The Relation between IPO Underpricing
and Non-GAAP Earnings Measures:
Evidence from Empirical Analysis
University of Amsterdam, Amsterdam Business School
MSc Business Economics, Finance track
Master Thesis
Yang, Guanlan - 10599614
Thesis supervisor: Todorov, Radomir
Abstract
The main purpose of this thesis is examining the relation between different kinds of non-GAAP earnings measures disclosure in IPO prospectus and the IPO underpric-ing. The data range covers those companies going to public from 2004 to 2013 in the United States. The information of Non-GAAP earning measures usage was hand-extracted from their IPO prospectus (S-1 form). This thesis also examines which kind of companies would like to disclose non-GAAP earnings measures and whether the lead underwriters and IPO auditor can affect the willingness of the company to dis-close non-GAAP earnings measures.
Keywords: non-GAAP; non-GAAP earnings measures; pro forma net income; IPO
Acknowledgement
I would like to express my sincere gratitude to my supervisor Dr. Todorov for the continuous support and advices of my master study and thesis. His guidance helped me through the entire period of the thesis procedure. During these couple of months’ guide, I have not only finished my master thesis, but also learnt the altitude to do the research.
Contents
1 Introduction . . . 4
2 Literature Review . . . 8
3 Data, Methodology and Results . . . 14
3.1 Data Collection . . . 14 3.2 Data Summary . . . 18 3.3 Method Design . . . 21 3.4 Results Explanation . . . 24 4 Conclusion . . . 30 5 Literature List . . . 33 6 Appendix . . . 37 6.1 Variables Definition . . . 37 6.2 Tables . . . 38
1
Introduction
The IPO of Twitter brought an argument among investors again. The leading actor of the argument is the so-called “Creative” performance metrics, which, in fact, has been widely used by listed companies in United States for a long time to state that the alternative figures reflect companies’ performance more accurately than the ones ac-cording to the GAAP. With that being said, such non-GAAP earnings measures are not something new. They had already been widely used even before the internet bubble and the methods used during the internet bubble was far more creative than the ones that this thesis investigates.
A non-GAAP earning measure is an alternative measure of the financial performance of a company. The managers of companies choose to disclose non-GAAP earning in addition to the required GAAP earnings when they consider the possibility that those alternative earnings measures better reflect the profitability of their companies. Thus, any number disclosed but not required by SEC can be regarded as a non-GAAP earning measure. As the result, there will be too many different types of non-GAAP earnings to identify. A practical way to study those measures is to restrict them within those commonly used non-GAAP earnings measures, such as adjusted EBITDA or non-GAAP EBITDA, since these kinds of non-GAAP earnings measures can reflect the information of the managers’ personal judgement on the companies’ status to some ex-tent.
IPO prospectus, Form S-1 in this thesis, is a SEC filling used by those companies plan-ning to go public to register their companies with the SEC. The IPO prospectus
con-tains the necessary business and financial information of the companies so that the investors can utilize them to consider their decision on how to invest in the stock mar-ket. During the other time, the investors can rely on the financial numbers disclosed by different financial intermediaries. But for those investors planning to purchase stocks in the IPO market, the IPO prospectus almost is the only and earliest resource that investors can get to analyze the companies’ value.
IPO, the abbreviation of initial public offering, is one of the methods for a company to raise fund from public instead of banks. By selling its stocks to the public investors, the IPO companies are able to expand the business further. Details of the information pro-vided to potential investors are included in the very IPO prospectus, which is audited by the audit firm hired by the board of directors of the IPO companies. One or some-times more investment banks are hired by the IPO companies as underwriters to sell stocks to the first batch of the investors. But when there is no enough investor coming to purchase stocks due to inappropriate offering price, these underwriters need to hold the position and bear the risk until aftermarket trading. Law firms are also involved in IPO process while they are quite discrete compared to audit firms and investment banks.
In Twitter’s case, the management team of Twitter not only disclosed the GAAP fig-ures but also the non-GAAP figfig-ures, the two most important figfig-ures of which are ad-justed EBITDA and adad-justed net income (loss). They successfully turned negative EBITDA as -14 million USD to a positive adjusted EBITDA as 21 million USD by ex-cluding the stock compensation. Thus they reduced their net loss from around -69
million USD to -27 million USD and they call it “non-GAAP net loss”. By processing this measure, the managers are able to convince the investors that their real opera-tional profit is positive and their real net earning is not far from profit.
Based on earlier research related to non-GAAP earning measures, high technology companies, especially the dot-com companies, prefer the non-GAAP earning measures in their annual press releases. That makes sense because usually the business model of a high technology company is relatively more creative so that the traditional account-ing principle may not be suitable from manager’s perspective. But when talkaccount-ing about whether the high technology companies also tend to disclose the non-GAAP earnings measures in IPO prospectus, we don’t know yet, so it is also a target of this thesis. On the other side, opponents claim that non-GAAP earnings measures are mostly used to furnish the financial statements. Regardless of the argument of critics, this study focuses on the market reaction just after the IPO to those companies using non-GAAP earning measures in their IPO prospectus compared with the market reaction to those companies not using them.
The main purpose of this thesis is to examine the relation between IPO underpricing and non-GAAP earning measures used in IPO prospectus. The research questions to be investigated are a) which kind of company tends to disclose non-GAAP earnings in their IPO prospectus? b) whether the IPO underpricing of the companies reporting non-GAAP earning measures smaller than them of those companies that only report-ing GAAP earnreport-ing measures at the first IPO day? c) whether high technology com-panies using non-GAAP earning measures have larger IPO underpricing than other
companies using the same measures?
The data samples selected are those companies finishing their IPO between January 2004 and December 2013 in United States of America. Removing financial companies and the IPO missing too many records, I got the final sample size around 650. Several regression approaches are used to investigate my hypotheses.
Linear probability model together with probit probability model and logit probability model are used to test which kind of companies are more willing to disclose differ-ent kinds of non-GAAP earnings measures. And OLS model with robust standard errors and WLS model are used to test the relation between the IPO underpricing and disclosure of non-GAAP earnings. The general results are that companies hiring lead underwriters with high reputation tend to disclose non-GAAP earnings and the dis-closure of non-GAAP earnings tend to decrease the IPO underpricing significantly.
The structure of the thesis will be as below: I will discuss the research related to both non-GAAP earning measures and the IPO underpricing in the second section, and also raise my hypotheses at the end of the second section. In the third section, I will first describe the data samples and the methodology designed to test the hypotheses, and then explain the results and make sense of it. Following the third section, a conclusion will be there as an overview.
This study is a good start to investigate the non-GAAP earnings measures in the IPO prospectus, and it can also be regarded as a reference for companies planning to go
public and for investors wanting to invest in the IPO market. This study can provide a control variable for the further research related to the IPO underpricing as well.
2
Literature Review
During the last dot-com bubble, the non-standard performance metrics were far more creative than the ones of recent years, but after the bubble was broken and especially after the SEC published the Regulation G on 2003 to regulate how should them to be reported, a significant decreasing was observed. Up to now, most of the papers in this field were to examine the effect of the regulation G and non-GAAP earning method through the accounting view or the regulation view but few from the market view or investment view. On the other hand, IPO underpricing has been examined for sev-eral decades, and disclosed information prior to IPO, underwriters’ reputation, IPO auditor can all explain some of the IPO underpricing, but no one has ever link the IPO underpricing to non-GAAP reporting method before.
Heflin and Hsu (2008) first briefly introduced the regulation G. It requires the compa-nies wishing to disclose non-GAAP earning measures to disclose a) the most directly comparable GAAP numbers as well. b) the reconciliation how the non-GAAP earning measures can be linked to GAAP numbers. They find that the regulation causes both declines in the frequency of different kinds of exclusions and a decline in exclusion magnitude. They also find a slight decline in the probability disclosed earnings to exceed forecasts and a decline in the link between returns and forecast errors is also found to be related to the regulation G. Considering the impact of Regulation G on
how companies choose to disclose non-GAAP earning measures, I decided to select the data samples after 2003, so that the test does not need to specially consider the im-pact of the regulation G. They also raise some thoughts on the reason why managers choose to disclose the non-GAAP earning measures. They think one of the reasons is that the managers can use non-GAAP earning measures to increase the appearance of their performance, and the second reason is that the managers regard non-GAAP earn-ings to be more informative. Bradshow and Sloan (2002) and Lougee and Marguardt (2005) also provide evidence to support this idea. Based on this idea, managers’ dis-closure of non-GAAP earning measures in IPO prospectus should have some impact on the IPO underpricing, negative or positive.
