Real Activities Management during Initial Public Offerings:
Evidence from R&D Expenditures
Tatiana Fedyk
University of San Francisco
[email protected]
Natalya Khimich
Drexel University
[email protected]
March 2015
Abstract
In this paper we demonstrate that an aggregate real earnings management measure used in prior
literature leads to conflicting results in the initial public offering (IPO) setting. To present
evidence of real activities management during IPOs, we study a specific component of real
activities management – R&D expenditures. We find both types of R&D management – R&D
underinvestment and overinvestment – with R&D overinvestment predominating. We show that
management decision varies systematically with firms’ cross sectional characteristics: 1) firms
that are at the growth stage, non-profitable, or belong to science driven industries are more likely
to overinvest in R&D; 2) firms with reduced accounting flexibility and firms that would report
losses in the absence of downward R&D management are more likely to underinvest in R&D.
We further demonstrate that both under- and overinvesting firms exhibit future
underperformance. Our findings that overinvesting in R&D firms experience higher abnormal
trading volume around lockup expiration date point to managerial opportunism as a dominating
motive for overinvestment.
The authors are grateful for comments from Patricia Dechow, Richard Sloan, Alastair Lawrence, Michal Michael,
Barbara Grein, Zvi Singer, Theodore Sougiannis, and Mark Vargus as well as the workshop participants at the
University of California at Berkeley, Arizona State University, Drexel University, and the conference participants at
the 2012 AAA annual meeting. We would like to thank Elizabeth Demers, Alexander Ljungqvist, and Jay Ritter for
sharing their data with us.
1.
Introduction
In this paper, we examine firms’ research and development (R&D) expenditures during
initial public offerings (IPOs). We demonstrate that aggregate real earnings management
measure used in prior literature (Roychowdhury, 2006; Gunny, 2010; Cohen and Zarowin, 2010;
Zang, 2012, Wongsunwai, 2013) leads to conflicting results in IPO setting. A detailed
examination of R&D expenditures, a large and increasingly important component of
discretionary expenses (Lev, 2001; Curtis et. al., 2015), allows us to provide novel evidence on
real activities management during IPOs. We demonstrate the presence of two distinct practices
regarding R&D spending – R&D underinvestment and R&D overinvestment (with
predominating R&D overinvestment) – and show that investment choice varies systematically
with firms’ cross sectional characteristics. We also show negative consequences of R&D underand overinvestment on IPO firms’ long-term future performance.
Extant accounting literature contains numerous studies of accrual earnings management
during IPOs, but lacks evidence of real activities manipulation around initial public offerings.1
In this paper we fill the gap and demonstrate that IPO firms engage in real activities
management, specifically in R&D management. An IPO provides both a motivation and an
opportunity for earnings management: the need to raise external financing, which is cited as an
important motivation for earnings management (Dechow et al., 1996; Healy and Whalen, 1999;
Dechow et al., 2011), is combined with high information asymmetry between insiders and
potential investors (Fan, 2007).
However, increased monitoring and enhanced regulatory
scrutiny around the IPO should substantially limit, if not prevent, firms from inflating earnings
via accrual earnings management (Ball and Shivakumar, 2008).
Real earnings management
(REM) is arguably an attractive alternative to the accrual earnings management as it cannot be
1
A non-exhaustive list of IPO studies includes Aharony et al. (1993), Friedlan (1994), Teoh et al. (1998a), Teoh et
al, (1998b), DuCharme et al. (2001), Darrough and Rangan (2005), Morsfield and Tan (2006), Fan (2007), Ball and
Shivakumar (2008), Armstrong et al. (2013), Cecchini et al. (2012).
1
easily detected by auditors and regulators (Graham et al., 2005; Cohen et al., 2008). However,
empirical evidence of real earnings management during IPOs is scarce. Wongsunwai (2013), in
his study of the effect of venture capitalist quality on earnings management during the IPO,
documents a negative mean real earnings management around the IPO lockup periods,
suggesting that IPO firms on average are less likely to engage in REM than more mature firms.
We attribute the lack of REM findings in the IPO setting to the aggregate measure of
REM used in most studies (Roychowdhury, 2006; Gunny, 2010; Cohen and Zarowin, 2010;
Zang, 2012, Wongsunwai, 2013). The aggregate REM measure includes three components:
abnormal level of cash flow from operations, abnormal production costs and abnormal
discretionary expenses (i.e., advertising, R&D, and SG&A expenses).
We argue that the
aggregate proxy of REM, originally developed by Roychowdhurry (2006) to capture REM
around zero earnings threshold, is not suitable in the IPO setting due to multiple, and sometimes
even opposite incentives related to different REM components.
Discretionary expense component is a special component of REM as it is a subject of two
opposing forces. On one hand, firm can manage discretionary expense downwards in order to
increase earnings. For example, some IPO firms – e.g., firms that face a danger of reporting
losses in the absence of discretionary spending reduction, or exhausted their accounting
flexibility to manage earnings through accruals – have strong incentives to manage earnings
upward through real activities management (Roychowdhury, 2006; Gunny, 2010; Cohen and
Zarowin, 2010). On the other hand, for some firms – e.g., high growth firms at early stages of
their lifecycle, firms that experience losses or operate in science driven industries – earnings are
not that informative in their valuation (Hayn, 1995; Bartov et al., 2002b, Callen et al., 2008).
Furthermore, such firms might have strong incentives to increase discretionary expenses if those
expenses are positively perceived by investors. In order to answer a question which firms – with
2
expense cutting or expense increasing strategy – dominate during the IPO, we study a special
component of discretionary expenses, R&D expenditures.
In this paper, we concentrate on R&D expense, a disaggregate component of
discretionary expense, for a number of reasons.
First and foremost, a link between R&D
expenditures and firm value is well established in the literature: R&D is an important accounting
item that is highly valued by investors (Lev and Sougiannis, 1996; Guo et al., 2005; Amir and
Lev, 1996; Lev et al., 2005; Andre et al., 2007; Joos and Zhdanov, 2008). Therefore, R&D
expenditures can be managed upward to have a positive impact on firm’s valuation or downward
to report higher earnings (Dechow and Sloan, 1991; Baber et al, 1991; Bushee, 1998; Darrough
and Rangan, 2005). Second, R&D is essential for modern IPO firms as it became an increasingly
important element of US and global economies over the past 30 years (Lev,2001 and Curtis et
al., 2015). Third, a focus on one component of discretionary expense allows us to construct a
refined expectation model of R&D expense and conduct an in depth analysis of R&D
management practices during the IPO.
We start our analysis with an examination of the real earnings management during IPOs
by using an aggregate real earnings management measure as in Roychowdhury (2006), Cohen
and Zarowin (2010), and Zang (2012). Our finding that the aggregate REM measure is
insignificantly different from zero suggests that IPO firms do not engage in real earnings
management. However, an analysis of the real earnings management components unveils a more
complex story. A significantly negative abnormal level of cash flow from operations and a
significantly positive abnormal production costs suggest that IPO firms engage in real earnings
management via offering price discounts to temporarily increase sales and overproduction to
report lower cost of goods sold (Roychowdhury, 2006). On the other hand, all discretionary
expenses (R&D, advertising and SG&A expense) are significantly positive, which is inconsistent
3
with real earnings management, suggesting that on average, IPO firms increase discretionary
expenses above the expected level despite the negative impact on earnings. Therefore, we find no
evidence of REM when using aggregate REM measure, and mixed REM results when using
different REM components.
In our main analysis, we focus on R&D component of discretionary expenses. We follow
prior studies (Berger, 1993; Gunny 2010; Fedyk et al., 2014) and augment the basic model of
Roychowdhury (2006) with lagged R&D expenditures, average level of cash holdings, change in
sales and Tobin’s Q. 2 In addition, we perform a performance matching suggested by Cohen et
al. (2014). 3 Specifically, we match all IPO firms with non-IPO firms on industry, year,
performance and life cycle stage. We find that discretionary R&D expenses are still on average
positive, but smaller in magnitude than discretionary R&D from the Roychowdhury (2006)
model, during the IPO year and one year after the IPO, and become insignificantly different from
zero thereafter.
In the next step, we explore the heterogeneity of our sample. We show that decision to
underinvest or overinvest in R&D varies systematically with firms’ ex-ante cross sectional
characteristics. We predict and find that probability of R&D underinvestment (i.e., earnings
management) increases for firms that would report losses in the absence of R&D reduction, and
for firms with reduced accruals management flexibility, captured by net operating assets (NOA).
The probability of R&D overinvestment is higher for growth firms, firms that experience
substantial losses, firms with zero sales, and for firms in the science-driven industries.
We examine the consequences of R&D under- and overinvestment to future performance
of IPO firms.
Both, R&D under- and overinvestment are suboptimal investment choices.
Therefore, a firm most likely will exhibit an inferior post-IPO performance if management
2
Tobin (1969).
For robustness analysis, we also repeat all our tests using changes in R&D expenditures scaled by average total
assets instead of discretionary R&D. All results hold.
3
4
deviates from the optimal level of investment for opportunistic or myopic reasons. Alternatively,
management can overinvest in R&D to signal a firm’s quality and future prospects to investors.
We do not find evidence in support of the signaling hypothesis. R&D overinvestment is
negatively associated with long-term (up to five years after the IPO) future industry adjusted
long-term ROA, future number of patents and patent quality, and positively associated with the
frequency of delisting due to poor performance. As for the R&D underinvestment, we find a
negative association of discretionary R&D with future long-term operating ROA, but not with
the frequency of delisting due to poor performance, and number of patents and patent quality.
In order to further distinguish between opportunistic and myopic reasons, we perform
trading volume analysis around lockup expiration. 4 We find that trading volume for the 3-day
period around the unlock day is significantly higher for firms that overinvest in R&D than for
firms that invest at the expected level. This finding speaks in favor of opportunistic motive for
overinvestment suggesting that management intentionally overinvest in R&D to increase the
firm’s market value and then cashes out its stock holding after the unlock date. We do not find
that trading volume of underinvesting firms is significantly different from that of the firms
investing at the expected level.
We make the following contributions to the literature. First, we contribute to the real
activities management literature by presenting evidence that an aggregate real earnings
management measure leads to conflicting findings in the IPO setting. Using disaggregated
measures, we find evidence of real earnings management through sales management and
overproduction. Second, by studying a special component of discretionary expenses, i.e., R&D
expenditures, we demonstrate two distinct strategies with regards to R&D spending during the
4
Lockups are voluntary agreements between the IPO firm and its underwriter, which prevent insiders who are
holding a company's stock from selling the stock for a period usually lasting 90 to 180 days after the company goes
public. Insiders include company founders, owners, managers, employees and venture capitalists. The lockup
expiration date, i.e., the unlock day, is specified in the IPO prospectus.
5
IPOs: real earnings management through underinvestment in R&D, and R&D management
through overinvestment. Third, we present novel evidence on the decline in the post-IPO
performance for up to five years after the IPO for firms overinvesting in R&D. While prior
literature documents negative effect of R&D underinvestment (Cohen and Zarowin, 2010), we
are the first to document long-term negative effects of R&D overinvestment on future operating
and innovation performance.
The remainder of this paper is organized as follows. Section 2 reviews the literature and
states our hypotheses. Section 3 discusses sample construction and presents empirical evidence
on real earnings management during the IPO. Section 4 presents empirical findings on underand overinvestment in R&D and discusses their cross-sectional determinants. Section 5 analyzes
post-IPO long-term economic performance associated with under- or overinvestment in R&D
and examines abnormal trading volume around the unlock day. Section 6 concludes the paper.
2.
Related literature and hypotheses development
Prior literature cites the desire to attract external financing at low cost (Dechow et al.
1996; Healey and Wahlen, 1999; Dechow et al. 2011) as an important motivation for earnings
manipulation. Beginning with Friedlan (1994), Teoh et al. (1998a), and Teoh et al. (1998b),
researcher have been concerned with whether firms manage reported earnings during IPOs,
whether there are market consequences to such activities, and whether investors and auditors can
recognize earnings management. Despite fairly voluminous literature, evidence on accrual based
earnings management during IPOs is still mixed. While some papers find evidence of earnings
management (Friedlan, 1994; Teoh et al., 1998a; Teoh et al., 1998b; Morsfield and Tan, 2006;
Fan, 2007; and Lee and Masulis, 2011), others do not support earnings management story
(DuCharme et al., 2004; Ball and Shivakumar, 2008; Armstrong et al., 2015). For example, Ball
6
and Shivakumar (2008) argue that IPO firms report more conservatively as they must meet
higher quality financial reporting standards and are subjects to extra scrutiny form auditors,
regulators, analysts, and the media.
In contrast to accrual-based earnings management, real earnings management involves
manipulation of real activities and is considered less likely to be scrutinized by auditors and
regulators than accrual-based earnings management. 5 Schipper (1989) defines real earnings
management as being “accomplished by timing investment or financing decisions to alter
reported earnings or some subset of it." A survey paper by Graham et al. (2005), which conducts
interviews with CFOs about earnings management decisions, supports the idea that managers
prefer real earnings management since auditors “cannot readily challenge real economic actions
to meet earnings targets that are taken in ordinary course of actions". Consistent with the idea
of shifting from accrual-based earnings management to real earnings management in the periods
of higher scrutiny, Cohen et al. (2008) provide evidence that the level of accrual-based
management declined, but the level of real earnings management increased significantly after the
passage of Sarbaney-Oxley (SOX) period. Therefore, given that the IPOs are associated with
high level of attention from auditors, regulators, and investment bankers (Ball and Shivakumar,
2008), it would be reasonable to presume that IPO firms engage in real activities manipulation.
