What factors determine the effect of dual class share structure on firms` operating performance? : the influence of technology, leverage, issues type and economic level

(1)

Zhelun Li 11021209 August 2016

Supervisor: Ryan van Lamoen Second reader:

What factors determine the effect of dual class share

structure on firms` operating performance?

The influence of technology, leverage, issues type and economic level

MSc Business Economics, Finance track

Master Thesis

(2)

1

Statement of Originality

This document is written by Student Zhelun Li, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

Abstract

This paper examines how factors, such as technology, leverage, issues type and economic level of a country, influence the effect of dual class share structure on firms` post-issuing operating performance. Specifically, a sample of 2989 firms worldwide is employed with either one-vote one-share standard or a dual class shares policy from 1990 to 2015. A two-step methodology is adopted to investigate the effects of a dual class share policy on operating performance. First, the post-issuing cumulative abnormal operating returns are estimated mainly by using companies` accounting data. The influences of each specific factor such as technology, leverage, issues type and economic level of a country is tested by

employing interaction terms. Results of both steps seem to be robust. Generally speaking, firms that are employing the dual class share structure appear to underperform compared to firms with the one-share one-vote structure. In particular, the empirical evidence supports that both high technology firms and high leveraged firms tend to deteriorate the function of a dual class share structure. However, an advanced economic level of a country typically

strengthens the value of such a structure. However, we find no significant proof that issues type – IPO or follow-on issues – matters for the operating performance under a dual class share structure. Finally, the conclusions of this study are of great importance to stock exchange regulators and government policy makers when firms request to issue dual class shares. The results of this study suggest that criteria can be built to exploit the beneficial side of the structure. For instance, restrictions could be formulated for high technology or high leveraged firms while flexible and circumstantial policies could be laid down for companies operating in advanced economies.

(4)

3

1. Introduction

In September, 2014, Alibaba Group Holding Limited, China`s dominant E-commerce

company, began its Initial Public Offering (IPO) on the New York Stock Exchange (NYSE), which is the biggest IPO in the U.S. stock market history. One of the hot topics debated in public before the IPO was Alibaba group`s proposed governance structure, namely the dual class shares structure, which has been rejected by Hong Kong Stock Exchange and thought broadly to be the key reason that Alibaba chose to go public in the United States. Briefly, the dual class shares structure allows a company to issue two classes of common stocks with different voting power, namely class A shares held by the public with only one vote per share and class B shares held by insiders of the company with more than one vote per share.

Usually class B shares are non-traded and have 10 votes per share. Such a new shareholder structure typically violates the “one share one vote” principle and generates a wedge between cash-flow rights and voting rights. Like a two-side sword, great public debates regarding the positive and negative impacts of the new capital structure on post-issuing operating

performance has been raised up for a long period. Generally, the dual class share structure is popular among internet, media and high technology industries as companies in those

industries depending heavily on innovations would hardly generate short-term benefits. Company owners` decisions with long-term vision would be badly violated by potentially fickle-minded and short-sighted public investors. Proponents of dual class shares structure would argue that such a governance structure would enable knowledge-embedded and

insightful management to induce projects which would benefit in the long run without fear of losing control and such decisions would be ultimately beneficial to all shareholders.

Examples of Google glass and Google driverless car could be cited as beneficiaries of the dual class share structure.

However, protectors of the traditional single class share structure also propose comments against the dual class share structure. Agency and entrenchment problems due to lack of market checks on managerial misconducts under such a structure have been frequently

mentioned as criticism in public. Through the separation between cash-flow rights and voting rights, the mechanism of one-share one-vote that allows shareholders to easily and efficiently

(6)

5 monitor performance of directors would fail to work. By transferring risks and potential negative consequences to other shareholders, management teams are more likely to make decisions less prudently by bearing less cost. Moreover, evidence from an empirical study by Alon Brav, Wei Jiang, and Lucian Bebchuk (2015) on the long-term effects of hedge fund activism shows that activist shareholders` interventions in nature help to improve operating performance of firms lagging behind peers consistently in the long run, which challenges previous supportive argument. An example is the case of Hollinger International`s financial and share performance that suffers under former CEO Conrad Black`s control, who owned 30% of the equity and 73% of the voting power.

