The relationship between the underpricing of seasoned equity offerings and the research & development expenditures

The relationship between the underpricing of

seasoned equity offerings and the research &

development expenditures

Bachelor Thesis

University of Amsterdam Economics and Business

Author: Yorick Hendricx Student number: 10751246 Hand-in date: 31-01-2018

Abstract

Seasoned equity offerings have been broadly researched by several researchers in the past. Corwin (2003) found a positive relationship between the underpricing and the proxy’s information asymmetry. This thesis will focus on a specific source of information asymmetry namely, the research and development expenditures. The seasoned equity offerings examined are between 2009 and 2013. Furthermore, the R&D intensity is split up in three different groups in order to explain the relationship between the underpricing and R&D intensity. This research finds differences between the cumulative abnormal returns for companies with and without R&D expenditures.

Index

1. INTRODUCTION ... 4

2. LITERATURE REVIEW ... 5

2.1SHORT-TERM UNDERPRICING AND R&D ... 5

2.3HYPOTHESIS ... 6

2.3.1SHORT-TERM UNDERPRICING AND R&D HYPOTHESIS ... 6

3. METHODOLOGY ... 8

3.1EVENT STUDY ... 8

3.1.1MARKET MODEL USED TO ESTIMATE THE BETA ... 9

3.2REGRESSION VARIABLES ... 9

3.2.1REGRESSION AND STATISTICAL HYPOTHESIS ... 12

3.3DATA ... 13

4. RESULTS... 15

4.1CUMULATIVE ABNORMAL RETURNS ... 15

4.2REGRESSION WITHOUT R&D DUMMIES... 16

4.3REGRESSION WITH DUMMIES... 17

5. CONCLUSION AND DISCUSSION ... 18

6. REFERENCE LIST ... 19

1. Introduction

There exists extensive research on the stock market. Two of the specific areas of interest within this are initial public offerings and seasoned equity offerings, which will be referred to as IPOs and SEOs from hereon out. An IPO is the process of selling a stock to the public for the first time (Berk & Demarzo, 2014). Any other process of selling stocks to the public after the initial

offering is referred to as an SEO. IPOs and SEOs both improve companies’ ability to access and create greater liquidity for investors.

Research regarding SEOs discovered two phenomena: short-term underpricing and long-term underperformance (Corwin, 2003; Loughran & Ritter, 1995). Over time, researchers have discovered various explanations for both phenomena. For example, short-term underpricing is partially due to information asymmetry (Myers & Majluf, 1984; Carlson, Fisher & Giammarino., 2006). Despite extensive research analyzing the determinants of both phenomena, neither has been completely explained (Guo, Lev and Shi, 2006).

This paper will focus on a major contributor of SEO uncertainty namely, the research and development activities of issuers (Guo et al.2006). R&D expenditures are the most extensively researched intangible investments in the fields of economics, finance and accounting (Guo et al. 2006). The reason for the study on R&D expenditures is the fact that these expenditures have to be disclosed in corporate financial reports. While Guo et al. (2006) found a positive relationship between the R&D activities and short-term underpricing for IPOs together with lower abnormal returns for the long-term performance for R&D-intensive companies compared to no-R&D companies, this thesis will seek to uncover the same evidence for SEOs. This leads to the following research question:

How does the R&D-intensity of a company influence short-term SEO underpricing?

The importance of identifying the source of SEO uncertainty can eventually help to reduce the uncertainty. Once the source of information asymmetry associated with the short-term underpricing is identified, several stakeholders of the company can act in order to reduce the asymmetry. Managers, for example, might want to disclose the nature of the R&D activities to reduce information asymmetry. Investors, on the other hand, will be motivated to do extensive research on the nature of the research and development activities to pinpoint the benefits of the se

activities. Lastly, it is important to identify the source of SEO uncertainty, because once it is known to the it enhances the market efficiency public (Guo et al. 2006).