However, research on real earnings management around IPOs is limited, and evidence is
scarce. Wongsunwai (2013) studies the effect of venture capitalist (VC) quality on earnings
management during initial public offerings using an aggregate REM measure as in Cohen and
Zarowin (2010). He reports a negative mean of the aggregate REM, and interprets his findings
5
Examples of real earnings management include: overproduction to report lower costs of goods sold, special
discount or incentive programs to increase sales when revenue targets are not met, myopic operation and investment
decisions, like cutting R&D expenditures, or postponing desirable investments.
7
that IPO companies on average are less likely to engage in real earnings management than more
mature companies in the same industry.
Our first objective is to demonstrate that IPO firms are involved in real activities
management. We posit that the use of aggregate measure of REM (Roychowdhury, 2006; Cohen
and Zarowin, 2010, Wongsunwai, 2013), which consists of three components: abnormal level of
cash flow from operations (resulting from additional price discounts or more lenient credit terms
in order to accelerate sales), abnormal production costs (to report lower cost of goods sold
through increased production), and abnormal discretionary expenses including advertising, R&D,
and SG&A expenses (to report higher earnings), is problematic in the IPO setting as different
components are driven by different and sometimes opposite incentives. The aggregate proxy
serves as an appropriate measure of REM when a firm faces a clear task of meeting or beating
certain earnings benchmarks like positive earnings or analysts’ expectations, in order to avoid
severe negative market reaction. However, an IPO firm presumably faces a different objective –
to achieve the highest possible market valuation from the investors. A link between earnings and
firms’ value is not uniform among IPO firms as prior literature demonstrates that earnings are
not always informative for the firm’s valuation. Hayn (1995) shows that negative earnings are
less informative than positive earnings. Bartov et al. (2002b) find that during the IPO, for
Internet firms earnings are not priced, while negative cash flows are priced perhaps because they
are viewed as investments. Callen et al. (2008) find that investors value loss firms based on their
sales growth rather than earnings. 6 Therefore, in order to detect real activities management in
the IPO setting, a researcher should study separately individual REM components that can be
subject to different incentives.
6
Almost 40 percent of IPO firms are still unprofitable in the IPO year (see Table 4 Panel A).
8
One specific component of real activities management that is subject to two opposite
forces is R&D. First, as demonstrated by prior literature, firms can reduce R&D expenditures as
part of real earnings management strategy to report higher earnings (Dechow and Sloan, 1991;
Baber et al, 1991; Bushee, 1998; Bens et al., 2002). 7 Second, prior literature links R&D
expenditures to firms’ value. Guo et al. (2005) find that initial pricing for science-based IPOs is
mainly determined by R&D expenditures and non-financial variables, such as the number of
products under development, the stage of product development, and the percentage of products
with patents. Andre
et al. (2007) find that for biotechnology firms, R&D expenditures are
value-relevant. Joos and Zhdanov (2008) further demonstrate that the quality of research and the
use of science to develop new products are key factors for science-based firms’ success. If R&D
is an important value-driver, firms can increase R&D expenditures to achieve a positive impact
on firm’s valuation. We expect that both types of R&D management are present among IPO
firms: R&D underinvestment and R&D overinvestment, and that decision to under- or overinvest will vary with firms’ cross-sectional characteristics.
Loss avoidance is cited as one of the leading objectives of earnings management
(Burgstahler and Dichev, 1997; Roychowdhury, 2006). Therefore, we anticipate IPO firms that
report profits, but would incur losses in the absence of R&D management, to be more likely to
underinvest in R&D. Barton and Simko (2002) show that managers are constrained in their
ability to engage in the accrual earnings management if firm’s net operating assets, NOA, are
already overstated. Based on this finding, it is reasonable to expect that firms with high NOA
attempt to manage earnings via real activities manipulations. Therefore, our firs hypothesis
regarding the probability of engaging in R&D underinvestment is:
7
Dechow and Sloan (1991) find that CEOs reduce spending on R&D toward the end of their tenure to increase
earnings. Baber et al. (1991) and Bushee (1998) also find evidence consistent with reduction of R&D expenditures
to meet positive or previous year’s earnings benchmarks. Bens et al. (2002) report that managers repurchase stock to
avoid EPS dilution, and further demonstrate that managers partially finance these repurchases by reducing R&D
expenditures.
9
H1a: Firms that would incur losses in the absence of earnings management, or have low
accounting management flexibility are more likely to underinvest in R&D.
A decision to overinvest in R&D can be motivated by the desire to increase firm’s
valuation during the IPO. Prior literature presents evidence that R&D expenditures are highly
valued by investors (Sougiannis, 1994; Lev and Sougiannis, 1996; Lev and Zarowin, 1998;
Chambers et al., 2002). Amir and Lev (1996), in their study on value-relevance of the nonfinancial information in the wireless communication industry, point out that value-relevance of
non-financial information and R&D expenditures overwhelm that of traditional financial
indicators, such as earnings and book value, for young, science-based and high-growth firms. 8
Shortridge (2004) examines the value relevance of R&D expenditures for the pharmaceutical
industry and finds that R&D expenditures are positively associated with stock prices. Darrough
and Ye (2006) provide evidence that R&D have a positive valuation multiplier for loss firms.
Furthermore, Franzen and Radhakrishnan (2009) demonstrate that positive valuation multiplier
for R&D for loss firms does not extend to profit firms. Based on the discussion above, our first
hypothesis regarding the probability of engaging in R&D overinvestment is
H1b: Firms that are at the growth stage of their life-cycle, non-profitable, or belong to science
driven industries are more likely to overinvest in R&D.
Our next objective is to examine where there is cost associated with R&D under- and
overinvestment. R&D is important driver of future operating performance (Sougiannis, 1994;
Lev and Sougiannis, 1996; Lev and Zarowin, 1998) and innovations (Griliches, 1984). A
negative association between underinvestment or overinvestment and future performance
supports prior research that suggests that opportunistic managers use operational discretion to
extract personal wealth benefits (Gunny, 2010; Cohen and Zarowin, 2010).
A negative
association is also consistent with myopic hypothesis that management has difficulties with
8
Anthony and Ramesh (1992) also show the difference in the value-relevance of financial variables across life cycle
stages of the company.
10
estimating the normal (optimal) level of R&D investment due to nature of the business or their
managerial abilities.
Alternatively, managers can increase R&D spending to signal future firm value (Leland
and Pyle, 1977; Fan, 2007). A positive association suggests that managers reduce or increase
R&D spending in an effort to attain benefits that allow a firm to perform better in the future. For
example, managers can decrease R&D to meet earnings benchmarks in an effort to enhance the
firm’s credibility and reputation with investors (Bartov et al., 2002a; Burgstahler and Dichev
1997). Managers can increase R&D spending to achieve higher market valuations which leads
to larger cash inflows in form of larger IPO proceeds. An IPO firm can benefit from under- or
overinvestment to the extent that the benefits exceed the costs. Lastly, it may be that managers
engage in R&D under- or overinvestment for several reasons (e.g., opportunistic, myopic or
signaling) and the combined effects on future performance offset on average. We argue that
opportunistic or myopic motives overweight the signaling one and state our second set of
hypotheses in the directional form:
H2a: Future economic performance of IPO firms is negatively associated with R&D
underinvestment.
H2b: Future economic performance of IPO firms is negatively associated with R&D
overinvestment.
We use both, accounting based and non-accounting based measures of future economic
performance. Specifically, we use long-term (up to five years) industry adjusted post-IPO
operating return on assets, frequency of delisting due to poor performance, and quality of future
innovation, measured by the number of patents and patent quality, as proxies for economic
performance of a firm.
Future performance tests do not allow us differentiate between myopic and opportunistic
motives of under- or overinvesting. To disentangle these alternatives we analyze the relation
11
between two types of R&D activity management and abnormal trading volume around lockup
expiration date (i.e., in a three-day window around the end of the lockup period day). The lockup
expiration date represents a well-defined and anticipated event that is associated with significant
selling by pre-IPO shareholders (Field and Hanka, 2001). An appealing aspect of such selling
from the standpoint of a researcher is that the lockup expiration dates are common and binding,
and they are well known and specified in the IPO prospectus. If managers opportunistically
manipulate R&D expenditures during IPO for personal gains, we expect a greater degree of
insiders’ selling at the lockup expiration date. Therefore, opportunistic motivation for under- or
overinvestment in R&D suggests that abnormal trading volume around the unlock date of underor overinvesting firms is higher than abnormal trading volume for firms investing in R&D at the
expected level. Alternatively, the myopic hypothesis predicts no association between over- and
underinvestment in R&D and abnormal trading volume around the unlock event.
H3a: Abnormal trading volume around the unlock period for IPO firms is positively associated
with R&D underinvestment.
H3b: Abnormal trading volume around the unlock period for IPO firms is positively associated
with R&D overinvestment.
3.
IPO sample and evidence of real earnings management during the IPO
3.1.
Data and sample selection
Our sample of domestic U.S. initial public offerings is from 1980 to 2006. The cutoff of
2006 is motivated by the need to have data on future long-term (up to five years) operating
performance after the IPO year and availability of patents data. Patents data are obtained from
the National Bureau of Economic Research (NBER) website (http://www.nber.org/patents/).
The list of IPO firms, data on underwriters, venture capitalists (VC), offer price, lock-up
expiration date, and shares issued are obtained from the Securities Data Corporation Global New
Issues Database by Thompson Financials. Financial accounting data comes from the
12
COMPUSTAT annual industrial database. The Center for Research Securities Prices (CRSP)
monthly files were used to obtain information about prices, stock returns, and shares outstanding.
Founding dates and underwriters’ reputation rankings are obtained from Jay Ritter’s website
(http://bear.warrington.ufl.edu/ritter/ipodata.htm). We follow the IPO literature and exclude from
the sample unit offerings, Real Estate Investment Trust (REIT) offerings, American Depository
Receipts (ADRs), closed-end funds and firms in regulated industries (SIC 4910–4939) and
financial institutions (SIC 6000–6999). To be included in the sample, an IPO firm must have an
offer price of at least $3 per share.
For the initial analysis of different real earnings management (REM) metrics, we use a
full sample of IPO firms. The full IPO sample consists of 3,695 firms. For our main analysis,
since our variable of interest is R&D expenditures, we require that each sample firm has R&D
expenditures during the IPO year, one year prior to the IPO year, and one year after the IPO.
Our final sample consists of 2,337 firms; we refer to this sample as the “IPO with R&D” sample
throughout the paper.
Table 1 reports descriptive statistics of our main IPO sample. Panel A shows the
distribution of IPOs over time. Consistent with the previous IPO research, there is clustering,
since many IPOs occurred during the 1990s. For example, 1,578 firms from of our total sample
of 2,337 firms went public during 1991–2000. This is not surprising, since the stock market
boomed during this period, and equity valuations were high. Since the market experienced a
significant downturn beginning in 2000, there have been fewer IPOs; but IPO activities
somewhat resumed in 2004.
Panel B reports the frequency of IPOs by industry. It indicates that IPOs occur in many
different types of businesses, however they are especially frequent in the high technology areas
such as electronic and other electric equipment, chemical products, instruments and related
13
products, and industrial machinery and equipment. Together, these four industries comprise more
than 50% of the sample. Business services is the biggest industry group which comprises around
27% of the sample. 9
3.2.
Real earnings management during the IPO
We start our analysis by examining the presence of real activities manipulations during the IPO
year (Year 0). 10 We utilize the following measures of REM from the prior literature: sales
manipulation estimated through abnormal cash flows from operations (Ab_CFO), overproduction
(Ab_Prod), and abnormal discretionary expenses (Ab_DiscExp), which is the sum of abnormal
advertising expenses (Ab_Adv), abnormal R&D expenditures (Ab_R&D), and abnormal SG&A
(Ab_SG&A). We also use three aggregate REM metrics currently employed in the literature
(Cohen et al., 2008; Cohen and Zarowin, 2010; and Zang, 2012). Specifically, we focus on:
REM_Total, which is the sum of abnormal cash flows multiplied by negative one and abnormal
production costs and abnormal discretionary expenses multiplied by negative one; REM_1,which
is the sum of abnormal production costs and abnormal discretionary expenses multiplied by
negative one; and REM_2, which is the sum of abnormal cash flows multiplied by negative one
and abnormal discretionary expenses multiplied by negative one. Finally, we introduce a new
aggregate measure, REM_3, defined as the sum of abnormal cash flows multiplied by negative
one and abnormal production costs. The real activities management metrics and models are
described in the Appendix B.
We estimate all measures of real activities manipulations on the full IPO sample
consisting of 3,695 firms. Abnormal levels of cash flows from operation, abnormal discretionary
9
The business service group consists of SIC codes 7371-7379 frequently classified as computers.
The general consensus of earnings management literature documents the presence of accrual earnings
management at the IPO year (Year 0), and absence of accrual earnings management at the pre-IPO year (Year -1).
The only two studies that do find evidence of earnings management in Year -1 are Friedlan (1994) and Ducharme et
al. (2001). But both studies use quite small samples (155 and 171 firms, respectively) and much earlier time periods
(1981-1984 and 1982-1987, respectively) which may not be as representative and robust.