Under the conflicts of publicly debated arguments and disagreements among previous empirical studies, Adams & Ferreira (2008) suggested that more work is still needed to identify the circumstances in which the effect of the dual class share structure on operating performance is positive and those in which it is not. In this empirical research, interaction terms would be employed to investigate how different factors such as technology, leverage, issues type and economic level influences the effect of the dual class share structure on post-issuing operating performance. The difference in the methodology compared with previous literatures is that we examine all these factors in one framework and meanwhile we take reference to the methodologies to estimate the normal price returns of stocks after certain events under different pricing models such as CAPM model, market model, mean returns model etc. (MacKinlay 1997). Besides, another contribution of this empirical research paper is the inclusion of the economic circumstance in a country as a factor that may influence the effect of dual class structure on operating performance. Furthermore, more recent data is used compared to other studies by expanding the data range to include new-issues that happened in recent years from 1990 to 2015. Therefore, the results become more relevant to making policies and regulations nowadays. The results would contribute to help stock exchanges regulators and government policy makers to identify restricting criteria for dual class share structure oriented at new issues. These restricting criteria could be based on specific firm characteristics or economic environment in which a firm operates in a high technology industry or a developed country. The main research question we would like to answer is whether factors such as technology, leverage, issues type and economic level influences the effect of dual class shares structure on the post-issuing operating performance.

The remainder of this paper is organized as follows. In section 2, the literature review and hypotheses that are examined in this study will be presented. Section 3 will contain the

(7)

6 sample design and descriptive statistics including the process of deriving the new-issues and relevant financial data. Section 4 will illustrate our two-step methodology and section 5 shows the empirical results together with analysis. In section 6, the robustness of our results will be proved. In section 7, the conclusions will be conducted and discussions on further research will be included. Finally, the limitations regarding the data source and intrinsic characteristics of the database will be raised up in section 8.

2. Literature Review and Hypotheses

2.1 Literature review

Jain & Kini (1994) found evidence that IPO firms where entrepreneurs retain higher

ownership generally demonstrate superior post-IPO operating performance relative to other issuing firms both before and after the adjustment for industry effects. However, the authors also indicate in the conclusion that it is not possible to tell by their empirical results if the relatively superior operating performance occurs as a result of lower agency conflicts when there is higher ownership retention. Nevertheless, the higher equity retention by the original entrepreneurs can be obtained by both the one-share one-vote structure and the dual class shares structure. The difference would be that the so-called insider controllers (the vote controllers of the firm) enjoy a relatively lower cost to maintain control through the latter structure. Therefore the research question in this paper starts with the doubt whether the dual class share structure, which comprised only a small part of the total sample firms with higher ownership retention, would violate the empirical results of the post-IPO operating

performance of superior ownership centralized firms (as stated in Jain & Kini`s conclusion). As discussed in the CFA Institute article written by Matt Orsagh (CFA, CIPM) in the year of 2014, related research has shown that companies with voting power controlled by a founder, a family or other entity perform better for minority shareowners and controlling shareowners under one-share one-vote structure. He also backing the argument by referring to the

conclusion from a study report published by the Investor Responsibility Research Center (IRRC) in which it is restated by Matt as “on average, and over time, companies with dual-class shares underperform those with a one-share, one-vote standard in which the owner’s

(8)

7 economic risk is commensurate with his voting power.” Matt Orsagh concludes that the nature of control matters and in his point of view, a structure that shares the same link between economic risks and control among all share owners works best for all parties. However, what this study would like to further investigate is the identification of the factors that actually drive the ambiguous effect of dual class share structure on operating

performance. To reach this objective, the relevant conclusions in the existing literature are examined.