This thesis is organized as follows: the next section will discuss previous literature on SEOs in general and their short-term underpricing. Furthermore, the relevance of R&D activities and the relationship with SEO underpricing will be illustrated. This will be followed by the methodology, the results and a conclusion.

2. Literature Review

This section focusses on previous SEO research, and takes a more in depth look at SEO underpricing. The relationship between R&D intensity and the short-term SEO underpricing explained. Lastly, the hypothesis for this thesis will be presented.

2.1 Short-term underpricing and R&D

Whereas IPO underpricing has been researched extensively, SEO pricing has gotten less attention until Smith’s research on the topic (1977). Smith (1997) was the first researcher to document the significant underpricing for SEOs. In a sample of 328 exchange listed firms that season new issues, it is found that the average close on the offer date returns are -0,0054 and the average return from the offer price to the close on the offer date is are 0,0082. In addition, Eckbo and Masulis (1992) report that the offer price for industrial firms is set at an average discount of 0,00044 from the closing price. Corwin (2003) also finds an increasing underpricing over time in his sample of seasoned equity offerings during 1980s and 1990s. In the period 1980-1989, an average underpricing of 0,0115 was found in his sample, whereas in the period 1990-1998 the average underpricing is 0,0292. This results in an average underpricing of 0,0022 during the 1980s and 1990s. Corwin (2003) finds a positive relationship between the underpricing and size and price uncertainty. The paper also finds evidence for a reliable relation between underpricing and the proxies for information asymmetry, such as firm size and the bid-ask spread.

Furthermore, Berk and Demarzo (2014) conclude that on average, the market greets the news of an SEO with a decline in stock price. They explain this using adverse selection principle, introduced by Akerlof (1978). Berk and Demarzo (2014) state that managers that know their firms’ securities are overvalued do not want to sell their securities. In contrast, if managers are aware of the undervaluation of their firm’s securities, they do want to sell the securities. Investors

are aware of this and due to adverse selection, investors will only pay a low price for the securities. This issue can be problematic for companies that want to issue equity. The costs to raise equity increases since investors are only willing to pay a discount to account for the possibility of bad news that is unknown to the investors at the time of issuance. Above example provide evidence for information asymmetry to be a source for short-term underpricing for a SEO, where managers tend to have more firm specific information than investors (Myers & Majluf, 1984).

The empirical research & development literature identified two phenomena that are important for this thesis. Firstly, R&D contributes to information asymmetry. Aboody and Lev (2000) documented that insiders of R&D-intensive companies are privy to the nature and

progress of the R&D projects. They find that from 1985 to 1997, insider gains in R&D-intensive companies are substantially larger than the insider gains for companies with low or no R&D expenditures. They point out that all corporate investments create information asymmetries because managers observe the productivity of the individual assets, but investors only get the aggregated productivity at different points in time and not continuously like managers do. However, they argue that the extent of information asymmetry associated with R&D is larger than the information asymmetry associated with tangible or financial assets. Aboody and Lev (2000) show the difference between information asymmetry for tangible and intangible assets. The failure in unique events of R&D activities will not always be shared with the public, but the depreciation of a companies’ real-estate is publically observable. This implies that information asymmetry associated with R&D activities is larger than the information asymmetry for tangible or financial assets.

2.3 Hypothesis

This part of the thesis will discuss the expectations for the outcome of the proposed regressions in the next section.

2.3.1 Short-term underpricing and R&D hypothesis

Several studies found that SEOs are underpriced. This short-term underpricing phenomenon is partially attributed to information asymmetry (Corwin, 2003). Corwin (2003) found that there is a positive relationship between information asymmetry and short-term underpricing. It is also known that the extent of information asymmetry associated with R&D is larger than information

asymmetry associated with tangible or financial assets (Aboody and Lev, 2000). Therefore, it is expected that R&D-intensive companies face a higher intensity for underpricing compared to no-R&D companies. Additionally, since information asymmetry partially contributes to both IPO and SEO short-term underpricing, the insights of Guo et al. (2006) provide valuable insights for the relationship between R&D-intensity and the magnitude of short-term underpricing (Corwin, 2003; Ritter & Welch, 2002). In line with the hypothesis, this paper also finds that the magnitude of underpricing is greater for R&D-intensive companies than for no-R&D companies that an issue an SEO.

