1
Underpricing of Initial Public Offerings (IPOs) and
Direct Public Offerings (DPOs)
Abstract
This paper examines the difference in underpricing between initial public offerings (IPOs) and direct public offerings (DPOs). The data set consists of 4,484 companies that went public between the 1st of December 2012 and the 1st of December 2017 in the U.S. For this sample I find no significant difference between the types of offering on the level of underpricing.
Keywords: IPO, DPO, initial return, underpricing.
Student: Martin van Dieten Number: 11022329
Program: Economics and Business Track: Economics and Finance Supervisor: Shivesh Changoer
Credits: 12
2
Statement of Originality
This document is written by Martin van Dieten who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
1. Introduction
An initial public offering (IPO) is an event where a company raises external capital from the public, by listing itself on the stock exchange (Berk & DeMarzo, 2013). When the price of the offering turns out higher than the offer price at the day of issuing, investors who purchases the shares can make significant returns. This scenario is called underpricing and reduces the capital raised by the issuing firm.
According to Ibbotson (1975), Baron and Holmstrom (1980) and Baron (1982) IPOs are on average underpriced. This underpricing is costly, due to underwriter fees, time and effort and results in less revenues for the issuing firm, including the shareholders (Ritter & Ibbotson, 1995). Therefore, different types of offerings exist to minimize these type of costs. One of these types is a direct public offering (DPO).
Direct public offerings are a type of offering on the stock exchange without an underwriter managing the offering (Gregoriou, 2008). DPOs have several advantages. The main advantage is that issuing firms are not forced by an underwriter to underprice the offering. Underwriters have namely the incentive to set a lower offer price, because it has a commitment of buying the leftover shares, which will provide substantial costs (Baron & Holmstrom, 1980). A second advantage of DPOs is that issuing companies can avoid paying high fees for underwriters and other commissions within the process (Gregoriou, 2008). Another advantage is that firms that are small, have insufficient economic success or have bad growth prospects have the possibility to list on the exchange via a DPO. Furthermore, companies that are rejected by underwriters to take the company public can use a DPO (Gregoriou, 2008).
To examine whether DPOs are less underpriced than IPOs I examine if there is a significant difference in underpricing between Initial Public Offerings (IPOs) and Direct Public Offerings (DPOs). The answer on this question concludes if a DPO is more advantageous than an IPO.
Data are obtained from the Thomson One database between the 1st of December 2012 and the 1st of December 2017 in the U.S stock market.
For my sample of 4,484 firms I find no significant difference in underpricing between IPOs and DPOs. This finding adds to prior research that there is no free lunch when conducting a DPO instead of an IPO and therefore, DPOs are not more advantageous even though no underwriter fees have to be paid.
Several studies have been done about the underpricing of IPOs. Studies from for example Loughran & Ritter (2004), Ibbotson (1975) and Baron & Holmstrom (1980) show that there is, on average, underpricing of initial public offerings. Nonetheless, there has been little
4
research done about DPO underpricing compared to IPO underpricing. This is because traditionally only a handful of small companies in industries such as food and biotech have gone public via DPO every year. Therefore, my research extends the research that has been done about IPO underpricing by testing if there is a significant difference in underpricing between an IPO and a DPO.
In the following section the literature and background information on IPOs are discussed. The third section contains the hypothesis and literature on DPOs. The method and data are described in section 4. The descriptive statistics, correlations, univariate and multivariate regressions are reviewed in section 5. In section 6 a sensitivity analysis is done. Section 7 contains the conclusion of the thesis.
2. Literature and background information 2.1.IPOs
An Initial Public Offering (IPO) is an event whereby a company offers its common stock for the first time on a public stock exchange (Berk & DeMarzo, 2013). Initial Public Offerings are managed by an investment banker called an underwriter, who manages the offering and outlines its structure (Berk & DeMarzo, 2013). Between the period of 1980 and 2001 there was more than 1 IPO per business day in the U.S (Ritter & Welch, 2002). However, this amount varied between a minimum of 100 and a maximum of 400 IPOs from year to year (Ritter & Welch, 2002). These IPOs had an average deal size of 78 million dollars, with a total of 488 billion dollars in gross proceeds (Ritter & Welch, 2002). According to Ritter & Welch (2002) the underpricing was on average 18.8 percent after one day of trading. Within 3 years the underpricing increased to 22.6 percent.