Ang & Megginson (1989) raised up two hypotheses for predicting the impact on firm performance from a dual class share structure. The two hypotheses are the Takeover Defense/Agency Cost Hypothesis (TDACH) and the Optimal Insider Control Hypothesis (OICH). The former one predicts that the insider controllers employ dual-class shares

structure primarily to keep their own positions safe from hostile takeovers. This will lead to a negative wealth effect, while the OICH predicts that restricted voting shares will be issued only by firms already dominated by wealth-maximizing insider controllers acting in the best interests of all shareholders. This will have non-negative wealth effects. The testing results from their empirical methodology tend to support the prediction of OICH and indicate that a dual class capitalization optimize the time needed by the firms` insider controllers to resolve asymmetric information problems. However, Wang, Xie, & Masulis (2009) used a sample of U.S. dual-class companies and found evidence to support the agency hypothesis which states that managers with excess control rights over cash flow rights are more prone to pursue private benefits at the shareholders` expense. One example would be making shareholder value-destroying acquisitions more often. Amoako-Adu & Smith (2001) conclude from their results by analyzing changes in capitalization and control of dual class share structure firms before and after the IPO that it is the combination of a large controlling shareholder with family interests rather than the management entrenching purpose that leads to dual class capitalization. Together with Amoako-Adu and Smith`s findings, Barontini & Caprio (2006) found no evidence to support the hypothesis that family control hampers firm performance in Continental Europe. Now that family control is not the key to the relatively worse post-issuing performance, what other factors drive the phenomenon?

Gompers, Ishii, & Metrick (2010) found evidence of a negative relation between firm value and the wedge between the voting rights and cash flow rights. Holmen & Nivorozhkin (2007) found that both the hazard rate of takeover and firm market value decline with the wedge between the families` voting rights and cash flow rights. Chemmanur & Jiao (2012) predicts

(9)

8 that dual class IPO firms are likely to outperform (underperform) single class IPO firms if the reputation of the incumbent is high (low) and the firm is operating in an industry where the difference in intrinsic values between the projects with high and low near-term uncertainty is large(small) in a theoretical paper. Even though both dual-class recapitalization and leveraged buyout can realize the controlling of votes in a company, Kenneth, Jeffry & Annette (1990) found evidence to support that firms with greater growth opportunities are more likely to consolidate control through dual-class recapitalizations. Michael & Eric (2008) conclude from their empirical results which based on a database of 613 Canadian firms from 1998 to 2005 that family owned firms with a single share class have similar market, accounting and leverage performance, but family owned firms with dual-class shares have valuations that are lower by 17% on average relative to widely held firms. Anete (2005) however found similar reasons for dual-class firms to switch back to one-share one-vote as they previously preferred going dual class share structure, i.e., the need to issue new equity and to defend a firm from a possible takeover, due to recent changes in the corporate governance environment. Anete also found evidence that the likelihood of unification decreases with the separation between control and cash flow rights and firm value increases after the unification. Scott and Chad (2003) found evidence that dual class IPOs are less underpriced than single class IPOs because management of dual class share structured firms have less incentives to increase outside ownership dispersion, combined with evidence that dual-class managers earn higher compensation, suggests that dual-class ownership structures protect private control benefits.

The main conclusion based on these papers is that the effect of dual class shares structure on firms` performance is ambiguous, other factors may influence the effects. In order to

investigate these factors, we made four hypotheses based on the existing arguments and theories where each hypothesis focuses on one factor. The four factors include technology, leverage, issues type and economic level. Moreover, most of the existing empirical literatures typically test the influencers individually and separately, this paper will test these factors in one framework. We will discuss them in details in the next section.

(10)

9

2.2 Hypotheses

Based on previous related literature and academic debates in public, we focus on four factors that are broadly thought to have a potential impact on the effect of a dual class share structure and post-issuing operating performance. Those factors are I. Technology, II. Leverage, III. Issues type, IV. Economic level of a country. Before we deeply investigate each factor through empirical methodologies, we first raise up four hypotheses regarding each factor based on related theories and public arguments.

Hypothesis I Dual class share structure has a positive effect on operating performance in a

high technology industry.