Lougee and Marquardt (2004), just as discussed above, provides evidence to support the idea that non-GAAP earning measures can be more informative. They also in-dicate that high technology companies with non-GAAP earning measures used have better performance in the future. Based on this idea, I decided to separate companies into companies in high technology industries and companies in other industries. Bhat-tacharya, Black and Christensen, Larson (2003) has similar results in this field. They study whether market participants and investors think non-GAAP financial measure to be more informative and persistent than GAAP operating income. They provided evidence that indicates market participants tend to regard non-GAAP earning mea-sures to be closer to real profitability of a company than the GAAP numbers. The result is in line with De Reis Nunes (2012), who believed that the most likely reason why the managers choose to disclose non-GAAP earning measures is to report to in-vestors the “core earnings”. De Reis Nunes (2012) also raised another reason why
managers chose to disclose non-GAAP earning measures. It is like an attempt to ad-just or drive investors’ valuation on the companies’ performance so that they can beat analysts’ forest or performance benchmark.
The results of the research of Doyle, Lundholm and Soliman (2003) showed that the concern about the purpose of non-GAAP earning measures may be right. They found that the items excluded from the GAAP numbers are closely tied to future stock re-turn. Being similar to Bradshaw and Sloan (2002) and Brown and Sivakumar (2001), non-GAAP earning is defined as the actual EPS reported by I/B/E/S. This definition does not work well for this thesis as there are no enough data in I/B/E/S for the time before or just at IPO. That’s reason why I decided to start directly from the original resource, which is the IPO prospectus.
Young (2013) claimed that clear reconciliation of non-GAAP earnings to the related GAAP numbers helps the investors understand the performance of the company even better. Only the reconciliation but without transparent explanation would still mislead some non-professional investors. Also due to the wide existence of non-professional investors in the market, such creative disclosure may cause risks in the financial sys-tem and threaten the solid of the reporting syssys-tem. This thesis is to examine how the non-GAAP earning measures have the impact on the IPO underpricing, and since the IPO price to some extent is determined by professional investors, the negative effect of the non-GAAP earning measures can be best eliminated while the market investors may still take the risk into their consideration when pricing the stock just finishing IPO.
The papers just discussed are all related to the discussion on non-GAAP field. On the other hand, IPO underpricing actually has been studied for several years and dif-ferent theories have been raised to explain how the IPO underpricing always hap-pens. Allen and Faulhaber (1989), Chemmanur(1993), Grinblatt and Hwang (1989), and Welch (1989) have a theory that managers want to push the IPO price lower to la-bel the companies as “High Quality” company so that they can extend the fundraising in the future. If the theory stands, the managers are able to influence the IPO under-pricing. But by which method can they manage to do that? IPO prospectus, the first batch of information, can be the one that managers exert their influence and express their understanding of the companies. If the managers are able to push down their IPO price to increase their IPO underpricing, they are also able to process it on the contrastive way, to earn the fair IPO price by squeezing the IPO underpricing.
According to Loughran and Ritter (2004), the large first day IPO underpricing during earlier 21st century is the result of offering quid pro quos1 from investors to
under-writers for receiving IPO allocations. One necessary condition of this judgement is that the issuing firms are willing to accept the fund raised by the IPO less than the market value. That condition was believed to be satisfied during the Internet Bubble, but not only the market condition has changed recently but also the issuing compa-nies are becoming far more rational than the ones at that time. Also, the prestige of the financial intermediaries was entirely damaged after the internet bubble and espe-cially after the previous financial crisis, so it is widely believed that nowadays issuing companies may not want to leave the money on the table any more, especially those
companies that are able to afford the underwriters with high reputation.
Hong, Scheinkman and Xiong (2006) introduced an explanation of how the bubble appears. The definition of heterogeneous belief, optimist, pessimist and therefore the winner’s curse can also be applied for this research to explain some part of the smaller IPO underpricing for non-GAAP method companies. Lin, Kao and Chen (2010) ex-plained how winner’s curse can happen in the IPO market and its relationship with the IPO underwriters. After the dot-com bubble was broken, the market has never recovered to the hill of around 2000, let alone the previous financial crisis. Thus al-most all the financial intermediaries are facing the profit pressure since then. As the result, the fierce competition on the financial intermediaries side does not allow them to keep pressing down the IPO price to leave an IPO underpricing free profit room. They may have to accept the fact that no matter whether there is IPO underpricing for them to earn the free money by aftermarket trading or to decrease their underwriting risk, they need to consider getting the job first so that at least they can earn the ser-vice fee to rebuild their reputation. When they do, the winner’s curse appears to work.
Non-GAAP earnings measures disclosed by the managers could contain more infor-mation compared with GAAP earnings measures only. That would directly adjust the information asymmetry between the issuers and the potential market investors to squeeze the profit room of underwriters. And at the same time, the market situation starting from the broken internet bubble does not allow the financial intermediaries to keep holding their ground to avoid the winner’s curse. They have to give up part of the profit usually earned from IPO underpricing especially for those issues choosing
to disclose non-GAAP earnings measures.
Thus, my hypotheses are:
H1: Companies with assistance of high reputation lead underwriters tend to disclose non-GAAP earnings measures.
H2: Companies selected Big4 as their IPO auditors tend to disclose non-GAAP earn-ings measures.
H3: High technology companies tend to disclose non-GAAP earnings measures com-pared to companies in other industries.
H4: Companies using non-GAAP earnings measures have smaller first day IPO un-derpricing than those companies only using GAAP earning measures.
H5: High technology companies using non-GAAP earning measures have larger first month IPO underpricing than other companies also using non-GAAP earning measures.
The existing literature is mainly related to the impact of regulation G and the market reaction to the non-GAAP earnings of quarterly or annually financial statements, but little literature is found to investigate the market reaction to the non-GAAP earnings on IPO prospectus on the first day of IPO and the following month while my study can fill in the blanks. My study can also contribute to separate the effect of non-GAAP on stock return adjusted by high technology companies and the other companies.
3
Data, Methodology and Results
3.1
Data Collection
I set the sample range as the companies finished IPO in United States markets from 2004 to 2013 because the Regulation G was published on 2003 and it had a significant impact on how should non-GAAP earnings measures be disclosed. I started from data with IPO date from 1st Jan 2004 to 31st Dec 2013 from COMPUSTAT, and downloaded the Total Assets (AT), Current Assets (ACT), Current Liabilities (LCT), Total Debt (DT), Shareholders’ Equity (SEQ), Net Income (NI) and Sales (SALE) as well from the lat-est entire fiscal year’s data in the COMPUSTAT, in accord with the data in the IPO prospectus. Then I calculate below four ratios, which are the control variables used by Razafindrambinina and Kwan (2013), and I will need these ratios later.
Return on Assets (ROA) measures the profitability relative to its total assets. It is cal-culated as the net income divided by total assets.
ROA = N et Income
T otal Assets (1)
Asset Turnover (TATO) is a financial ratio to measure the turnover ability and asset management of a company relative to its total assets. It is calculated as total sales divided by total assets.
T OT A = Sales
Current Ratio (CR) is a short term liquidity ratio to measure a company’s ability to pay back short term obligations. It is calculated as current assets divided by current liabilities.
CR = Current Assets
Current Liabilities (3)
Debt to Equity Ratio (DER) is a measure of a company’s financial leverage indicating what ratio of equity and debt the company uses to finance assets and financial risk. It is calculated as total debt divided by total shareholders’ equity.
DER = Short T erm Debt + Long T erm Debt
T otal Shareholders0Equity (4)
I also created other two control variables, one of which is called REP and the other is called Big4. REP measures the reputation of the lead underwriters of each IPO com-pany and Big4 measures whether an IPO comcom-pany selected one of the four biggest international audit firms2as its IPO auditor.