10
14
expenses, abnormal advertising expenses, abnormal R&D expenditures, and abnormal SG&A
expenses are reported after being multiplied by negative 1 so that positive values represent
positive earnings management and negative values represent negative earnings management. The
results are presented in Table 2. Consistent with real earnings management, the mean values of
abnormal cash flows and abnormal production costs are positive and statistically significant
(0.09 and 0.07, respectively) suggesting that IPO firms engage in real earnings management via
offering price discounts to temporarily increase sales and overproduction to report lower cost of
goods sold. On the other hand, the mean of discretionary expenses is negative and statistically
significant at 1% level (-0.12). This finding is inconsistent with real earnings management,
suggesting that on average IPO companies overspend on discretionary expenses. Moreover, the
means of all separate components of abnormal discretionary expenses, i.e., abnormal advertising
expenses, abnormal R&D expenditures and abnormal SG&A are also negative and statistically
significant (-0.01, -0.05, and -0.04, respectively).
We get similar results using aggregate
measures: all aggregate measures involving abnormal discretionary expenses, i.e., REM_Total,
REM_1, and REM_2, are inconsistent with real earnings management at the IPO. The mean of
REM_Total is equal to 0.03 and statistically insignificant; and the means of REM_1 and REM_2
are negative -0.05 and -0.04, respectively, and statistically significant at the 1% level. Only
aggregate measure REM_3, which does not include any discretionary expenses, is consistent
with real earnings management.
Overall, we find mixed evidence of real activities manipulations during the IPO. To sum,
abnormal cash flows and abnormal production costs point to the presence of real earnings
management during the IPO. However, abnormal discretionary expenses, on average, exhibit the
sign opposite of what expected if firms managed earnings upward. We posit that such mixed
results are driven by multiple reporting objectives of IPO firms. Therefore, a finer analysis
15
should be done to detect presence of real activities manipulation.
In further analysis we
investigate R&D expenditures, a specific component of real activities management.
4.
Empirical evidence on R&D activities management during the IPO
4.1.
Discretionary R&D measure
We apply the following procedure to construct discretionary performance-adjusted R&D
expenditures (DRDi,t) for an IPO firm i at time t. First, we estimate the expectation model of
R&D expenditures on the entire sample of Compustat firms (excluding IPO firms). Next, we use
coefficient estimates from the expectation model to calculate discretionary R&D, which is the
difference between actual and expected R&D expenditures, for firms in our “IPO with R&D”
sample. Finally, we calculate performance-adjusted discretionary R&D (DRD) by subtracting
discretionary R&D of the matched non-IPO firm from discretionary R&D of the IPO firm
(Kothari et al., 2005; Cohen et al., 2014). Specifically, we match each IPO firm with a non-IPO
firm by industry (2-digit SIC), year, and life cycle stage. 11 Then we find the firms whose return
on assets (ROA) differs from that of the IPO firm by less than 30%. Lastly, we pick the matching
firm that is the closest in total assets (TA) to the IPO firm. As the result of our procedure we
were able to estimate discretionary R&D for 2,173 firms out of 2,337 firms form our IPO-R&D
sample. We refer to performance adjusted discretionary accruals as discretionary accruals
throughout the paper.
To determine non-discretionary (expected) R&D expenditures we estimate model (1),
which is similar to Berger (1993), Gunny (2010), and Fedyk et al. (2014), cross-sectionally, by
industry and year for all non-IPO firms available on COMPUSTAT with non-missing variables
11
While the majority of prior studies mainly match sample firms on industry, year and performance, we also match
on life cycle stage given that our main variable of interest is R&D expenditures, and R&D expenditures heavily
depend on industry and life-cycle stage (Thietart and Vivas, 1984; Klepper, 1996; Audretsch, and Feldman, 1996).
We measure life-cycle stage using cash flow patterns, as in Dickinson (2011).
16
required to estimate this regression. An industry j is defined by 3-digits Standard Industrial
Classification (SIC) code. If for a particular year the 3-digits SIC industry has fewer than 15
observations the regression (1) is estimated for 2-digits SIC industry-group.
RDi , j ,t
TAi , j ,t
= a 0, j ,t + a1, j ,t
RDi , j ,t −1
CASH i , j ,t
DSalesi , j ,t
1
+ a 2, j ,t
+ a 3, j ,t
+ a 4, j ,t
+ a 5, j ,t Qi , j ,t + e i ,t
TAi , j ,t
TAi , j ,t
TAi , j ,t
TAi , j ,t
(1)
The subscript i stands for an individual firm, j stands for the industry classification, and t stands
for the year of the observation. Prior year R&D expenditures (RDi,,j,t-1) capture the persistence of
R&D activities. Average cash and marketable securities (CASHi,j,t), a proxy for internal funds
available for R&D investment, are measured as the average of cash and marketable securities at
the beginning and at the end of the year. Change in sales (ΔSalesi,j,t) is used to control for growth
in R&D spending due to a product life cycle. To capture the growth opportunity available for the
firm we use Tobin’s Q (Qi,,j,t), which is measured as the ration of the sum of market value of
equity and book value of debt to book value of total assets. 12 All variables except for Tobin’s Q
are scaled by average total assets (TAi,j,t) in years t-1 and t.
Table 3 Panel A reports the average coefficient estimates of model (1) and t-statistics
from standard errors across industry-years. All coefficient estimates, except inverse total assets,
are positive and significant at the 1% level. Prior R&D expenditures exhibit strong persistence:
the coefficient on RDt-1/TA is 0.895. Average cash and marketable securities (CASH) is
positively associated with R&D expenditures (coefficient 0.017). Change in sales, a control for
the growth in R&D spending due to product life cycle, and Tobin’s Q, a proxy for growth
opportunity, are also positively associated with the level of R&D expenditures (coefficients of
0.007 and 0.002, respectively). The mean adjusted R2 of the regression (1) is 85.94%. Panel B of
Table 3 presents the mean values of discretionary R&D scaled by average total assets (DRD) for
2,337 sample firms for the IPO year (Year 0) and four years thereafter (Year 1, Year 2, Year 3
12
Tobin, 1969.
17
and Year 4, respectively). The mean of DRD for the IPO year is 0.021, and significant at 1%
level. We find that the mean of discretionary R&D one year after the IPO is 0.019, and is also
significant, but starting from the second year after IPO, the means of discretionary R&D are not
significantly different from zero.
Overall, the results reported in Panel B of Table 3 suggest that on average firms do not
reduce R&D spending at the IPO to achieve higher earnings. These findings are consistent with
the results in the Table 2 obtained from the less sophisticated R&D expectation model of
Roychowdhury (2006). The mean of discretionary R&D estimated from the augmented model
(1), 0.021, is lower than that from the Roychowdhury (2006) model, 0.05, which is not surprising
as augmented model (1) and matching on industry, performance and life-cycle reduce estimation
errors in discretionary R&D proxy. As a robustness check we repeat all the analysis conducted
in this paper using changes in R&D expenditures scaled be average total assets, a simplified
proxy of discretionary expenditures that reflects the persistence in R&D spending demonstrated
by the model (1). All the results of this paper hold. We untabulate these results due to space
constraint.
4.2
R&D activity management during the IPO - empirical evidence
To identify firms engaging in under- and overinvestment we partition firms in our IPO-
R&D sample into quintiles based on the discretionary R&D (DRD) at the IPO year. We denote
the lowest DRD quintile, combined three medium DRD quintiles, and the highest DRD quintile
as LDRD, MDRD, and HDRD, respectively. The negative (positive) mean of discretionary R&D
of one of the groups would provide evidence that firms in that group, on average, invest in R&D
below (above) the expected level, i.e., under-(over-) invest in R&D. Note, that because the R&D
expectation model is estimated on the non-IPO sample and adjusted by the discretionary R&D of
the performance matched firms, the mean of discretionary R&D for the entire IPO sample does
18
not have to be equal to zero a priori. Moreover, the mean of discretionary R&D for the LDRD
(HDRD) quintile does not have to be negative (positive) a priori.
Panel C of Table 3 reports mean discretionary R&D (DRD) for LDRD, MDRD and
HDRD groups. The mean of DRD of the lowest DRD quintile (LDRD group) is equal to -0.12
and significant at the 1% level. The mean of DRD of the highest DRD quintile (HDRD group) is
equal to 0.20 and also statistically significant at 1% level. The mean of MDRD group is equal to
0.01 and statistically significant. We label the firms in the lowest and highest quintiles as underand overinvesting groups respectively. We decrease the probability that the estimated abnormal
R&D expenditures are attributable to the measurement error by taking two extreme quintiles
from each side of the distribution.
Figure 1 illustrates the dynamics of discretionary R&D after the IPO year by DRD group.
It plots the time-series of the mean of discretionary R&D of the firms assigned to the lowest,
medium and highest quintiles (LDRD, MDRD, and HDRD, respectively) based on their level of
discretionary R&D at the IPO year. Immediately after the IPO year, the level of discretionary
R&D of LDRD (HDRD) group rises (drops) and converges to the zero level two years (one year)
after the IPO.
This pattern demonstrates that low- and high- discretionary R&D are
unsustainable. Reversion to zero after the IPO indicates that firms increase (decrease) R&D
levels only during stock issuance and helps to mitigate the concerns that our finding of presence
of R&D over-(under-) investment is attributable to the estimation errors in the expected level of
R&D expenditures.
To summarize, we present evidence of R&D underinvestment (earnings management)
among one group of firms, as well as R&D overinvestment (R&D management) among another
group of IPO firms. Our next objective is to detect factors associated with firms’ decisions to
either over- or underinvestment in R&D.
19
4.3.
Cross-sectional determinants of R&D management strategies for IPO firms
Table 4 reports descriptive statistics of cross sectional characteristics of firms in the
LDRD, MDRD and HDRD groups. A few observations are noteworthy. Firms that either
underinvest or overinvest in R&D are smaller in size: their total assets are significantly lower
than those of the firms in the MDRD group. They are on average younger, less profitable, and
have lower quality of underwriters than the firms in MDRD group. Also, their book-to-market
ratio is significantly lower than that of the firms in the MDRD group. Approximately 15% (7%)
of the firms overinvesting (underinvesting) in R&D exhibit the absence of Sales compared to
only 3% of the firms in the MDRD group. The ratio of cash proceeds from the IPO to the
average total assets of the firms in the LDRD and HDRD groups are on average larger than that
of the firms in the MDRD group: 64% and 60%, respectively versus 47%. Eleven percent of the
firms in the underinvestment group would incur losses if they did not lower their R&D
expenditures below the expected level, but only four percent of such firms are present in the
MDRD group.
The ratio of the number of secondary shares sold at IPO by the pre-IPO
shareholders to the total number of shares offered at IPO (Participation) is only 0.11 and 0.07 for
the underinvesting and overinvesting firms respectively, which is substantially lower than the
ratio of 0.14 for the firms in the MDRD group.
Next, we use a multinomial logit model (2) of the R&D investment choice – underinvest
in R&D (LDRD), overinvest in R&D (HDRD), or invest at the expected level (MDRD) – to test
our hypotheses H2a and H2b:
20
R&D Investment Choice = β0 + β1Indic_switch + β2NOA + β3Indic_loss_RDadj + +
β4Indic_zero_sales + β5Indic_science + β6 Log_age + β7BM + β8 Size + β9UW+ β10VC +
(2)
β11Participation + β12Op_ROA+ β13CAPEX+ β14RD+ β15Cash +e
Consistent with H1a and H1b, our explanatory variables are net operating assets at the
fiscal year prior to IPO deflated by total assets (NOA), an indicator variable of losses in the
absence of earnings management (Indic_switch), an indicator variable that is equal to one if a
firm has negative earnings before extraordinary items adjusted by R&D expenditures in the year
prior to IPO (Indic_loss), indicator variable of the absence of Sales (Indic_zero_sales), book-tomarket ratio (BM), natural logarithm of firm age (Log_age), and an indicator variable of science
driven industry (Indic_science). We use firm age, book-to-market ratio and the absence of sales
as the empirical proxies for growth.
Other control variables include a natural logarithm of the market value of the firm on the
day after the IPO (Size); an indicator variable of the underwriters’ quality (UW); an indicator
variable of whether a firm is backed by venture capitalists (VC); the ratio of secondary shares
sold at IPO to the total number of shares offered at IPO (Participation); operating ROA
(OP_ROA), computed as operating income before depreciation and R&D expenditures in the
year prior to IPO divided by total assets; capital expenditure incurred in the year prior to IPO
scaled by total assets (CAPEX); R&D expenditures incurred during the year prior to IPO scaled
by average total assets (RD); and cash and cash equivalents in the year prior to IPO scaled by
total assets (Cash).
We include underwriters’ quality (UW) and the presence of venture
capitalists financing (VC) to control for third party certification and monitoring. Recent studies
find that VCs and high quality underwriters play a monitoring role and constraint incomeincreasing accruals management (Morsfield and Tan, 2006; Katz, 2009; Hochberg, 2012).
OP_ROA and Cash provide controls for the financial conditions of an IPO firm.
R&D
21
expenditures (RD) and capital expenditures (CAPEX) control for importance of such
expenditures to the firm’s business strategy.
Table 5 presents results of the multivariate test of H1a and H1b. When reporting results
from the multinomial logit model (2) we use a choice of “invest at expected level”, MDRD, as
the base case. Thus, parameter estimates from the multinomial logit should be interpreted as the
effect of explanatory variables on the choice to underinvest or overinvest over the choice to
invest at the expected level. Accordingly, the parameter estimates are labeled as “LDRD vs.
MDRD” and “HDRD vs. MDRD”.