First of all, as broadly discussed in public, high technology firms are more likely to employ the dual class share structure by claiming that this special ownership structure tends to protect their knowledge-embedded decision of proceeding certain research programs from the short-vision speculators, examples are Google and Facebook. As internet or technology companies usually have innovations that would hardly generate short-term benefits, but would probably bring better future lives (Wang, 2016). The dual class share structure, which gives less cost to maintain the control, would allow knowledge-embedded and insightful management to induce projects without fear of losing control. If the argument is true and the motivation to introduce the dual class share structure by the management of the high technology firms is consistent with what they claimed, it is reasonable to predict that new dual-class equities issued by firms operating under technology-oriented industries would give impact of a superior cumulative post-issuing operating returns than the ones issued by firms in other industries. If the Hypothesis I can be proved to be true by our regression results analyzed in section 5, we would suggest to support the high technology firms to issue the dual-class new equites as they are beneficial to improve the operating performance and profit to all

(11)

10

Hypothesis II Dual class share structure has a negative effect on operating performance in a high leveraged industry.

Secondly as mentioned by Martin and Eugene (2007), firms adopting the dual class share structure are less likely to have high leverage compared to firms with the one-share one-vote structure. As Stulz (1988) pointed out, the potential explanation to such a phenomenon could be that debt can be used as a device that allows current owners to retain control of their firms, however the management team – the vote controllers, of firms with the dual class shares structure can take advantage of their prior voting rights to realize the need to control

decisions rather than relying on debt. As the dual class shares structure can already offer the less-costly way to access to control, if we believe that the management team pursues to control the firm only for processing their considerable and insightful ideas rather than exacting private benefits, there should be no motivations from them to remain a higher leverage ratio to realize the control under the dual class share structure. The potential reason that these firms issue new equities could be that they want to use the new fund to cover the debt rather than improving the performance. Therefore if we observe firms with the dual class share structure together with a high leverage ratio, it is rational to doubt the purpose of the new-issues. There is the possibility that the firm issues the new equities mainly aiming at raising equity instead of improving operating performance. Therefore, we would expect a relatively lower post-issuing abnormal operating performance. In such a case it can be

recommended that restrictions should be put on dual-class new issues requested by firms with high leverage.

Hypothesis III Dual class shares structure has a negative effect on operating performance

if the new equity is issued as follow-on.

Third, as what Nicole Sandford, partner, Deloitte & Touche LLP, the leader of the National Governance Services practice, argued in the article of Dual-class Share Structure: Weighing the Risks and Rewards from The Wall Street Journal, under the requirements of the major stock exchanges, a dual-class structure should be put in place at the time of the IPO rather than afterwards. Because the dual class shares structure tend to weaken the role of board of directors and hamper its ability to execute its fiduciary responsibility to shareholders,

(12)

11 members. If the bad influences indeed happened, firms issuing dual-class IPO will typically suffer lower performance afterwards due to the relatively low quality of the board members. If our empirical results displayed in section 5 shows to be consistent with Hypothesis III, we would suggest to imply more restricted principles when regulating these dual-class IPOs.

Hypothesis IVDual class share structure has a positive effect on operating performance in

an advanced economy.

Finally, as so far no other literature has reached the field of economic level of a country, but the developing degree of the economic environment under which the dual class shares

structure being implemented has large potentials to have impact on the post-issuing operating performance. As most of the economists believe that highly developed economies tend to have more comprehensive regulations and protection laws on the management team to exploit the advantages of the dual class shares structure. Therefore, we predict that firms being

regulated in advanced economies would benefit from issuing the dual-class new equities compared to less advanced economies. Our research would be the first one to test the

empirical evidence of the topic. If Hypothesis IV can be proved to be true with the regression results, we would suggest to support firms being regulated in advanced economies to issue dual-class new equities since this structure tend to improve their performance and benefit to all shareholders.