The two biggest advantages of IPOs, with respect to staying private, are an increase in liquidity and better access to capital (Berk & DeMarzo, 2013). Furthermore, founders and shareholders are given the possibility to convert their capital into cash at a future date (Ritter & Welch, 2002). The fact that companies increase their publicity with an IPO plays mostly a minor role (Ritter & Welch, 2002). Besides that, there are advantages when comparing an IPO to a DPO. According to Gregoriou (2008) issuing firms can signal their quality by the chosen underwriter. Another advantage is that underwriters can stake out investors or confirm the quality of the securities (Anand, 2003).
The disadvantages consist of direct and indirect costs. These direct costs include mostly legal, auditing and underwriter expenses (Ritter & Ibbotson, 1995). The indirect costs are the time and effort from conducting the IPO (Ritter & Ibbotson, 1995). Another disadvantage
5
according to Ritter and Ibbotson (1995), is the possibility for dilution of shares, because the shares are being sold to the public at, on average, a lower offering price. As mentioned before, some companies find these costs expensive. One of the types that does not incur these cost is a direct public offering, which will be elaborated in section 3.
2.2 IPO underpricing
Underpricing of an initial public offering is one of the most researched phenomena regarding IPOs, see for example Loughran & Ritter (2004), Ibbotson (1975) and Baron & Holmstrom (1980). The mentioned studies show that there is, on average, underpricing of initial public offerings. This means that the issuing firms leave money on the table (Loughran & Ritter, 2004). If there is a positive initial performance it indicates that offerings are underpriced. This means that the share price turned out greater than the offer price set by the underwriter (Ibbotson, 1975).
According to Baron and Holmstrom (1980) the main reason why IPOs are underpriced are incentives of the underwriter. Issuing companies asks underwriters to set an offer price, because the underwriter has better information on the demand of the securities and the condition of the capital market (Baron & Holmstrom, 1980). The underwriter uses this freedom to set a lower offer price, because it has a commitment of buying the leftover shares, which will provide substantial costs (Baron & Holmstrom, 1980). This way the underwriter has an incentive to limit these costs and advice an offer price below the interest of the issuer. This ensures that the shares will be sold easily and will limit the loss for the underwriter (Baron & Holmstrom, 1980).
Loughran and Ritter (2002) further investigate the conflict between the freedom of the underwriter and the issuing firm. As mentioned before, the discretion in share allocation of the underwriter is not used in the best interest of the issuer. According to Loughran and Ritter (2002) underwriters deliberately underprice the offering to sell shares to favoured buy-side clients. This to gain quid pro quos (Loughran & Ritter, 2002).
Baron (1982) supports this theory. He argues that, when the underwriter has better information regarding market demand prior to the contracting, the underwriter sets the price below the first-best offer price to limit the costs and risks for the underwriter (Baron D. P., 1982).
Benveniste and Spindt (1989) also assumes that there is asymmetric information. They argue that, when investors are more informed than the issuers underwriters can use bookbuilding to obtain information from investors. To truthfully reveal the demand of the investors, underwriters offer to underprice offerings (Benveniste & Spindt, 1989).
6
Another theory is based on the assumption that there is asymmetric information. It is assumed that issuers are better informed than investors. This causes rational investors to fear a lemons problem (Ritter & Welch, 2002). This implies that only issuing firms with a below average quality want to sell the shares at an average price (Ritter & Welch, 2002). Therefore, high-quality issuers try to signal their quality by underpricing.
There are also theories based on symmetric information. Hughes and Thakor (1992) argue that issuing firms wants to underprice the offering to reduce their legal liability. This is because overvalued IPOs have a higher chance to be sued (Hughes & Thakor, 1992). Therefore, to avoid subsequent lawsuits issuers want to leave money on the table. (Hughes & Thakor, 1992). Besides that, Boehmer and Fishe (2001) argue that the underpricing is more severe when the trading volume in the aftermarket is higher.
Ritter (1984) shows that high risk IPOs are more underpriced than low risk IPOs. Carter and Manaster (1990) find similar results. Low risk firms want to show their low risk characteristics to the public. Because prestigious underwriters want to maintain their reputation they only market IPOs of these low dispersion firms (Carter & Manaster, 1990). Therefore, prestigious underwriters are affiliated with lower risk and leave less money on the table. However, Loughran and Ritter (2001) contradicts this finding and argues that during the internet bubble prestigious underwriters left enormous amounts of money on the table.