Confirming hypothesis H1a, coefficient estimates of
Indic_switch (1.51) and NOA (0.21) are positive and significant at the 1% level. Coefficient
estimates of Indic_loss (0.35), Indic_zero_sales (0.52), Indic_Science (0.29), Log_age (-0.30),
and BM (-2.13) are also in the predicted direction and statistically significant, confirming
hypothesis H1b.
Negative and significant coefficients on Size for LDRD and HDRD indicate that smaller
firms are more likely to engage in R&D expenditures management. The coefficient estimate on
Participation is negative and significant for the choice of “overinvest in R&D” over “invest at
expected level” and insignificant for the choice of “underinvest in R&D”. Darrough and Rangan
(2005) interpret a negative coefficient on managerial selling when they perform a linear
regression of change in R&D spending on the number of shares sold by management at IPO as
evidence of greater managerial selling when they reduce R&D spending. In line with Darrough
and Rangan (2005) findings, we also find a negative association between discretionary R&D and
managerial selling of stocks at IPO. However, our results reveal a different story: insider holders
tend to sell less shares if they spend more on R&D, but they do not sell more shares when they
22
under invest in R&D. 13 Positive and significant coefficient estimates on R&D expenditures in
the year prior to IPO for both R&D under- and overinvestment suggest that firms tend to engage
in R&D expenditures manipulations – either positive or negative – when R&D activities are
important to their business strategy. The coefficient estimate on operating ROA in the year prior
to IPO is insignificant for LDRD and positive and significant for HDRD. Coefficient estimates
on capital expenditures (CAPEX) are insignificant for LDRD and HDRD. We do not find
significant results for underwriters’ quality (UW) and the presence of venture capitalists
financing (VC). The coefficients estimates on Cash are positive and significant for both LDRD
and HDRD choices.
Panel B of Table 5 displays the marginal effects of the dependent variables in the
regression (2) and allows us to interpret economic significance of the estimated parameters. The
marginal effect of a continuous variable in Panel B is the change in outcome probability when
the continuous variable is increased from one standard deviation below its mean value to one
standard deviation above its mean value. The marginal effect of a dummy variable is the change
in outcome probability when the dummy variable is increased from zero to one. Marginal effects
in the boldface are statistically significant at the 90% level.
One standard deviation increase in NOA increases the probability of engaging in the real
earnings management by underinvesting in R&D by 25.2%. A firm that reports profits, but
would have been unprofitable if it did not underinvest in R&D below the expected level is 37.7%
more likely to underinvest in R&D. These results are consistent with H2a and demonstrate its
economic significance. The economic significance of different proxies of growth in the H2b
varies. The effect of the book-to-market ratio is the most economically significant: one standard
13
As will be demonstrated by abnormal volume around the unlock date (Table 8), abnormal volume is significantly
higher around the unlock date for firms overinvesting in R&D. Given that discretionary R&D is calculated over the
first fiscal year after the IPO, our findings are consistent with strategic R&D management (overinvestment) during
the IPO.
23
deviation increase in the book-to-market ratio decreases the probability of engaging in R&D
activities manipulation and overinvesting in R&D by 37.5%.
Additionally, one standard
deviation increase in age decreases the probability of overinvestment in R&D by 2.8%; and if a
firm has zero sales, that firm is 3.7% more likely to overinvest in R&D. The economic effect of
losses and science-based industry are less profound. A firm that belongs to a science driven
industry is 2% more likely to overinvest in R&D. A firm that experiences losses is 1.2% more
likely to overinvest in R&D. Overall, results presented in Table 5 Panels A and B support
predictions formulated in H1a and H1b.
5.
Effect of R&D management on long-term post-IPO performance
5.1
Effect of R&D expenditures management on long-term operating performance
In this section we test our hypotheses H2a and H2b, relating the presence of R&D under-
and overinvestment to future operating performance. We use operating return on assets
(OP_ROA) as the measure of future operating performance of an IPO firm. 14 OP_ROA is defined
as operating income before depreciation (OIBDP) adjusted for R&D expenditures divided by
total assets (TA). We adjust the income measure by adding R&D expenditures back. An
adjustment for R&D expenditures is performed to avoid mechanical dependency between
operating performance and R&D intensity because current GAAP requires expensing R&D
expenditures once they have occurred, creating a negative relationship between earnings and
R&D intensity. 15 To account for variation of operating ROA across industries, we adjust it by
14
We used ROA and ROE adjusted for R&D expenditures as alternative measures of operating performance and
obtained qualitatively similar results that are not reported due to space constraint, but are available upon request.
15
For example, Lev and Sougiannis (1996) suggest that R&D expenditures are value relevant for the investors and
should be capitalized and amortized.
24
subtracting 2-digits matching industry medians operating ROA from firms’ own operating
ROA: 16
OPROA_ind_adj = OP_ROA – industry median OP_ROA.
(3)
We aggregate future operating performance into two measures: short- and long- term
performance. Short-term operating performance, OPROA_ind_adj12, is measured as an average
industry adjusted operating ROA for two years after the IPO year.
Long-term operating
performance, OPROA_ind_adj35, is measured as average industry adjusted operating ROA for
3-5 years after the IPO year.
Panel A of Table 6 reports univariate analysis of short-term (OPROA_ind_adj12), and
long-term operating performance (OPROA_ind_adj35). Both short- and long-term operating
performance are lower for the firms that either underinvest (lowest DRD quintile) or overinvest
in R&D (highest DRD quintile). The difference between mean OPROA_ind_adj12 of the firms
in the MDRD group and firms in the lowest (highest) quintile is 0.10 (0.12), positive and
statistically significant. The difference between mean OPROA_ind_adj35 of the firms in MDRD
quintiles and firms in the lowest (highest) quintile is 0.06 (0.08), positive and significant.
Next, we perform a regression analysis of the future operation performance by estimating
the following regression:
OPROA_ind_adj = β0 + β1Indic_LDRD + β2Indic_HDRD + β3Indic_loss+ β4Log_age +
β5UW + β6VC + β7Participation + β8OP_ROA+ β9CAPEX + β10Sales + β11NOA +
(4)
β12RD + e
where
the
dependent
variable
OPROA_ind_adj
is
either
OPROA_ind_adj12
or
OPROA_ind_adj35. The main explanatory variables are Indic_LDRD and Indic_HDRD. The
16
As an alternative specification we computed industry-medians over 10-years rolling window. The results
remained qualitatively similar and are not reported due to space constraint, but are available upon request.
25
coefficient estimates on Indic_LDRD (Indic_HDRD) measure the difference in the future
operating performance between firms in lowest (highest) quintile of discretionary R&D and the
medium group. We control for firms’ characteristics correlated management choice to under- or
overinvest and with future operating performance. The control variables include operating ROA
in the year prior to IPO (OP_ROA), indicator loss (Indic_loss), a natural logarithm of firms’ age
(Log_age), an indicator variable of underwriters quality (UW), an indicator of whether a firm is
backed by venture capitalists (VC), the ratio of secondary shares sold at IPO to the total number
of shares offered (Participation) that reveals a level of participation in the offering of existing
stockholders, capital expenditures at the year prior to IPO (CAPEX), R&D expenditures at the
year prior to IPO (RD), total sales at the year prior to IPO (Sales), and net operating assets at the
year prior to IPO (NOA) – are defined in the same way as in model (3). All continuous variables
except Participation are deflated by total assets at the year prior to IPO.
Panel B of Table 6 reports results of the regression (4). We find that the coefficient on
Indic_LDRD is negative and significant at 1% level (-0.038) for the short-term operating
performance captured by the dependent variable OPROA_ind_adj12, and negative and
significant at 10% level (-0.023) for the long-term operating performance captured by the
dependent variable OPROA_ind_adj35.
These findings suggest that firms with downward
management of R&D expenditures experience weaker future operating performance, especially
within two years after the IPO, than the firm investing in R&D at the expected level, controlling
for prior operating performance and other control variables covariate with future performance.
For both independent variables, OPROA_ind_adj12 and OPROA_ind_adj35, the coefficient on
Indic_HDRD is negative and statistically significant at the 1% level, and is equal to -0.052 and 0.040, respectively. Negative coefficient on Indic_HDRD indicates that overinvesting in R&D
firms exhibit lower short- (up to two years) and long–term (three-five years) future operating
26
performance than the firms investing in R&D at the expected level even after controlling for
other observable variables influencing future operating performance. In untabulated analysis we
also repeated estimation of regression (4) using continuous variables DRD_negative and
DRD_positve, which represent the amount of R&D under- and overinvestment, instead of
indicator variables Indic_LDRD and Indic_HDR, respectively.
Results are similar to those
reported in Panel B of Table 6: both types of discretionary R&D are associated with weaker
future performance, with slightly stronger economic effect of R&D overinvestment that of R&D
underinvestment. 17 Therefore, we find a support of H2a and H2b using operating performance as
a proxy for future economic performance.
5.2.
Effect of R&D management – evidence from delisting
Supplementing the analysis of future operating performance, we test H2a and H2b using
the frequency of the delisting due to poor performance in five year period after the IPO as
another proxy for future firms’ performance. Table 7 reports the delisting frequency due to poor
performance. 18 We find that firms in the highest discretionary R&D quintile (HDRD) have the
highest rate of delisting due to poor performance (61 firms or 14% of all IPO firms in the HDRD
group are delisted during the five year period after the IPO). However, delisting rates in the
lowest and medium discretionary R&D quintiles look similar: 10% and 8% of the firms in the
lowest and medium DRD groups, respectively, are delisted during the five year period after the
IPO. The difference in the delisting rate between the HDRD and MDRD group is significantly
positive, while the difference in the delisting rate between the LDRD and MDRD groups is
17
As a robustness check we estimate two-stage model applying Heckman (1979) and Wooldridge (2000) approach
to address the possibility of endogeneity of the self-selection to under-(over-) invests in R&D. We find negative and
statistically significant effect of overinvestment in R&D (HDRD) on both short- and long-term operating
performance, OPROA_ind_adj12 and OPROA_ind_adj35. However, we find negative and significant effect of
underinvesting in R&D only on short-term operating performance, OPROA_ind_adj12, but not on long-term,
OPROA_ind_adj35. These results are available upon request and are not reported in the paper due to space
constraint.
18
We identify the delisting due to bad performance by the delisting code between 400 and 599 on the CRSP
database.
27
insignificant. Therefore, we find the support of H2b, but not of H2a when we use the delisting
frequency due to poor performance as a proxy for future economic performance.
Higher frequency of delisting due to poor performance of the firms in the highest
discretionary R&D quintile suggests that overinvestment in R&D at IPO is associated with poor
future performance of the firm. This finding is consistent with the results obtained from the
regression analysis of future operating ROA.
5.3
Effect of R&D expenditures management on firm’s future innovation
In this section we test our H2a and H2b hypotheses using future innovation efficiency as
an indicator of future economic performance. We measure innovation efficiency as firm’s ability
to generate patents and patent citations per dollar of research and development (R&D)
investment. US firms have increasingly recognized the necessity to patent their innovations and,
hence, have been especially active in patenting activities since the early 1980s (Hall and
Ziedonis, 2001; Hall, 2005). Pakes et al. (1998) and Hirshleifer et al. (2013), demonstrate that
the number of patents represents a valuable measure of innovation. Therefore, we use the number
of patents as a first proxy for innovation efficiency. Moreover, prior literature shows that
patents’ technological and economic impact exhibits high variability. Patents that received a
large number of citations are associated with substantial future value and widely used as a proxy
for patents’ quality (Hall et al., 2005; Jaffe et al., 1993; Jaffe et al., 2000; Ahuja et al., 2005; Guo
et al., 2005; Heeley et al., 2009; and Hirshleifer et al., 2013). Therefore, our additional two
proxies for innovation efficiency are based on the number of citations made to a firm’s patents.
In particular, we use the following three proxies of future innovation: Npat_15, which is
defined as the total number of patents filed by a firm with US PTO (US Patent and Trademark
Office) in the years 1 through 5 after the IPO; Ncite_15, which is defined as the total number of
citations received by patents filed by a firm in the years 1 through 5 after the IPO; and Nstar_15,
28
which is defined as is the total number of “star” patents filed by a firm in the years 1 through 5
after the IPO. 19 Similar to the concept of “star scientists”, i.e., most productive scientists, we
define a patent as a star patent if it belongs to the 90th citations percentile of the patents in the
same industry and year, i.e., belongs to the most valuable patents. 20 Following Hirshleifer et al.
(2013) we scale all measures of innovation by the sum of R&D expenditures in years 1 through 5
after the IPO.
Panel B of Table 7 reports all three measures of innovation for LDRD, MDRD and HDRD
groups. Panel B demonstrates that firms overinvesting in R&D exhibit lower future innovation
efficiency than firms investing in R&D at the expected level. The mean differences between
HDRD and MDRD firms of -0.12, -0.21, and -0.04 in Npat_15, Ncite_15, and Nstar_15,
respectively, are all negative and statistically significant.
All three measures of future
innovations for firms underinvesting in R&D, i.e., LDRD firms, are not statistically different
from those for MDRD firms. Therefore, we find support for H2b, but not for H2a when future
performance is proxied by innovation efficiency.
To sum, we find full support for H2b, i.e., evidence that future economic performance of
IPO firms is negatively associated with the presence of R&D overinvestment; and we find only
partial support for H2a, i.e., limited evidence that future economic performance of IPO firms is
negatively associated with the presence of R&D underinvestment. Therefore, we can conclude
that overinvestment in R&D results in more severe decline in post-IPO performance than
underinvestment in R&D.