3. Sample design and Data

3.1 The new-issues sample

The new-issues sample consists of two subsamples, one subsample concerns new issues with the dual class share structure and the other subsample concerns a non-dual class share

structure. Both of them are constructed from the Global New Issues Database of SDC Platinum under Thomson One with a date range from 1st January 1990 to 31st December 2015. The Global New Issues Database has been tracking corporate new issues activity since 1970 and flagging those issuing additional classes of common stock who already have

(13)

12 existing separate classes of common stocks. The raw sample includes both IPO issues and follow up and flags those IPO issues. This study identifies different companies by using the Ticker Symbol and subsequently merge the financial data, we will describe the process of obtaining the financial data more in details in the next section. To further reach the actual list of new issues available for this study, we exclude LBO firms, Closed-end Fund/Trust, Unit Investment Trust and Fund or Trust Issues, as these entities have the potential to violate the conclusions, we refer the excluding process to how most of the other relevant literatures proceed, for example, Paul, Joy and Andrew (2010), Scott, Ramabhadran and Chad (2008) and Valentin and Prem (2006). Besides, the sample selected in our research allows multiple new-issues by the same firm. The new-issues events are identified by grouping sets of data rather than firm entities.

Since the abnormal operating performance is measured by the cumulative abnormal operating return on asset, which is computed by the difference between the real operating return and the predicted normal operating return during the issuing year based on the regression analysis, a balanced database is necessary for the prediction model. We first drop all new issues with financial data with less than 15 consecutive years previous to the issuing year and less than consecutive 3 years after the issuing year. The ex-issuing year data are used for predicting normal operating return while the post-issuing year data are used for computing cumulative abnormal operating return. Then we drop the new issues as neither IPO nor follow-on.

3.2 Financial data

The financial data are obtained from WRDS-Compustat (global) for the fiscal years 1985-2016. In addition to the regular elements to compute EBIT (operating income) on income statement, we also include data of total assets, total liabilities, employees, acquisitions and SIC codes. The financial data is collected 15 years earlier previous to the starting point of the new-issues sample because we need 15 years` data before the new-issue event of firms` characteristics to determine the normal operating performance during the event window if no such an event happened. More about the process is described in section 4.1, Methodology Step 1.

(14)

13 Based on the raw financial data, three of the variables are constructed by the following

computations: 𝑂𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 𝑜𝑛 𝑎𝑠𝑠𝑒𝑡𝑠: 𝑂𝑅𝑂𝐴 = 𝐸𝐵𝐼𝑇 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡 (1) 𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑎𝑡𝑖𝑜: 𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒 = 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 − 𝑡𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠 (2) 𝑓𝑖𝑟𝑚 𝑠𝑖𝑧𝑒: 𝑠𝑖𝑧𝑒 = ln (𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠) (3)

In order to avoid the potential violation from outliers, we manually exclude data sets with extremes value by each variable. Since the subsample of new issues with dual class share structure is relatively small and where our research questions focus on, ideally we would like to select a corresponding new issues sub-sample of single class share structure with similar size and firms characteristics for comparison. However, the reality is that it is difficult to obtain such a perfect corresponding sub-sample without violation of random selection rule. What we can do based on the situations is only excluding data sets with variable values out of the range within dual-class sub-sample. In this way, we can on the one hand minimize the violations from outliers and on the other hand contract the sub-sample size of single-class firms relative to dual-class sub-sample. Besides, since we intend to employ the econometric methodology of firm level estimations with loops to estimate the post-issuing normal

operating returns for each event (time series analysis), a balanced database is essential as the data in each loop is limited.

3.3 Descriptive statistics

In the final sample in our design, following the selection criteria described in the last two sections, we have 2989 sets of data in total for the regression analysis, each set corresponds to one event of the new-issues, and the financial data collected for each event starts from the minus 15th year before the issuing year until the 3rd year after the issuing year consecutively. The average values for each variable within each issues event are shown in table 1 Panel A

(15)

14 below. In Panel B, C and D, we show the summary statistics of the cumulative abnormal operating performance by different event windows. As we can see that the mean of the cumulative abnormal operating performance increases more than double with the size of event window, we can roughly conclude that the magnitudes of the abnormal operating performance show up to be even larger in the later years

