University of Amsterdam Amsterdam Business School
Master Thesis
Long-Term Performance Following Open Offers in the US
- In light of the recent financial crisis -
Name: Peereboom, T.T.T. (timster1990@hotmail.com; tim.peereboom@student.uva.nl) Student number: 6054765
Programme: MSc Business Economics, Finance Number of credits thesis: 15 ECTS
Name of supervisor: Dr. Ligterink, J.E. Month of completion: January 2014
Acknowledgment
The author would like to thank Dr. Ligterink, J.E. for his valuable comments.
Abstract
This study shows evidence of significant long-term overperformance in terms of stock returns following open offers made during the total period of the bull cycle (2002 – 2006) & bear cycle (July 2007 – July 2009) in the US compared to non-issuing size-industry-matched benchmark firms. However, long-term operating overperformance is only partly confirmed by the long-term average abnormal results of the operating fundamentals used in this study. No significant differences are found in long-term abnormal performance of companies making open offers during the bull cycle (2002 – 2006) versus the bear cycle (July 2007 – July 2009).
Keywords: open offers, seasoned equity offerings, long-term performance, operating performance, financial crisis, bear cycle, bull cycle
Table of Contents
I. Introduction………...………....…….………4
II. Theoretical Framework………..………...………7
A. Seasoned Equity Offerings & Market Efficiency………...………7
B. Trade-Off Theory, Pecking Order Theory and Market Timing Hypothesis…..………...8
C. Equity Issue Methods………...………….……....………...………..…9
D. Open Offer Technique...10
E. Previous Studies of Long-Term Performance of SEOs & Possible Explanations………….10
F. Market Sentiment………...14
III. Hypotheses……….……….……….……..…...15
IV. Research Design………...…...………...………..16
A. Data Sample………..………..16
B. Methodology………..………..22
V. Empirical Results & Analysis………..………...………35
A. Post-Event ABHARs for Firms that Conduct Open Offers………35
B. Post-Event Evaluation of Operating Performance for Firms that Conduct Open Offers...42
C. Crisis Dummy Regressions………….……….………48
VI. Summary & Conclusions……….………...……….………55
References………...………..………57
Appendices………..………...……….………..59
I. Introduction
There are four methods of seasoned equity offers (SEOs) by listed firms: (1) rights issue, (2) firm-commitment offer, (3) private placement and (4) open offer. Most of the previous studies on long-term performance following seasoned equity issues are done regarding SEOs in general and not regarding one of the four mentioned issue techniques. Especially literature regarding long-term performance following open offers is scarce. The only study that can be found is that of Ngatuni, Capstaff and Marshall (2007). Their study of open offers in the UK during the period 1991-1995 showed significant overperformance over a 5-year post-issue period compared to non-issuing benchmark firms: (a) size-matched, (b) size-industry-matched and (c) size-B/M-matched. This benchmark matching method is also used for this study. It is interesting to show whether significant long-term overperformance following open offers in the US will be observed, in order to extend the existing scarce literature.
First, the open offer technique is discussed in more detail. The buying agreements in an open offer are most of the time verbal agreements. The contracts are signed on the day of announcement and sometimes shortly after the announcement day. The shares are offered in proportion (pro rata) to the existing shareholders on the announcement day. There is a two week during offer period. The ex-rights day is the day of announcement and sometimes it is the day after announcement. Unlike in a rights issue, the rights cannot be traded. So these are worth nothing unless the shareholders want to buy new shares. The share issue is conditional on approval of shareholders at the EGM (extraordinary general meeting), if an EGM is needed which is usually the case. Most of the time the issues are approved at the EGM.
There is more literature available regarding the long-term performance of equity issuing firms in general. Previous studies on seasoned equity offerings in general have unanimously shown a significant underperformance for periods up to five years after the event. Possible explanations for the long-term underperformance could be: (1) the underreaction hypothesis, (2) equiprobability arguments of Fama (1998), (3) poor risk controls when conducting a matched-firm technique and results also may be affected by including IPO matched-firms in the matched-matched-firm sample, and (4) as a result of the new listing bias, rebalancing bias and skewness bias the long-term buy-and-hold abnormal returns are generally negatively biased, when using a reference portfolio as benchmark.
This study also takes into account the financial crisis. Taylor (2009) suggests that the classic explanation of financial crises is that they are caused by frequent monetary excesses which lead to a boom and an inevitable bust. In the recent crisis there was a housing boom and
bust which in turn led to financial turmoil in the United States and some other countries. This crisis ended mid-2009 (Financial Crisis Inquiry Commission, 2011).
This thesis examines the long-term three year stock performance and operating performance following open offers made in the US during the bull cycle (2002 – 2006) and bear cycle (July 2007 – July 2009). For this study a total sample of 93 open offers is used, whereof 54 open offers are made during the bull cycle and 39 open offers are conducted during the crisis period.
As mentioned, in this study a closer look is taken at the difference between high sentiment (bull cycle) and low sentiment (bear cycle) periods. There are also other reasons why a company will conduct an SEO, e.g. liquidity needs and investment opportunity. So market sentiment plays a role in the decision to issue equity; yet it is only a consideration. It is not the driving force behind the decision to conduct an SEO.
Deng, Hrnjíc and Ong (2012) state that managers conduct SEOs to benefit from the inflated share price. Investors realize that managers behave opportunistically. The authors conjecture that the mentioned behavior is stronger during times of high sentiment and that a higher probability of conducting SEOs and a higher overpricing of companies that make SEOs are expected during these times. Investors realize this and adjust for this behavior. Therefore, announcement of an SEO is even more negative news during times of high sentiment (Deng, Hrnjíc & Ong, 2012, p. 10). As previously mentioned, during periods of high sentiment higher overpricing is expected of firms that make SEOs. When the sentiment deteriorates, valuations will go back to the intrinsic value over the long-term and the reversion to the intrinsic value is higher during these periods of high sentiment. Then the post-issue long-term return is lower. Deng, Hrnjíc and Ong (2012, p. 12) state that there is a negative correlation between post-issue long-term return and investor sentiment.
However, Ngatuni, Capstaff and Marshall (2007, p. 60) state that the overvaluation exploitation argument does not apply to the open offers because the shares in open offers are offered to the existing shareholders. Therefore, the conjecture of Deng, Hrnjíc and Ong (2012, p. 10) of higher overpricing of firms that make SEOs during times of high sentiment does not hold for open offers. Therefore the expectation of this thesis is that long-term performance of firms making open offers during bull cycle does not significantly differ from firms making open offers during bear cycle, when compared to non-issuing peers.1
1 The non-issuing benchmark firms are chosen on the basis of (i) size, (ii) size-industry and (iii)
size-book-to-market. See section IV. Research Design (subsection B).
5
The contribution of this thesis is twofold. Firstly, this study shows robust evidence that there is post-issue long-term overperformance in terms of stock returns compared to non-issuing size-industry-matched benchmark firms. Significant average buy-and-hold abnormal returns of 20.70% and 22.68% are reported for the k = 24 and k = 30 months periods, respectively. This is in line with the findings of long-term overperformance of companies engaging in open offers reported in previous research (Ngatuni, Capstaff & Marshall, 2007).