19
We also considered all three measures for a shorter time period of 3 years after the IPO and obtained qualitatively
similar results. Finally, we considered three measures of innovation, Npat_15, Ncite_15 and Nstar_15 that were
weighted (by industry-class) and adjusted for the number of citations received in the earlier and later years. Obtained
results were qualitatively similar.
20
Hall et al. (2005) show that while the average and below average patent quality does not affect market value at all,
patents with more than 20 citations boost market value by 50-75 percent. Trajtenberg (1990) also shows that patent
citation data is highly skewed with almost half the patents never cited Almeida and Kogut (1999) study the top 25
percent most highly cited patents to investigate the localization of technological knowledge.
29
5.4
Abnormal trading volume around the unlock day.
In this section we test our third set of hypotheses to differentiate whether myopic or
opportunistic behavior is responsible for R&D under- and overinvestment. We examine
abnormal trading volume around the IPO lockup expirations date – the first day when insiders
are allowed to trade after the IPO.
If management opportunistically manipulates R&D
expenditures during IPO for personal rent extraction we expect that trading volume at the lockup
expiration date is higher for under- and overinvesting firms than for firms investing at the
expected level.
Abnormal daily trading volume on a given day is measured relative to each firm’s preunlock mean daily trading volume measured over the [-50, -6] window, where day zero is the
lockup expiration day (Field and Hanka, 2001). Specifically, abnormal volume for a specific firm
i on date T, where T belongs to {-1, 0, 1}, is determined by (5):
AbVoli ,T =
Vi ,T
1 −6
∑Vi,T
45 t =−50
−1
(5)
Average abnormal trading volume, AbVolt-1,t+1, is the mean abnormal trading volume
measured over the [-1,+1] window, where day zero is the lockup expiration day.
Figure 2 depicts abnormal volume around the unlock date (from -50 to +50 days around
the unlock date) of firms in the highest discretionary R&D quintile (HDRD), in the medium
discretionary R&D quintiles (MDRD), and in the lowest discretionary R&D quintile (LDRD).
As can be seen from Figure 2, abnormal volume around the lockup date is the lowest for the
MDRD group, while abnormal volume around the lockup date is the highest for the HDRD
group.
Panel A of Table 8 reports abnormal trading volume around the unlock date for each
DRD group. Consistent with prior literature (Field and Hanka, 2001; Bradley et al., 2001; Brave
30
and Gompers, 2003), abnormal volume is positive for every group around the unlock date (4.89
percent, 4.15 percent and 22.23 percent for LDRD, MDRD, and HDRD groups, respectively).
However, there is substantial difference between the groups. Firms in high discretionary R&D
group have significantly more positive abnormal volume than firms in the medium discretionary
R&D group (mean difference between the groups is -18.08 percent and strongly significant). On
the other hand, abnormal volumes for the firms in the low discretionary R&D group are not
statistically different from those for the medium discretionary R&D group (the mean difference
between abnormal volumes is -0.74 percent and insignificant). Interestingly, our analysis reveals
that well-documented abnormal trading volume around the lockup expiration day is mainly
driven by firms overinvesting at R&D during the IPO.
Panel B of Table 8 reports abnormal volume, AbVolt-1,t+1, around the unlock date after
controlling for variables known to be correlated with abnormal volume around the unlock date
(Field and Hanka, 2001). We run the following regressions:
log (1 + AbVolt −1,t +1 ) = α 0 + α 1 Indic _ HDRD + α 2 Indic _ LDRD + α 3VC + α 4 Indic _ NYSE +
α 5 Runup _ return + α 6 Size + α 7 Abn Re tt −1,t +1 + e i ,t
The variables of interest are Indic_HDRD
(6)
and Indic_LDRD, which are indicator
variables equal to one if a firm belongs to the HDRD and LDRD group respectively, and zero
otherwise. Control variables that are correlated with abnormal trading volume include
Indic_NYSE, which is an indicator variable equal to one if a firm is traded on NYSE, and zero
otherwise, VC, which is an indicator variable equal to one if a firm is backed by venture
capitalist(s) at the time of the IPO, and zero otherwise, Runup_return, which is market adjusted
buy and hold return over the window starting from the issue date of the IPO and ending one day
before the start of the announcement quarter during which the lockup expires, and Size, which is
the natural logarithm of the market value of a firm computed as total number of shares
31
outstanding after IPO multiplied by the stock price on the first day after IPO. Abnormal returns,
AbRett-1,t+1, are calculated as mean market adjusted buy and hold returns over the [-1,+1]
window, where day zero is the lockup expiration day.
Multivariate results in Panel B of Table 8 show that coefficient estimate on Indic_HDRD
is equal to 0.2, positive and statistically significant. The coefficient estimate on Indic_LDRD is
insignificantly different from zero. These results suggest that abnormal trading volume of firms
overinvesting in R&D is higher than the abnormal trading volume of firms investing at the
expected level. These findings speak in favor of opportunistic motive for overinvestment
suggesting that managers overinvest in R&D to increase market value of the firm and then cash
out their stock holdings after the unlock day. We cannot make a conclusion that managerial
opportunistic behavior is responsible for underinvestment. To summarize, we find the support for
hypothesis H3b, but not for hypothesis H3a.
6.
Conclusion.
Prior literature on earnings management around initial public offerings (IPOs) presents
ample evidence of accrual earnings management (e.g, Friedlan, 1994; Teoh et al., 1998a; Teoh
et al, 1998b; Darrough and Rangan, 2005; Morsfield and Tan, 2006; Fan, 2007; Lee and
Masulis, 2011), but is lacking any pervasive evidence of real earnings management. The lack of
real earnings management is especially surprising in light of and managerial preferences for real
earnings management over accrual earnings management in periods of higher scrutiny (Graham
et. al., 2005; Cohen et al., 2008) excised by regulators and investors (Ball and Shivakumar,
2008.
In this paper we demonstrate that aggregate real earnings management measure used in
prior literature (Roychowdhury, 2006; Gunny, 2010; Cohen and Zarowin, 2010; Zang, 2012,
32
Wongsunwai, 2013) might lead to conflicting results in the IPO setting, while disaggregated
measure allows us to detect different types of real activities manipulation. Thus, we examine
one specific component of real activities manipulation – R&D management. We show that both
types of R&D activities management are present for the IPO firms - underinvestment and
overinvestment - and that the overall level of R&D overinvestment is higher than that of R&D
underinvestment. We show that the decision to underinvest or overinvest in R&D varies
systematically with firms’ cross sectional characteristics. We demonstrate that growth, nonprofitable firms, firms with no sales, or firms that belong to science driven industries are more
likely to overinvest in R&D. On the other hand, firms with reduced accounting flexibility (as
captured by net operating assets), and firms that would report losses in the absence of R&D
management are more likely to underinvest in R&D.
Analysis of future performance reveals that firms overinvesting in R&D experience a
decline in short-term (up to two years) and long-term (up to five years) future operating
performance, have higher frequency of delisting due to underperformance, and have lower
quality of future innovation. Firms that underinvest in R&D demonstrate decline in future shortand long-term operating performance, but do not exhibit higher frequency of delisting and a
decline in innovation. Our analysis of the future performance of IPO firms does not support a
conjecture that signaling about firm’s future prospects serves as a motivation for R&D
overinvestment. Moreover, our finding that overinvesting in R&D firms have significantly
higher abnormal volume around the unlock period than firms investing at the expected level
points to management opportunism as a driving force for R&D overinvestment.
33
APPENDIX A: Variable Definition
AbnRett-1,t+1
abnormal stock return around the lockup expiration date, which is calculated as mean
market adjusted buy and hold returns over the three day [-1,+1] window, where day
zero is the lockup expiration day.
AbnVolt-1,t+1
abnormal treading volume around the lockup expiration date, which is calculated as
the average daily abnormal volume in a three day window [-1, +1], where day zero is
the lockup expiration day. Daily abnormal volume is measured relative to each
firm’s average three-day trading volume in days -50 to -6 before the unlock day:
AbVoli ,T =
Vi ,T
1 −6
∑Vi,T
45 t =−50
−1
Adv
advertising expenses (COMPUSTAT item XAD), in millions.
Age
the difference between the foundation date of the company and the date of the IPO.
Founding
dates
are
obtained
from
Jay
Ritter’s
website
(http://bear.warrington.ufl.edu/ritter/ipodata.htm).
Log_age
the natural logarithm of firm’s age computed as the difference between the
foundation date of the company and the date of IPO.
BM
book-to-market ratio computed as book value of the firm divided by market value of
the firm.
BV
book value of common equity of the firm (COMPUSTAT item CEQ), in millions.
CAPEX
capital expenditure (COMPUSTAT item CAPX) incurred in the year prior to IPO
scaled by total assets.
Cash
cash and cash equivalent holding COMPUSTAT item CHE) at the end of the fiscal
year prior to IPO scaled by total assets.
CFO
operating cash flows (COMPUSTAT item OANCF), in millions.
COGS
cost of goods sold (COMPUSTAT item COGS), in millions.
DRD
performance adjusted discretionary research and development expenditures of an
IPO firm calculated as the difference between discretionary R&D of an IPO firm and
discretionary R&D of a matched non-IPO firm. Every matched non-IPO firm is at
the same life-cycle stage, belongs to the same industry (2-digit SIC), reports at the
same fiscal year, has the absolute value of difference in return of assets (ROA) less
than 30% between itself and corresponding IPO firm, and has the closest in total
assets value to the IPO firm. Discretionary R&D of an IPO firm is calculated by
applying coefficient estimates of model (1) in the text estimated on the entire sample
of non-IPO Compustat firms by industry and year to the corresponding variables of
IPO firms.
an indicator variable that equals to one if a firm is traded on NYSE, and zero
otherwise.
Indic_NYSE
34
Indic_HDRD
an indicator variable that is equal to one if a firm has discretionary R&D in the
highest quintile and zero otherwise.
Indic_LDRD
an indicator variable that is equal to one if a firm has discretionary R&D in the
lowest quintile and zero otherwise.
Indic_loss
an indicator variable that is equal to one if a firm has negative earnings before
extraordinary items (COMPUSTAT item IB) in the year prior to IPO, and zero
otherwise.
an indicator variable that is equal to one if a firm has negative earnings before
extraordinary items (COMPUSTAT item IB) adjusted by R&D expenditures
(COMPUSTAT item XRD) in the year prior to IPO, and zero otherwise.
Indic_loss_RD_adj
Indic_science
an indicator variable that is equal to 1 if a firm belongs to a science driven industry
with the SIC codes 2830 - 2839, 3840-3859, or 8070-8072, and zero otherwise.
Indic_switch
an indicator variable that is equal to 1 if a firm reports positive income before
extraordinary items at IPO year (COMPUSTAT item IB), but its income before
extraordinary items adjusted for discretionary R&D expenditures, DRD, is negative,
and zero otherwise.
Indic_zero_sales
an indicator variable that is equal to 1 if a firm reports zero sales (COMPUSTAT
item SALE) or sales are missing, and zero otherwise.
∆Inv
annual change in inventory (COMPUSTAT item INVT), in millions.
MV
total number of shares outstanding (COMPUSTAT item CSHO) multiplied by the
stock price at the end of first annual financial statements after the IPO year, in
millions.
NOA
net operating assets computed as total current assets (COMPUSTAT item ACT) –
cash and short term investments (COMPUSTAT item CHE) – total current liabilities
(COMPUSTAT item LCT) + debt in current liabilities (COMPUSTAT item DLC) +
taxes payable (COMPUSTAT item TXP. NOA is reported in millions.
Ncite_15
number of citations of the patents filed in the years 1-5 after the IPO scaled by the
total R&D expenditures in the years 1-5 after the IPO.
Npat_15
number of patents filed in the years 1-5 after the IPO scaled by the total R&D
expenditures in the years 1-5 after the IPO.
Nstar_15
number of the star patents filed in the years 1-5 after the IPO scaled by the total
R&D expenditures in the years 1-5 after the IPO. We define a patent as a star patent
if it belongs to 90th citations percentile of the patents in the same industry and year.
OP_ROA
operating ROA is defined as the sum of operating income before depreciation
(COMPUSTAT item OIBDP) and R&D expenditures (COMPUSTAT item XRD)
divided by total assets (COMPUSTAT item AT).
OPROA_ind_adj12
average industry-adjusted operating ROA for two years after the IPO year. Industry
adjusted OP_ROA computed as the difference between a firm’s OP_ROA and
industry median of OP_ROA, where industry is determined by two- digit SIC code.
35
OP_ROA is operating ROA defined as operating income before depreciation
(COMPUSTAT item OIBDP) plus R&D (COMPUSTAT item XRD) expenditures
divided by total assets.
OPROA_ind_adj35
average industry-adjusted operating ROA for years 3 to 5 after the IPO year.
Industry adjusted OP_ROA computed as the difference between a firm’s OP_ROA
and industry median of OP_ROA, where industry is determined by two- digit SIC
code. OP_ROA is operating ROA defined as operating income before depreciation
(COMPUSTAT item OIBDP) plus R&D (COMPUSTAT item XRD) expenditures
divided by total assets.
Participation
the ratio of number of secondary shares sold at IPO to the total number of shares
offered at IPO.
Proceeds
the sum of new shares (primary) shares and secondary (old) shares sold at IPO
multiplied by the offer price, in millions.
RD
annual research and development expenditures (COMPUSTAT item XRD), in
millions.
ROA
a return on asset which is computed as net income before extraordinary items
(COMPUSTAT item IB) divided by total assets (COMPUSTAT item AT).
Runup_return
is market adjusted buy and hold return over the window starting from the issue date
of the IPO and ending one day before the start of the announcement quarter during
which the lockup expires.