Table 1. Summary Statistics

Panel A: Summary statistics for financial data (average value of each issuance)

variables Obs Mean Std. Dev. Min. Value Max. Value

EBIT/total asset 2989 0.09 0.08 -0.88 0.32 Leverage 2989 2.47 2.57 -20.28 21.94 Total liabilities 2989 2575.72 6036.31 0.96 51992.37 Total assets 2989 4367.89 10394.67 5.53 93209.31 Firm size 2989 6.48 2.06 1.69 11.23 Sales 2989 4362.88 10651.62 5.28 94479.05

Cost of goods sold 2989 3067.00 8349.13 1.81 86822.36

Cost of selling, general &administration 2989 677.20 1766.69 0.81 18936.74 Employees 2989 15.28 31.23 0 685.06 Depreciation & amortization 2989 187.66 477.49 0.15 5399.26 Acquisition 2989 111.98 234.81 -27.97 2273.11

(16)

15 Panel B: Summary statistics for sum abnormal return, T= +1

Cum_ab_OROA 2989 0.01 1.37 -20.96 23.09

Panel C: Summary statistics for sum abnormal return, T= +2

Cum_ab_OROA 2989 0.05 2.01 -21.69 28.36

Panel D: Summary statistics for sum abnormal return, T= +3

Cum_ab_OROA 2989 0.07 2.96 -38.56 54.80

In order to test the 4 hypotheses discussed in section 2.2, we need to classify the new issues in the sample based on criteria that they are issued by high or low technology firms, by high or low leveraged firms, in advanced or developing economies and as IPO or follow-on issuing. We will describe briefly the process how we identify high technology firms, high leveraged firms, issues of initial public offers and advanced economies.

First of all, the high technology firms are identified based on the SIC codes, SIC is short for standard industrial classification, which is widely known by public and popularly used in academic research. This study will follow the classification of Hecker (1999), in which high technology industries are defined thoroughly. Appendix I List of High Technology Industry provides an overview of the classification. We then recognize the high leveraged firms as the leverage ratio lying within the top 25% among the new issues sample. Finally, we distinguish advanced economies according to GDP size in the country in which the company is

(17)

16 regulated. A full list of economies classification is available in Appendix II List of Advanced Economies. IPO issues have been automatically labelled in the raw database.

Table 2 below displays the factors distributions among sample, as we can see, in the total new-issues sample, 6.35% of them are issued under dual class shares structure, around one third are issued by high technology firms and one forth approximately are issued by high leveraged firms. Roughly 95% of the new issues are issued in advanced economies and about 7% are issued as first time IPO. Basically, the distributions of each factor among new-issues sub-sample of dual class shares structure and non-dual class share structure approximately maintain at the similar level. Specifically, non-dual sample includes more high technology and advanced economies-oriented new issues while dual sample contains more issues by high leveraged firms and IPO issues.

Table 2. Factors Distributions

Dual issues High_tech High_leverage IPO issues Advanced economy % of total sample 6.46 32.12 24.62 6.76 95.55 % of dual sample 100 20.73 33.68 8.29 82.38 % of non-dual sample 0 32.90 24.00 6,65 96.46

(18)

17

4. Methodology

Generally, the econometric methodology of Event Study will be employed as the main research technique in this paper. The main focus of this research is on the new-issues of equities between 1990 and 2015, either by initial public offers or follow-on (secondary

issues) and either in the form of dual class or non-dual class shares structure. We set the event year as the time during which the new stocks were issued, to be clearer, we call the event year as the issuing year as well through the discussion in the paper. The event windows will be opened for one, two and three years just after the issuing year of each new-issue for sensitivity analysis and the estimation windows will be opened for up to 15 years exactly previous to the issuing year. In this paper, we will raise up the concept of the effective post-issuing period, which basically corresponds to the different lengths of the event windows. As a matter of fact, the methodology of the Event Study we employed in this paper comes from the ideas of the methodology to estimate the abnormal stock prices by stock pricing models such as CAPM model, which is commonly used in many other relevant empirical literatures, e.g. Yen-Sheng and Ralph (1987). MacKinlay discusses the methodology in details in the paper “Event Studies in Economics and Finance” (1997). The only shortcoming of applying such a methodology to our research is that the available size of the financial data before and after the event time – it would be the issuing year in our case and the event date in the stock pricing case – is limited as our financial data will be collected yearly instead of daily, we will discuss more about the drawbacks of this approach in the last section – Limitations.