3. Methodology

In the following session, there will be an in depth analysis of the methods used in this thesis. There will be an explanation about the even study, data and regressions. The accounting data is collected from CRSP and Compustat. The SEO information is collected from the Thomson one database.

3.1 Event study

Since SEOs do not occur all simultaneously, an appropriate estimation method must be used. The estimation method used in this thesis is the event study. An event study solves the

non-simultaneous occurrence of SEOs and exists out of two windows, the estimation window and the event window. The estimation window is assumed as the time period in a “normal” environment, whereas the event window is the period in which the effect of the event is expected to occur. The estimation window is chosen prior to the event window and has a length of 1 year. The event window starts two months prior the event date and ends 5 days after the issue date. In this thesis, the event date is the issue date from the Thomson one database. The announcement date is not used as the event date since this information was not available in the Thomson one database. However, to capture the announcement effect, the event window begins two months prior to the issue date. It is assumed that every month includes the same number of trading days and exists out of 4 weeks, where weekends are excluded. This leads to 252 yearly trading days. Dividing the yearly trading days by 12 months and rounding them for the sake of simplicity leads to 21 trading days per month. It was previously mentioned that the estimation window has a length of one-year prior the start of the even window which leads to an estimation window ranging from -274 to -22. The event window will range from -21 to +10.

3.1.1 Market model used to estimate the beta

The research of Guo et al. (2006) serves as a guideline paper for this thesis. They use cumulative abnormal returns to explain short-term underpricing. In order to compute the cumulative

abnormal returns, one needs to predict the normal expected returns. This will be done using the capital asset pricing model (Fama and French, 2004). The data from the estimation window, together with the capital asset pricing model, will compute the alpha and beta which eventually will be used for the event window.

E(Ri) = α + β ∗ E(Rm) where i = 1, . . . . , N

Where E(Ri) is the expected return of firm i that will be predicted by this model, Rm is the return of the market portfolio. The standard and Poor’s index (S&P 500) is used as market portfolio since all SEOs take place in the United States. The most used market proxy for the United States is the S&P500 and therefore this index is used in this thesis. The alpha’s and beta’s will be estimated over time.

3.2 Regression variables

In this section, the used regression will be explained and motivated. The previous section explained that an event study will be used in this thesis. In order to determine whether there exists a difference between the returns ex and ex-ante the event, the cumulative abnormal returns have to be created. The cumulative abnormal returns are created as follows:

CARi= ∑(Ract , it− Rpred , it

n t=1

)

The CAR is the cumulative abnormal return and serves as dependent variable. Where, Ract, it is the return of stock i at time t and Rpred is the predicted normal return using the CAP-model of stock i at time t. Since the CAR is computed for the event window consisting of 252 days, it follows from the summation of all of the abnormal returns in the event window per company.

Main independent variable

RD_SALES = ∑ (𝑇𝑜𝑡𝑎𝑙 𝑅&𝐷 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒 𝑖𝑡 𝑇𝑜𝑡𝑎𝑙 𝑆𝑎𝑙𝑒𝑠 𝑖𝑡 )

n t=1

The RD_SALES variable represents the R&D over sales ratio of the fiscal year prior to the event date. Thus for the fiscal year of the event date, t, the R&D over sales ratio will computed from the end year values of R&D expenditures and Sales in t-1. This is done because it is expected that the most recent financial reports will be used by investors to value the stock price. This main independent variable will be split up in three dummy variables: high-R&D, low-R&D and no-R&D, which will be explained in the following section.