Sales
annual sales (COMPUSTAT item SALE), in millions.
∆Sales
sales growth calculated as difference between current and last period sales
(COMPUSTAT item SALE) divided by the last period sales.
Size
is the natural logarithm of the market value of a firm (in millions) computed as total
number of shares outstanding after IPO multiplied by the stock price on the first day
after IPO.
SG&A
selling, general and administrative expenses (COMPUSTAT item XSGA), in
millions.
TA
total assets of the firms (COMPUSTAT item AT), in millions.
Tobin Q
the sum of market value of equity and book value of debt (COMPUSTAT item
DLTT) divided by average total assets (TA) between year year t-1 and t, where year t
is the IPO year.
UW
an indicator variable of underwriter reputation ranking set to 1 if the underwriter
reputation rank, which ranges from 1 to 9 with 9 being the highest rank, is greater or
equal to 8, and 0 otherwise.
VC
an indicator variable set to 1 if the firm is backed by venture capitalist(s) at the time
of IPO, and 0 otherwise.
36
APPENDIX B: Real Activities Management Metrics
1. Sales manipulation (acceleration of the timing of sales through increased price discounts and
more lenient credit terms).
Empirical proxy is the abnormal level of operating cash flows (Ab_CFO) computed as the
difference between operating cash flows and the expected level of operating cash flows of an
IPO firm. The expected level of cash flows is computed from the coefficient estimates of the
following cross-sectional regression estimated by industry-year on the sample of all non-IPO
COMPUSTAT firms with non-missing variables required to estimate the regression:
Salesi ,t
CFOi ,t
∆Salesi ,t
1
(I)
+ e i ,t
+ β3
+ β2
= β1
TAi ,t −1
TAi ,t −1
TAi ,t −1
TAi ,t −1
2. Overproduction (reporting of lower cost of goods sold through increased production).
Empirical proxy is abnormal level of production costs (Ab_Prod) computed as the difference
between production cost (Prod) and the expected level of production cost of an IPO firm.
Production costs (Prod) are defined as sum of cost of goods sold (COGS) and change in
inventory (∆Inv) during the year. The expected level of production costs is computed from the
coefficient estimates of the following cross-sectional regression estimated by industry-year on
the sample of all non-IPO COMPUSTAT firms with non-missing variables required to estimate
the regression: 21
Pr od i ,t
TAi ,t −1
= β1
Salesi ,t
∆Salesi ,t
1
+ β2
+ β3
+ e i ,t
TAi ,t −1
TAi ,t −1
TAi ,t −1
(II)
3. Discretionary expenses (decreases in discretionary expenses including advertising, R&D and
SG&A expenses).
Empirical proxy is abnormal level of discretionary expenses, Ab_DiscExp, computed as the
difference between discretionary expenses (DiscExp) and the expected level of discretionary
expenses of an IPO firm. Discretionary expenses (DiscExp) are measured as the sum of
advertising expenses (Adv), R&D expenditures (RD), and SG&A expenses (SGA). The normal
level of discretionary expenses computed from the coefficient estimates of following crosssectional regressions estimated by industry-year on the sample of all non-IPO COMPUSTAT
firms with non-missing variables required to estimate the regression: 22
Advi ,t
TAi ,t −1
= β1
Salesi , t −1
1
+ β2
+ e i ,t
TAi ,t −1
TAi ,t −1
(III a)
21
Due to data availability for years before the IPO, we only include current change in sales.
We follow Cohen and Zarowin (2010) who use lagged sales instead of current sales in their study of REM around
SEOs. Cohen and Zarowin (2010) claim that modeling discretionary expenses as a function of current sales might
create a mechanical problem if firms manage earnings upward to increase reported earnings in a current year.
22
37
R & Di ,t
TAi ,t −1
= β1
Salesi , t −1
1
+ β2
+ e i ,t
TAi ,t −1
TAi ,t −1
(III b)
Salesi , t −1
1
(III c)
+ β2
+ e i ,t
TAi ,t −1
TAi ,t −1
TAi ,t −1
We also separately estimate the normal advertising expenses, Ab_Adv, using equation (III a); the
normal level of R&D expenditures, Ab_R&D, using equation (III b); and the normal level of
SG&A, Ab_SG&A, using equation (III c).
SG & Ai ,t
= β1
4. Total aggregate measure of three real activities manipulation measures, REM_Total, defined
as the sum of abnormal cash flows multiplied by negative one and abnormal production costs
and abnormal discretionary expenses multiplied by negative one: 23
REM_Total=Ab_CFO*(-1) + Ab_Prod + Ab_DiscExp*(-1)
(IV)
5. First aggregate measure of two real activities manipulation, REM_1, defined as the sum of
abnormal production costs and abnormal production costs and abnormal discretionary expenses
multiplied by negative one:
REM_1=Ab_Prod+Ab_DiscExp*(-1)
(V)
6. Second aggregate measure of two real activities manipulation, REM_2, defined as the sum of
abnormal cash flows multiplied by negative one and abnormal discretionary expenses multiplied
by negative one:
REM_2=Ab_CFO*(-1)+Ab_DiscExp*(-1)
(VI)
7. In addition to six existing measures, we introduce a new aggregate measure of real activities
manipulations, REM_3, defined as the sum of abnormal cash flows multiplied by negative one
and abnormal production costs (i.e., sales management and overproduction, excluding
discretionary expenses):
REM_3=Ab_CFO*(-1)+Ab_Prod
(VII)
Please see Appendix A for variable descriptions.
23
Abnormal CFO and abnormal discretionary expenses are multiplied by negative one, so that higher amounts mean
higher earnings management.
38
References
Aharony, J., C. Lin, and M. Loeb. 1993. Initial public offerings, accounting choices, and
earnings management. Contemporary Accounting Research 10(1): 61-81.
Ahuja, G., R. Coff, and P. Lee. 2005. Managerial foresight and attempted rent appropriation:
Iinsider trading on knowledge of imminent breakthroughs. Strategic Management Journal
26: 791-808.
Almeida, P., and B. Kogut. 1999. Localization of Knowledge and the mobility of engineers in
regional networks. Management Science 45(7): 905-917.
Amir, E. and B. Lev. 1996. Value-relevance of nonfinancial information: The wireless
communications industry. Journal of Accounting and Economics 22(1): 3–30
Andre, P., M. Magnan, and B. Xu. 2007. The stock market valuation of R&D information in
biotech firms. Contemporary Accounting Research 24(4): 1291-1318.
Anthony, J. H., and Ramesh, K. 1992. Association between accounting performance measures
and stock prices: A test of the life cycle hypothesis. Journal of accounting and Economics
15(2): 203-227.
Armstrong, C., G. Foster, and D. Taylor. 2015. Abnormal accruals in newly public companies:
Opportunistic misreporting or economic activity? Journal of Accounting and Economics
(2015): Forthcoming
Audretsch, D. and M. Feldman. 1996. Innovative clusters and the industry life cycle. Review of
Industrial Organization 11 (2): 253-273
Baber, W. R., Fairfield, P. M., & Haggard, J. A. 1991. The effect of concern about reported
income on discretionary spending decisions: The case of research and development.
Accounting Review 66: 818-829.
Ball, R., and L. Shivakumar. 2008. Earnings quality at initial public offerings. Journal of
Accounting and Economics 45(2-3): 324-349.
Barton, J., and Simko, P. J. 2002. The balance sheet as an earnings management constraint. The
accounting review 77(1): 1-27.
Bartov, E., D. Givoly, and C. Hayn. 2002a. The rewards to meeting or beating earnings
expectations. Journal of Accounting and Economics 33 (2): 173–204.
Bartov, E., P. Mohanram, and C. Seethamraju. 2002b. Valuation of Internet stocks—an IPO
perspective. Journal of Accounting Research 40(2): 321–346.
Bens, D., Nagar, V., Franco Wong, M.H., 2002. Real investment implications of employee stock
option exercises. Journal of Accounting Research 40 (2): 359–393.
Berger, P. 1993. Explicit and implicit tax effects of the R&D tax credit. Journal of Accounting
Research 31 (2): 131–71
Bowen, R., A. Davis, and S. Rajgopal. 2002. Determinants of revenue-reporting practices for
Internet firms. Contemporary Accounting Research 19(4): 523–562.
Bradley, D. J., B. D. Jordan and H. Yi. 2001. Venture Capital and IPO Lockup Expiration: An
Empirical Analysis. Journal of Financial Research 14: 465 - 492.
Brave, A. and P. A. Gompers. 2003. The Role of Lockups in Initial Public Offerings. Review of
Financial Studies 16: 1-29.
Burgstahler, D., and Dichev, I. 1997. Earnings management to avoid earnings decreases and
losses. Journal of accounting and economics 24(1): 99-126.
Bushee, B. J. 1998. The influence of institutional investors on myopic R&D investment
behavior. Accounting review 73(3): 305-333.
Callen, J., S. Robb, and D. Segal. 2008. Revenue manipulations and restatement by loss firms.
Auditing: A Journal of Practice and Theory 27(2): 1-29.
39
Cecchini, M., S. Jackson and X. Liu. 2012. Do initial public offering firms manage accruals?
Evidence from individual accounts. Review of Accounting Studies 17(1): 22-40.
Chambers, D., R. Jennings, R. B. Thompson. 2002. Excess returns and R&D intensive firms.
Review of Accounting Studies 7: 133-158.
Chan, Louis K. C., Josef Lakonishok, and Theodore Sougiannis. 2001. The stock market
valuation of research and development expenditures, Journal of Finance 56: 2431-2457.
Cohen, D., Dey, A., and T. Lys. 2008. Real and accrual based earnings management in the Pre
and Post Sarbanes Oxley periods. The Accounting Review 83: 757–787.
Cohen, D. and P. Zarowin. 2010. Accrual-based and real earnings management activities around
seasoned equity offerings. Journal of Accounting and Economics 50: 2–19.
Cohen, D., S. Pandit, C. Wasley, and T. Zach. 2014. Measuring real activity management.
Working Paper, University of Texas in Dallas, University of Illinois at Chicago, University
of Rochester, and Ohio State University.
Curtis, A., McVay, S., and S.,Toynbee, 2015. The Changing Implications of Research and
Development Expenditures for Future Profitability. Working paper, University of
Washington.
Darrough, M., and S. Rangan. 2005. Do insiders manipulate earnings when they sell their shares
in an Initial Public Offering? Journal of Accounting Research 43(1): 1-33.
Darrough, M., Ye, J., 2007. Valuation of loss firms in a knowledge-based economy. Review of
Accounting Studies 12: 61–94.
Dechow, P. M., and Sloan, R. G. 1991. Executive incentives and the horizon problem: An
empirical investigation. Journal of accounting and Economics, 14(1): 51-89.
Dechow, P. M. and D. J. Skinner. 2000. Earnings management: reconciling the views of
accounting academics, practitioners, and regulators. Accounting Horizons 14: 235-250.
Dechow, P. M., Sloan, R. G., & Sweeney, A. P.,1996. Causes and Consequences of Earnings
Manipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC.
Contemporary Accounting Research, 13(1), 37–47.
Dechow, P. M., Ge, W., Larson, C. R., & Sloan, R. G., 2011. Predicting Material Accounting
Misstatements. Contemporary Accounting Research, 28(1), 17–82.
Demers, E., and B. Lev. 2001. A rude awakening: Internet shakeout in 2002. Review of
Accounting Studies 6(2-3): 331–359.
Dickinson, V. 2011. Cash flow patterns as a proxy for firm life cycle. The Accounting Review
86(6): 1969-1994.
DuCharme, L., P. Malatesta, and S. Sefcik. 2001. Earnings management: IPO valuation and
subsequent performance. Journal of Accounting, Auditing and Finance 16 (4): 369–396.
DuCharme, L., P. Malatesta, and S. Sefcik. 2004. Earnings management, stock issues, and
shareholder lawsuits. Journal of Financial Economics 71: 27–49.
Fan, Q. 2007. Earnings management and ownership retention for initial public offering firms:
Theory and evidence. The Accounting Review 82(1): 27–64.
Fedyk, T., Z. Singer and M. Soliman. 2014. The sharpest tool in the shed. An examination of
financial management tools during IPOs. Working paper, University of San Francisco,
McGill University, and University of South California.
Field, L. C. and G. Hanka. 2001. The Expiration of IPO Share Lockups, Journal of Finance 56:
471 – 500.
Franzen, L., and Radhakrishnan, S. 2009. The value relevance of R&D across profit and loss
firms. Journal of accounting and public policy 28(1): 16-32.
Friedlan, J. 1994. Accounting choices of issuers of initial public offerings. Contemporary
Accounting Research 11(1): 1–31.
40
Graham, J.R., Harvey, C.R., and S. Rajgopal. 2005. The economic implications of corporate
financial reporting. Journal of Accounting and Economics 40: 3–73.
Griliches, Z. (editor). 1984. R&D, patents and productivity. University of Chicago Press.
Gunny, K. 2010. The relation between earnings management using real activities manipulation
and future performance: evidence from meeting earnings benchmarks. Contemporary
Accounting Research 27(3): 855-888.
Guo, R., B. Lev, and N. Zhou. 2005. The valuation of bio-tech IPOs. Journal of Accounting,
Auditing and Finance 20(4): 423-459.
Hall, B. H., and Ziedonis, R. H. 2001. The patent paradox revisited: an empirical study of
patenting in the US semiconductor industry, 1979-1995. RAND Journal of Economics 32(1):
101-128.