Besides, the post-issue excess operating performance of open offer firms is compared to non-issuing benchmark firms. Four accounting indicators are used to measure operating performance: change in turnover, change in earnings before tax, return on assets and change in fixed assets. Return on assets indicates the profitability of a firm relative to its total assets. It shows the efficiency of managers to generate earnings by using their assets. It is calculated by dividing earnings before tax by total assets. The results present that only the average abnormal ROA for the total sample of 93 open offer firms shows significant results for all the periods t = 0,
t + 1, 2, 3 years (size-industry-matched and size-B/M-matched). They show negative results for
all periods. However, an improvement can be observed for all periods when looking at the size-industry-matched average abnormal ROA. This means that US managers of open offer firms become less inefficient in generating earnings by using their assets compared to their non-issuing matching companies.
However when looking at the size-B/M-matched average abnormal ROAs for the total sample of 93 open offers, a three year deterioration in average abnormal ROA between t = 0 and
t + 3 years can be observed. Also note that the other operating performance indicators only show
a few significant positive results for the long-term period but no real trends can be observed due to a lack of significant results. In conclusion, these findings are not fully in line with a post-issue long-term operating overperformance compared to non-issuing benchmark firms.
Secondly, in order to demonstrate differences in long-term abnormal performance of firms conducting open offers during bull cycle versus companies making open offers during bear cycle, crisis dummy regressions are run. However, for the t + 3 years’ post-event period no significant results are found for the crisis dummy variables. So no significant differences in long-term abnormal performance of companies making open offers during the pre-crisis period versus firms conducting open offers during the crisis period are found. Market sentiment plays no significant role in the long-term abnormal performance of open offer firms.
Firstly, in the next section a theoretical framework is discussed. This framework consists of (a) SEOs and market efficiency, (b) trade-off theory, pecking order theory and market timing hypothesis, (c) equity issue methods, (d) open offer technique, (e) previous studies of long-term 6
performance of SEOs and possible explanations and (f) market sentiment. The two hypotheses of this study are described in the third section and the research design is presented in the fourth section. This fourth section is split into (a) data sample and (b) methodology. The fifth section shows the empirical results and analysis: (a) event ABHARs for open offer firms, (b) post-event evaluation of operating performance for open offer firms and (c) crisis dummy regressions. Finally, the summary and conclusions are presented in the sixth section of this thesis.
II. Theoretical Framework
In this section the theoretical framework is discussed. Subsection A discusses general theory of seasoned equity offerings (SEOs) and market efficiency. Subsection B explains trade-off theory, pecking order theory and market timing hypothesis, respectively. Subsection C mentions several equity issue methods and subsection D explains the open offer technique in more detail. Subsection E examines previous studies of long-term performance of seasoned equity offerings and gives possible explanations for the long-term effects. Finally, the influence of market sentiment on the long-term effect of SEOs is discussed in the last subsection of this theoretical framework.
A. Seasoned Equity Offerings & Market Efficiency
A seasoned equity offering (SEO) is also known as a follow-on offering or seasoned issue. When a seasoned equity offering takes place, additional securities are issued from a company whose securities already trade in the secondary market. It is important to explain what the foundational reason is why most firms conduct SEOs. DeAngelo, DeAngelo and Stulz (2010) state that market timing cannot be the primary motive for selling stock. They state that the foundational reason most firms conduct SEOs is to meet a near-term cash need and, conditional on such a need, SEO decisions reflect the firm’s lifecycle stage and market timing motives. It is noticed that the lifecycle effect (i.e. there is a need for external financing, or owners may want to cash out) is the empirically stronger one of these two ancillary motivations. The next two paragraphs explain the differences between SEOs and IPOs, and market efficiency respectively.
Initial public offerings (IPOs) could be considered in the same way as SEOs. However in the case of SEOs as mentioned, the shares that are issued by a company already trade in the equity market. This is not in the case of IPOs. The second difference between SEOs and IPOs is that SEOs are made by companies that have matured beyond the initial public offering. These companies have a significant track record of operating and financial performance at time of the SEO (Chemmanur, He & Hu, 2009, p. 385).
Modigliani and Miller (1958) argued that under a market price process, the classical random walk, in an efficient market and in the absence of bankruptcy costs, agency costs, taxes and asymmetric information, the firm value is not affected by the way of financing. It does not matter if the capital is raised by the sale of debt or by the issuance of stocks. So there should not be any abnormal returns as financing decisions do not matter. However, in an inefficient market, some stocks will be underpriced and others will be overpriced (Hsu, 2004, p. 4). So some investors could lose more than guaranteed by the level of exposure to risk, while others can make excess returns. If it is assumed that markets are completely efficient, these threats and opportunities would not exist. The reason for this is that the market price would go quickly to match a stock’s true value as a change is observed (Hsu, 2004, p. 4). However, crashes that were market-wide seem to lead to some kind of inefficiency within markets (Greenwald & Stein, 1991, p. 445). For example, crashes like the dotcom bubble and the recent financial crisis.
B. Trade-Off Theory, Pecking Order Theory & Market Timing Hypothesis
Publicly listed firms could raise additional capital from new investors or from existing shareholders through SEOs. The trade-off theory of capital structure states that a firm chooses a proportion of equity finance and debt finance by balancing the benefits and costs. This hypothesis goes back to Kraus and Litzenberger (1973). They considered a balance between tax saving benefits of debt and dead-weight costs of bankruptcy.
This theory can be seen as a competitor theory to pecking order theory of capital structure. The pecking order theory states that the cost of financing increases with asymmetric information. This theory is popularized by Myers and Majluf (1984). They mention that financing comes from three sources: internal funds, debt and equity. Firms prefer internal funds to debt to equity. If it is necessary to use external financing, debt is preferred over equity. Issuing equity means that external ownership is brought into the company. Myers and Majluf argue that managers know more about the true conditions of the company than investors. When managers issue new equity, investors could believe that managers think that the company is overvalued. Investors realize that managers make SEOs to take advantage of the share price that is inflated and the investors revise the firm’s valuation downwards after announcement of the seasoned equity offering (Deng, Hrnjíc & Ong, 2012).
The market timing hypothesis is often contrasted with the trade-off theory and the pecking order theory. Baker and Wurgler (2002) argue that companies do not care whether they finance with equity or debt, but they choose a financing form that at a specific point in time
seems to be overvalued by financial markets. This theory assumes that firms can detect mispricing better than markets can.
C. Equity Issue Methods
Armitage (2013) mentions that there are four methods of seasoned equity issues by listed firms: (1) rights issue, (2) firm-commitment offer, (3) private placement and (4) open offer.
In a rights issue, a firm offers the rights to buy new shares to its existing shareholders. This is done in proportion (pro rata) to the existing shares owned by these shareholders before the ex-rights day. This ex-rights day is the beginning of the issue period and it lasts for three weeks. The existing shares cease to have an entitlement to buy new shares during the offer period; the existing shares trade ex-rights. The rights to buy new shares can be traded just like the existing shares. If the offer price of the new shares is smaller than the market price of the existing shares, the rights to buy new shares have value. The right holder must pay the firm the price of the offer before closing of the offering period in order to buy a new share. If the rights to any shares are not subscribed during the offer period, these rights are sold to other investors who want to subscribe. However, if they still cannot be sold they are taken up by underwriters, otherwise they expire. In case no EGM (extraordinary general meeting) of the firm is required, the day after announcement of the offer is the ex-rights day. Usually there is an EGM and this meeting is held three weeks after announcement. In this case the ex-rights day is the day after the EGM.
Another method is used for larger offerings, the firm-commitment offer. There is a one month book-building period after announcement of the offering. During this period, a syndicate finds applications for the new shares. In a shelf offering, the SEC (Securities and Exchange Commission) already registers the new shares. Determination of the offer price takes place the day before the new shares are issued. Most of the time, the offer price is set the same as the market price. However, sometimes the offer price is set at a relatively small discount to the market price. ‘Best efforts’ offer is another name for a non-underwritten firm-commitment offer.
