Seasoned Equity Offerings in Germany. Determinants of Short- and Long-run Abnormal Return

Table of Contents

Acknowledgement

List of Tables

1 Introduction

2 Literature review
2.1 Short-run Analysis
2.2 Long-run Analysis
2.3 German Evidence

3 Data

4 Methodology
4.1 Abnormal Announcement Return
4.2 Abnormal Buy-and-Hold Return
4.3 Company Characteristics
4.4 Robustness Tests
4.4.1 Industry-adjusted Variables
4.4.2 Outlier Robustness

5 Results
5.1 General Findings
5.2 Short-run Findings
5.3 Long-run Findings
5.4 Robustness Tests

6 Conclusion

Bibliography

Appendix

ABSTRACT. This paper studies the abnormal returns of seasoned equity offerings over short- and long-run horizons in Germany and their determining company characteristics. Contrary to previous findings for the German market, I find that the abnormal returns around the announcement are significantly negative with Run Up, Volatility, Firm Age and Earnings per Share as explanatory variables. The long-run abnormal returns are also significantly negative. The determinants of abnormal returns in the long-run are Run Up, Firm Age, Transaction Size, Size, Leverage and Profit Margin. My findings suggest that there is a structural break in the German market in 2002/2003. Furthermore, the theoretical explanations suggested in prior research on the U.S. market are also valid for the German market.

Submission Deadline: 04/09/2013

Keywords: Seasoned Equity Offerings in Germany, SEO, Buy-and-Hold Return, CAR, BHAR

Acknowledgement

I would like to express my sincere gratitude to my supervisor Dr Marie Dutordoir for her patience, her inspiring ideas and guidance. Thanks to the Friedrich Naumann Stiftung for the scholarship and thanks to Maik Hille and Andre Zachmann for an unforgettable time.

First and foremost, I would like to thank my family for the endless support and love. Thanks for giving me the chance to fulfil my dreams. This work is dedicated to you.

List of Tables

Table 1 Major Research on SEO Announcement Effects for the U.S. Stock Market

Table 2 Major Research on SEO long-run Effects for the U.S. Stock Market

Table 3 Distribution of SEOs and Firm Characteristics

Table 4 Distribution of Market-to-Book and Size Portfolios

Table 5 Variable Description

Table 6 Abnormal Return Overview

Table 7 Predictions based on U.S. literature

Table 8 Regression Results Abnormal Announcement Returns

Table 9 Regression Results Abnormal Buy-and-Hold Return 36 months

Table 10 Robustness Test 1 Variables versus industry-adjusted Variables

Table 11 Robustness Test 2 Outlier Robust Regressions CAR / BHAR 36

Appendix 1 Abnormal Return Distribution by Year

Appendix 2 Pearson pairwise Correlation Matrix

Appendix 3 Variance Inflation Factors

Appendix 4 Shapiro-Wilk Test Results

Appendix 5 Residuals of optimised CAR Regression versus Normal Distribution

Appendix 6 Cook-Weisberg Test Results

Appendix 7 Residuals of optimised CAR Regression versus fitted Values

Appendix 8 Regression Results Abnormal Buy-and-Hold Return 12 months

Appendix 9 Regression Results Abnormal Buy-and-Hold Return 24 months

Appendix 10 Regression Results abnormal short-run Returns on industry-adjusted Variables

Appendix 11 Statistical Leverage versus Normalized squared residuals for CAR regression

Appendix 12 Robustness Test 2 Outlier Robust Regressions BHAR 12 / BHAR 24

1 Introduction

“Investing in firms issuing stocks is hazardous to your wealth“

Tim Loughran and Jay R. Ritter in their striking paper “The new Issues Puzzle” from 1995

Commerzbank’s recent capital increase on 28/05/2013 has raised heated discussions about wealth destruction and dilution following equity offerings in Germany. Commerzbank AG conducted their 8th issue of new equity capital within a 5-year period, raising public awareness about such corporate events. Although Germany is one of the world’s leading economies, academic and practical research on the short-run performance around the announcement and the long-run performance after the offer­ing is relatively rare and outdated.