The basic idea how to define the reputation of the lead underwriters is in line with Jelic and Briston (2001), which is to classify the reputation based on the frequency of the financial intermediaries acting as lead underwriters. I collected all the lead un-derwriters of the IPO companies in the samples, and sorted them by their appearing
2BIG4, usually is known as the four most famous international audit firms – Deloitte, Ernst & Young,
frequency. When a specific lead underwriter appeared more than 30 times in past 10 years, I gave it “2” as the value of the Reputation. If the frequency of a lead under-writer was between 10 and 30, I gave it “1” as the value of the Reputation, otherwise, “0” was the value of the Reputation. Then for each IPO, I added up the reputation of all the lead underwriters and got the final reputation for that IPO. Big4 was easy to get. I just marked the value of Big4 as “1” if one company selected Big4 as its IPO auditor; otherwise, I marked it as “0”.
Using their IPO date and ticker (TIC), I got their IPO date opening price and closing price in CRSP. With the IPO date for those companies, I also downloaded the market return (Value-Weighted Return) in CRSP and combined it into previous file, and then calculated the first day IPO underpricing as the return of certain stock in first IPO day minus the market return. This method was introduced by Quintana, Luque and Isasi (2005).
RET1stD = (
Pc− Po
Po
) − M arket Return (5)
Regarding the non-GAAP earnings usage information, I manually downloaded all the IPO prospectus (to be specific, the S-1 form) from official SEC and NASDAQ websites with the assistance of Python3. I defined four binary variables for each IPO company,
NonG, NonG1, NonG2, NonG3 and NonG4. Whenever a company clearly mentioned “non-GAAP” in its IPO prospectus, no matter in what circumstance and no matter which kind of non-GAAP earning measure the company was used, I set 1 as the value
3Python is a widely used programming language, and often used as a scripting language. Here I
of NonG1, otherwise, 0 as the value of NonG1 for that company. NonG2 represents whether a company disclosed adjusted EBITDA or non-GAAP EBITDA, and if the company did, then the value of NonG2 is 1, otherwise it is 0. NonG3 represents whether a company disclosed adjusted net income (loss) or non-GAAP net income (loss), and if the company did, the value of NonG3 is 1, otherwise it is 0. NonG4 rep-resents whether a company disclosed pro forma net income (loss), and if the company did, the value of NonG4 is 1, otherwise it is 0. Whenever a company at least had an “1” in NonG1, NonG2, NonG3 or NonG4, I set the value of NonG as 1, otherwise, I set it as 0.
Since there is no clear non-GAAP earnings measures list and a company can disclose any kind of non-GAAP number as long as it can be reconciled to GAAP number, and also given the high number of the IPO prospectus, this thesis was not able to cover all the non-GAAP earnings measures. But based on manual browsing, adjusted EBITDA, non-GAAP EBITDA, adjusted net income (loss), non-GAAP net income (loss) and pro forma net income (loss) are the most common non-GAAP earnings measures, and this thesis has already covered them. If a company did not choose one of them and it did not even mention non-GAAP in its IPO prospectus, this thesis treats it as a company not disclosing any non-GAAP earnings measure in its IPO prospectus. This might be one limitation of this thesis.
I also created two dummy variables called HiTech and Utility to indicate whether an IPO company is a high technology company, utility company or not. The definition of high technology company is based on the company’s SIC code and high technology
SIC codes list summarized by Lee (2000)4. And the definition of the utility company is
based on the sic code between 4900 and 5000.
Fama-French and Liquidity Factors including SMB, HML, MKTRF, RF and UMD were selected as the market representatives for each IPO date. SMB (Small Minus Big) is the average return on the three small portfolios minus the average return on the three big portfolios. HML (High Minus Low) is the average return on the two value port-folios minus the average return on the two growth portport-folios. MKTRF (or Rm Minus Rf) is the excess return on the market. It is calculated as the value-weight return on all NYSE, AMEX, and NASDAQ stocks minus the one-month Treasury bill rate. RF is one month treasury bill rate and UMD is the momentum factor.
Companies with SIC code between 6000 and 6999 were directly excluded as those are financial companies and they had different regulation on how to disclose their perfor-mance. At the same time, companies missing information in COMPUSTAT or CRSP are also excluded, which gave me the final data samples of around 650 IPOs.
3.2
Data Summary
Table 1 shows the summary of different variables for those companies that went to public during 2004 and 2013. The table also shows that around 62% companies chose to disclose at least one kind of non-GAAP earning measures for the past ten years. And the average IPO underpricing of those IPO was 1.17% for past ten years. The
data also reflect that 84% of the companies selected Big4 as their IPO auditors and they also tended to select more than one lead underwriters and the lead underwriters with high reputation.
Table 2 shows the summary of IPO underpricing and non-GAAP earning measures (Variable NonG) for each year from 2004 to 2013. The data shows that for the past ten years, the average IPO underpricing for each year changed quite a lot. Especially for the past three to four years, there is no significant IPO underpricing compared to first couple of years. It occurs to me that IPO underpricing slightly went down with the in-creasing of the percentage of the companies choosing to disclose non-GAAP earning measures. As the result, managers might want to use non-GAAP earning measures disclosure to decrease the IPO underpricing, in another word, to get the fair amount of fund raised by IPO instead of letting the potential price rising room be left to the second market.
Table 3 contains the frequency of different financial intermediaries playing role as lead underwriters in past ten years and therefore the definition of their reputation. Per table 3, there are fourteen financial intermediaries acting as lead underwriters more than 30 times, ten financial intermediaries acting as lead underwriters between 10 and 30 times and there are forty-three financial intermediaries acting as lead underwriters less than 10 times. Table 3 is sorted by the frequency of their acting as lead underwrit-ers, so I am not surprised that those financial intermediaries on the top of the list are all extremely famous ones, though some of which either bankrupted or were merged during the previous financial crisis, such as that Merrill Lynch was merged into
in-vestment department of BOA becoming BAML and Lehman Brothers went bankrupt. Compared to the high and median financial intermediaries, there are also many rela-tive small financial intermediaries working as lead underwriters for past ten years and the number of which is larger than the total number of both high and median reputa-tion financial intermediaries. Therefore, there is still some living room in this market for those small financial intermediaries though not so much.
Table 5 reflects the correlation among different independent variables. Per the table, the correlation among NonG, NonG1, NonG2, NonG3 and NonG4 is relative high, indicating that whenever a company decided to disclose some non-GAAP earnings measures, they often not only disclosed one. The high correlation among different NonG does not affect the test as they don’t appear in the same regression run. REP has a positive correlation with Big4, which might indicate that whenever a company has high expectation on its IPO or has enough cash allowance, it would like to spend more money on its underwriters and IPO auditors. Whenever the company would like to spend money on either underwriters or IPO auditors, it would also like to spend money on the other. REP also has a positive correlation with the ROA but negative cor-relation with CR. This fact is very interesting, which tells that companies with higher profitability tend to hire lead underwriters with higher reputation but when they are not facing a liquidity pressure, the eager decreases. A positive correlations between Big4 and HiTech is observed, indicating high technology companies tend to hire Big4 audit firm as their IPO auditors. There are also some positive correlation observed between financial ratios and market ratios.
3.3
Method Design
For hypothesis 1, hypothesis 2 and hypothesis 3, I first used below linear probabil-ity approach to test. Dependent variables are NonG, NonG1, NonG2, NonG3 and NonG4 respectively. Independent variables are REP, Big4 and HiTech as indicated in the previous section. Control variables are Utility, ROA, TATO, CR, DER and AT. In the model, β1 measures the effect of the reputation of lead underwrites selected by
the IPO companies on the decision to disclose the non-GAAP earnings measures; β2
measures the effect of whether the IPO companies selected Big4 as their IPO auditors on the decision to disclose non-GAAP earnings measures; β3 measures the effect of
whether the IPO companies are high technology companies or not on the decision to disclose non-GAAP earnings measures in IPO prospectus.