Hall, B. H. 2005. The financing of innovation. The Handbook of Technology and Innovation
Management, 409-430.
Hall, B., A. Jaffe, and M. Trajtenberg. 2005. Market value and patent citations, Rand Journal of
Economics 36: 16-38.
Hand, J. 2003. Profits, losses and the non-linear pricing of Internet stocks. In: Intangible assets:
Values, measures and risks, eds. J.R.M. Hand and B. Lev. Oxford: Oxford University Press.
Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 20:
125–153.
Healy, P. M. and J. M. Whalen. 1999. A review of the earnings management literature and its
implications for standard setting. Accounting Horizons 13: 365-383.
Heckman, J. 1979. Sample selection bias as a specification error". Econometrica 47(1): 153–61.
Heeley, M., S. Matusik, and N. Jain. 2009. Innovation, appropriability and the underpricing of
Initial Public Offerings. Academy of Management Journal 50: 209-225.
Hirshleifer, D., Hsu, P.-H., and D. Li. 2013. Innovative efficiency and stock returns. Journal of
Financial Economics, 107(3): 632–654
Hochberg, Y. 2012. Venture capital and corporate governance in the newly public firm. Review
of Finance 16(2): 429–80.
Jaffe, A., M. Trajtenberg, and M. Henderson. 1993. Geographic localization of knowledge
spillovers as evidenced by patent citations. The Quarterly Journal of Economics 108(3): 577598
Jaffe, A., M. Trajtenberg, and M. Fogarty. 2000. Knowledge spillovers and patent citations:
evidence from a survey of investors. The American Economic Review 90(2): 215-218.
Joos, P., & Zhdanov, A. 2008. Earnings and Equity Valuation in the Biotech Industry: Theory
and Evidence. Financial Management 37(3): 431–460.
Katz, S. P. 2009. Earnings Quality and Ownership Structure: The Role of Private Equity
Sponsors. The Accounting Review 84(3): 623–658.
Klepper, S. 1996. Entry, Exit, Growth, and Innovation over the Product Life Cycle. The
American Economic Review 86 (3): 562-583.
Kothari, S., A. Leone, and C. Wasley. 2005. Performance matched discretionary accrual
measures. Journal of Accounting & Economics 39(1): 163–197.
Lee, G., & Masulis, R. W. 2011. Do more reputable financial institutions reduce earnings
management by IPO issuers? Journal of Corporate Finance 17(4): 982-1000.
Leland, H. E. and D. H. Pyle. 1977. Informational asymmetrics, financial structure, and financial
intermediation. Journal of Finance 32 (May): 371–387.
Lev, B., 2001. Intangibles: Management, measurement and reporting. The Brookings Institution
Press.
41
Lev, B., Sarath, B., and T. Sougiannis. 2005. R&D Reporting Biases and Their Consequences.
Contemporary Accounting Research 22(4): 977–1026.
Lev, B. and T. Sougiannis. 1996. The capitalization, amortization and value-relevance of R&D.
Journal of Accounting and Economics 21: 107–138.
Lev B., and Zarowin, P. 1998. The market valuation of R&D expenditures. Unpublished
manuscript, New York University.
Morsfield, S. G. and C. E. L. Tan. 2006. Do venture capitalists influence the decision to manage
earnings in initial public offerings? The Accounting Review 81: 1119-1150.
Pakes, A., B. Hall, and Z. Griliches. 1998. The value of patents as indicators of inventive
activity, National Bureau of Economic Research (NBER) Working paper No.2083.
Roychowdhury, S. 2006. Earnings management through real activities manipulation. Journal of
Accounting and Economics 42 (3): 335–370.
Schipper, K. 1989. Commentary on earnings management. Accounting Horizons 3: 91-102.
Shivakumar, L. 2000. Do firms mislead investors by overstating earnings before seasoned equity
offerings? Journal of Accounting and Economics 29: 339–371.
Shortridge, R. T. 2004. Market Valuation of Successful versus Non‐successful R&D Efforts in
the Pharmaceutical Industry. Journal of Business Finance & Accounting 31(9-10): 13011325.
Sougiannis, T. 1994. The accounting based valuation of corporate R&D. Accounting review
69(1): 44-68.
Teoh, S., T. Wong, and G. Rao. 1998a. Are accruals during initial public offerings opportunistic?
Review of Accounting Studies 3(1): 175–208.
Teoh, S., I. Welch, and T. J. Wong. 1998b. Earnings management and the long-run market
performance of initial public offerings. Journal of Finance 53: 1935–1974.
Thietart R.A. and R. Vivas. 1984. An Empirical Investigation of Success Strategies for
Businesses Along the Product Life Cycle, Managemnet Science 30(12): 1405 – 1423.
Tobin, J. 1969. A general equilibrium approach to monetary theory. Journal of Money, Credit
and Banking 1: 15–19.
Trajtenberg. M. 1990. A penny for your quotes: patent citations and the value of innovations.
Rand Journal of Economics 21(1): 172-187.
Wongsunwai, W. 2013. The effect of external monitoring on accrual-based and real earnings
management: evidence from venture-backed initial public offerings? Contemporary
Accounting Research 30(1): 296-324.
Woolridge, J. M. 2000. Instrumental Variables Estimation of the Average Treatment Effect in
the Correlated Random Coefficient Model, mimeo, Michigan State University, Dept. of
Economics.
Zang, A. 2012. Evidence on the tradeoff between real activities manipulation and accrual-based
earnings management. The Accounting Review 87(2): 675–703
42
FIGURE 1
Dynamics of Discretionary R&D expenditures in Post-IPO Years
Figure 1 shows the dynamics of discretionary R&D expenditures of firms in the highest
discretionary R&D quintile (HDRD), in three medium discretionary R&D quintiles (MDRD),
and in the lowest discretionary R&D quintile (LDRD) during five-year period after the IPO.
FIGURE 2
Abnormal Trading Volume around the Unlock Day
Figure 2 shows abnormal trading volume around the unlock day (from -50 to +50 days around
the unlock date) of firms in the highest discretionary R&D quintile (HDRD), in three medium
discretionary R&D quintiles (MDRD), and in the lowest discretionary R&D quintile (LDRD).
43
TABLE 1
Descriptive Statistics for IPO Sample Firms
Panel A: Distribution of IPO sample firms by year
Year
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Total
Number of
issuers by year
19
53
23
138
35
34
100
81
28
35
44
115
149
156
123
182
233
135
76
173
192
23
15
17
63
44
51
2337
Percentage of
issuers by year
0.81
2.27
0.98
5.91
1.50
1.45
4.28
3.47
1.20
1.50
1.88
4.92
6.38
6.68
5.26
7.79
9.97
5.78
3.25
7.40
8.22
0.98
0.64
0.73
2.70
1.88
2.18
100
Cumulative
percentage
0.81
3.08
4.07
9.97
11.47
12.92
17.20
20.67
21.87
23.36
25.25
30.17
36.54
43.22
48.48
56.27
66.24
72.02
75.27
82.67
90.89
91.87
92.51
93.24
95.93
97.82
100
Panel B: Industry (SIC) distribution of IPO sample firms
Industry
Business services
Electronic and other electric equipment
Chemical and applied products
Instruments and related products
Industrial Machinery and Equipment
Engineering and management services
Others (each one < 2% of total IPO
sample)
Total
Two-digit SIC codes
73
36
28
38
35
87
Freq
634
403
359
324
282
74
%
27.13
17.24
15.36
13.86
12.07
3.17
261
11.15
2337
100
44
This table displays the distribution of IPOs overtime and across the industries. Sample consists
of all IPO firms that incurred R&D expenditures during the IPO year, one year prior to the IPO
year, and two years after the IPO year. IPO sample contains 2337 firms conducting IPOs during
1980 -2006.
45
TABLE 2
Real Activities Management during the IPO
Variable
Individual REM
measures
Ab_CFO*(-1)
Ab_Prod
Ab_DiscExp*(-1)
Ab_Adv*(-1)
Ab_R&D*(-1)
Ab_SG&A*(-1)
Aggregate REM
measures
REM_Total
REM_1
REM_2
REM_3
Full IPO sample
N
Mean
t-stat
2899
3587
3580
3587
3580
3587
0.09***
0.07***
-0.12***
-0.01***
-0.05***
-0.04***
-15.03
11.41
15.60
7.26
8.29
15.55
2877
3570
2886
2890
0.03
-0.05***
-0.04***
0.16***
1.97
-4.09
-4.54
14.53
This table reports mean abnormal cash flows from operation, Ab_CFO; mean abnormal
production costs, Ab_Prod; mean abnormal discretionary expenses, Ab_DiscExp; mean abnormal
advertising expense, Ab_Adv; mean abnormal R&D expenditures, Ab_R&D; mean abnormal
SG&A expense, Ab_SG&A; total real activities management, REM_Total, and three aggregate
measures REM_1, REM_2, and REM_3 for a full sample of 3,695 IPOs.
Please refer to Appendix B for the models that estimate normal levels of cash flows from
operation, production costs, discretionary expenses, advertising expenses, R&D expenditures,
and SG&A expenses. Abnormal levels of cash flows from operation, abnormal discretionary
expenses, abnormal advertising expenses, abnormal R&D expenditures, and abnormal SG&A
expenses are reported after being multiplied by negative 1 so that positive values represent
positive earnings management and negative values represent negative earnings management.
Aggregate measures of real activities manipulations are defined as follows (positive values
represent positive earnings management and negative values represent negative earnings
management):
REM_Total=Ab_CFO*(-1) + Ab_Prod + Ab_DiscExp*(-1)
REM_1=Ab_Prod+Ab_DiscExp*(-1)
REM_2=Ab_CFO*(-1)+Ab_DiscExp*(-1)
REM_3=Ab_CFO*(-1)+Ab_Prod
(IV)
(V)
(VI)
(VII)
***, **, **
denotes significance at < .01, < .05, and <.10 levels, respectively, for two-tailed t-tests of
differences in means.
46
TABLE 3
R&D Management Tests
Panel A: Coefficient estimates of R&D expectation model
Variable
Coefficient
Intercept
1/TA
RDt-1/TA
CASH/TA
DSales/TA
Tobin Q
Adjusted R2
0.000
-0.002
0.895
0.017
0.007
0.002
85.94%
***
***
***
***
t-stat
0.23
-0.27
46.09
10.82
8.85
10.65
***, **, * Denotes significance at < .01, < .05, and <.10 levels, respectively, for two-tailed t-tests of differences in
means.
Panel B: Performance-adjusted discretionary R&D by years
Variable
Year 0
Year 1
Year 2
Year 3
Year 4
mean t-stat mean t-stat mean t-stat mean t-stat mean t-stat
Discretionary R&D (DRD) 0.021 6.37 0.019 7.52 0.000 0.12 0.002 0.53 0.002 0.60
Number of observations
2173
2128
2091
1809
1591
Panel C: Performance-adjusted discretionary R&D by quintiles
Variable
Discretionary R&D (DRD)
Number of observations
LDRD
(Quintile 1)
mea
t-stat
n
-0.12
-17.81
MRDR
(Quintiles 2,3&4)
HDRD
(Quintile 5)
mean
t-stat
mean
t-stat
0.01
8.28
0.20
19.28
424
1318
431
Panel A reports coefficient estimates of R&D expectation model (1)
RDi , j ,t
aveTAi , j ,t
= a 0, j ,t + a 1, j ,t
RDi , j ,t −1
CASH i , j ,t
DSales i , j ,t
1
+ a 2 , j ,t
+ a 3, j , t
+ a 4 , j ,t
+ a 5, j ,t Qi , j ,t + e i ,t (1)
aveTAi , j ,t
aveTAi , j ,t
aveTAi , j ,t
aveTAi , j ,t
The model (1) is estimated for every industry-year on the sample of all non-IPO firms available
on COMPUSTAT with non-missing variables required to estimate this regression. An industry j
is defined by 3-digits Standard Industrial Classification (SIC) code. If for a particular year the 3digits SIC industry has fewer than 15 observations the regression (1) is estimated for 2-digits SIC
industry-group. The table reports the mean coefficient across all industry-years and t-statistics
calculated using the standard error of the mean across industry-years. The table also reports the
mean R2 (across industry-years) for each of these regressions.
47
Panel B reports the mean values of performance-adjusted discretionary R&D (DRD), at the IPO
year (Year 0), and four years thereafter (Year 1, Year 2, Year 3, and Year 4). Performanceadjusted discretionary R&D of an IPO firm is calculated as difference between discretionary
R&D of an IPO firm and discretionary R&D of matched non-IPO firm. Every matched non-IPO
firm is at the same life-cycle stage, belongs to the same industry (2-digit SIC), reports at the
same fiscal year, has the absolute value of difference in return of assets (ROA) less than 30%
between itself and corresponding IPO firm, and has the closest in total assets value to IPO firm.
Panel C exhibits the mean values of performance-adjusted discretionary R&D at the IPO year for
firms grouped into quintiles by discretionary R&D expenditures: the lowest discretionary R&D
quintile (LDRD), medium discretionary R&D quintiles (MDRD), and in the highest discretionary
R&D quintile (HDRD).
Please see Appendix A for variable descriptions.
48
TABLE 4
Descriptive Statistics of the IPO Sample Firms Grouped by Quintiles of Discretionary R&D
Firm characteristics
DRD
TA
BV
MV
B/M
Proceeds
Sales
NOA
CAPEX
OP_ROA
Cash
RD
Age
UW
VC
Participation
Indic_loss_RD_adj
Indic_switch
Indic_zero_sales
Indic_science
N obs
LDRD, Quintile 1
Mean Median Std. dev.