Prior research also shows that the level of underpricing varies with other characteristics. For instance, Ritter (1991) shows that the age of the firm has a negative influence on the level of underpricing. He argues that younger firms tend to have higher initial returns and thus more underpricing. This can be explained due to the fact that older firms are less risky, so that the value of the firm can be predicted more accurately (Ritter, 1991). Ritter and Loughran (2004) argue that classification as a technological company adds to the risk composition. By increasing the risk composition, the underpricing becomes more severe. Moreover, According to Hayes (1971), larger IPOs are done by more prestigious underwriters, which have the capacity to take on larger equity offerings. Furthermore, there is less information available for smaller IPOs, which causes information asymmetry. (Lowry, Officer, & Schwert, 2010). This implies that there will be less underpricing when the deal size is larger.
7
3. Hypothesis
Following from the literature and background information it can be concluded that there is substantial evidence for underpricing of IPOs. This underpricing comes at a cost. The issuer needs to pay underwriter fees and puts in time and effort. Besides that, the incentives of the underwriter causes the stock to be underpriced. This results in less revenues for the issuing firm including the shareholders. Therefore, different types of offerings exist to minimize these type of costs. As mentioned before, one of these types is a direct public offering.
DPOs
Direct Public Offerings (DPOs), also called Direct Listings, are a type of offering on a public stock exchange without the involvement of an underwriter (Gregoriou, 2008). Companies then directly list its shares on an exchange to raise debt or equity finance (Gregoriou, 2008). The main advantage of a DPO is the independence from the underwriter. This way the incentives of the underwriter to underprice an offering also disappear. Using a DPO a company can avoid paying high transaction costs, mostly underwriter fees (Gregoriou, 2008). According to Sjostrom (2001, p. 531), an DPO can save up to 13 percent in underwriter fees and commissions within the process.
DPOs also have a number of disadvantages. First of all, a DPO could be less efficient than an IPO, because firms cannot signal their quality by the chosen underwriter. Secondly, there is no underwriter who can stake out investors or confirm the quality of the securities (Anand, 2003). However, it is not clear if the benefits from the DPO outweigh the costs.
To say something about these net benefits I am going to research if there is a difference in the level of underpricing between an IPO and a DPO. Because the issuing firm can set its own offer price and there is no underwriter that wants to keep the price as low as possible in order to avoid a possible loss, I expect that DPOs are less underpriced than IPOs. Therefore, the following hypothesis will be tested:
8
4. Research design 4.1 Method
To test the hypothesis, I use ordinary least squares (OLS) to estimate the following model: 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛 = 𝛽0+ 𝛽1∗ 𝐷𝑃𝑂 + 𝛽2∗ 𝐿𝑁𝑎𝑔𝑒 + 𝛽3 ∗ 𝐿𝑁𝑑𝑒𝑎𝑙𝑠𝑖𝑧𝑒 + 𝛽4∗ 𝐴𝑔𝑟𝑖𝑐𝑢𝑙𝑡𝑢𝑟𝑒 + 𝛽5∗ 𝑀𝑖𝑛𝑖𝑛𝑔 + 𝛽6∗ 𝐶𝑜𝑛𝑠𝑡𝑟𝑢𝑐𝑡𝑖𝑜𝑛 + 𝛽7∗ 𝑀𝑎𝑛𝑢𝑓𝑎𝑐𝑡𝑢𝑟𝑖𝑛𝑔 + 𝛽8∗ 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡𝑎𝑡𝑖𝑜𝑛 + 𝛽9∗
𝑊ℎ𝑜𝑙𝑒𝑠𝑎𝑙𝑒𝑡𝑟𝑎𝑑𝑒 + 𝛽10∗ 𝑅𝑒𝑡𝑎𝑖𝑙 + 𝛽11∗ 𝐹𝑖𝑛𝑎𝑛𝑐𝑒 + 𝛽12∗
𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝑠 + 𝐸𝑟𝑟𝑜𝑟 (1) Initial Return is the percentage difference between the closing price and the offer price (Lowry, Officer, & Schwert, 2010). The initial return is calculated by subtracting the offer price from the closing price and dividing this amount by the offer price. This results in the percentage difference, when multiplying it by 100, between the offer price and the closing price. The closing price after 180 days is taken for the calculation. When the closing price is greater than the offer price, the underpricing increases. The initial return is therefore positively related to the level of underpricing. DPO is a dummy variable which is equal to 1 when it is a DPO and zero when it is an IPO. Because I expect that the level of underpricing is lower for a DPO, the coefficient (β1) on this variable is expected to be negative.