P r(N onGi =1|REP, Big4, HiT ech) = β0+ β1· REP + β2· Big4 + β3· HiT ech
+β4· U tility + β5· ROA + β6· T AT O + β7· CR + β8· DER + β9· AT
(6)
While the dependent variables are all binary variables, only using linear probability approach may not give satisfied results. So besides the linear probability approach, I also reported the results of probit regression and logit regression and estimated the marginal effects of those two non-linear models. The independent and dependent variables stay the same. Hence the probit probability model is:
P r(N onGi =1|REP, Big4, HiT ech) = Φ · (β0+ β1· REP + β2 · Big4 + β3· HiT ech
+β4· U tility + β5· ROA + β6· T AT O + β7· CR + β8· DER + β9· AT )
where Φ is the cumulative standard normal distribution function. And the logit prob-ability model is:
P r(N onGi =1|REP, Big4, HiT ech) = F · (β0+ β1· REP + β2· Big4 + β3· HiT ech
+β4· U tility + β5· ROA + β6· T AT O + β7 · CR + β8 · DER + β9· AT )
= 1
1 + e−(β0+β1·REP +β2·Big4+β3·HiT ech+β4U tility+β˙ 5·ROA+β6·T AT O+β7·CR+β8·DER+β9·AT )
(8)
Although the coefficients of probit probability approach and logit probability approach are not easy to interpret, the sign and the significance can be good references and sup-plements to the result of the linear probability approach and it is also possible to esti-mate the average marginal effect of these two non-linear models.
For hypothesis 4, I used below OLS approach to test. Dependent variable is the first day IPO underpricing, which is defined as the first IPO day stock return minus the market return. Independent variables are NonG, NonG1, NonG2, NonG3 and NonG4 while I did not put them all together in the same regression running because of the high correlation among them. I ran the regression several times with different combi-nation of control variables. Control variables are ROA, TATO, CR, DER and AT, which are the same as the control variables in the regression of hypothesis 1, hypothesis 2 and hypothesis 3. I also added REP, Big4, HiTech, Utility and several market factors as the control variables in the regression. The result can also show the effect of market situation on the IPO underpricing. In this approach, β1,imeasures the effect of whether
the IPO companies disclosed some kind of non-GAAP earnings measure on the first day IPO underpricing respectively.
RET1stD = β0+β1,i· N onGi+ β2· ROA + β3· T AT O + β4· CR
+β5· DER + β6· REP + β7 · Big4 + β8· HiT ech
+β9· M KT RF + β10· SM B + β11· HM L + β12· RF + β13· U M D
(9)
For the last hypothesis, the approach is designed to be similar to the one for hypothesis 4. The main difference is that five subgroups were created out of the total data samples. These five subgroups are the subgroup with NonG as “1”, NonG1 as “1”, NonG2 as “1”, NonG3 as “1” and NonG4 as “1” respectively. Thus, as below regression shows, the β1 can measure within a certain subgroup, whether high technology companies
bear the higher IPO underpricing than other companies. Also as the above regression shows, the control variables include ROA, TATO, CR, DER, AT, REP, Big4, MKTRF, SMB, HML, RF and UMD.
RET1stD = β0+β1· HiT ech + β2· ROA + β3· T AT O + β4· CR
+β5· DER + β6· AT + β7· REP + β8 · Big4
+β9· M KT RF + β10· SM B + β11· HM L + β12· RF + β13· U M D
3.4
Results Explanation
Table 6 is the result of the linear regression to test hypothesis 1, 2 and 3. The purpose of the regression is to test which kind of characteristic can affect the willingness of a company to disclose one or some non-GAAP earnings measures. NonG is the sum-mary of all the non-GAAP earnings measures, so to some extent it can be regarded as delegate of NonG1, NonG2, NonG3 and NonG4. This thesis use not only the linear probability approach to test the hypothesis but also the probit probability and logit probability approaches. Overall, when first looking at the summary variable NonG, reputation has a significant positive effect on the willingness to disclose non-GAAP earnings measures, which means the higher reputation of lead underwriters selected by the IPO companies is, more willingness have IPO companies to disclose the non-GAAP earnings measures, if not specifying which kind of non-non-GAAP earnings mea-sure they actually choose. And the results of probit and logit approaches are in line with the linear probability approach about the effect of the REP. Although significant coefficients are also found for TATO and DER in the linear probability approach, the numbers are too tiny to reflect a strong relationship especially when the coefficients of TATO and DER are no longer significant in probit and logit probability approaches. The result is in line with my hypothesis 1.
When specifying the kind of the non-GAAP earnings being disclosed by the IPO com-panies, the results become different. For NonG1, measuring the willingness of an IPO company clearly mentioned “non-GAAP” in its IPO prospectus, the coefficient of the reputation of lead underwriters is still significantly positive, while we can also see that the return on assets (ROA), current ratio (CR) and debt to equity ratio (DER) have a
significant effect on the willingness for the IPO companies to mention “non-GAAP” in their IPO prospectus. The effect of ROA and DER on the willingness to clearly men-tion “non-GAAP” in IPO prospectus is positive while the effect of CR is negative. The results of probit and logit approaches also support this idea while the DER no longer shows a significant effect on the IPO underpricing but the coefficient of TATO becomes significantly negative.
Similar result is found for NonG2, which measures whether an IPO company dis-closed “adjusted EBITDA” or “non-GAAP EBITDA” in their IPO prospectus. The only difference between the results of NonG2 and NonG1 is the coefficient of DER is not significant in result of NonG2 in the linear model. From the results of the regressions to test of NonG1 and NonG2, we still see a significantly positive effect of reputation of lead underwriters. Also, the larger the profitability an IPO company had, the more willingness it had to clearly mention “non-GAAP” and disclose “adjusted EBITDA” or “non-GAAP EBITDA” in its IPO prospectus. If we regard the companies with high ROA as good quality companies, they sure have no fear or concern to disclose their true value of the companies. It is the icing on the cake for their IPO. On the other hand, when an IPO company had higher CR, it did not eagerly want to clearly men-tion “non-GAAP” and disclose “adjusted EBITDA” or “non-GAAP EBITDA” in its IPO prospectus. Companies having higher current ratio faced smaller liquidity pressure; hence they do not need to treat IPO as a method to solve their liquid problem but to develop their future. Similar results are found in probit and logit approaches making the results more solid.
When talking about the results for NonG3, all the coefficients are not significant. But for NonG4, Big4 and DER have significantly positive coefficients while TATO and AT have significantly negative coefficients. Big4 measures whether a company hired one of the four biggest international auditors firms as its IPO auditor. Big4 has a signifi-cantly positive effect on NonG4, whether an IPO company disclosed “pro forma net income” in its IPO prospectus. Compared to other “creative” performance measures, pro forma earnings have been used even widely and in a relatively standardized for-mat, and big 4 auditor firms definitely are good assistants for the IPO companies to disclose pro forma earnings.
Besides the linear probability model, this thesis also provides the estimation of the average marginal effect of probit and logit probability models. As table 8 shows, the average marginal effect of those variables having significant coefficients in the regres-sion is not far away from the effect of the linear model. This result gives a strong support to the finding of the ordinary linear model. As a summary of the test to hy-pothesis 1, 2 and 3, the reputation of the lead underwriters does have a significantly positive effect on the willingness to disclose non-GAAP earnings measures no matter whether we look at the overall results or some of the details. While the Big4 only has the significantly positive effect on the willingness to disclose “pro forma net income”
After checking the characteristics affecting the willingness to disclose non-GAAP earn-ings, this thesis next processes the “core” test, which is the test designed to check the relationship between the first day IPO underpricing and different kind of non-GAAP earnings measures. The OLS regression model is very easy to understand, of which the
dependent variable is the first day IPO underpricing, being measured as first day stock return minus the corresponding market return. The independent variables are differ-ent kinds of non-GAAP earnings measures that whether the IPO companies disclosed or not. Also, considering the potential heteroskedasticity problem, this thesis selected heteroskedasticity-robust standard errors (estimated by Huber-White sandwich esti-mators) to replace the normal standard errors. Considering the heteroskedasticity-robust standard errors may not be able to eliminate the heteroskedasticity problem completely, I further used WLS model to estimate the relationship and attached the result. I have run the regression several times to include different kinds of the non-GAAP earnings measures. Besides the basic running, I also added industry variables or ran the regression within subgroups as robustness test to check whether the signif-icance of the coefficients still stand.