Previous research on the German market in various fields has always been inspired by the difference in market structure between Germany and the U.S. Especially, the differences of bank-based versus market-based financing and the presence of Civil Law instead of Common law provide an environment to inspect the universality of U.S. findings. Moreover, Germany’s stock market has a much higher concentration in ownership than the U.S., with a high number of influential block holders present in many firms. Regarding SEO research, a close relationship to the principal bank (Hausbank) as one of the underwriters and a higher concentration in ownership are factors that reduce asymmetric information. Therefore, abnormal returns around an equity offering may be less severe for German firms than for U.S. firms. Studies on German SEOs over sample periods before 2000 indeed report positive abnormal returns, contrary to U.S. studies.

However, the German market structure has significantly changed in the last 10 years. After the burst of the dot-com bubble, German investors lost trust in the market and firms needed to raise foreign capital. Through the introduction of improved transparency standards in the equity and debt market and by reinforcing shareholder protection, firms positioned themselves to attract international investors. Fundamental milestones were the change in reporting language to English, the adoption of international accounting standards, the introduction of the Prime / General Standard in 2003 and various legal changes to facilitate investing in Germany.

In subsequent years, German firms have partially uncoupled the close Hausbank relation, the average free float for CDAX members increased from 38% in 2002 to almost 70% nowadays and the security regulations became aligned to the U.S. system.^[1] Additionally, the liquidity has increased in response to enhanced market making, the foreign holdings in the stock market increased from below 30% in 2001 to almost 70% in 2010 and in the case of an offering, the flotation method is nowadays the same as in the U.S.^[2]

In summary, I assert a structural market break in Germany around 2003 and claim that the stock market environment is similar to the U.S. nowadays. Therefore, I hypothesise that the previous findings for the German market may not hold under the current market environment. Nonetheless, the ownership structure, the maturity of a firm conducting an SEO and primarily the mind-set of German investors still remains different and an investigation is worthwhile.

Regarding my analysis, I examine the short- and long-run abnormal returns and the potential determining factors. The simultaneous investigation of both time horizons is a unique approach and I expect to gain a superior insight about the immediate and the delayed effects of various factors. Particularly, with regard to theories that prevail over both horizons but reveal their actual effect in the long run,^[3] I anticipate substantial findings. From an economic point of view, it may be also beneficial to observe how firms could manage their operational measures to absorb large price drops to prevent bad press, as in the case of Commerzbank.

The dissertation is organised as follows: Section 2 summarises the previous work on short- and long-run abnormal returns for seasoned equity offerings. In Section 3, I will present my data selection and the final sample. Section 4 contains the calculation methodology for short- and long-run abnormal returns, the motivation for the chosen explanatory variables and regression methods including robustness tests. In section 5, I will display my initial predictions, my major findings and the potential explanations. Various robustness tests are performed to test my findings. Section 6 contains my own conclusion, the limitations of my work and motivation for further research.

2 Literature review

Research on seasoned equity offerings is numerous for the U.S. stock market. The line of SEO research can be traced back to the 60s with a couple of short-run studies such as Stigler (1964) or Friend & Longstreet (1967). The early research focuses exclusively on announcement effects, while long-term performance studies have their origin in later years. Up to now, researchers have always distinguished between short- and long-run abnormal return analysis. Subsequently, the major work and crucial theories for both time horizons will be presented.

2.1 Short-run Analysis

Market reactions to offering announcements are well-examined and tackled from diverse perspectives. Consistently, negative abnormal returns around the announcement are found and assigned to the signalling effect.

Exemplified by Akerlof’s market for lemons (1970), the determination of quality in stocks is a striking example for asymmetric information. Naturally, a firm’s management is better informed about the current business and potential investments. Leland & Pyle (1977) and Myers & Majluf (1984) have developed models that deal with asymmetric information and the signalling that occurs when a firm becomes active in the capital market. Especially the Myers and Majluf pecking-order model is highly accepted in the academic world. In their model, insiders have advantageous information and aim to maximize the existing shareholder’s wealth. In this regard, the management chooses to issue equity when it is overvalued. Hence, an equity offering signals the overvaluation and the market reaction is negative. Particularly, the use of proceeds can be interpreted in different ways.