LNage is the natural logarithm of the age of the firm at the time of the offering. It is calculated by subtracting the issue date from the date founded. The logarithm ensures a symmetric distribution of the residuals, so that it compensates for a skewed distribution. Because prior research shows that younger firms tend to be more underpriced, I expect that LNage is negatively related with the Initial Return.
Previous research also included the logarithm of the deal size (LNdealsize) as a control variable, where the deal size is expressed as the gross proceeds from the offering (Carter, Dark, & Singh, 1998). It is the amount of money the firm brings in for selling the shares. Again, the logarithm is used to compensate for a skewed distribution. Because prior research suggests that there will be less underpricing when the deal size is larger, I expect a negative relation between LNdealsize and the Initial Return (Lowry, Officer, & Schwert, 2010).
Agriculture is a dummy variable that equals 1 if the issuing firm is from the agriculture, forestry or fishing sector and 0 otherwise. If the primary SIC code is between 01 and 09 the firm is indeed from the agriculture, forestry or fishing sector. Mining is a dummy variable that equals 1 if the issuing firm is from the mining sector and 0 otherwise (SIC code between 10 and
9
14). Construction is a dummy variable that equals 1 if the issuing firm is from the construction sector and 0 otherwise (SIC code between 15 and 17). Manufacturing is a dummy variable that equals 1 if the issuing firm is from the manufacturing sector and 0 otherwise (SIC code between 20 and 39). Transportation is a dummy variable that equals 1 if the issuing firm is from the transportation, communications, electric, gas or sanitary service sector and 0 otherwise (SIC code between 40 and 49). Wholesaletrade is a dummy variable that equals 1 if the issuing firm is from the wholesale trade sector and 0 otherwise (SIC code between 50 and 51). Retail is a dummy variable that equals 1 if the issuing firm is from the retail trade sector and 0 otherwise (SIC code between 52 and 59). Finance is a dummy variable that equals 1 if the issuing firm is from the finance, insurance or real estate sector and 0 otherwise (SIC code between 60 and 67). Services is a dummy variable that equals 1 if the issuing firm is from the services sector and 0 otherwise (SIC code between 70 and 89). If all sector dummies are zero, then it resembles the 10th sector public administration, where the primary SIC code is 99. These control variables are added to correctly estimate the effect of my variable of interest (DPO) on the dependent variable (Initial Return). This is because control variables reduce the effect of omitted variable bias.
4.2.Data
For this research, I focus on initial public offerings and direct public offerings from the U.S stock market between the 1st of December 2012 and the 1st of December 2017 are used. The U.S stock market is chosen, because it is the biggest stock market in the world and has therefore the most listings on the stock exchange. This time period is chosen because it contains recent data and does not include recessions and booms, which may confound the results. The data set is collected from the Thomson One database. TheThomson One database includes: The name of the company, primary SIC code, issue date, date founded, issue type, offer price, closing price, number of underwriters, name of the underwriter and the gross proceeded amount. Offerings without an identifier and offerings with one of the above variables missing were deleted. Besides that, SIC codes which started with 18, 19, 68, 69, 98 and 99 are not assigned to any industry and are therefore deleted. Because of this and some missing information, selection bias can occur. In summary, the final sample consists of 4,484 companies. In table 1 I give the descriptive statistics.
The average value of Initial Return from the selected sample is 40.56%. This suggests that there is on average an increase of the share price of 40.56% between the first day offer price and the closing price after 180 days.
10
logarithm of the age to the age without the logarithm, I find that the average value is 17.832. This means that the age of the firm at the time of the offering was on average 17.832 years.
The average value of LNdealsize from the selected sample is 3.749. Notice that the minimum value is -4.605. When converting the logarithm of the deal size into the deal size without the logarithm, I find that the average value is 175.394 and the minimum value is 0.010. This suggests that the gross proceeds from the offering are on average 175.394 million dollars and have a minimum value of 1000 dollars. Because DPOs are mostly small tech firms, the sample is not filtered for a minimum offer price or minimum offer value. The average value of DPO is 0.229. This means that there are more IPOs than DPOs in my sample which makes sense. In total there are 1,026 non-underwritten firms and 3,458 underwritten firms.