-0.12
-0.09
0.14
76.15
40.08
217.55
58.45
30.02
199.06
446.68
118.04 1298.85
0.26
0.25
0.15
0.60
0.56
0.35
1.10
1.08
1.00
0.01
0.30
3.22
0.09
0.06
0.11
0.00
0.15
0.53
0.34
0.29
0.29
0.11
0.07
0.45
9.08
6.00
10.46
0.29
0.00
0.46
0.64
1.00
0.48
0.11
0.00
0.18
0.45
0.00
0.50
0.11
0.00
0.31
0.07
0.00
0.25
0.21
0.00
0.41
424
MRDR, Quintiles 2,3&4
Mean Median Std. dev.
0.01
0.01
0.03
155.91
50.68
552.07
80.39
37.40
286.25
497.97
142.37 2324.06
0.29
0.27
0.19
0.47
0.43
0.26
1.24
1.28
0.81
0.20
0.42
6.18
0.09
0.06
0.10
0.15
0.22
0.38
0.27
0.19
0.26
0.09
0.08
0.07
12.59
7.00
17.09
0.36
0.00
0.48
0.62
1.00
0.49
0.14
0.00
0.21
0.27
0.00
0.44
0.04
0.00
0.19
0.03
0.00
0.18
0.17
0.00
0.37
1318
HDRD, Quintile 5
Mean Median
Std. dev.
0.20
0.14
0.22
59.99
32.40
147.23
46.02
23.94
126.21
289.93
102.75
684.59
0.23
0.22
0.12
0.64
0.55
0.38
0.98
0.62
1.61
0.18
0.23
0.50
0.11
0.08
0.11
0.00
0.03
0.50
0.38
0.35
0.31
0.27
0.21
0.23
7.32
5.00
7.86
0.29
0.00
0.45
0.69
1.00
0.46
0.07
0.00
0.13
0.52
1.00
0.50
0.00
0.00
0.00
0.15
0.00
0.36
0.32
0.00
0.47
431
Please see Appendix A for variable descriptions.
49
TABLE 5
Cross-sectional Determinants of Discretionary R&D Strategies in IPO Year
Panel A: Coefficient estimates of the multinomial logit regression testing H2a and H2b
Alternatives (base status - MDRD)
LDRD vs. MDRD
HDRD vs. MDRD
Estimat
Wald
Wald
PP-value Estimate
2
2
e
Chi
Chi
value
Intercept
-1.04
7.82
0.01
0.41
1.10
0.29
Indic_switch
1.51
43.96 <.0001
-13.93
0.00
0.96
NOA
0.21
33.95 <.0001
0.23
5.62
0.02
Indic_loss_RDadj
0.65
13.67
0.00
0.35
4.16
0.04
Indic_zero_sales
-0.33
1.33
0.25
0.52
4.39
0.04
Indic_science
-0.09
0.33
0.56
0.29
3.50
0.06
Log_age
-0.15
3.29
0.07
-0.30
11.82
0.00
BM
-0.43
0.99
0.32
-2.13
18.01 <.0001
Size
-0.10
3.48
0.06
-0.29
25.48 <.0001
UW
-0.09
0.41
0.52
0.04
0.09
0.76
VC
-0.12
0.78
0.38
0.21
2.18
0.14
Participation
0.42
1.33
0.25
-1.08
5.80
0.02
OP_ROA
-0.03
0.03
0.87
0.51
7.27
0.01
CAPEX
-0.20
0.10
0.75
0.64
1.22
0.27
RD
1.10
41.94 <.0001
1.09
40.12 <.0001
Cash
0.56
4.17
0.04
0.52
3.42
0.06
2
Av. likelihood ratio Chi
381.27
Av. maximum likelihood R2 19.49%
Number of observations
2110
Number of LDRD firms
415
Number of HDRD firms
414
Panel B: Marginal effects of the multinomial logit regression (2).
Alternatives (base status - MDRD)
LDRD vs. MDRD
HDRD vs. MDRD
Indic_switch
-18.7%
37.7%
NOA
25.2%
9.1%
Indic_loss
10.2%
1.2%
Indic_zero_sales
-6.4%
3.7%
Indic_science
-2.0%
2.0%
Log_age
-3.7%
-2.8%
BM
-4.2%
-37.5%
Size
-3.5%
-4.5%
UW
-1.5%
0.4%
VC
-2.4%
1.4%
Participation
3.4%
-2.8%
OP_ROA
-0.8%
2.1%
CAPEX
-0.9%
0.8%
RD
24.2%
6.6%
Cash
5.1%
1.3%
50
Panel A presents the results of test of H2a and H2b. The determinants of firms’ investment in
R&D are modeled using multinomial logit regression:
R&D Investment Choice = β0 + β1Indic_switch + β2NOA + β3Indic_loss_RDadj + +
β4Indic_zero_sales + β5Indic_science + β6 Log_age + β7BM + β8 Size + β9UW+ β10VC +
β11Participation + β12Op_ROA+ β13CAPEX+ β14RD+ β15Cash +e
(2)
The dependent variable, R&D Investment Choice, is equal to 0, 1, and 2 if a firm belongs to the
LDRD group (lowest quintile of discretionary R&D), MDRD group (quintiles 2,3&4 of
discretionary R&D), and HDRD group (highest quintile of discretionary R&D), respectively.
Panel B reports the marginal effects of the independent variables of multinomial logit regression
(2). The marginal effect of a continuous variable is the change in outcome probability when the
continuous variable is increased from one standard deviation to below its mean value to the one
standard deviation above its mean value. The marginal effect of a dummy variable is the change
in outcome probability when the dummy variable is increased from zero to one. Marginal effects
in the boldface are statistically significant at the 90% level.
Please see Appendix A for variable descriptions
51
TABLE 6
Operating Performance of IPO firms
Panel A: Short- and long-term operating performance by discretionary R&D quintiles
LDRD
OPROA_ind_adj12
OPROA_ind_adj35
Mean
Std
-0.13
-0.09
0.32
0.26
MDRD
Mea
n
-0.03
-0.03
Std
0.24
0.21
Difference
between
MDRD and
LDRD
Difference
between
MDRD and
HDRD
Std
Mean t-stat
0.35
0.32
0.10
0.06
Mea
n
0.12
0.08
HDRD
Mea
n
-0.15
-0.11
7.02
4.91
t-stat
8.14
5.85
Panel B: Regression analysis of short- and long-term future operating performance
Dependent variable
OPROA_ind_adj12
OPROA_ind_adj35
Parameter estimate
t-stat
Parameter
t-stat
estimate
Intercept
-0.181 ***
-8.47
-0.160 ***
-7.16
Indic_LDRD
-0.038 ***
-2.98
-0.023 *
-1.72
***
***
Indic_HDRD
-0.052
-4.03
-0.040
-2.99
Indic_loss
-0.040 ***
-2.61
-0.034 **
-2.11
***
***
Log_age
0.027
4.32
0.024
3.66
UW
0.049 ***
4.70
0.040 ***
3.69
VC
0.057 ***
5.46
0.057 ***
5.21
Participation
0.145 ***
5.13
0.095 ***
3.21
OP_ROA
0.332 ***
18.99
0.184 ***
9.89
CAPEX
0.025
0.52
0.167 ***
3.30
Sales
-0.017 **
-2.39
-0.009
-1.24
NOA
0.002
1.35
0.003 *
1.81
RD
0.005
0.90
-0.001
-0.19
Adj R2
40.64%
21.13%
Number of observations
2,109
1,917
Panel A displays short-term (first two years after the IPO), OPRAO_ind_adj12, and long-term
(years three to five after the IPO), OPRAO_ind_adj35, industry adjusted operating performance
by discretionary R&D quintiles. LDRD group contains firms in the lowest quintile of
discretionary R&D, MDRD group contains firms in quintiles 2,3&4 of discretionary R&D, and
HDRD group contains firms in the highest quintile of discretionary R&D.
Panel B exhibits coefficient estimates of the regression analysis of short-term (first two years
after the IPO), OPRAO_ind_adj12, and long-term (years three to five after the IPO),
OPRAO_ind_adj35, future industry adjusted operating performance of firms underinvesting in
R&D (LDRD) and overinvesting in R&D (HDRD):
52
OPROA_ind_adj = β0 + β1Indic_LDRD + β2Indic_HDRD + β3Indic_loss+ β4Log_age +
β5UW + β6VC + β7log_age + β8Participation + β9OP_ROA+ β10CAPEX + β11Sales +
β12NOA + β13RD + e
(4)
***, **, * Denotes significance at < .01, < .05, and <.10 levels, respectively, for two-tailed t-tests
of differences in means.
Please see Appendix A for variable descriptions.
53
TABLE 7
Long-term (up to five years after the IPO) Performance of IPO firms
Panel A: Frequency of delisting due to poor performance by discretionary R&D quintiles.
Number of firms delisted Percent of firms delisted
Difference
due to poor performance due to poor performance between MDRD
and LDRD
Chi
LDRD MDRD HDRD LDRD MDRD HDRD
P-value
square
First year after the IPO
Second year after the IPO
Third year after the IPO
Fourth year after the IPO
Fifth year after the IPO
Total of five years
Number of observations
1
3
11
21
7
43
424
0
9
23
30
40
102
1,318
0
11
17
14
19
61
431
0%
1%
3%
5%
2%
10%
0%
1%
2%
2%
3%
8%
0%
3%
4%
3%
4%
14%
2.43
0.12
Difference
between MDRD
and HDRD
Chi
P-value
square
15.81
<.0001
Panel B: Future patent quality by discretionary R&D quintiles.
LDRD
Npat_15
Ncite_15
Nstar_15
N obs.
Mean
0.34
0.78
0.16
51
Std
0.92
3.50
0.46
MDRD
Mean
0.24
0.41
0.09
166
Std
0.42
1.34
0.12
HDRD
Mean
0.12
0.21
0.04
48
Std
0.14
0.39
0.06
Difference between
MDRD and LDRD
Mean
0.10
0.36
0.07
t-stat
0.75
0.72
0.83
Difference between
MDRD and HDRD
Mean
-0.12***
-0.21*
-0.04***
t-stat
-3.08
-1.76
-2.92
Panel A presents frequency of delisting due to poor performance by discretionary R&D quintiles. Delisting due to bad performance is
identified by the delisting code between 400 and 599 on the CRSP database. LDRD group contains firms in the lowest quintile of
discretionary R&D, MDRD group contains firms in quintiles 2,3&4 of discretionary R&D, and HDRD group contains firms in the highest
quintile of discretionary R&D.
54
Panel B displays future patent quality by discretionary R&D quintiles. LDRD group contains firms in the lowest quintile of discretionary
R&D, MDRD group contains firms in quintiles 2,3&4 of discretionary R&D, and HDRD group contains firms in the highest quintile of
discretionary R&D.
***, **, * Denotes significance at < .01, < .05, and <.10 levels, respectively, for two-tailed t-tests of differences in means.
Please see Appendix A for variable descriptions.
55
TABLE 8
Abnormal Trading Volume around the Unlock Day
Panel A: Univariate analysis of abnormal trading volume around the unlock day.
Difference
Difference
between
between
LDRD
MDRD
HDRD
MDRD and
MDRD and
LDRD
HDRD
Mean Std
Mean
Std
Mean
Std
Mean t-stat
Mean t-stat
AbVol t-1,t+1 4.89%
0.72 4.15% 1.02 22.23% 2.87 -0.74% -0.09 -18.08%*** -2.17
Panel B: Regression analysis of abnormal trading volume around the unlock day.
Log of (1+ Three-day Abnormal Volume)
Dependent variable
Intercept
Indic_HDRD
Indic_LDRD
VC
Indic_NYSE
Runup_ return
Size
AbnRett-1,t+1
N obs
R2
Parameter estimate
-1.01
0.20
0.00
0.32
-0.23
-0.14
0.17
-1.22
1,216
6.94%
***
**
***
***
***
t-stat
-6.67
2.47
0.02
4.76
-1.50
-0.56
5.58
-3.79
Panel A presents the results of univariate analysis of abnormal trading volume (AbVol t-1,t+1)
measured over days -1 to +1 around the unlock day. Abnormal volume is relative to each firm’s
average three-day trading volume in days -50 to -6. LDRD group contains the firms in the lowest
discretionary R&D quintile; MDRD group contains firms in three medium discretionary R&D
quintiles; and HDRD group contains firms in the highest discretionary R&D quintile.
Panel B displays the coefficients estimates of the regression (6) of abnormal volume (AbVol t1,t+1) in a three day window (-1, +1) around the unlock day on Indic_HDRD, Indic_LDRD, and
control variables. Indic_HDRD (Indic_LDRD) is an indicator variable that equals to 1 is a firm
belongs to HDRD (LDRD) group, and zero otherwise. LDRD group contains the firms in the
lowest discretionary R&D quintile; MDRD group contains firms in three medium discretionary
R&D quintiles; and HDRD group contains firms in the highest discretionary R&D quintile.
log (1 + AbVolt −1,t +1 ) = α 0 + α 1 Indic _ HDRD + α 2 Indic _ LDRD + α 3VC + α 4 Indic _ NYSE +
α 5 Runup _ return + α 6 Size + α 7 Abn Re tt −1,t +1 + e i ,t
(6)
***, **, * Denotes significance at < .01, < .05, and <.10 levels, respectively, for two-tailed t-tests
of differences in means.
Please see Appendix A for variable descriptions.
56