Control variables

For the control variables, the article of Fama and French (1992) is used as a guideline. This article finds significant evidence for the cross-sectional variation in average stock returns. Therefore, it is expected that the following variables control for the variance in the cumulative abnormal returns. The following variables summed up are: leverage, book-to-market ratio, size and earning-price ratios. Furthermore, industry and R&D dummies will be added.

The first control variable is leverage:

Leveragei = ∑ (𝐴𝑠𝑠𝑒𝑡𝑠𝑖𝑡 𝐵𝑉𝑖𝑡 )

n t=1

Where leverage is the book value of leverage for every firm i, in year t-1. The assets are the total asset value for every firm i at the end of the fiscal year t-1, coming from the Compustat database. The BV is the book value of equity per firm i at the end of the fiscal year t-1. The BV is

computed as follows: Common equity minus deferred tax on the balance sheet for firm i and at the end of the fiscal year t-1.

The second control variable is the book-to-market ratio:

Book − to − market ratioi = ∑ (BVit

MVit)

n t=1

The book-to-market ratio is the book-to-market value for every firm i in the year t-1. The book value is referred to as the book value of equity per firm i at the end of the fiscal year t-1 and computed as follows: common equity minus deferred tax on the balance sheet for firm i and at the end of the fiscal year t-1. The market value is referred to as the end fiscal year, t-1 market

capitalization for every firm i and computed as follows: the amount of end fiscal year shares outstanding multiplied by the end of the year closing price.

The third control variable is size:

Sizei= ∑(Shares outstandingit∗ closing priceit)

n t=1

The size is the size value for every firm in period t-1. The shares outstanding are referred to as the end fiscal year number of shares outstanding for firm i at the end of the fiscal year t-1. The closing price is referred to as the end fiscal year price of the common stock for firm i at the end of the fiscal year t-1.

The fourth control variable is the earnings-to-price ratio: E P= ∑ ( Earnings it Price it ) n t=1

The earnings represent the earnings of firm i at the end of the fiscal year t-1 and price represents the closing price of firm i at the end of fiscal year, t-1.

Where Earnings are computed as follows:

Earnings = Income Before Extraordinary Items + Income Taxes Deferred − Dividends Preferred

The last control variables are industry and R&D dummies.

The industry dummies are to indicate the different outcomes for underpricing per industry. Guo et al. (2006) argue that there is a difference between the disclosure of R&D expenditures per

industry. The pharmaceutical industry, for example, will show provide more information about the nature of the R&D expenditures and therefore it is expected that investors in these companies experience less information asymmetry. In table 3.2.1.1 in the appendix, the sic industries are tabulated.

Lastly, the main independent variable is split up into three different dummies: high-R&D, low-R&D and no-R&D. Companies that do not have R&D expenditures have a corresponding dummy no-R&D. The distinction created for high and low-R&D firms is based on the median. Companies above or equal to the median of RD-sales ratio have high-R&D is equal to 1 and companies below the median of RD-sales ratio have low-R&D is equal to 1. The median is: 0.120373

3.2.1 Regression and statistical hypothesis

In order to find the effect of R&D expenditures on the cumulative abnormal returns following an SEO, the below regression is used. To conclude that the R&D expenditures have an effect on the cumulative abnormal returns the following regression will be:

CAR = β0+ β1∗ log (RD_Sales) + β2∗ log(BTM) + β3∗ log(LEV) + β4∗ log (SIZE) + β5∗E

P + β6∗ Ind_dummies

The second regression will divide the R&D – sales ratio into the three different dummy variables as mentioned in section 3.2.