The third offering technique is the private placement. In a private placement, a block of shares is sold to only one or two investors for a relatively small amount (a couple of million dollars) and no prospectus is required. Most of the time, the private placements are made at a discount.
The last offering technique is the open offer. In the same way as the rights issue, shareholders retain preemption rights (i.e. rights the shareholders have to refuse the new shares; the shareholders can sell the rights to the stock market during a specific period if he or she does not want to buy the new shares). However, the most important difference is that the rights cannot 9
be traded. So these are nothing worth unless the shareholder wants to buy new shares. The next subsection clarifies the open offer technique.
D. Open Offer Technique
This subsection explains the open offer technique in more detail and the key differences between an open offer and a rights issue are mentioned.
Armitage (2010, pp. 1-2) mentions that before an open offer is announced, the new shares are sold by private bookbuilding or negotiation. The buying agreements are most of the time verbal agreements. The contracts are most of the time signed on the day of announcement and sometimes shortly after the announcement day. The shares are offered pro rata (i.e. allocated proportionately) to the existing shareholders on the day of announcement. There is a two week during offer period. The ex-rights day is the day of announcement and sometimes it is the day after announcement. Unlike in a rights issue, the entitlements cannot be traded. So these are worth nothing unless the shareholders want to buy new shares. The share issue is conditional on approval of shareholders at the EGM (extraordinary general meeting), if an EGM is needed which is usually the case. Most of the time the issues are approved at the EGM. How many shares the placee will finally receive is dependent upon the number of shares taken up by the existing shareholders.
Armitage (2010, p. 2) also states that open offers differ from rights issues in three ways. First, a placing takes place before the open offer. The placees are not solely underwriters. They expect to get some shares because they are entitled to the shares by virtue of being a shareholder. Armitage (2002) mentions that in open offers, on average half of the shares are not allocated to the existing shareholders on pro rata basis. Second, the entitlements cannot be traded in an open offer. Third, the ex-rights day is usually the announcement day in an open offer. ‘Placing with clawback’ is an informal name for an open offer. Existing shareholders can ‘claw back’ the shares that are initially placed, on pro rata basis. The name ‘open offer’ is misleading because it can be thought that this is a public offer to investors in general.
E. Previous Studies of Long-Term Performance of SEOs & Possible Explanations
The issuance of new equity represents a decisive market signal regarding the abilities of the managers and the future potential of the firm. An investor should immediately adjust his or her valuations and individual beliefs to reflect this new information set. Besides, prices should immediately go to a new equilibrium. Under the informational efficiency paradigm, stock prices for firms that engage in SEOs should adjust with the revaluations that are required. These
revaluations are not completed later than the end of the period of announcement. When a new equilibrium is reached, the stock performance of the firms after the SEO should on average not perform worse or better than the rest of the market (Andrikopoulos, 2009, p. 191).
However, earlier studies on the long-term performance of SEOs in US have revealed different outcomes. Table 1 shows that earlier studies conclude that SEOs are followed by significant long-term underperformance. A mispricing pattern can be observed. Firms that issue equity outperform on average the benchmarks. However, during the period of announcement this is followed by a significant mean reversion and a long-term underperformance can be observed after the event.
This table also shows that a recent study (Ngatuni, Capstaff & Marshall, 2007) on the long-term performance of open offers in the UK reveals a significant long-term overperformance. No other studies regarding long-term performance following open offers can be found. Korteweg and Renneboog (2003) mentioned possible reasons why companies might prefer open offers, e.g. good growth prospects and illiquid market for rights issues. Ngatuni, Capstaff and Marshall (2007) studied the 5-year post-issue period of open offers in the UK. They used a sample of 132 open offers that covered the period 1991-1995 and they reported an overperformance of 11.22% annually. This was calculated in average compounding terms for companies that conducted open offers compared to firms that did not issue with similar size & book-to-market value. The authors conclude that in their study good performing companies chose the open offer technique and that these firms had superior growth perspectives (Ngatuni, Capstaff & Marshall, 2007, pp. 60-61).
So, earlier studies on US seasoned equity offerings in general showed underperformance in the long-term post-issue period. Therefore, it is questioned whether the market is informationally efficient. The long-term underperformance can be interpreted in the following way: overpriced stocks exist after, prior to and during the period of SEO and this is combined with sub-optimal investment decision-making because investors underreact to the news (Andrikopoulos, 2009, p. 191).
There is still a discussion about the cause of this post-issue long-term underperformance. A possible explanation could be the underreaction hypothesis. Pre-SEO stocks could be overpriced because before the SEO, both investors and managers are overoptimistic with regards to future performance of the firm. This overoptimistic period creates opportunities for managers to announce the SEO and to exploit the overvaluation. The adverse selection model of Myers and Majluf (1984) gives an explanation for the direct fall in security prices at the beginning of the announcement (Choe, Masulis & Nanda, 1993; Eckbo & Masulis, 1995; Levis, 1995; Slovin,
Table 1
Panel A: Results of previous studies on SEOs in the United States
Study Date Methodology Sample n Pre-SEO: Post-SEO:
Period EW VW EW VW
Loughran & Ritter 1995 BH/a/s 1970-1990 3702 72.30% - -9.10% -
Loughran & Ritter 2000 CTAR/3FF 1973-1996 6461 - - -5.64% -3.84%
Brav, Geczy & Gompers 2000 BH/sb 1975-1992 3775 - - -3.90% -3.40%
Eckbo, Masulis & Norli 2000 BH/sb 1963-1995 3315 - - -4.80% -2.20%
-//- -//- CTAR/3FF -//- -//- - - -1.44% -2.04%
Jegadeesh 2000 BH/sb 1970-1993 2992 - - -4.90% -
-//- -//- CTAR/3FF -//- -//- - - -5.40% -
-//- -//- CTAR/4FM -//- -//- - - -3.72% -
Mitchell & Stafford 2000 BH/sb 1961-1993 4439 17.35% -0.55% -2.70% -1.10%
-//- -//- CTAR/3FF -//- -//- 15.36% 2.25% -3.96% -0.36%
Eberhart & Siddique 2002 BH/a/sb 1980-1992 189 - - -5.01% -
DeAngelo, DeAngelo & Stulz 2010 BH/sb 1975-2001 4291 - - -6.98% -
Panel B: Result of previous study on open offers in the United Kingdom
Study Date Methodology Sample n Pre-SEO: Post-SEO:
Period EW VW EW VW
Ngatuni, Capstaff & Marshall 2007 BH/a/s/si/sb 1991-1995 132 15.34% - 11.22% -
This table shows the results of earlier studies on SEOs in the US and a recent study on open offers in the UK. This table is partly retrieved from table 1 of Andrikopoulos, 2009, p. 192. No other studies regarding long-term performance following open offers can be found. The returns are annualised. This is based on assumptions of Ritter (2003, p. 267). BH: Buy and Hold abnormal returns; CTAR: calendar time abnormal returns; a: firm matching approach; 3FF: Fama and French three-factor model; 4FM: Carhart four-factor model; s: size benchmarks; si: size/industry benchmarks; sb: size/book-to-market value of equity benchmarks; EW: equal-weighted portfolio; VW: value-weighted portfolio; n: number of sample firms in study.