Heinkel (1982) and Masulis (1983) argue that if proceeds are used to retire debt, the firm is actually in need to decrease leverage to balance extended business risk. In contrast, proceeds used to fund capital expenditures reveal that existing assets cannot generate sufficient funds for new investments (Myers & Majluf, 1984). Accompanied by the agency model, which indicates that managers occasionally pursue their own interests, the signalling model is the basis for most explanation of short-run underperformance (Jensen, 1986; La Porta et al., 2000). Researchers have used various proxies for asymmetric information and the related signalling effect.

Following the introduction of the Myers & Majluf model, SEO announcement research became popular.^[4] Asquith & Mullins (1986), Masulis & Korwar (1986) and Mikkelson & Partch (1986) report negative abnormal returns for their SEO announcement analysis, delivering convincing arguments for the signalling effect. They use proxies such as transaction size, decrease in management’s shareholdings and use of proceeds and find significant results for these variables. In succeeding years, Korajczyk et al. (1990) report abnormal returns of almost -3%, using the time gap between the last company report and the announcement as an indicator for asymmetric information. In the same year, Hansen & Crutchley (1990) have adapted the full information model by Fama & Miller (1970) and confirmed the signalling model also for times of earning downturns and the subsequent need for external financing.

Further evidence for the signalling effect is adduced by Dierkens (1991). The author uses volatility, number of offerings and the trading frequency as attributes for asymmetric information. Subsequently, Eckbo & Masulis (1992) deliver new insight on announcement effects, using an analytical model that combines asymmetric information proxies such as offer size, percentage of outstanding shares, or shareholders concentration with listing aspect like offering costs and flotation method. The authors aim to distinguish between listing costs and temporary overvaluation and conclude that the signalling effect still persists, even when controlling for related costs of the issue.

Another stream of explanations dates back to Howe & Lin (1992) who examine the market making process and found lower bid-ask-spreads for dividend payer than for non-payer. The authors conclude that dividend payers incur lower asymmetric information. Loderer & Mauer (1992) were the first who test the dividend status on its explanatory power for SEO returns. However, the authors report insignificant results, leaving an unsolved puzzle.

A completely new aspect is picked up by Jagadeesh et al. (1993). In their analysis, they link SEOs to the preceding initial public offering. The authors conclude that high-value firms that knowingly under-price their IPO have relatively higher abnormal returns after the IPO. These higher returns have positive effects on subsequent SEOs. However, the economic significance of the findings is low. Adapting Jagadeesh et al.’s approach, Slovin et al. (1994) report a higher significance on the same variables for a different time period and a smaller sample.

In the same year, Denis (1994) linked the profitability of investment opportunities to abnormal announcement effects. The author uses only firms that state capital expenditure/investments as the use of proceeds. The author confirms the Myers & Majluf prediction that the lack of internal funds for investments has negative effects on the announcement return.

Another potential explanation is introduced by Bayless & Chaplinsky (1996). In addition to various accounting measures, the authors incorporate a dummy variable for “hot and “cold” markets to explain the announcement effects. Hot markets are characterised by rising stock markets and a favourable macroeconomic environment but especially by high volumes of equity market activity. The authors report that the reaction to SEO announcements is less severe in “hot” than in “cold” markets. Furthermore, investor’s concerns about firm specific information asymmetries are stronger in “cold” markets.

Moreover, D’Mello & Ferris (2000) use the number of analysts to represent the degree of asymmetric information. Consequently, with a decreasing number of analysts the abnormal returns are more negative around the announcement. In a subsequent paper, D’Mello et al. (2003) also examine the effect of the frequency of SEOs on the announcement effect. The authors report diminishing asymmetric information for each issue of a firm, resulting in less negative abnormal returns for subsequent offers. Besides, there is evidence for opportunistic behaviour of the firms. The companies tend to increase the offer size and shorten the intervals between the SEOs.

More recently, Kim & Purnanandam (2009) investigate corporate governance impacts on announcement returns. The authors report more negative market reactions in weaker governance environments. The reaction is amplified in cases where firms had M&A activity after previous equity offerings.

Based on Gibson et al. (2004), Demiralp et al. (2011) provide evidence that institutional ownership has positive impact on announcement returns. Active institutional investors promote monitoring and decrease consequently the degree of asymmetric information.