Table 9 contains the results of these multiple regressions to test the relationship be-tween first day IPO underpricing and different disclosure of non-GAAP earnings with the first part of the robustness test. The result is part in line with my hypothesis that only NonG2 has a significantly negative effect on the first day IPO underpric-ing, with or without the industry variables. And the coefficient of NonG2 stays sig-nificant in WLS running. The effect of all the other kinds of disclosure of non-GAAP earnings measures on the first day IPO underpricing is not significantly different from zero. The coefficient of NonG2 is -1.548 if not including the industry control variables, which means whenever an IPO company discloses “adjusted EBITDA” or “non-GAAP EBITDA” in its IPO prospectus, its IPO underpricing would decrease by 1.548% on av-erage. This result does not change a lot when including the industry control variables
or changing the method to WLS. Also a significantly negative effect of MKTRF on the IPO underpricing is observed across the several regression running.
Table 10 contains the second part of the robustness test. First I ran the regression within two subgroups, high technology companies (HiTech subgroup) and companies in other industries (non-HiTech subgroup). The coefficients of NonG2 in non-HiTech subgroup becomes positive and not significant while in the HiTech subgroup, the ef-fect of NonG2 is -2.671%. With that being said, the IPO fund raised by the high tech-nology companies can even be higher on the first IPO day and the underwriters or IPO investors may suffer the loss, but for those issuers of companies in other industries, non-GAAP earnings do not have clear effect on the first day IPO underpricing. Then I removed the utility companies from HiTech subgroup, the result did not change a lot.
Although it is not easy to just say it is the decreasing of market value or the increas-ing of IPO price that cause the IPO underpricincreas-ing to shrink because of the disclosure of the “adjusted EBITDA” or “non-GAAP EBITDA”, we do see the significantly in-verse correlation between the first day IPO underpricing and NonG2. It also can be the combination of those two effects. On the issuers - lead underwriters side, after the internet bubble was broken, the NASDAQ Composite index has never reached its historic highs, above 5000 points around year 2000 and the number of IPOs in NAS-DAQ sharply decreased, especially in technology and internet sector since then. If also considering the impact of the 2008 financial crisis, the position of the underwriters is no longer that strong. The consequence is that the underwriters are not able to force down the price to avoid the winner’s curse. Especially when the underwriters face
those issuers who would like to disclose non-GAAP earnings measures to reflect the true value of their companies, it is even more difficult for the underwriters to negoti-ate a relative low IPO price compared to the market value. From the market investors’ perspective, given the fact that during the last internet bubble non-GAAP earnings measures were abused quite often, investors may tend to regard the companies dis-closing non-GAAP earnings measures as high risk companies in short term. Hence, the market investors do not want to purchase the stock just finished IPO at a rela-tive high price. Adjusted EBITDA (or non-GAAP EBITDA), the adjusted version of EBITDA, the best representative of companies’ operation ability actually pulled the trigger of those two effects.
Alternatively, Boehmer and Fishe (2000) provided another theory to explain the IPO underpricing, which can also be used to explain the negative relation in this result. They believe the IPO underpricing can be considered as the cost of providing liquid-ity in the IPO market. Lower IPO price can attract more IPO investors to purchase the stocks. But when non-GAAP earnings measures really are more informative than GAAP earnings measures, the issuers do not need to bear this “cost” to attract enough investors. When the issuers believe the non-GAAP earnings measures do have the function to deliver the information to potential investors, the issuers would like to raise the IPO price to fund more money.
Table 11 contains the result of test that whether a company is a high technology com-pany affects the effect of the disclosure of non-GAAP earnings measures. This test actually is a supplement to the previous test. But the result turns out that when
spec-ifying which kind of non-GAAP earnings that has been disclosed, the first day IPO underpricing has no significantly different between high technology companies and companies from the other sectors. That means whenever an issuer decides to dis-close a specific non-GAAP earnings (based on previous test, actually NonG2 has a significant effect on the first day IPO underpricing), the first day IPO underpricing is determined and not affected by which industry is the company in.
4
Conclusion
This thesis is mainly checking the relation between the first day IPO underpricing and different kinds of non-GAAP earnings measures. Besides this, this thesis also checks the characteristics that can affect the willingness of IPO companies to disclose different kinds of non-GAAP earnings measures and also how the first day IPO underpricing differs between different industry sectors if the disclosure of non-GAAP earnings mea-sures was decided.
I set the data samples within the companies that finished the IPO between 2004 and 2013 because the Regulation G issued by SEC on 2003 could change the willingness and the way to disclose the non-GAAP earnings measures. I also remove the financial companies as the financial companies have the different requirement to disclose their performance. After removing the samples without enough data, I got my final sam-ples with size around 700 IPO companies. First day IPO underpricing, different kinds of non-GAAP earnings measures, reputation of the lead underwriters, whether they hired four biggest international audit firms, whether they are high technology
compa-nies, return on assets, total asset turnover, current ratio and debt to equity ratio are the important variables that I extracted from either database or their IPO prospectus. After several tests, we have generally find some facts about the relationship between non-GAAP earnings measures and the first day IPO underpricing, and also the char-acteristics that can affect the willingness of the issuers to disclose non-GAAP earnings measures.
Generally speaking, reputation of lead underwriters and whether the IPO companies hire four biggest international auditors can both affect the willingness of IPO compa-nies to disclose some kind of non-GAAP earnings measures. Based on the test, the IPO companies hiring lead writers with higher reputation have larger probability to clearly mention “non-GAAP” and disclose “adjusted or non-GAAP EBITDA” in their IPO prospectus. The IPO companies hiring one of the four biggest international au-dit firms tend to disclose “pro forma net income (loss)” in their IPO prospectus. But high technology companies do not have significant difference compared to compa-nies in other industries regarding the willingness to disclose the non-GAAP earnings measures. Above findings are in line among the linear probability approach, probit probability approach and logit probability approach. The tests also found that the profitability has positive effect on the willingness of IPO companies to clearly mention “non-GAAP” and disclose “adjusted or non-GAAP EBITDA” while the current ratio has negative effect on them.
Regarding the effect of disclosure different kinds of non-GAAP earnings measures on the first day IPO underpricing, the tests find that disclosing “adjusted or non-GAAP
EBITDA” has a significantly negative effect on the first day IPO underpricing. The coefficient is around -1.548% if not including industry dummy variables, and -1.536% if including them while the average first day IPO underpricing for past ten years is only 1.17%. Although the coefficient of disclosure “adjusted or non-GAAP EBITDA” is around -2.67% within the high technology companies, from my last test, we have no evidence to say that disclosure makes high technology companies different from companies in other industries on the first day IPO underpricing.
The limitation of this thesis is that I was not able to quantize the non-GAAP earnings measures, especially the “adjusted or non-GAAP EBITDA”. Now that I have found the negative effect of disclosure of “adjusted or non-GAAP EBITDA” on the first day IPO underpricing, I think it is a good direction to continue the research by quantizing “adjusted or non-GAAP EBITDA”. Also, another potential direction to continue is to study the relationship between the “adjusted or non-GAAP EBITDA” and the first day IPO underpricing before year 2003, and compare it to the finding in this thesis.