Sushka & Lai, 2000). Investors perceive the SEO event as a signal of overvaluation. So after the SEO announcement, market prices will fall back again to the intrinsic level. If the price decreases even some years after the event, the assumption of market efficiency is violated because it could be perceived as a sign of underreaction on behalf of the investors.
Kabir and Roosenboom (2003, p. 21) mention that the information asymmetry hypothesis can explain the investors’ signal of overvaluation of an equity issue. Firm managers have private information on either investment opportunities or assets of the company. Therefore it is believed that the stock market knows less about the specific firm than the firm managers. The issuance of equity contributes to the convey of private information to the capital market. Myers and Majluf 12
(1984) show that managers with private information are incentivized to issue equity when stock is overvalued. Investors know that managers will avoid the issuance of shares that are undervalued. Therefore investors perceive the SEO event as a signal of overvaluation. In the literature this is known as information asymmetry hypothesis (Kabir & Roosenboom, 2003, p. 21).
The second possible explanation of underperformance in the post-issue period is based on the equiprobability arguments of Fama (1998). Fama (1998, p. 284) argues that the long-term overreaction and underreaction phenomena are not inconsistent with market efficiency. He states that the market overreacts as frequently as it underreacts. He also mentions that the expected value of long-term abnormal returns is zero. However, chance generates anomalies. These anomalies are deviations from zero. These deviations from zero can be in both directions (Fama, 1998, p. 286). Mitchell and Stafford (2000) show strong stock returns and earnings in the three year pre-SEO period. Stock prices tend to be too high at time of the seasoned equity offering if the market cannot understand that the growth in earnings will mean revert. This overreaction to growth in past earnings is corrected slowly in future periods if the market recognizes its errors (Fama, 1998, p. 286).
Eckbo, Masulis and Norli (2000) show a third alternative explanation. They test SEOs made by US firms during the period 1963-1995. They use macroeconomic variables in a six-factor model and they generate expected returns that are risk-adjusted. The authors document that the long-term underperformance of firms conducting seasoned equity offerings reflects that equity issuing companies are less exposed to systematic risk relative to the non-issuing matching firms. This can be explained as follows: leverage levels of issuing companies decrease, their default risk and exposures to inflation that is unexpected decrease, thus this leads to a decrease in the stocks’ expected returns of the issuing firms relative to the non-issuing matching firms. Furthermore, Eckbo, Masulis and Norli (2000, p. 251) state that SEOs significantly increase turnover (stock liquidity). This also lowers the expected returns of the SEO companies relative to the non-issuing matching companies. The authors criticize earlier work of Loughran and Ritter (1995). They state that the new equity puzzle is the result of poor risk controls when conducting the matched-firm technique. However, Jegadeesh (2000) mentions that the results also may be affected by including IPO firms in the matched-firm sample. IPOs historically lead to lower long-term returns than SEOs (Andrikopoulos, 2009, p. 193).
Finally, the method of measuring long-term abnormal returns plays an important role. Barber and Lyon (1997, p. 342) document that using a portfolio as benchmark to calculate long-term buy-and-hold abnormal returns will lead to some biases. The three observed biases are: (1)
new listing bias, (2) rebalancing bias and (3) skewness bias. In the studies of long-term abnormal stock returns, the benchmark portfolio includes new firms that start to trade subsequent to the event, while the issuing companies generally have a certain history of stock returns. This is called the new listing bias. A market index that is equally weighted is an example of a benchmark portfolio. The returns of this benchmark portfolio are generally calculated with the assumption of periodic rebalancing. However, the returns of the issuing companies are compounded without periodic rebalancing. This is the rebalancing bias. At last, the skewness bias is observed: long-term buy-and-hold abnormal returns are positively skewed.
Barber and Lyon (1997, p. 370) state that as a result of these three mentioned biases the long-term BHARs (buy-and-hold abnormal returns) are generally negatively biased, when using a reference portfolio as benchmark. Barber and Lyon (1997) conclude that using non-issuing benchmark firms with similar size & book-to-market ratios leads to test statistics that are well-specified. Matching on basis of benchmark companies with similar characteristics (e.g. size & book-to-market ratios) alleviates the skewness bias (abnormal returns are reasonably symmetric when these abnormal returns are calculated on basis of the benchmark firm approach), the rebalancing bias (the returns of the benchmark companies and issuing firms are compounded in the same way), and the new listing bias (benchmark companies and issuing companies must be listed in the month of event). Therefore, the benchmark method of this thesis is also based on the control firm approach of Ngatuni, Capstaff and Marshall (2007). They compared issuing firms to non-issuing benchmark firms: (a) size-matched, (b) size-industry-matched and (c) size-B/M-matched.
F. Market Sentiment
Deng, Hrnjíc and Ong (2012) mention that managers make seasoned equity offerings to take advantage of the share price that is inflated. Investors realize this and they revise the firm’s valuation downwards after announcement of the SEO. This impacts the share price negatively and negative returns will occur at the announcement day. During periods of high sentiment there are more overoptimistic investors and managers have more incentives to act opportunistically. Deng, Hrnjíc and Ong (2012, p. 10) conjecture that the mentioned behavior is stronger during times of high sentiment. During these times a higher probability of making SEOs and a higher overpricing of firms that make seasoned equity offerings are expected. Investors realize this and adjust for this behavior. Therefore, announcement of seasoned equity offering is even more bad news during times of high sentiment (Deng, Hrnjíc & Ong, 2012, p. 10). As mentioned earlier, during periods of high sentiment higher overpricing is expected of firms that conduct seasoned 14
equity offerings. When the sentiment deteriorates, valuations will go back to the intrinsic value over the long-term and the reversion to the intrinsic value is higher during these times of high sentiment. Then the post-issue long-term return is lower. Deng, Hrnjíc and Ong (2012, p. 12) state that there is a negative correlation between post-issue long-term return and investor sentiment.
III. Hypotheses
This section describes the two hypotheses of this research. As previously mentioned, there is sufficient literature regarding the long-term performance of firms conducting seasoned equity offerings in general. However, the literature on the long-term performance of companies using the issue technique open offer is scarce. The study of Ngatuni, Capstaff and Marshall (2007) is the only study of the long-term effect of open offers that can be found. They studied the 5-year post-issue period of open offers in the UK and their results showed long-term positive buy-and-hold abnormal returns. Besides, they found that firms making open offers instead of rights issues had good growth perspectives. Because there is no other literature available regarding the long-term performance of companies that choose open offers, the first hypothesis of this study is based on their results:
H1: Firms conducting open offers overperform in the long-term compared to non-issuing
peers.2
This study also takes a closer look at the difference between high and low sentiment periods. Note that there are also other reasons why a firm will make an SEO, e.g. need for liquidity and investment opportunity. So market sentiment plays a role in the decision to issue equity; yet it is only a consideration. It is not the driving force behind the decision to make an SEO. In this study, the bull cycle 1st of January 2002 to 31st of December 2006 is taken for the high sentiment period. For the low sentiment period, the bear cycle period 1st of July 2007 to 31st of July 2009 is taken.3
Deng, Hrnjíc and Ong (2012) state that managers conduct SEOs to benefit from the inflated share price. Investors realize that managers behave opportunistically. They revise the company’s valuation downwards after the announcement of the seasoned equity offering. This
2 The non-issuing benchmark firms are chosen on the basis of (i) size, (ii) size-industry and (iii)
size-book-to-market. See the next section (subsection B).
3
The reason for taking this crisis period is explained in the next section (subsection A).