CAR = β0+ β1∗ HighRD + β2 ∗ low_RD + β3∗ log(BTM) + β4∗ log(LEV) + β5∗

log (SIZE) + β6∗E_P+ β7∗ Ind_dummies

In order to conclude that there is an effect of R&D on the cumulative abnormal returns, the following 0 hypothesis has to be rejected:

H1: β1≠ 0

The following hypotheses have been conducted in order to highlight the difference between the underpricing for high-R&D and low-R&D companies:

𝐻0: 𝛽1 = 𝛽2= 0

𝐻1: 𝛽1≠ 0 𝑜𝑟 𝛽2≠ 0

If this paper finds that the 0 hypothesis can be rejected, the effect of the relative R&D expenditures can be explained. It is important to find whether the coefficient of 𝛽1 or 𝛽2 is positive or negative in order to explain the effect of R&D intensity on the cumulative abnormal return and thus the underpricing for SEOs.

3.3 Data

The sample consists of SEOs during the period between 2009-2013, and the event dates are obtained from the Thomson One database. The period 2009-12/31/2013 was chosen because it is expected that the event window will have little to no influence from the financial crisis in 2009, the year prior. The stock and accounting data were obtained from the CRSP and Compustat database.

In order to be able to work with the data obtained from the databases, the data have to meet all of the following criteria: (1) The SEO has to be a follow-on offering as in the Thomson One database, (2) the offering needs to consist entirely of common stock, (3) the offering cannot be too small, therefore the proceeds of the offering should exceed 1 million dollar, and finally, (4) all the offerings should be US based. Respecting the aforementioned criteria allows for a total of 8,752 SEOs to work with. However, the offering should also be independent, because it is expected that other offerings possibly create biases in the cumulative abnormal returns which eventually creates bias for the entire research. An SEO is independent when, after the SEO issue date, there is no consecutive issue within the event-window. After dropping the non-independent SEOs, there are 5,837 SEOs remaining. In order to prevent biased results, multiple SEOs by companies are excluded, which results in a total of 1,764 companies that have done an SEO. Furthermore, there needs to be sufficient accounting data to work with, dropping incomplete data

results in 658 issues. Financial companies have to be excluded from the dataset since they tend to create biases. Similarly, companies with SIC between 6000 and 6999 will also be excluded from the sample which results in a sample of 470 SEOs. Lastly, ticker VC is omitted as it is an outlier.

Table 3.3.1 Without winsorizing (no outliers removed)

(1) (2) (3) (4) (5)

VARIABLES N mean sd min max

CAR 469 0.00163 1.095 -16.46 8.899 E/P 469 -0.253 0.679 -7.391 2.863 Log(size) 469 5.614 1.831 1.334 12.19 Log(btm) 469 -0.903 1.075 -6.828 2.821 Log(leverage) 469 0.866 0.796 0.0168 6.488 RD_Sales 469 36.73 588.0 0 12,522

4. Results

This section will present and discuss the empirical results of the regression. Firstly, the overall effect of R&D-intensity on the cumulative abnormal returns is discussed and afterwards, the R&D-expenditures are split up in 3 dummies, namely high-R&D, low-R&D, no-R&D. 4.1 Cumulative abnormal returns

In order to identify whether the magnitude of R&D expenditures have an effect on the cumulative abnormal returns caused by SEOs, the CARs have to be different from zero. To find whether the cumulative abnormal returns in this sample are not equal to zero, the CAR’s will be tested. In table 4.1.1, the statistic of the CARs are presented.

Table 4.1.1 After winsorizing for 5%

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Unfortunately, the coefficient of the CAR is not significantly different from zero for a 5 or 10% significance level. Therefore, this thesis will not provide additional evidence for the short-term underpricing phenomenon of seasoned equity offerings. One of the explanations for the non-significant CAR’s is the length of the event window. Since it was not possible to obtain the announcement date of the seasoned equity offering in the Thomson One database, capturing the announcement in the event window was attempted. However, since the underpricing phenomenon is short-term, it can be argued that the event period after the issue date of 10 days was too long.