5
Literature List
Allen, F and Faulhaber, G.R., 1989, Signaling by Underpricing in the IPO Market, Jour-nal of Financial Economics, Vol. 23, Issue 2, p. 303-323 1989
Bhattacharya, N., Black, E.L., Christensen, T.E., and Larson, C.R., 2003, Assessing the relative informativeness and permanence of pro forma earnings and GAAP op-erating earnings, Journal of Accounting and Economics 36 (2003) 285–319
Bradley, D.J., and Jordan, B.D., 2002, Partial Adjustment to Public Information and IPO Underpricing, Journal of Financial and Quantitative Analysis Vol.37 No.4 December 2002
Bradshaw, M.T., and Sloan, R.G., 2002, GAAP versus The Street: An Empirical Assess-ment of Two Alternative Definitions of Earnings, Journal of Accounting Research Vol.40, No.1 (Mar., 2002), pp 41-66
Chemmanur, T., 1993, The pricing of initial public offerings: a dynamic model with information production, Journal of Finance 48, 285-304
Dos Reis Nunes, F., 2012, The Effect of Corporate Governance Quality on The Strategic Use of Non-GAAP Disclosures to Beat Earnings Benchmarks
Doyle, J.T., Lundholm, R.J., and Soliman, M.T., 2003, The Predictive Value of Expenses Excluded from Pro Forma Earnings, Review of Accounting Studies, 8, 145–174, 2003
Fishe, R.P.H. and Boehmer, E., 2000, Do Underwriters Encourage Stock Flipping? A New Explanation for the Underpricing of IPOs
Grossman, S., and Hart, O.D, 1980, Takeover Bids, The Free Rider Problem and the Theory of the Corporation, Bell Journal of Economics, Spring, 42-64
Heflin, F., and Hsu, C., 2008, The impact of the SEC‘s regulation of non-GAAP disclo-sures, Journal of Accounting and Economics 46 (2008) 349-365
Hong, H., Scheinkman, J., and Xiong, W., 2006, Asset Float and Speculative Bubbles The Journal of Finance Vol.LXI, No.3, June 2006
Jelic, R., Saadouni, B. and Briston, R. 2001 Performance of Malaysian IPOs: Underwrit-ers Reputation and Management Earnings Forecasts Pacific-Basin Finance Journal No. 9, pp. 457-486
Katrina Ellis, Roni Michaely and Maureen O’Hara, 2000, When the Underwriter Is the Market Maker: An Examination of Trading in the IPO Aftermarket The Journal of Finance Vol. 55, No. 3 Jun, pp. 1039-1074
Lee, C., 2000, Master List of SIC Codes Considered High Technology by One or More Sources; SIC Codes used in GCCC’s High Tech Database.
http://www.selfcraft.net/ATSELF/HTSICs.htm
Sub-scriptions with Investors’ Withdrawal Options. Asia-Pacific Journal of Financial Studies, 39: 3–27.
Lougee, B.A., and Marquardt, C.A., 2004, Earnings Informativeness and Strategic Dis-closure: An Empirical Examination of “Pro Forma” Earnings, The Accounting Review, Vol. 79, No. 3 (Jul., 2004), pp. 769-795
Loughran, T. and Ritter, J.R., 2004, Why Has IPO Underpricing Changed Over Time? Financial Management Vol. 33, No. 3
Loughran, T., Ritter, J.R., and Rydqvist, K., 2008, Initial Public Offerings: Interna-tional Insights Pacific-Basin Finance Journal Vol.2, pp.165-199, June 1994, updated Novermber 18, 2008
Marques, A., 2006, SEC Interventions and The Frequency And Usefulness of Non-GAAP Financial Measures, Rev Acc Stud (2006) 11:549–574
Quintana, D., Luque, C., and Isasi, P., 2005, Evolutionary Rule-Based System for IPO Underpricing Prediction, Association for Computing Machinery 2005. p. 983 - 989
Razafindrambinina, D., Kwan, T., 2013, The Influence of Underwriter and Auditor Reputation on IPO Under-pricing, European Journal of Business and Management Vol.5, No.2, 2013
Welch, I., 1989, Seasoned Offering, Imitation Cost and the Underpricing of Initial Pub-lic Offerings, Journal of Finance 44, 421-449
Young, S., 2013, The Drivers, Consequences and Policy Implications of Non-GAAP Earnings Reporting
6
Appendix
6.1
Variables Definition
Variables Description Source
AT Total Assets. Compustat
Big4 Measures whether the issuer hired one of four biggest audit firms as its IPO auditor. IPO prospectus Those four firms are Deloitte, Ernst & Young, KPMG and PwC.
CR Current Ratio = Current Assets / Current Liabilities. Calculation / Compustat
DER Debt to Equity Ratio = Total Debt / Total Equity. Calculation / Compustat
HiTech Measures whether the issuer is a high technology company. SIC
HML High Minus Low. Fama-French
Average return on the two value portfolios minus the average return on the two growth portfolios.
MKTRF Market Return Minus Risk Free Rate. Fama-French
NonG A summary variable of NonG1, NonG2, NonG3 and NonG4. Whenever an issuer have one positive IPO prospectus value in any of those four variables, its NonG is “1”.
NonG1 Measures whether the issuer clearly mentioned “non-GAAP” in its IPO prospectus. IPO prospectus
NonG2 Measures whether the issuer disclosed “non-GAAP or adjusted EBITDA” in its prospectus. IPO prospectus
NonG3 Measures whether the issuer disclosed “non-GAAP or adjusted net income (loss)” in its prospectus. IPO prospectus
NonG4 Measures whether the issuer disclosed “pro forma net income (loss)” in its prospectus. IPO prospectus RET1stD Fist Day IPO Underpricing = (Opening Price - Closing Price) / Opening Price - Market Return Calculation / CRSP
REP Measures the reputation of the lead underwriters hired by the issuer. Definition / NYSE / NASDAQ
RF Risk Free Rate Fama-French
ROA Return on Assets = Net Income / Total Assets Calculation / Compustat
SMB Small Minus Big. Fama-French
Average return on the three small portfolios minus the average return on the three big portfolios. Fama-French
TATO Total Assets Turnover = Total Sales / Total Assets Calculation / Compustat
UMD Momentum Factor Fama-French
Utility Measures whether an IPO company is a utility company. SIC
6.2
Tables
Table 1: Summary of Data Samples
This table shows the summary of all the final data samples. All the accounting numbers were extracted from the latest entire fiscal year data before IPO date from COMPUSTAT.
ROA is return on assets, which is calculated as net income divided by total assets; TATO is assets turnover, which is calculated as sale divided by total assets; CR is current ratio, which is calculated as current assets divided by current liabilities; DER is debt to equity ratio, which is calculated as total debt divided by total shareholder’s equity. REP is the reputation of lead underwriters and Big4 measures whether a company selected Big4 as its IPO auditor. NonG1, NonG2, NonG3 and NonG4 represents whether a company mentioned non-GAAP, disclosed adjusted or non-GAAP EBITDA or adjusted or non-GAAP net income (loss) or Pro Forma net income (loss) in IPO prospectus. NonG represents whether a company fulfilled any one of NonG1, NonG2, NonG3 or NonG4.
The data shows that around 62% companies chose to disclose some non-GAAP earning measures and the average IPO underpricing for past ten years is 1.17%.
Variable Obser. Num. Mean Std. Dev. Min Max
Current Assets 688 164.13 632 0 13948 Current Liabilities 688 117.59 521 0 12384 Total Assets 700 577.49 2061.29 0.11 43211 Total Debt 696 250.75 759.61 0 8551 Total Equity 700 152.12 996.07 -4934 23264 Sale 695 699.51 6993.03 -26.48 182133 Net Income 695 6.48 200.23 -1053 4775 EBITDA 695 58.71 205.09 -253.42 3482 ROA 695 -0.2 1.19 -27.34 1.14 TATO 648 9.2 77.18 -10.56 1536 CR 687 2.38 4.09 0 90.27 DER 696 0.32 43.63 -817.43 691.61 REP 696 3.39 1.86 0 8 Big4 704 0.84 0.37 0 1 NonG1 704 0.23 0.42 0 1 NonG2 704 0.26 0.44 0 1 NonG3 704 0.09 0.29 0 1 NonG4 704 0.45 0.50 0 1 NonG 704 0.62 0.49 0 1 IPO Underpricing (%) 697 1.17 8.62 -40.32 36.62
Table 2: Summary of IPO underpricing and non-GAAP method usage for Each Year
This table shows the IPO underpricing and non-GAAP earning measures usage for past ten years on annual basis. IPO underpricing is calculated as the first IPO date return for one stock minus the the corresponding market return. Non-GAAP earning measures usage is measured by whether a company mentioned non-GAAP in its IPO prospectus or disclosed adjusted EBITDA, non-GAAP EBITDA, ad-justed net income (loss), non-GAAP net income (loss), pro forma net income (loss) in its IPO prospectus. The data reflects a phenomenon that it seems the more companies chose to disclose some non-GAAP earning measures, the smaller the average IPO underpricing is for that year. Especially for 2009, the following year of the financial crisis, only 45% companies chose to disclose the non-GAAP earnings measures, and the corresponding IPO underpricing suddenly increased. As the result, managers might want to use non-GAAP earning measures disclosure to decrease the IPO underpricing, in other word, to get the fair amount of fund raised by IPO instead of letting the potential price rising room be left to the second market.