15
has a negative impact on the share price and negative returns will be observed at the day of announcement. There are more overoptimistic sentiment investors during periods of high sentiment and managers are more incentivized to act opportunistically. The authors (Deng, Hrnjíc & Ong, 2012, p. 10) conjecture that the mentioned behavior is stronger during times of high investor sentiment. During periods of high investor sentiment, the authors expect higher probability of conducting seasoned equity offerings and higher overpricing of companies that make SEOs. Investors realize this and they adjust for this behavior. Therefore, announcement of seasoned equity offering is even more bad news during times of high sentiment (Deng, Hrnjíc & Ong, 2012, p. 10).
However, Ngatuni, Capstaff and Marshall (2007, p. 60) state that the overvaluation exploitation argument does not apply to the open offers because the shares in open offers are offered to the existing shareholders. Therefore, the conjecture of Deng, Hrnjíc and Ong (2012, p. 10) of higher overpricing of firms that make SEOs during times of high sentiment does not hold for open offers. This leads to the second hypothesis of this study:
H2: The long-term abnormal performance of firms conducting open offers during bull cycle
does not significantly differ from firms making open offers during bear cycle, when compared to non-issuing peers.
IV. Research Design
In this section the research design is presented. The data used for this study is described in subsection A. The sample characteristics and the research period are discussed as well. The empirical methodology is described in subsection B.
A. Data Sample
As the base sample, seasoned equity offerings listed on the database Thomson ONE, accessible at the Pierson Révész library of the University of Amsterdam, are selected. This is done for two time frames: 1st of January 2002 to 31st of December 2006 (bull cycle, pre-crisis period) and 1st of July 2007 to 31st of July 2009 (bear cycle, crisis period). Ryan (2008) states that the credit crunch did not really begin until the second wave. This second wave started in July 2007. However, this crisis ended mid-2009 (Financial Crisis Inquiry Commission, 2011). Therefore, the crisis period July 2007 – July 2009 is taken and compared with the pre-crisis period 2002 – 2006.
Seasoned equity offerings are selected that meet the following three main criteria: (1) the offerings are made in exchange for cash, (2) the issue was of ordinary shares of NASDAQ, NYSE or NYSE Amex listed firms and (3) offered to existing shareholders. Besides these main criteria, some specific exclusions are also made in Thomson ONE:
- Initial Public Offerings (IPOs); - Rights issues;
- Unit offerings;
- American Depositary Receipts (ADRs), American Depositary Shares (ADSs), Depositary Receipts (DRs);
- Shelf filings;
- Real Estate Investment Trusts (REITs); - Simultaneous issues of any kind; - Private placements;
- International offerings; - Firm-commitment offerings.
Furthermore, all SEO firms that were either delisted or taken over in the 3-year post-offering period are excluded. The companies of the initial sample are intersected with the CRSP and Compustat databases to obtain all needed data. Compustat and CRSP database (available at the Pierson Révész library of the University of Amsterdam) provide data on company characteristics and returns, respectively. All observations with missing values (‘’bad records’’) are deleted from the sample. This will lead to some kind of sample selection bias. Besides, seasoned equity offerings by the same company during the three years after an SEO that is in the sample are excluded, following Healy and Palepu’s (1990) procedure. Finally, firms that conduct open offers are included if the company has not made any kind of seasoned equity offering or IPO within the previous three years (Loughran & Ritter, 1997, p. 1826).
This leads to a total sample of 93 open offers, during the periods 2002 – 2006 and July 2007 – July 2009. From these 93 open offers, 54 offerings were announced during the bull cycle 2002 – 2006 and 39 offerings were announced during the bear cycle July 2007 – July 2009. Please refer to figure 1 for the distribution of the sample of open offers under examination. This table shows that 26 of the 39 open offers during the bear cycle are made in the period January 2009 – July 2009. So two-thirds of the open offers during the crisis period July 2007 – July 2009 are made during the relative small time frame January 2009 – July 2009. Furthermore, it can be seen that only 35 of the total 93 open offers are conducted during the period 2004 – 2008.
The final sample of 93 open offers during both periods is categorized into seven industrial sectors, the sample of 54 offerings during the bull cycle also into seven sectors and that of 39 offerings during the bear cycle into five sectors. These industrial classifications are based on SIC (Standard Industrial Classification) codes. Please view figure 2, 3 and 4 for the industrial categorization of the three mentioned samples. These figures show that the three samples of this study consist mostly of firms in the finance, insurance, real estate sector and in the manufacturing sector. The three samples consist for about two-thirds of firms in these two sectors. However, fewer firms that make open offers are active in the mining, retail trade, services and construction sector. For the total 93 sample firms this is just 15.06%. For the 54 offerings during the bull cycle this is 16.66%. The 39 offerings during the bear cycle are categorized into five sectors and just 12.82% consists of firms active in the mining or construction sector. For more specific classifications, see appendix I and II. Appendix III, IV and V also show the frequencies of the industry sectors of the open offer firms.
In panel A of table 2, there are 93 open offers that satisfy the mentioned criteria for the periods 2002 – 2006 & July 2007 – July 2009. In panel A of table 3 and 4, 54 open offers can be seen for the period 2002 – 2006 and there are 39 open offers for the period July 2007 – July 2009. Furthermore, panel A shows the distribution of the open offers (NYSE or NASDAQ). Panel B presents the summary statistics. The median offer price for open offer firms during the periods 2002 – 2006 & July 2007 – July 2009 is $19.89, during the period 2002 – 2006 is $25.00 and during the period July 2007 – July 2009 is $12.68. The median shares offered for open offer firms during the periods 2002 – 2006 & July 2007 – July 2009 is 3,307,582, during the period 2002 – 2006 is 4,076,550 and during the period July 2007 – July 2009 is 2,716,350. The median offer amount for open offer companies during the periods 2002 – 2006 & July 2007 – July 2009 is $60.65 million, during the period 2002 – 2006 is $75.53 million and during the period July 2007 – July 2009 is $31.58 million. Table 2, 3 and 4 also present the mean offer price, mean shares offered and mean offer amount of the three samples of this study.