Since Loderer & Mauer (1992) were not able to find significant results for dividend status as an explanatory variable, Booth & Chang (2011) have revisited the approach for a different sample period and a larger sample size. The authors reconcile the puzzle by reporting a positive relationship between dividend payment and abnormal return around the announcement.

Most recently, Hull et al. (2012) examine the consequences of insider ownership on announcement returns. The authors report a negative relation between returns and insider holdings. Hence, higher insider ownership levels worsen the returns around the announcement.

Table 1 summarises the aforementioned research and displays the examined period, the sample size and the abnormal return. All reported abnormal returns are significant at the 1% level.

Abbildung in dieser Leseprobe nicht enthalten

Table 1 Major Research on SEO Announcement Effects for the U.S. Stock Market

The table reports the major work on seasoned equity offerings and the abnormal returns around the announcement for the U.S. stock market. The table is arranged by date of publishing. The abnormal returns are calculated either over a two days [-1, 0] or a three days [-1, +1] window. The list of equity offering is restricted to “regular” offerings, omitting private placements, standby rights and shelf offerings. The table is loosely based on Eckbo et al. (2007). All abnormal returns are statistically significant at the 1% level.

2.2 Long-run Analysis

In comparison to the short-run literature, researchers are less conclusive about long-run abnormal performance. While most researchers agree on long-run abnormal underperformance of firms undergoing an SEO, others dispute the phenomenon and assign the findings to a wrong calculation methods or erroneous benchmark selection. Researchers that are convinced about the abnormality of returns do at least somehow link their explanations again to the signalling effect but extend the reasoning to other theories. These theories include but are not limited to delayed reactions to signalling, the theory of “windows of opportunities”, earnings management and the overconfidence hypothesis. Since the argumentation is less consecutive in the chronological sense the review will be ordered theory-wise.

Windows of opportunities

Classical efficient market models fail to explain abnormal stock price behaviours related to equity offerings. Therefore, researchers develop behavioural theories such as the “windows of opportunities” model to support their findings. Loughran & Ritter (1995) argue for the existence of windows, in which firms can exploit temporary overvaluation, without facing an immediate share drop as suggested by Myers & Majluf (1984). During these windows, investors tend to be overoptimistic about the overall market performance and the expectations for future performance of SEO firms. Hence, firms neglect the pecking order hypothesis and issue equity rather than internal funds or debt. The over optimism during these periods impede an immediate price drop but causes the firm to be revaluated when the market perception normalises. Research on “windows of opportunities” is conducted by Loughran & Ritter (1995), Spiess & Affleck-Graves (1995) Brav et al. (2000) and Clarke et al. (2001). Closely related to this field is the research by Lee (1997) and Kahle (2000) who examine executive trading and the exploitation of temporary overvaluation. The authors report that insiders do knowingly exploit specific patterns by selling company holdings previous to new equity issues. Furthermore, the hypothesis is confirmed by the findings of Bayless & Chaplinksy (1996), who report less severe announcement effects in “hot” periods and Loughran & Ritter (1997), who backed their previous hypothesis in a different setting.

Great insight is also brought by Daniel et al. (1998) in a more formal model loosely related to the timing-hypothesis of Loughran & Ritter. The authors argue that investors are overconfident about their private information and thus underreact to new public information. This argumentation could support the explanation for delayed market reaction to equity issue announcements.

Earnings Management

Another stream of explanations for long-run underperformance is the earnings management theory. The concept has been developed as a response to Loughran & Ritter’s findings about operational measures peaking around an SEO and dropping in subsequent years. The perspective suggests that earnings management is conducted by a firm prior to an announcement to milder the short-run effects. Teoh et al. (1998) and Rangan (1998) show that firms who actively adjust discretionary accruals before an announcement do face relatively lower long-run returns. Logically, the subsequent revelation of unadjusted earnings causes downward adjustments by the market. Later studies do also report significant earnings management before SEOs. However, the researchers are not able to confirm the previous findings about the relation between stock performance and discretionary accruals in the long term (Shivakumar, 2000; Hribar & Collins, 2002; Di et al., 2012).^[5]