Panel A: IPO Underpricing (%)
IPO Year Obser. Num. Mean Std. Dev. Min Max
2004 93 1.83 7.93 -13.79 26.39 2005 82 0.82 7.39 -20.40 21.13 2006 111 1.20 7.68 -29.32 36.62 2007 114 2.69 9.62 -15.21 36.02 2008 16 -0.37 8.83 -16.85 25.10 2009 33 1.98 9.14 -13.75 31.79 2010 63 0.91 8.94 -18.97 28.86 2011 69 0.37 8.69 -40.32 20.50 2012 74 0.13 8.39 -18.81 28.05 2013 42 -0.26 10.95 -23.37 27.64
Panel B: Non-GAAP Earning Measures Usage (NonG)
IPO Year Obser. Num. Mean Std. Dev. Min Max
2004 95 0.6211 0.4877 0 1 2005 85 0.6235 0.4874 0 1 2006 112 0.5357 0.5010 0 1 2007 115 0.5391 0.5006 0 1 2008 16 0.5000 0.5164 0 1 2009 33 0.4545 0.5056 0 1 2010 63 0.6667 0.4752 0 1 2011 69 0.6377 0.4842 0 1 2012 74 0.7838 0.4145 0 1 2013 42 0.7857 0.4153 0 1
Table 3: Reputation of Different Underwriters
This table reflects the frequency of different intermediaries act IPO lead underwriters for the past 10 years, and the definition of the reputation. When a specific lead underwriter appeared more than 30 times in past 10 years, I gave “2” to the value of the Reputation. If a frequency of a lead underwriter was between 10 and 30, I gave “1” to the value of the Reputation, otherwise, “0” was the value of the Reputation.
Lead Underwriters Freq. Repu. Lead Underwriters Freq. Repu.
Morgan Stanley 158 2 Leerink 2 0
J.P. Morgan 144 2 Leerink Swaan 2 0
Goldman Sachs 137 2 Maxim 2 0
Credit Suisse 120 2 Newbridge 2 0
Citigroup 97 2 Robert W. Baird 2 0
Deutsche Bank 75 2 Think Equity 2 0
Lehman Brothers 71 2 Renaissance 2 0
Merrill Lynch 61 2 Roth 2 0
BAML 57 2 SunTrust R.H. 2 0
UBS 52 2 Allen 1 0
Piper Jaffray 44 2 Apollo 1 0
Barclays 36 2 BB&T 1 0
Jefferies 34 2 Canaccord Genuity 1 0
Banc of America 30 2 Craig-Hallum 1 0
Stifel Nicolaus 15 1 Dahlman Rose 1 0
William Blair 15 1 Dawson James 1 0
Bear, Stearns 14 1 Desjardins 1 0
RBC 14 1 Dominick & Dominick 1 0
SG Cowen 14 1 Dundee 1 0
Thomas Weisel 13 1 Ferris, Baker Watts 1 0
Wachovia 12 1 FTN Midwest 1 0
Wells Fargo 12 1 I-Bankers 1 0
CIBC World 11 1 JMP 1 0
Cowen 10 1 KeyBanc 1 0
Oppenheimer 8 0 Ladenburg Thalmann 1 0
Raymond James 8 0 Legg Mason W.W. 1 0
A.G. Ed. & Sons 6 0 Macquarie 1 0
Anderson & Strudwick 6 0 MDB 1 0
Lazard 6 0 Mer. Cur. Ford 1 0
Robert W.Baird 6 0 National Securities 1 0
Feltl 5 0 Pacific Crest 1 0
FBR 5 0 Paulson 1 0
Leerink Swann 5 0 Roert W.Baird 1 0
W.R. Hambrecht 5 0 Sandler O’Neill 1 0
McDonald 3 0 Simmons 1 0
Needham 3 0 Stanford 1 0
Pacific Growth 3 0 Stephens 1 0
Rodman & Renshaw 3 0 Taglich Bro. 1 0
BMO 2 0 Thomas Capital 1 0
D.A. Davidson 2 0 Thomas Weisel Par. 1 0
FBR 2 0 Wedbush Morgan 1 0
First Albany 2 0 WestPark 1 0
Table 4: Definition of HiTech
This table shows the definition of HiTech, which is based on the SIC code of each IPO company. Whenever the SIC code of an IPO company appears in below table, the HiTech variable is “1”. Otherwise, it is “0”.
SIC Description of Business SIC Description of Business
12xx Coal mining 361x Electrical transmission, etc. 131x Crude petroleum & gas operations 362x Electrical industrial apparatus 132x Natural gas liquids 363x Household appliance manufacture 138x Oil & gas field drilling, etc. 364x Electric lighting & lighting equipment 201x Food & kindred product manufacturing 365x Video equipment & audio recording 2087 Flavorings 366x Communications equipment
211x Cigarette manufacturing 367x Electronic components & accessories 229x Misc. textile goods 369x Electrical machinery, n. e. c.
261x Pulp mills 371x Motor vehicles & equipment 267x Misc. converted paper products 372x Aircraft and parts
281x Industrial inorganic chemicals 373x Ship & boat building & repairing 282x Plastics, synthetic resins 374x Railroad equipment
283x Drugs & related product manufacture 375x Motorcycles, bicycles & parts 284x Soaps, cleaners, toilet goods 376x Guided missiles and space vehicles 286x Industrial organic chemicals 379x Misc. transportation equipment 287x Agricultural chemicals 381x Search, detection, etc.
289x Misc. chemical products 382x Laboratory, analytical, etc. 2911 Petroleum refining 384x Surgical, medical, etc. 299x Misc. petroleum & coal products 3851 Ophthalmic goods
30xx Rubber & misc. product manufacturing 386x Photographic equipment & supplies 3264 Porcelain electrical products 3873 Watches & clocks
335x Nonferrous rolling & drawing 3999 Misc. manufacturing industries 336x Nonferrous die-castings 48xx Communication services 348x Ordnance, small arms, ammunition 49xx Electric, gas & sanitary services 349x Industrial valves, fluid power valves, etc. 5045 Computers, peripherals, software sales 351x Engines & turbines 737x Computer & data-processing services 353x Industrial equipment & machinery 7389 Business services n. e. c.
354x Machine tools, metal cutting, etc. 781x Motion picture production, etc. 355x Special industrial machinery 8700 Engineering & management services 356x General industrial machinery, n. e. c. 871x Engineering & architectural services 357x Computer & office equipment 873x Research & testing services
358x Machines, including env. & eco. products 874x Management & public relations 359x Industrial machines, n. e. c.. 899x Services n. e. c.
Table 5: Correlation of Different Independent Variables
Table 4 reflects the correlation among different independent variables. Despite the high correlation among different non-GAAP independent variables, other correlation is acceptable. Also the high correlation between the lead underwriters reputation and the Big4 IPO auditors indicates that whenever a company has higher expectation on spend money on either underwriters or IPO auditors, it would also like to spend money on the other.