Figure 1
Distribution of sample of open offers under examination
18 15 17 6 4 12 4 9 26 0 5 10 15 20 25 30 2002 2003 2004 2005 2006 July 2007
Figure 2
Industry breakdown of 93 open offer firms in bull cycle (2002 – 2006) and bear cycle (July 2007 – July 2009)
Figure 3
Industry breakdown of 54 open offer firms in the bull cycle (2002 – 2006)
19 1,85% 25,93% 38,89% 3,70% 1,85% 9,26% 18,52%
Industrial Classification
ConstructionFinance, Insurance and Real Estate Manufacturing
Mining Retail Trade Services
Transportation, Communications, Electric, Gas and Sanitary service 3,23% 33,33% 36,56% 5,38% 1,07% 5,38% 15,05%
Industrial Classification
ConstructionFinance, Insurance and Real Estate Manufacturing
Mining Retail Trade Services
Transportation, Communications, Electric, Gas and Sanitary service
Figure 4
Industry breakdown of 39 open offer firms in the bear cycle (July 2007 – July 2009)
Table 2
Number of open offers by year or specific period (bull cycle 2002 – 2006 & bear cycle July 2007 – July 2009)
Panel A
Total Open Offers Open Offers Open Offers
Year Open Offers Percentage of Sample NYSE AMEX NASDAQ
2002 15 16.1% 9 0 6 2003 17 18.3% 9 0 8 2004 6 6.5% 4 0 2 2005 4 4.3% 0 0 4 2006 12 12.9% 1 0 11 July 2007 – Dec 2007 4 4.3% 0 0 4 2008 9 9.7% 3 0 6 Jan 2009 – July 2009 26 27.9% 13 0 13 Total 93 100% 39 0 54 Panel B
Variable Mean Median
Offer Price ($) 22.08 19.89
Shares Offered 6,361,299 3,307,582
Offer Amount ($ million) 117.84 60.65
20 5,13% 43,59% 33,33% 7,69% 10,26%
Industrial Classification
ConstructionFinance, Insurance and Real Estate
Manufacturing
Mining
Transportation, Communications, Electric, Gas and Sanitary service
Table 3
Number of open offers by year (bull cycle 2002 – 2006)
Panel A
Total Open Offers Open Offers Open Offers
Year Open Offers Percentage of Sample NYSE AMEX NASDAQ
2002 15 27.8% 9 0 6 2003 17 31.5% 9 0 8 2004 6 11.1% 4 0 2 2005 4 7.4% 0 0 4 2006 12 22.2% 1 0 11 Total 54 100% 23 0 31 Panel B
Variable Mean Median
Offer Price ($) 26.04 25.00
Shares Offered 8,103,145 4,076,550
Offer Amount ($ million) 156.70 75.53
Table 4
Number of open offers by year or specific period (bear cycle July 2007 – July 2009)
Panel A
Total Open Offers Open Offers Open Offers
Year Open Offers Percentage of Sample NYSE AMEX NASDAQ
July 2007 – Dec 2007 4 10.2% 0 0 4
2008 9 23.1% 3 0 6
Jan 2009 – July 2009 26 66.7% 13 0 13
Total 39 100% 16 0 23
Panel B
Variable Mean Median
Offer Price ($) 17.90 12.68
Shares Offered 3,949,513 2,716,350
Offer Amount ($ million) 64.04 31.58
Table 2, 3 and 4 show the number of open offers by year or specific period of the three samples of this study. Panel A shows the sample of open offers conducted during the specific period. Panel B presents some descriptive statistics of the issuing companies.
Appendix VI, VII and VIII give a comparison of the summary statistics based on open offer firms and matched companies. The median pre-issue market capitalization for open offer firms during the periods 2002 – 2006 & July 2007 – July 2009 is $589.62 million. The median change in turnover is -4.45%, the median change in EBT is 16.75%, the median ROA is 1.81% and the median change in fixed assets is 7.64%. These tables also show other size measures: book equity, total assets, total cash, plant, property & equipment and capital expenditure. The summary statistics are given for all three samples of this study. Note that outliers could lead to significant differences between the mean and median values.
B. Methodology
Long-term market performance (three years after SEO announcement) for the issuers and benchmark firms is estimated using the BHRs (buy-and-hold) returns methodology. This methodology is based on Andrikopoulos’ (2009) methodology with a different benchmark approach. Instead of a benchmark portfolio approach, a benchmark firm approach is used. In the same manner as Ngatuni, Capstaff and Marshall (2007), three benchmarks are used for the non-issuing matching firms: (i) size, (ii) size & industry and (iii) size & book-to-market.
The SEO buy-and-hold return (BHR) for security i is calculated as:
BHRi,τ,k = ∏ (1 + 𝑟𝑘𝑡=𝜏 i,t) – 1 (1)
In which:
- BHRi,τ,k denotes the SEO buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- ∏𝑘𝑡=𝜏 denotes the product from t = τ to k = 6, 12, 18, 24, 30, 36 months (i.e. holding period) - ri,t denotes the raw return on security i in month t.
Similarly the buy-and-hold returns of the matching firms (i) size, (ii) size & industry and (iii) size & book-to-market are calculated as:
M(i)BHRi,τ,k = ∏ (1 + 𝑟𝑘𝑡=𝜏 i,t) – 1 (2)
M(ii)BHRi,τ,k = ∏ (1 + 𝑟𝑘𝑡=𝜏 i,t) – 1 (3)
M(iii)BHRi,τ,k = ∏ (1 + 𝑟𝑘𝑡=𝜏 i,t) – 1 (4)
In which:
- M(i)BHRi,τ,k denotes the (i) size-matched buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- M(ii)BHRi,τ,k denotes the (ii) size-industry-matched buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- M(iii)BHRi,τ,k denotes the (iii) size-B/M-matched buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- ∏𝑘𝑡=𝜏 denotes the product from t = τ to k = 6, 12, 18, 24, 30, 36 months (i.e. holding period) - ri,t denotes the raw return on security i in month t.
The BHAR (buy-and-hold abnormal return) is defined as the difference between the BHR of the open offer firm and the BHR of the matching firm (i) size, (ii) size & industry and (iii) size & book-to-market:
BHAR(i)i,τ,k = BHRi,τ,k - M(i)BHRi,τ,k (5)
BHAR(ii)i,τ,k = BHRi,τ,k - M(ii)BHRi,τ,k (6)
BHAR(iii)i,τ,k = BHRi,τ,k - M(iii)BHRi,τ,k (7)
In which:
- BHAR(i)i,τ,k denotes the (i) size-matched buy-and-hold abnormal return on security i over the holding period τ to k = 6, …, 36 months
- BHAR(ii)i,τ,k denotes the (ii) size-industry-matched buy-and-hold abnormal return on security
i over the holding period τ to k = 6, …, 36 months
- BHAR(iii)i,τ,k denotes the (iii) size-B/M-matched buy-and-hold abnormal return on security i over the holding period τ to k = 6, …, 36 months
- BHRi,τ,k denotes the SEO buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- M(i)BHRi,τ,k denotes the (i) size-matched buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- M(ii)BHRi,τ,k denotes the (ii) size-industry-matched buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months
- M(iii)BHRi,τ,k denotes the (iii) size-B/M-matched buy-and-hold return of security i for a holding period of k = 6, 12, 18, 24, 30, 36 months.
The ABHAR (average buy-and-hold abnormal return) for N firms over the holding period is calculated as:
ABHARτ,k = 1
𝑁∑𝑁𝑖=1𝐵𝐻𝐴𝑅i,τ,k (8)
In which:
- ABHARτ,k denotes the average buy-and-hold abnormal return for N firms over the holding period
- N denotes the number of securities
- ∑ Ni=1 denotes the sum from security i = 1 to security N
- BHARi,τ,k denotes the buy-and-hold abnormal return on security i over the holding period τ to
k = 6, …, 36 months.
To test for both, significantly negative and positive buy-and-hold abnormal returns, it is useful to conduct a two-sided t-test:
H0: ABHAR(τ; k) = 0 H1: ABHAR(τ; k) ≠ 0
For each holding period τ to k = 6, 12, 18, 24, 30, 36 months is tested whether the average buy-and-hold abnormal return is significantly different from zero. The following t-value formula is used to conduct a basic two-sided t-test:
tABHARt,k = 𝐴𝐵𝐻𝐴𝑅𝜏,𝑘∗ √𝑁𝜎𝐵𝐻𝐴𝑅𝑖,𝜏,𝑘 ~ t[df = N – 1] (9)
In which:
- tABHARt,k denotes the observed t-statistic for the ABHAR in period k months across N securities
- ABHARt,k denotes the cross-sectional mean of the BHARs in period k months for the sample
of N securities
- N denotes the number of securities
- σBHARi,τ,k denotes the cross-sectional standard deviation of the BHARs in period k months for the sample of N securities
- df = N – 1 means that there are N minus one degrees of freedom.