(1) VARIABLES Significance test

CAR

Constant -0.0121

(0.0100)

Observations 469

However, it is interesting to see that the average cumulative abnormal returns found by Corwin (2003) during the 1980s and 1990s are -1.15%.

Despite the CAR’s not being significant, the average cumulative abnormal return is negative. This shows that there is, on average, underpricing for SEOs but again, this is not significant.

4.2 Regression without R&D dummies

This section will discuss the regression results. Firstly, the regression that includes the total R&D – sales ratio will be explained. Table 4.2.1 provides the coefficients of the first regression as discussed in the methodology. The coefficient of the main independent variable is not significant for the 10% significance. In addition, the value of the log(RD_Sales) coefficient is equal to 0.00926. This implies that when the R&D-sales ratio increases by 1%, the CAR will increase by 0.000926 * 0.01, or 0.0000926. This amount of increase is very close to zero, thus we can conclude that the R&D-sales ratio has no effect on the CAR.

Furthermore, the coefficient of ep is -0.0101. It is interesting to notice that the coefficient of ep is negative. This is counterintuitive since current earnings are used as a proxy for future earnings. An increase in earnings would normally result in a positive effect on cumulative

abnormal returns and not a negative one, as as seen in this regression. However, it is important to note that more than of companies 50% in this sample are companies with negative earnings. This might cause a bias in the sample, but removing companies with negative earnings would result in too few results. In addition, if size increases by 1%, the cumulative abnormal return increases by 0.0034*100% is 0.34%. This implies that the bigger the company, the bigger the corresponding abnormal returns. The third control variable is book-to-market ratio, which identifies the over or under valuation of a company (Fama & French,1992). If the book-to-market value of a company is above 1, it’s implied that the company is undervalued. Therefore, the positive coefficient of the book-to-market ratio indicates that a 1% increase causes the cumulative abnormal returns to increase by 0.0107*100%, which is 1.07%. Undervalued companies tends to have high cumulative abnormal returns. However, the coefficient is not significant and thus it is not possible to state this with 95% certainty. The last non-dummy independent variable is leverage. The coefficient of leverage is not significant from zero and implies that a 1% increase in leverage causes the cumulative abnormal returns to increase by 0.00918 *100%, or 0.918%. Lastly, the

industry dummies will be discussed. As mentioned previously, companies operating in the financial services industry (sic 6000-6999) were deleted in order to prevent biased results. In table 4.2.1, the agriculture dummy has been omitted in order to prevent the omitted variable bias. There are six industry dummies that are significant from zero using a 5% significance level. The most interesting dummy to look at is the manufacturing dummy (sic 2000-2999) since this dummy includes pharmaceutical companies. Guo et al. (2006) mentioned that pharmaceutical companies generally face less information asymmetry since they tend to provide investors with more information about the nature of their R&D activities and thus it is expected that these companies should face less cumulative abnormal returns in comparison to other companies. However, the coefficient of the sic2000_2999 dummy is 0.0671. This implies that companies in this industry affect the cumulative abnormal returns on average 6.71% more than companies in the industry of the omitted variable, which is the agriculture industry.

4.3 Regression with R&D dummies

This section will discuss the regression where the RD-sales ratio is split up into dummy variables. Regression 1 from table 4.2.2 divides the RD-sales ratio in three different dummies: the high_rd, low_rd and no_rd. High_rd is equal to 1 if the RD-sales ratio of this company are above the median of RD_Sales and low_rd is equal to 1 if the RD-sales ratio are below the median. No_rd is equal to 1 if RD-sales ratio is equal to zero. For this regression, the control variables that are not industry dummies are all non-significant and can be interpreted the same as in the regression without the R&D dummy variables. In table 4.2.2 we see that the low_rd coefficient is significant from zero with a 5% significance level and the high_rd coefficient is also significant from zero for a 10% significance level. Both of the coefficients are negative, implying that companies with R&D expenditures face negative cumulative abnormal returns compared to companies with no R&D expenditures. This is in line with the hypothesis and the findings of Guo et al. (2006) about the information asymmetry in combination with R&D expenditures causing negative abnormal returns. For companies with high R&D expenditures, we find 0.104 lower abnormal returns compared to companies without R&D expenditures. In addition, we find that companies with low R&D expenditures have 0.120 lower abnormal returns than no R&D expenditures companies that do a SEO.