Benchmark and Calculation Issues

Although, the “windows of opportunity” as well as the earnings management are both logical models, some researchers do disagree on these explanations and neglect the occurrence of abnormalities in return. Mostly, the critics assign potential abnormalities in the return to misspecified methodologies. Researchers such as Eckbo et al. (2000) and Jagadeesh (2000), argue that the calculation of long-term abnormal performance, using matched firms or matched portfolios as the benchmark, does not appropriately incorporate a decrease in leverage and the implied reduction of systematic risk. Moreover, Mitchell & Stafford (2000) propose that a bootstrapping procedure and the use of calendar-time abnormal returns can mitigate potential return abnormalities. Later research such as Kothari & Warner (1997), Lyon et al. (1999) and Li & Zhao (2006) suggest complex methods for long-run return calculation and benchmarking. However, with increasing complexity, the models become less transparent and traceable.

Table 2 summarises the aforementioned research and displays the examined period, the sample size, the holding period and the abnormal return. All abnormal returns are significant at the 1% level. Research with incomparable methodology is left out.

Abbildung in dieser Leseprobe nicht enthalten

Table 2 Major Research on SEO long-run Effects for the U.S. Stock Market

The table reports the major work on seasoned equity offerings and the abnormal returns in the long-run for the U.S. stock market. The table is arranged by date of publishing. The abnormal returns are calculated using a standard buy-and-hold methodology. The list of equity offering is restricted to “regular” offerings, omitting private placements, standby rights and shelf offerings. The table is loosely based on Eckbo et al. (2007). All abnormal returns are significant at the 1% level.

Abbildung in dieser Leseprobe nicht enthalten

2.3 German Evidence

As stock markets are heterogeneous cross-nationally, I will limit international literature to findings for the German market. In a nutshell, international research is mostly limited to adapt U.S. market findings on the national stock market. Considering the German market, research is rare and out-dated.^[6] The first work on announcement effects is done by Brakmann (1993). Brakmann reports significant positive abnormal returns and is among the first who tests the U.S. market findings on a foreign exchange. Inspired by the opposing findings, Padberg (1995) examines 479 sample firms and confirms the significant positive abnormal return for the German market. As a proxy for information transparency, the market membership is used. The latest research on announcement effects dates back to 2001. Gebhardt et al. (2001) use a more homogeneous sample of 190 firms over the period from 1981 to 1990 to review the previous findings. In agreement with the earlier results, the authors report positive abnormal returns of 0.60%. As mentioned in the introduction, the authors ascribe the opposing findings to differences in the market structure, flotation methods, ownership structures and corporate governance methods.

Regarding the long-run SEO research, the German market is even less studied.^[7] Stehle et al. (2000) are the only ones who investigate the German stock market. In their combined study of IPOs and SEOs they adapt the Barber & Lyon (1997) approach and report an average negative abnormal return of -9.01% for a 3 years holding period. Their sample period ranges from 1960 to 1992.

In summary, seasoned equity offerings in the German market are little examined regarding their short- and long-run abnormal returns. Research for periods after the euro introduction and the stock market restructuring is completely missing.

3 Data

The SEO sample is sourced from Bloomberg and Thomson One Banker and includes all trading seasoned equity offerings by German firms listed in the Prime or General Standard respectively their predecessor “Amtlicher Markt” between 01/07/1995 and 01/07/2010.^[8] The selected time period allows for a 3-year holding period. Stocks listed in the Prime and General Standard must meet minimum requirements in terms of size, disclosure and trading volume. In this respect, information asymmetries might be slightly reduced compared to the preceding periods and international comparability facilitated. Consistent with prior literature, secondary share offerings, private equity exits, venture capital exits, IPOs and offerings by REITS, closed-end funds and investment trusts are excluded (Eckbo et al., 2000; Booth & Chang, 2011; Bilinski et al., 2012). These criteria result in an initial sample of 829 transactions.

The raw dataset is reduced by 149 offerings, where the firm’s IPO dates within the 12-months estimation period of the market model. Additionally, 30 transactions are deleted because the firm was delisted as a result of bankruptcy or M&A activity within the 36-months buy-and-hold period after the announcement.^[9] To be included in the final sample following items must be reported by Bloomberg: Stock price -12 months to +36 months around the announcement, industry classification, outstanding shares, total liabilities, financial leverage, book value of equity, total assets, return on assets, profit margin, 90-day volatility, sales growth and net income.^[10] These restrictions lead to a final data set of 416 transactions. The sample includes offers by 258 firms, with 168 firms that made one seasoned offer, 54 with two offers, 21 with 3 offers, 8 with four offers and 7 firms with five or more offers.