In table 2, I show the breakdown of my sample by industry. 38.4% of the firms operate in the manufacturing sector (Manufacturing). 21.4% of the firms operate in finance, insurance or real estate sector (Finances). The Manufacturing and Finances variables have an average value of 0.396 and 0.214 respectively. This means that the most issuing firms from the selected sample operates in these industries. Besides that, 0.2% of the firms operate in the agriculture, forestry or fishing sector (Agriculture). The average value of Agriculture is 0.002. This means that the fewest issuing firms from the selected sample operates in this industry. The 9 industries consist of 4,452 of the 4,484 observations. This implies that the omitted category public administration consist of the remaining 32 observations.
11 Table I
Descriptive Statistics
The descriptive statistics includes 4,484 offering observations from the 1st of December 2012 until the 1st of December 2017.
Data is collected from the Thomson One database. The sample is not filtered for a minimum offer price or minimum offer value. Initial Return is the percentage difference between the closing price and the offer price after 180 days. DPO is a dummy variable which is equal to 1 when it is a DPO and zero when it is an IPO. LNage is the natural logarithm of the age of the firm at the time of the offering. LNdealsize is the natural logarithm of the gross proceeds from the offering. Agriculture, Mining, Construction, Manufacturing, Transportation, Wholesaletrade, Retail, Finance and Services are dummy variables that equal 1 if the issuing firm is from the mentioned sector and 0 otherwise.
Variable Obs. Mean Std.Dev. Min Max
Initial Return (%) 4,484 40.561 607.391 -430.000 21150.000 DPO 4,484 0.229 0.420 0.000 1.000 LNAge (years) 4,484 2.354 1.149 -2.162 4.751 LNdealsize ($ millions) 4,484 3.749 2.190 -4.605 8.937 Agriculture 4,484 0.002 0.045 0.000 1.000 Mining 4,484 0.086 0.281 0.000 1.000 Construction 4,484 0.009 0.095 0.000 1.000 Manufacturing 4,484 0.396 0.489 0.000 1.000 Transportation 4,484 0.060 0.238 0.000 1.000 Wholesaletrade 4,484 0.019 0.137 0.000 1.000 Retail 4,484 0.052 0.222 0.000 1.000 Finance 4,484 0.214 0.410 0.000 1.000 Services 4,484 0.171 0.377 0.000 1.000 Table II
Observations per industry
This table consists for each industry in the sample the number of offerings and the respective percentage, with respect to the total offerings. The sample contains 4,484 offerings from the 1st of December 2012 until the 1st of December 2017. Data is
collected from the Thomson One database.
Number of offerings Percentages
Total Agriculture Mining Construction Manufacturing Transportation Wholesaletrade Retail Finances Services 4,484 9 386 41 1724 270 86 230 958 74 100% 0.2% 8.6% 0.9% 38.4% 6.0% 1.9% 5.1% 21.4% 16.7%
12
5. Results
5.1 Correlations
First the Pearson correlation coefficients are estimated, to see if there are variables highly correlated which each other. These coefficients represent the level that variables are associated with each other and in which direction they move together. The value will be between -1 and 1, where 0 stands for no correlation and (-)1 stands for perfect correlation. The results are shown in Table 3 below.
Table III
Correlation Coefficients
The correlation coefficients are taken from a sample that includes 4,484 offering observations from the 1st of December 2012
until the 1st of December 2017. Data is collected from the Thomson One database. The star indicates that the correlations are
13
According to the research of Farrar and Glauber (1967) correlations should be smaller than 0.8, otherwise multicollinearity could exist. Looking at the table above, none of the variables has a correlation with another variable above 0.8 and is significant at the 5 percent level. The highest correlation is between LNdealsize and DPO, namely -0.683, which is also significant at the 5 percent level. According to the research, this should not be a problem for multicollinearity. However, a VIF test is done to exclude any doubts. This will be done in the next paragraph.
The correlation coefficient between LNdealsize and the main explanatory variable DPO makes sense. There is a strong negative linear relation between the two variables. It implies that the LNdealsize is lower when an issuing company goes public through a DPO instead of an IPO. Because DPOs are mostly done by small firms, the LNdealsize could also be lower. Therefore, it makes sense that they are negatively related.
From the correlation matrix above it can be concluded that there is a small negative relation between LNdealsize and LNage. This is against the findings of Rossotto (2010), who describes that the deal size is higher for older firms, because of longer operation and finance history. However, the correlation of -0.036 is not significant at the 5 percent level.
Furthermore, there is a small positive relation between the main explanatory variable DPO and LNage. This finding can be explained by the fact that companies that perform an DPO are more independent and more experienced and therefore older. However, the correlation of 0.055 is not significant at the 10 percent level.