5. Conclusion and Discussion

This thesis aimed to find a possible explanation for the short-term underpricing of SEOs. The research question was prompted by the findings of Guo et al. (2006). They found that the short-term underpricing after IPOs is partially attributed to the information asymmetry associated with the R&D intensity of a company. Where Guo et al. (2006) focused on IPOs, this thesis focused on SEOs.

From a sample of 469 SEOs, ranging from the year 2009 until the end of 2013, it was found that the CARs measured were not significantly different from zero. As a result, it can be argued that the research performed on the relationship between the negative abnormal returns after a SEO did not yield definite results However, the regression was still performed and attempted to find potential sources for the abnormal returns. Eventually, it was found that there exists significant evidence for the relationship between the abnormal returns in the sample and the R&D expenditure of a company. It was found that companies who spend money on R&D tend to have negative abnormal returns compared to companies that do a SEO but do not spend money on R&D. This is an interesting insight, but since the CAR’s in this sample are not significant, it will not be as explanatory as hoped for in this thesis.

An interesting takeaway from this thesis is the fact that the 99 percentile of the RD-sales ratio consisted solely of companies in the pharmaceutical industry. The corresponding CAR’s from these companies were all positive. Since it is known that pharmaceutical companies elaborate more on the nature of their R&D expenditures, the information asymmetry associated with their R&D expenditures is lower than for companies that do not elaborate as extensively on the nature of their R&D expenditures.

Despite this thesis not finding clear evidence pertaining to the relationship between R&D intensity and the SEO underpricing, it would be interesting to research the topic further.

Researchers could do so by examining a broader time period, for instance. Secondly, researchers might want to identify the announcement date instead of the issue date for SEOs as event date. One can argue that once it is made public that a company intends to issue an SEO in the future, the market will adjust the stock price at that given point in time. Lastly, another market model can be used to estimate the returns. For example, one could use the Fama and French 5 factor model, which builds on the capital asset pricing model used in this thesis.

6. Reference list

Aboody, D., & Lev, B. (2000). Information asymmetry, R&D, and insider gains. The journal of

Finance, 55(6), 2747-2766.

Akerlof, G. A. (1978). The market for “lemons”: Quality uncertainty and the market mechanism. In Uncertainty in Economics (pp. 235-251).

Berk, J., & DeMarzo, P. (2014). Corporate Finance. Harlow: Person Education Limited.

Carlson, M., Fisher, A., & Giammarino, R. (2006). Corporate investment and asset price dynamics: Implications for SEO event studies and long‐run performance. The Journal of

Finance, 61(3), 1009-1034.

Corwin, S. A. (2003). The determinants of underpricing for seasoned equity offers. The Journal

of Finance, 58(5), 2249-2279.

Eckbo, B. E., & Masulis, R. W. (1992). Adverse selection and the rights offer paradox. Journal of

financial economics, 32(3), 293-332.

Fama, E. F., & French, K. R. (1992). The cross‐section of expected stock returns. the Journal of

Finance, 47(2), 427-465.

Fama, E. F., & French, K. R. (2004). The capital asset pricing model: Theory and evidence. Journal of economic perspectives, 18(3), 25-46.

Guenther, D. A., & Rosman, A. J. (1994). Differences between COMPUSTAT and CRSP SIC codes and related effects on research. Journal of Accounting and Economics, 18(1), 115-128.

Guo, R. J., Lev, B., & Shi, C. (2006). Explaining the Short‐and Long‐Term IPO Anomalies in the US by R&D. Journal of Business Finance & Accounting, 33(3‐4), 550-579.