NonG NonG1 NonG2 NonG3 NonG4 REP Big4 HiTech Utility ROA TATO CR DER AT NonG 1 NonG1 0.4278* 1 (0) NonG2 0.4709* 0.4540* 1 (0) (0) NonG3 0.2494* 0.1468* 0.1480* 1 (0) (0.0001) (0.0001) NonG4 0.7200* 0.1993* 0.1744* 0.1083* 1 (0) (0) (0) (0.004) REP 0.1298* 0.1251* 0.1871* 0.0851* 0.0523 1 (0.0006) (0.0009) (0) (0.0247) (0.1685) Big4 0.1098* -0.0446 0.0019 0.0194 0.1564* 0.3072* 1 (0.0035) (0.2373) (0.9594) (0.6066) (0) (0) HiTech 0.0333 0.0071 -0.1235* 0.0041 0.1100* 0.0356 0.0757* 1 (0.3783) (0.8516) (0.001) (0.9131) (0.0035) (0.3488) (0.0446) Utility -0.0506 0.0245 0.066 -0.0264 -0.0595 0.1215* 0.0549 0.1028* 1 (0.18) (0.5171) (0.0799) (0.4843) (0.115) (0.0013) (0.1457) (0.0063) ROA 0.0498 0.0892* 0.1082* 0.0368 -0.0076 0.1534* 0.0373 -0.1146* 0.0362 1 (0.1894) (0.0186) (0.0043) (0.3325) (0.8408) (0.0001) (0.3257) (0.0025) (0.3406) TATO -0.0651 -0.0525 -0.0571 -0.0294 -0.0354 -0.0682 -0.0431 0.0594 -0.01 -0.1946* 1 (0.098) (0.182) (0.1465) (0.4557) (0.3678) (0.0844) (0.2735) (0.1309) (0.7992) (0) CR -0.0611 -0.1022* -0.1077* -0.0312 -0.0043 -0.0967* -0.0367 0.0354 -0.0605 -0.0119 0.2285* 1 (0.1099) (0.0073) (0.0047) (0.4141) (0.9109) (0.0116) (0.3367) (0.3544) (0.1132) (0.7561) (0) DER 0.0681 0.0772* -0.0088 0.0172 0.0744* 0.0086 0.02 -0.0173 -0.0328 -0.0017 -0.0001 -0.0041 1 (0.0725) (0.0418) (0.8159) (0.6497) (0.0498) (0.8218) (0.5989) (0.6486) (0.387) (0.9641) (0.9985) (0.9158) AT 0.0234 0.0788* 0.0768* 0.0363 -0.0311 0.2802* -0.0049 -0.029 0.1178* 0.0498 -0.0224 -0.0527 0.012 1 (0.5361) (0.0371) (0.0422) (0.3378) (0.411) (0) (0.8979) (0.4442) (0.0018) (0.19) (0.57) (0.1674) (0.7512) Significance level in parentheses
*p<0.05
Table 6: Non-GAAP Earnings Usage and Different Characteristics Linear Regression
This table contains the results of the test about which characteristic can affect the willingness of IPO companies to disclose non-GAAP earnings measures. NonG, NonG1, NonG2, NonG3 and NonG4 are the dependent variables while REP, Big4 and HiTech are the independent variables that we want to observe. Utility, ROA, TATO, CR, DER and AT are the control variables.
The results turn out that the reputation of lead underwriters not only positively affect the NonG1 and NonG2, but also positively affect the summary variable NonG. On the opposite side, this thesis provides no evidence of whether an IPO company is a high technology company can affect its willingness to disclose any kind of non-GAAP earnings measures. Whether an IPO company hired big4 as its IPO auditor can only positively affect its willingness to disclose “pro forma net income (loss)”.
Pr(NonG=1) Pr(NonG1=1) Pr(NonG2=1) Pr(NonG3=1) Pr(NonG4=1) Linear REP 0.0317** 0.0288** 0.0353** 0.00937 0.0165 (0.0117) (0.0110) (0.0108) (0.00790) (0.0118) Big4 0.0532 -0.0837 -0.0355 -0.00497 0.155** (0.0572) (0.0531) (0.0524) (0.0348) (0.0564) HiTech 0.0260 0.0570 -0.0859 0.0181 0.0698 (0.0460) (0.0411) (0.0453) (0.0290) (0.0467) Utility -0.160 -0.0138 0.0769 -0.0869 -0.126 (0.122) (0.113) (0.124) (0.0628) (0.109) ROA -0.00578 0.130** 0.162*** 0.00426 -0.0849 (0.0575) (0.0398) (0.0346) (0.0330) (0.0606) TATO -0.000359** -0.0000264 0.0000172 -0.0000855 -0.000341* (0.000132) (0.0000633) (0.0000514) (0.0000565) (0.000151) CR -0.00144 -0.0271*** -0.0210** -0.00127 0.0100 (0.0107) (0.00732) (0.00741) (0.00678) (0.0112) DER 0.000698*** 0.000743** -0.000122 0.0000503 0.000821*** (0.000127) (0.000272) (0.000291) (0.000101) (0.000136) AT -0.0000129 -0.00000600 0.0000221 0.0000143 -0.0000419** (0.0000154) (0.0000128) (0.0000183) (0.0000130) (0.0000144) CONS 0.473*** 0.244*** 0.300*** 0.0534 0.219*** (0.0646) (0.0550) (0.0551) (0.0342) (0.0599) adj.R-sq 0.016 0.041 0.066 -0.003 0.039 N 625 625 625 625 625
Heteroskedasticity-robust standard errors in parentheses *p<0.05, **p<0.01, ***p<0.001
Table 7: Non-GAAP Earnings Usage and Different Characteristics Logit Regression and Probit Regression
In this test, since the dependent variables are binary variables, only using linear model may not give satisfied results. As the result, we also report the results of probit probability model and logit probability model. And we find the coefficients which are significant in the linear model are still significant in either logit model or probit model.
Pr(NonG=1) Pr(NonG1=1) Pr(NonG2=1) Pr(NonG3=1) Pr(NonG4=1) Logit Probit Logit Probit Logit Probit Logit Probit Logit Probit REP 0.139** 0.0848** 0.155** 0.0916** 0.183*** 0.110*** 0.113 0.0538 0.0740 0.0449 (0.0528) (0.0319) (0.0597) (0.0346) (0.0553) (0.0331) (0.0860) (0.0431) (0.0512) (0.0313) Big4 0.219 0.135 -0.439 -0.263 -0.165 -0.100 -0.0404 -0.0106 0.667** 0.416** (0.239) (0.149) (0.273) (0.163) (0.272) (0.163) (0.430) (0.214) (0.256) (0.156) HiTech 0.117 0.0728 0.360 0.215 -0.373 -0.220 0.219 0.106 0.299 0.188 (0.199) (0.123) (0.229) (0.134) (0.214) (0.128) (0.345) (0.171) (0.201) (0.124) Utility -0.659 -0.410 -0.0800 -0.0307 0.297 0.186 -1.108 -0.479 -0.519 -0.331 (0.507) (0.308) (0.573) (0.332) (0.551) (0.323) (1.176) (0.537) (0.535) (0.320) ROA -0.0286 -0.0154 0.932** 0.544** 1.262*** 0.792*** -0.0122 -0.0162 -0.349 -0.218 (0.248) (0.153) (0.344) (0.193) (0.283) (0.169) (0.431) (0.214) (0.253) (0.155) TATO -0.00200 -0.00126 -0.0233* -0.0145* -0.00921 -0.00523 -0.0177 -0.00876 -0.00215 -0.00134 (0.00109) (0.000671) (0.0114) (0.00700) (0.0113) (0.00648) (0.0171) (0.00787) (0.00127) (0.000808) CR -0.00262 -0.00149 -0.221* -0.125** -0.159* -0.0920* 0.0113 0.00762 0.0475 0.0290 (0.0483) (0.0295) (0.0887) (0.0467) (0.0760) (0.0420) (0.0960) (0.0456) (0.0502) (0.0304) DER 0.00521* 0.00315* 0.00781 0.00452 -0.000549 -0.000355 0.000639 0.000399 0.0126 0.00765 (0.00253) (0.00137) (0.00455) (0.00257) (0.00141) (0.000940) (0.00116) (0.000705) (0.00805) (0.00475) AT -0.0000562 -0.0000349 -0.0000305 -0.0000162 0.0000943 0.0000575 0.000115 0.0000603 -0.000219* -0.000127* (0.0000650) (0.0000403) (0.0000678) (0.0000415) (0.0000813) (0.0000454) (0.0000845) (0.0000472) (0.000104) (0.0000535) CONS -0.142 -0.0823 -1.057** -0.651*** -0.858** -0.535** -2.795*** -1.580*** -1.213*** -0.755*** (0.270) (0.168) (0.329) (0.190) (0.296) (0.179) (0.472) (0.234) (0.284) (0.173) pseu. R-sq 0.0239 0.0239 0.0608 0.0617 0.0764 0.0788 0.0203 0.0194 0.0436 0.0436 N 625 625 625 625 625 625 625 625 625 625 Heteroskedasticity-robust standard errors in parentheses
*p<0.05, **p<0.01, ***p<0.001