The observed t-values are compared with the critical t-values4 with the following degrees of freedom: N – 1. The null hypothesis is rejected if the observed t-value is greater than the positive critical t-value or lower than the negative critical t-value.
Long-term performance is measured by taking the difference between the returns of offering companies and the returns of matched non-offering companies. Matching non-issuing firms are only included if they have not participated in any SEO or IPO activity during the t – 3 years prior to the offering. Besides, all matching firms have to survive the three year post-issue period. The non-issuing matching firms that were either delisted or taken over in the 3-year post-offering period are excluded. They are replaced by the second closest non-issuing matching firm according to the three benchmarks used in this study:
(i) size;
(ii) size & industry;
(iii) size & book-to-market.
Ad (i) Size-matched benchmark
In the same way as Afleck-Graves and Page (1996), a non-issuing matched company with a market capitalization higher than the SEO company is chosen. The reason for this is that the expectation is that the size of the SEO firm will increase in the post-offering and pre-offering period. This increase in size is due to the SEO firm’s value of the issued shares and the pre-offering stock performance. So each SEO firm is matched with a non-issuing company that has the closest market capitalization at the year-end prior to the SEO. However, the market capitalization of the non-issuing firm has to be greater than that of the SEO firm.
4 The critical values at the 5 percent significance level for the total 93 cash offering firms during the periods 2002 –
2006 and July 2007 – July 2009, the 54 offerings during the period 2002 – 2006 and the 39 offerings during the period July 2007 – July 2009 are +/- 1.986, +/- 2.006, +/- 2.024 respectively. Those at the 1 percent significance level are +/- 2.630, +/- 2.672, +/- 2.712 respectively.
25
Ad (ii) Size-industry-matched benchmark
The same size criterion as above is adopted. However, the SEO firms are also matched by industry. There is only one exception; the firm with the second largest market capitalization in a specific industry (based on the four digit SIC code) is chosen if the SEO company has the highest market capitalization in that industry.
Ad (iii) Size-B/M-matched benchmark
There is a positive relationship between stock returns and book-to-market ratios in the United States (Fama & French, 1992, 1993). Therefore it is expected that SEO firms have higher market-to-book ratios than the non-issuing matched companies if there were pre-issue increases in share prices (e.g. Loughran & Ritter, 1995). Barber and Lyon (1997) found that the control firm approach based on market-to-book ratios and size was robust in all the considered situations of their research. The same approach is adopted in this study, and a market capitalization higher than 30% of the SEO firm’s market capitalization is chosen for the non-issuing matched companies.5 Then the non-issuing matching firms are ranked by market-to-book ratios.6 The SEO companies are matched with non-issuing firms on basis of an M/B ratio closest to but higher than that of the issuing firms. The matching is based on a minimization of the sum of the absolute percentage differences between M/B ratios and sizes of the SEO companies and the matched non-issuing companies.7
‘Level’ and ‘change’ models are used to estimate operating performance based on Barber and Lyon’s (1996) methodology. Expected operating performance for SEO companies that use ‘level’ type models is:
E(Pi,t) = PBji,t (10)
5 Barber and Lyon (1997) found that the 30% margin showed well-specified t-statistics. In this study this 30%
margin is also used. Using this criterion is not expected to lead to biased results since the matching is based on a minimization of the sum of the absolute percentage differences between M/B ratios and sizes of the SEO companies and the matched non-issuing companies. The M/B ratios are retrieved from Compustat. The market-to-book ratios are defined as the company’s market capitalization to net book value (net tangible assets). Book value of equity is defined as the difference between ordinary shareholder’s equity and total intangible assets.
6 The ratios at the accounting end are used for the matching. There are some differences in the accounting
year-end. These differences were ignored. It is not expected to lead to biased results since the randomization of the mismatch in accounting year-ends.
7 Another method is to start with the B/M ratio first and then to match on a size basis. Both methods are done by
Barber and Lyon (1997). However, the former method showed well-specified t-statistics in all the considered situations.
26
In which:
- Pi,t denotes the performance of the issuing firm i at time interval t
- PBji,t denotes the performance of the non-issuing benchmark firm of company i at time t and for the three different benchmark methods j = A, B, C (A = size, B = size & industry, C = size
& book-to-market)
- E(.) denotes the expectations operator.
Abnormal operating performance is the difference in annual performance for SEO firms and the benchmark firms:
AOPLi,t = Pi,t – PBji,t (11)
In which:
- AOPLi,t denotes the abnormal operating performance of company i at time t - (.)L denotes the superscript L to clarify this is a ‘level’ type model
- Pi,t denotes the performance of the issuing firm i at time interval t
- PBji,t denotes the performance of the non-issuing benchmark firm of company i at time t and for the three different benchmark methods j = A, B, C (A = size, B = size & industry, C = size
& book-to-market).
Similarly, a ‘change’ model is used to estimate abnormal operating performance. This is the difference between yearly changes in operating performance of the SEO companies and that of the non-issuing benchmark firms:
AOPCi,t = 𝑃𝑖,𝑡 − 𝑃𝑖,𝑡−1 𝑃𝑖,𝑡−1 - 𝑃𝐵𝑗,𝑖,𝑡−𝑃𝐵𝑗,𝑖,𝑡−1𝑃𝐵𝑗,𝑖,𝑡−1 (12)
In which:
- AOPCi,t denotes the abnormal operating performance of company i at time t - (.)C denotes the superscript C to clarify this is a ‘change’ type model - Pi,t and Pi,t-1 denote the performance of SEO firm i at time t and t – 1 year
- PBji,t and PBji,t-1 denote the performance of the non-issuing benchmark firm of company i at time t and t – 1 year and for the three different benchmark methods j = A, B, C (A = size, B =
size & industry, C = size & book-to-market).
Consistent with the ‘level’ model (8), the hypothesis tested is AOPCi,t = 0. This means that there
is no difference in the changes of operating performance between the SEO firms and the non-issuing matched benchmark firms. The key assumption is that the expected operating performance of SEO company i at time t should be equal to the lagged performance of the company and the operating performance change of the non-issuing matched benchmark company
ΔPBj i,t:
E(Pi,t) = Pi,t-1 𝑃𝐵𝑗,𝑖,𝑡 − 𝑃𝐵𝑗,𝑖,𝑡−1𝑃𝐵𝑗,𝑖,𝑡−1 + 1 (13)
E(Pi,t) = Pi,t-1(ΔPBji,t + 1) (14)
In which:
- Pi,t and Pi,t-1 denote the performance of SEO firm i at time t and t – 1 year
- PBji,t and PBji,t-1 denote the performance of the non-issuing benchmark firm of company i at time t and t – 1 year and for the three different benchmark methods j = A, B, C (A = size, B =
size & industry, C = size & book-to-market)
- ΔPBji,t denotes the relative change in operating performance of the non-issuing matched
benchmark company
- E(.) denotes the expectations operator.
Four accounting indicators are used to evaluate operating performance of the SEO firms. Please view table 5. The yearly change in turnover assesses the issuers’ potential future growth and the ability to generate cash inflow. The SEO firms’ profitability is also assessed by using ΔEBTi,t
(annual change in Earnings Before Tax) and ROA (Return on Assets). ROA is calculated by dividing EBT by the total book value of assets. This study also compares the yearly change in
fixed assets of SEO companies with the change of their non-issuing benchmark companies. This
is done to assess aggressive expansion of the firms.
From the mentioned four indicators, abnormal change in turnover, EBT and fixed assets are based on the ‘change’ type model, while excess ROA is assessed using the ‘level’ type model.