Table 3 reports the sample distribution by year and summary statistics for key company characteristics. Obviously, there is cyclicality in the numbers of SEOs per year, clustering in the middle and late 2000s and accounting for almost 80 % of the sample. This clumping is mostly due to the limited data coverage in the early years of the analysis but might be also explained by the “hot-market” approach by Bayless and Chaplinsky (1996). In their Paper they deliver factors on macroeconomic and firm level that help to explain the clustering effect of market activity. Especially macroeconomic criteria (see Choe et al., 1993; Moore, 1980) and broad market stock returns (Alti & Sulaeman, 2012) appear to be legit explanations for the slump in the early 2000s.^[11]

Abbildung in dieser Leseprobe nicht enthalten

Table 3 Distribution of SEOs and Firm Characteristics

The table lists the number of offers in Germany by on the left-hand side. The sample comprises 416 observations and is uneven distributed over the period. On the right hand side, firm characteristics of the relevant firms are displayed. Accounting measures have been extracted from the last company report prior to the transaction. Firm Age is the number of months from the IPO to the announcement date. Market Capitalization is the share price 30 days prior to announcement multiplied with total shares outstanding. Relative transaction size is the number of offered shares divided by the number of outstanding shares prior to announcement. Volatility is the annualized standard deviation of the relative price change for the 90 days prior of the announcement. Leverage on MV is total liabilities divided by the sum of total liabilities and market capitalization. Leverage on BV is average total assets divided by average total common equity. Average is the average of the beginning and ending balance. Total common equity is share capital, additional paid in capital and retained earnings. Source: Bloomberg, Thomson One Banker and Thomson Datastream.

Abbildung in dieser Leseprobe nicht enthalten

On the right side of the table, decisive firm characteristics are displayed. The accounting measures have been extracted from the latest company report prior to the transaction,^[12] firm age is the time period between IPO and announcement date and market capitalization and volatility are calculated based on data 1 month prior to the transaction.^[13] A closer inspection of the variables discloses the strong deviation between the average and median. Along with the high standard deviation, there are evidently some outliers in the characteristics that might distort the regression later on. Previous literature often omits this issue even when statistics about discrepancies between average and median are displayed (e.g. Corwin, 2003; Elliot et al., 2009; Booth & Chang, 2011). In the methodology part, this issue will be picked up again and discussed in more detail.

[...]

---

^[1] Source: Bloomberg; S&P 500 has 92% average free float, Russell 2000 78%

^[2] Source: Bloomberg, Spiegel, Thomson Reuters. Both in Germany and in the U.S. an underwriter is involved. Before 1995, there were differences in the flotation method. (Stehle et al., 2000)

^[3] “Windows of Opportunity”, Earnings Management

^[4] The following list of equity offering is restricted to “regular” offerings. Private placements, standby rights and shelf offerings are omitted

^[5] Di et al.’s approach is discussed again in 4.3

^[6] Additionally, the access to the literature is limited. Statements about the findings have to be handled carefully and I apologize for any misrepresentations

^[7] Even extensive research in English and German databases delivered just the noted work by Stehle et al. (2001)

^[8] Withdrawn or postponed transactions are excluded; The Prime and General Standard segments were introduced in 2003 and guarantee international transparency and disclosure standards such as quarterly reports in English and German, application of international accounting standards, obligatory analyst conferences and ad-hoc disclosure

^[9] Due to the small number of deleted transactions, a survivorship bias is neglected

^[10] Bloomberg fields in aforementioned order: last_price, industry_sector, bs_sh_out, bs_tot_liab2, fncl_lvrg, total_equity, bs_tot_asset, return_on_asset, prof_margin, volatility_90d, sales_growth and net_income

^[11] The leading index DAX dropped by ca. 7%, 20% and 43% in 2000,2001 and 2002

^[12] Quarter, half-year or annual report

^[13] The date choice for the variable market capitalization will be explained and deliberated in the methodology part