Although the correlations are small, I perform a variance inflation factor (VIF) test to examine if there is multicollinearity. Multicollinearity arises when two dependent variables are strongly correlated with each other. The number indicates how much of the variance is increased by the collinearity (O'Brien, 2007). The results can be found in table 4 below.
14
Table IV
Variance inflation factor
The variance inflation factors of the variables are taken from a sample that includes 4,484 offering observations from the 1st of
December 2012 until the 1st of December 2017. Data is collected from the Thomson One database.
Variable VIF 1/VIF
Manufacturing 29.501 0.034 Finances 20.987 0.048 Services 17.994 0.056 Mining 10.435 0.096 Transportation 7.676 0.130 Retail 5.753 0.174 Construction 2.088 0.479 LNdealsize 1.980 0.505 DPO 1.907 0.524 Wholesaletrade 1.640 0.610 Agriculture 1.243 0.805 LNAge 1.015 0.985 Mean_VIF 8.518 0.985
Looking at the table above, it can be concluded that there are 4 variables that have a value of 10 or more. This implies that there is strong evidence for multicollinearity. However, the correlation between these dummies are all below 0.8. Furthermore, these variables are dummy control variables and not dependent or main explanatory variables. Therefore, these variables are not excluded from the regression.
5.2 Regression results
The results of the OLS regression on the Initial Return are presented in table 5. Table 5 shows the results of the regressions. The first regression in this table is an univariate regression between Initial Return and the main explanatory variable DPO. In the second multivariate regression the control variables LNage and LNdealsize are added to the first regression. The third and last regression is between the independent, the dependent and all the control variables.
15
Table V Regressions
The summary statistics includes 4,484 offering observations from the 1st of December 2012 until the 1st of December 2017.
Data is collected from the Thomson One database. The sample is not filtered for a minimum offer price or minimum offer value. Initial Return is the percentage difference between the closing price and the offer price after 180 days. DPO is a dummy variable which is equal to 1 when it is a DPO and zero when it is an IPO. LNage is the natural logarithm of the age of the firm at the time of the offering. LNdealsize is the natural logarithm of the gross proceeds from the offering. Agriculture, Mining, Construction, Manufacturing, Transportation, Wholesaletrade, Retail, Finance and Services are dummy variables that equal 1 if the issuing firm is from the mentioned sector and 0 otherwise. P-values are computed using heteroscedasticity-consistent standard errors. These standard errors are clustered by the primary SIC codes, *, **, and *** indicate significance at 5%, 1%, and 0.1%, respectively
(1) (2) (3)
Initial Return Initial Return Initial Return
DPO 1.257*** 0.283 0.260 (0.359) (0.192) (0.186) LNAge -0.264 -0.247 (0.149) (0.141) LNdealsize -0.285* -0.319* (0.112) (0.129) Agriculture -2.187 (1.242) Mining -2.133 (1.176) Construction -2.023 (1.143) Manufacturing -2.155 (1.162) Transportation -1.561 (1.151) Wholesaletrade -0.626 (0.782) Retail -0.667 (1.113) Finances -1.743 (1.150) Services -1.909 (1.161) Intercept 0.118*** 2.030* 4.024* (0.025) (0.834) (1.595) N 4,484 4,484 4,484 R2 0.008 0.016 0.019
Standard errors in parentheses
16
In the first, second and third regression, I find a positive coefficient on DPO, which suggests that the level of underpricing is greater for DPOs. This finding does not support the hypothesis that the level of underpricing is greater for IPOs. However, in both multivariate regressions the positive coefficient is not significant at the 5 percent level.
I find, as expected and consistent with the research of Lowry, Officer and Schwert (2010) a negative coefficient on control variable LNage. This finding suggests that the level of underpricing is greater for younger firms. However, this finding is not significant at the 5 percent level.
The coefficient on LNdealsize is significantly negative. This finding suggests that the level of underpricing is lower when the gross proceeds from the offering are higher. This is consistent with prior research of Hayes (1971). He explains that larger IPOs are done by more established underwriters and have therefore less underpricing.
The coefficients on Agriculture, Mining, Construction, Manufacturing, Transportation, Wholesaletrade, Retail, Finance and Services are all negative. This finding suggests that the level of underpricing is lower when the issuing firm is from one of the mentioned sectors, relative to the omitted sector public administration. However, the coefficients are not significant at the 5 percent level.