Loughran, T., & Ritter, J. R. (1995). The new issues puzzle. The Journal of finance, 50(1), 23-51. Myers, S. C., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have. Journal of financial economics, 13(2), 187-221.

Schipper, K., & Smith, A. (1986). A comparison of equity carve-outs and seasoned equity offerings: Share price effects and corporate restructuring. Journal of Financial Economics, 15(1-2), 153-186.

7. Appendix

Table 3.2.1.1 Standard industry codes

*Guenther and Rosman (1994)

Standard Industry Code Industry

0-999 1000-1999 2000-2999 3000-3999 4000-4999 5000-5999 6000-6999 7000-7999 8000-8999 9000-9999 Agriculture Mining & Construction

Manufacturing Manufacturing

Transportation and Public utilities Wholesale trade

Finance, insurance, real estate Services

Services Public Administration

Table 4.2.1 𝐶𝐴𝑅 = 𝛽0 + 𝛽1 ∗ log (𝑅𝐷_𝑆𝑎𝑙𝑒𝑠) + 𝛽2 ∗ 𝑙𝑜𝑔(𝐵𝑇𝑀) + 𝛽3 ∗ 𝑙𝑜𝑔(𝐿𝐸𝑉) + 𝛽4 ∗ 𝑙𝑜𝑔 (𝑆𝐼𝑍𝐸) + 𝛽5 ∗𝐸_𝑃+ 𝛽6 ∗ 𝐼𝑛𝑑_𝑑𝑢𝑚𝑚𝑖𝑒𝑠 CAR winsorized 5% (1) VARIABLES Coefficient Log(RD_sales) 0.00926 (0.00630) Log(BTM) Log(leverage) Log(size) 0.0107 (0.0128) 0.00918 (0.0157) 0.00340 (0.00654) E/P -0.0101 (0.0283) Sic1000_1999 0.215** (0.0930) Sic2000_2999 0.0671** (0.0341) Sic3000_3999 0.114*** (0.0293) Sic4000_4999 0.0811 (0.0631) Sic5000_5999 0.107** (0.0512) Sic7000_7999 0.0978** (0.0386) Sic8000_8999 0.0282 (0.0468) Sic9000_9999 0.181* (0.105) Constant -0.111*** (0.0250) Observations 469 R-squared 0.027

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 4.2.2 𝐶𝐴𝑅 = 𝛽0 + 𝛽1 ∗ 𝐻𝑖𝑔ℎ_𝑅𝐷 + 𝛽2 ∗ 𝐿𝑜𝑤_𝑅𝐷 + 𝛽3 ∗ 𝑙𝑜𝑔(𝐵𝑇𝑀) + 𝛽4 ∗ 𝑙𝑜𝑔(𝐿𝐸𝑉) + 𝛽5 ∗ 𝑙𝑜𝑔 (𝑆𝐼𝑍𝐸) + 𝛽6 ∗𝐸_𝑃+ 𝛽7 ∗ 𝐼𝑛𝑑_𝑑𝑢𝑚𝑚𝑖𝑒𝑠 CAR winsorized 5% (1) VARIABLES Regression 1 High_rd -0.104* (0.0578) Low_rd -0.120** (0.0539) Log(BTM) 0.000451 (0.0125) Log(leverage) -0.00338 (0.0156) Log(size) 0.000725 (0.00657) E/P -0.0164 (0.0280) sic1000_1999 0.192** (0.0815) sic2000_2999 0.0859** (0.0364) sic3000_3999 0.125*** (0.0314) sic4000_4999 0.0900 (0.0650) sic5000_5999 0.0382 (0.0636) sic7000_7999 0.0985** (0.0394) sic8000_8999 0.00511 (0.0475) sic9000_9999 0.215** (0.104) Constant -0.0161 (0.0588) Observations 469 R-squared 0.031

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1