Table 5
Models of excess operating performance and accounting measures Panel A: models of excess operating performance
Model Models of excess operating performance Description
1 AOPCi,t = ΔPi,t – ΔPBAi,t Difference between yearly change in
sample firms’ performance and yearly change in matching size benchmark firms
2 AOPCi,t = ΔPi,t – ΔPBBi,t Difference between yearly change in
sample firms’ performance and yearly change in matching size & industry benchmark firms
3 AOPCi,t = ΔPi,t – ΔPBCi,t Difference between yearly change in
sample firms’ performance and yearly change in matching size & book-to- market benchmark firms
4 AOPLi,t = Pi,t – PBAi,t Level difference to matching size
benchmark firms
5 AOPLi,t = Pi,t – PBBi,t Level difference to matching size &
industry benchmark firms
6 AOPLi,t = Pi,t – PBCi,t Level difference to matching size &
book-to-market benchmark firms Panel B: accounting measures of operating performance
Accounting indicator Description
Change in Turnover ΔTi,t = (Ti,t – Ti,t-1)/Ti,t-1, the change in
turnover between t – 1 and t = 0
Change in Earnings Before Tax ΔEBTi,t = (EBTi,t – EBTi,t-1)/EBTi,t-1, the
change in earnings after interest and before tax between t – 1 and t = 0
Return on Assets (ROA) ROAi,t = EBTi,t/TAi,t, the EBTi,t divided by
the book value of assets of firm i at time t
Change in Fixed Assets ΔFAi,t = (FAi,t – FAi,t-1)/FAi,t-1, the change
in the total fixed assets of firm i between time t – 1 and t = 0; total fixed assets is defined as the sum of all fixed asset investments, property, intangible fixed assets and tangible fixed assets
Table 5 shows the models of excess operating performance and the accounting measures used in this study. Both ‘change’ and ‘level’ type models are used to capture the difference in performance between the firms that issue equity and the non-issuing benchmark firms. This is based on the methodology of Barber and Lyon (1996). Models
1 to 3 are based on the formula AOPCi,t = ΔPi,t – ΔPBji,t. In this formula, AOPCi,t is the abnormal operating
performance for SEO firm i at time t; ΔPi,t is the yearly change in operating performance of SEO firm i for time
interval t – 1 year to t = 0, and ΔPBji,t is the yearly change in operating performance for benchmark firm type j = A,
B, C. The benchmark firm types j = A, B, C correspond to the size, size & industry and size & book-to-market matched firms, respectively. Models 4 to 6 are used to calculate the level of operating performance at different times between the SEO firms and the corresponding benchmark firm types j = A, B, C.
Finally, based on Kabir and Roosenboom (2003) and Loughran and Ritter (1997), statistical significance of the difference in operating performance between SEO firms and non-issuing benchmark firms is determined by conducting the Wilcoxon signed-rank sum test:
H0: median difference between the pairs is zero (the related samples are drawn from the
same distribution)
H1: median difference between the pairs is not zero (the related samples are not drawn from
the same distribution)
N is the number of pairs. For i = 1,…, N let x2,i and x1,i denote the measurements of the SEO
firms and non-issuing benchmark firms, respectively.
1. For i = 1,…, N, |x2,i - x1,i| and sgn(x2,i - x1,i) are calculated, where sgn denotes the sign
function.
2. Exclude pairs with |x2,i - x1,i| = 0. Nr denotes the reduced sample size.
3. Remaining pairs are denoted by m. They are ordered from smallest to largest absolute difference |x2,i - x1,i|.
4. The pairs are ranked. The smallest rank is 1. Ties are equally ranked to the average of their ranks. Ri denotes the rank.
5. The test statistic W is calculated:
W = |∑ [𝑠𝑔𝑛𝑚𝑖=1 (x2,i - x1,i) * Ri]| (15)
In which:
- W denotes the absolute value of the sum of the signed ranks - ∑ mi=1 denotes the sum from pair 1 to pair m
- sgn denotes the sign function
- x2,i and x1,i denote the measurements of the SEO firms and non-issuing benchmark firms, respectively
- Ri denotes the rank.
6. The distribution of W converges to a normal distribution as Nr increases. Thus, for
Nr≥ 10, a z-score is calculated as follows:
z = 𝑊−0.5
𝜎𝑤 (16)
where
σw = �𝑁𝑟(𝑁𝑟+1)(2𝑁𝑟+1)6
In which:
- z denotes the observed z-statistic
- W denotes the absolute value of the sum of the signed ranks - σw denotes the standard deviation of W
- Nr denotes the reduced sample size.
The observed z-values are compared with the critical z-values.8 The null hypothesis is rejected if the observed z-statistic is greater than the positive critical z-value or lower than the negative critical z-value.
Finally, in order to look for a significant difference between the crisis firms and the pre-crisis firms, the following ordinary least squares (OLS) regressions are run for the total sample of 93 cash offerings and for all three types of benchmark firms (A = size, B = size & industry, C = size & B/M):9, 10
8 The critical values at the 5 percent significance level for all samples are +/- 1.960. The critical values at the 1
percent significance level for all samples are +/- 2.576.
9
The regressions are run with software program STATA12 and the standard error is corrected for heteroscedasticity.
10 Summarizing, the regressions are based on the following assumptions (Unesco, 2013):
• There is a linear relationship between the dependent variable and the independent variables. • The expected value of the error term is zero.
• The variance of the error term is constant for all the values of the independent variables. This is the assumption of homoscedasticity.
• There is no autocorrelation.
• The independent variables are uncorrelated with the error term.
31
Table 6
Classification of industry dummies used in the regressions
Industry Industry Dummy of SEO firm
Mining (1000-1499) D1
Construction (1500-1799) D2
Manufacturing (2000-3999) D3
Transportation, Communications, Electric, D4
Gas and Sanitary service (4000-4999)
Retail Trade (5200-5999) D5
Finance, Insurance and Real Estate (6000-6799) D6
Services (7000-8999) D7
This table shows the definition of the industry dummy variables used in the regressions. Four digit SIC codes are reported in parentheses. BHARi = β0 + β1Dcrisisi + β2D1i+ β3D2i+ β4D3i+ β5D4i+ β6D5i+ β7D6i+ β8D7i + ui, i = 1, 2, 3 years (17) AbnormalChangeInTurnoveri = β0 + β1Dcrisisi + β2D1i+ β3D2i+ β4D3i+ β5D4i+ β6D5i+ β7D6i+ β8D7i + ui, i = 0, 1, 2, 3 years (18) AbnormalChangeInEBTi = β0 + β1Dcrisisi + β2D1i+ β3D2i+ β4D3i+ β5D4i+ β6D5i+ β7D6i+ β8D7i + ui, i = 0, 1, 2, 3 years (19) AbnormalROAi = β0 + β1Dcrisisi + β2D1i+ β3D2i+ β4D3i+ β5D4i+ β6D5i+ β7D6i+ β8D7i + ui, i = 0, 1, 2, 3 years (20) AbnormalChangeInFixedAssetsi = β0 + β1Dcrisisi + β2D1i+ β3D2i+ β4D3i+ β5D4i+ β6D5i+ β7D6i+ β8D7i + ui, i = 0, 1, 2, 3 years (21) In which:
- i denotes the timeframe 0, 1, 2 or 3 years
- BHAR denotes the buy-and-hold abnormal return of SEO firms and non-issuing benchmark firms
- AbnormalChangeInTurnover denotes the excess change in turnover of SEO firms and non-issuing benchmark firms
• The error term is normally distributed.
32