In table 5 the R2 is measured for all the multivariate regressions. This coefficient only increases if an added variable improves the regression. It can be concluded from table 5 that the coefficient of determination increases from the first until the third regression. The R2 is the highest for the last multivariate regression where the independent, dependent and all control variables are added and has a value of 0.019. This is in line with other theses, which have also a R2 below 0.1. However, the main goal of this thesis was not to explain the total level of underpricing, but to research the effect of the type of offering on the level of underpricing. Therefore, a low R2 is not a problem
With previous discussed results, the hypothesis that there is significantly less underpricing for DPOs than for IPOs in the U.S stock market can be answered. The results show that there is no significant negative effect of the offering dummy on the level of underpricing at the 5 percent significance level. However, there is a small positive relation which is not significant in the multivariate analysis but significant in the univariate analysis. This is against the expectations mentioned before and therefore, the null hypothesis is not rejected at the 5 percent significance level. At last, the model fits best at the third multivariate regression where all the variables are added.
17
6. Sensitivity analysis
A sensitivity analysis is performed to examine if there is a difference of the coefficient DPO on the level of underpricing. In the previous regression the Initial Return was measured by the percentage difference between the closing price and the offer price (Lowry, Officer, & Schwert, 2010). Here the closing price after 180 days was taken for the calculation. However, for the sensitivity analysis a closing price after 90 days is taken, to examine if the coefficient DPO changes. The data consist again of 4,484 offerings from the U.S stock market between the 1st of December 2012 and the 1st of December 2017.
The results of the OLS regression on the Initial Return are presented in table 6. Table 6 shows the results of the regressions. The first regression in this table is an univariate regression between Initial Return and the main explanatory variable DPO. In the second multivariate regression the control variables LNage and LNdealsize are added to the first regression. The third and last regression is between the independent, the dependent and all the control variables.
In all the regressions, I find again a positive coefficient on DPO, which suggests that the level of underpricing is greater for DPOs. This finding does not support the hypothesis that the level of underpricing is greater for IPOs. However, in all regressions the positive coefficient is not significant at the 5 percent level. Therefore, there is no difference of the main explanatory variable DPO on the level of underpricing, when the dependent variable Initial Return changes.
The coefficients on LNage and LNdealsize are again negative. This finding suggests, that the level of underpricing decreases when the firms are older of have higher gross proceeds from the offering. However, in the sensitivity analysis these coefficients are not significant at the 5 percent level.
The coefficients on Agriculture, Mining, Transportation, Wholesaletrade, Retail, Finance and Services are all positive. This finding suggests that the level of underpricing is higher when the issuing firm is from one of the mentioned sectors, relative to the omitted sector public administration. The coefficients on Construction and Manufacturing are negative. This finding suggests that the level of underpricing is lower when the issuing firm is from one of the mentioned sectors, relative to the omitted sector public administration. However, all the coefficients are not significant at the five percent level.
It can be concluded from table 6 that again the coefficient of determination increases from the first until the third regression. The R2 is the highest for the last regression where the independent, dependent and all control variables are added and has a value of 0.013.
18
Table VI Sensitivity Analysis
The summary statistics includes 4,484 offering observations from the 1st of December 2012 until the 1st of December 2017.
Data is collected from the Thomson One database. The sample is not filtered for a minimum offer price or minimum offer value. Initial Return is the percentage difference between the closing price and the offer price after 90 days. DPO is a dummy variable which is equal to 1 when it is a DPO and zero when it is an IPO. LNage is the natural logarithm of the age of the firm at the time of the offering. LNdealsize is the natural logarithm of the gross proceeds from the offering. Agriculture, Mining, Construction, Manufacturing, Transportation, Wholesaletrade, Retail, Finance and Services are dummy variables that equal 1 if the issuing firm is from the mentioned sector and 0 otherwise. P-values are computed using heteroscedasticity-consistent standard errors. These standard errors are clustered by the primary SIC codes, *, **, and *** indicate significance at 5%, 1%, and 0.1%, respectively
(1) (2) (3)
Initial Return Initial Return Initial Return
DPO 0.318 0.183 0.161 (0.262) (0.150) (0.146) LNAge -0.0990 -0.0991 (0.0919) (0.0925) LNdealsize -0.0420 -0.0464 (0.0385) (0.0401) Agriculture 0.792 (0.830) Mining 0.0193 (0.121) Construction -0.0824 (0.147) Manufacturing -0.0674 (0.116) Transportation 0.0605 (0.131) Wholesaletrade 0.0645 (0.0843) Retail 0.0218 (0.115) Finances 0.0178 (0.121) Services 0.302 (0.368) Intercept 0.114*** 0.536 0.524 (0.0109) (0.384) (0.349) N 4,484 4,484 4,484 R2 0.007 0.012 0.013
Standard errors in parentheses
19
7. Conclusion
I examine the difference in underpricing between DPOs and IPOs. My prediction is the level of underpricing is greater for IPOs. To test this hypothesis, I examine a final sample of 4,484 companies from the U.S stock market between the period of 1st of December 2012 and the 1st of December 2017.
