
Σ (Cov(Rf ,Ri)t×MCQi,t)
i MCQt
where Rf is the firm monthly
return and Ri is the industry
monthly return index
net_incomei
revenuei
% WIND
Continued on next page
Continued from previous page
Variable Deftnition Formula Unit Sources
ROA1,2 Return on assets, annualized % WIND
ROE1,2 Return on equity, annualized % WIND
ROI1,2 Return on invested capital % WIND
OROC1 Total operating revenue / % WIND
total operating cost
2×net_incomei
T Ai,bob+T Ai,eob
2×net_inc_shareholderi
CEi,bob+CEi,eob
2×net_incomei
Inv_capi,bob+Inv_capi,eob
Operating_revenuei
Operating_costi
Control variable
SHANGHAI A dummy equal to 1 if the firm WIND
is listed in Shanghai and 0 for
Shenzhen
SOE A dummy equal to 1 if the firm WIND
is a state-owned enterprise and
0 for others.
MKTCAP Market capitalization of firm, RMBbn WIND
before suspension
LEV1 Leverage, average three years Ratio WIND
before the suspension
PBOOK1 Price to book ratio, average Ratio WIND
three years before the
suspension
SDEBIT Standard deviation of earnings WIND
before interest and taxes/total
assets over three years before
the suspension
D12 When it is 2012 suspension, CSRC
equal to 1
D15 When it is 2015 suspension, CSRC
equal to 1
1 Σ2 total_assetst−i
3 i=0 total_equityt−i
1 Σ2 price_valuet−i
3 i=0 book_valuet−i
,
var( EBITt )
total_assetst
1Winsorized 5% at each end. 2When these variables are used in the regressions, a constant term is added
to each varaible so that there are no negative values.
REFERENCES
Akhigbe, A., S. Borde, and A. M. Whyte (2003). Does an industry effect exist for
initial public offerings? The Financial Review 38 (4), 531–551.
Baker, M. and J. Wurgler (2000). The equity share in new issues and aggregate stock
returns. Journal of Finance 55 (5), 2219–2257.
Baschieri, G., A. Carosi, and S. Mengoli (2015). Local IPOs, local delistings, and the firm
location premium. Journal of Banking and Finance 53, 67–83.
Boeh, K. and C. Dunbar (2014). IPO waves and the issuance process. Journal of
Corporate Finance 25, 455–475.
Bradley, D. and X. Yuan (2013). Information spillovers around seasoned equity offer- ings.
Journal of Corporate Finance 21, 106–118.
Braun, M. and B. Larrain (2009). Do IPOs Affect the Prices of Other Stocks? Evi-
dence from Emerging Markets. Review of Financial Studies 22 (4), 1505–1544.
Chang, C., Z. Liu, M. M. Spiegel, and J. Zhang (2019). Reserve requirements and op- timal
Chinese stabilization policy. Journal of Monetary Economics 103 (1), 33–51.
Chemmanur, T. and J. He (2011). IPO waves, product market competition, and the going
public decision: Theory and evidence. Journal of Financial Economics 101,
382–412.
Chod, J. and E. Lyandres (2011). Strategic IPOs and product market competition.
Journal of Financial Economics 100, 45–67.
Cong, L. W. and S. Howell (2019, December). Policy Uncertainty and Innovation:
Evidence from IPO Interventions in China. SSRN-ID 2911221.
Henderson, B., N. Jegadeesh, and M. Weisbach (2006). World markets for raising new capital.
Journal of Financial Economics 82 (1), 63–101.
Hong, H., J. Kubik, and J. Stein (2008). The only game in town: stock-price conse-
quences of local bias. Journal of Financial Economics 90 (1), 20–37.
Hsu, H.-C., A. Reed, and J. Rocholl (2010). The new game in town: competitive effects
of IPOs. Journal of Finance 65 (2), 495–528.
Ismail, A., S. Oh, and N. Arsya (2015). Split ratings and debt signaling in bond
markets: A note. Review of Financial Economics 24, 36–41.
Li, Y., S. Qian, and T. Shu (2018). The impact of IPO approval on the price of
existing stocks: Evidence from China. Journal of Corporate Finance 50, 109–127.
Liu, Z., M. M. Spiegel, and J. Zhang (2020). Optimal capital account liberalization
in China. Journal of Monetary Economics (forthcoming).
Loughran, T. and J. Ritter (1997). The operating performance of firms conducting
seasoned equity offerings. Journal of Finance 52 (5), 1823–1850.
McGuilvery, A., R. Faff, and S. Pathan (2012). Competitive valuation effects of
Australian IPOs. International Review of Financial Analysis 24, 74–83.
Nguyen, T., N. Sutton, and D. Pham (2014). Intra-Industry Effects of IPOs on Stock
Repurchase Decisions of Rival Firms. Journal of Accounting and Finance 14 (4), 61–
82.
Packer, F. and M. Spiegel (2016). China’s IPO activity and equity market volatility.
FRBSF Economic Letter (2016-18).
Pagano, M., F. Panetta, and L. Zingales (1998). Why do companies go public? an
empirical analysis. Journal of Finance 53 (1), 27–64.
Ritter, J. (1991). The long-run performance of initial public offerings. Journal of
Finance 46 (1), 3–27.
Shi, S., Q. Sun, and X. Zhang (2018). Do IPOs Affect Market Price? Evidence from
China. Journal of Financial and Quantitative Analysis 53 (3), 1391–1416.
Spiegel, M. and H. Tookes (2020). Why Does an IPO Affect Rival Firms? The Review
of Financial Studies 33, 3205–3249.
Tian, L. (2011). Regulatory underpricing: Determinants of Chinese extreme IPO
returns. Journal of Empirical Finance 18 (1), 78–90.