For testing the relation between the type of offering and the Initial Return, the descriptive statistics and pearson correlation coefficients were computed at first. After that, univariate and multivariate OLS regression are done to answer the main question.
I find a positive but insignificant relation between the Initial Return and the DPO dummy. Hence, I cannot conclude that the level of underpricing is greater for IPOs. These findings suggest that DPOs are not more advantageous than IPOs and that no free lunch can be obtained from conducting a different type of offering. This can, for example, be explained by additional transaction costs that are incorporated in the offer price due to offering without an underwriter.
Limitations of the research are possible selection bias and outliers due to limited availability of DPOs. Besides that, the sample focuses on only one country and has a short sample period.
A suggestion for further research is to analyse additional effects that differ between IPOs and DPOs that could have an influence on the level of underpricing, like the additional transaction costs. At last, research between different time periods is suggested.
20
References
Anand, A. I. (2003). The Efficiency of Direct Public Offerings. Journal of small and
emerging business law, 433-464.
Baron, D. P. (1982). A Model of the Demand for Investment Banking Advising and Distribution Services for. The Journal of Finance, 955-976.
Baron, D., & Holmstrom, B. (1980). Abstract: The Investment Banking Contract for New Issues Under Asymmetric Information: Delegation and the Incentive Problem. The
Journal of Financial and Quantitative Analysis, 1115-1138.
Benveniste, L. M., & Spindt, P. A. (1989). How investment bankers determine the offer price and allocation of new issues. Journal of Financial Economics, 343-362.
Berk, J., & DeMarzo, P. (2013). Corporate Finance. Boston: Pearson Education Limited. Brau, J. C. (2006). Initial Public Offerings: An Analysis of Theory and Practice. The Journal
of Finance, 399-436.
Carter, R. B., Dark, F. H., & Singh, A. K. (1998). Underwriter Reputation, Initial Returns, and the Long-Run Performance of IPO Stocks. The Journal of Finance, 285-311.
Carter, R., & Manaster, S. (1990). Initial Public Offerings and Underwriter Reputation. The
Journal of Finance, 1045-1067.
Farrar, D. E., & Glauber, R. R. (1967). Multicollinearity in Regression Analysis: The Problem Revisited. The Review of Economics and Statistics, 92-107.
Gregoriou, G. N. (2008). Encyclopedia of Alternative Investments. London: Taylor & Francis group.
Hayes, S. L. (1971). Investment banking: power structure in flux. Harvard Business Review, 136-152.
Hughes, P. J., & Thakor, A. V. (1992). Litigation risk, intermediation, and the un derpricing of initial public offerings. Review of Financial Studies, 709-742.
Ibbotson, R. (1975). Price Performance of Common Stock New Issues. Journal of Financial
Economics, 235-272.
Loughran, T., & Ritter, J. (2004). Why Has IPO Underpricing Changed over Time? Financial
Management, 5-37.
Loughran, T., & Ritter, J. R. (2002). Why don't issuers get upset about leaving money on the table in IPOs? Review of Financial Studies, 413-443.
Lowry, M., Officer, M. S., & Schwert, G. W. (2010). The Variability of IPO Initial Returns.
The Journal of Finance, 40.
O'Brien, R. M. (2007). A Caution Regarding Rules of Thumb for Variance Inflation Factors.
21
Ritter, J. R. (1984). The "Hot Issue" Market of 1980. The Journal of Business, 215-240. Ritter, J. R. (1991). The Long-Run Performance of Initial Public Offerings. The Journal of
Finance, 3-27.
Ritter, J. R., & Ibbotson, R. G. (1995). Handbooks in Operations Research and Management
Science. Elsevier BV.
Ritter, J. R., & Welch, I. (2002). A Review of IPO Activity, Pricing, and Allocations. The
Journal of Finance, 1795-1828.
Sjostrom, W. K. (2001). Going Public through an Internet Direct Public Offering: A Sensible Alternative for small companies. Florida law review, 530-569.