Overall, two steps will be followed up during the process of empirical analysis. We will first obtain the cumulative post-issuing abnormal operating returns in each effective post-issuing period by computing the difference between the actual returns and the estimated normal returns. As mentioned in section 3.2 Financial data, the normal return is defined as the return if we assume the new-issues event did not happen, which will be estimated by the previous 15 years` data of firms` characteristics. Only after generating the cumulative abnormal

returns, we are able to continue to proceed the second step, in which the cumulative abnormal returns derived from step 1 will be used as the dependent variable. In the second step, we can tell how investigated factors affect the cumulative abnormal returns through the estimates of the coefficients in the OLS regressions with cross sectional data analysis. Step 2(A) will

(19)

18 illustrate how we test the general effects of each factor on the cumulative abnormal returns. Step 2(B) will explain the methodology we used to answer the core research question of this paper – the dual-class effects of each factor on the cumulative abnormal returns, to be clear of the concept, the dual-class effects mentioned in the discussion refer to the effects of factors on the cumulative abnormal returns if the new equities are issued under dual class share structure.

4.1 Step 1: Estimating the cumulative abnormal operating performance

Specifically, we have balanced 15 years of financial data for each firm previous to the issuing year in our database, we use these data to estimate the post-issuing normal operating

performance of firms for each new-issues event during each post-issuing effective period (event window). The performance will be measured by the operating return on assets, which is called OROA for short through the paper. The formula to generate OROA in the database is shown as below:

𝑂𝑅𝑂𝐴_𝑖,𝑡 = 𝐸𝐵𝐼𝑇_𝑖,𝑡 / 𝑡𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠_𝑖,𝑡 (4)

Both EBIT and total Assets are published in firms` financial reports, EBIT can be obtained from income statements, which represents the earnings before income and tax and total Assets can be obtained from balance sheets.

Based on accounting rules, we estimate the operating returns on assets by elements that commonly appear in income statements to compute the operating income (EBIT). In addition, we also include the leverage ratio, which is measured as total assets normalized by total equity, and total liabilities into the estimation. The general estimating equation for normal OROA in the effective post-issuing periods are shown as below:

𝑂𝑅𝑂𝐴𝑖,−𝑡= α + β1𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖,−𝑡+ β2𝑡𝑜𝑡𝑎𝑙 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑖𝑒𝑠𝑖,−𝑡+ β3𝑠𝑎𝑙𝑒𝑠 +

β4𝑐𝑜𝑠𝑡 𝑜𝑓 𝑔𝑜𝑜𝑑𝑠 𝑠𝑜𝑙𝑑𝑖,−𝑡+ β5𝑐𝑜𝑠𝑡 𝑜𝑓 𝑠𝑒𝑙𝑙𝑖𝑛𝑔, 𝑔𝑒𝑛𝑒𝑟𝑎𝑙 𝑎𝑛𝑑 𝑎𝑑𝑚𝑖𝑛𝑖𝑠𝑡𝑟𝑎𝑡𝑖𝑜𝑛𝑖,−𝑡+

β6𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑒𝑠𝑖,−𝑡+ β7𝑑𝑒𝑝𝑟𝑒𝑐𝑖𝑎𝑡𝑖𝑜𝑛 𝑎𝑛𝑑 𝑎𝑚𝑜𝑟𝑡𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖,−𝑡+ β8𝑎𝑐𝑞𝑢𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖,−𝑡+ ε𝑖,−𝑡

(20)

19 In which, 𝑖 represents each issuance event and – 𝑡 represents t years before the issuance. For each firm who issued new stocks, we use the 15 years of time series data to estimate normal 𝑂𝑅𝑂𝐴𝑖,+𝑡 during the issuing years, +𝑡 represents t years after the issuing year and 1 ≤ 𝑡 ≤ 𝑇

, 𝑇 represents the effective post-issuing periods (event windows) with value of +1, +2 or +3. The normal 𝑂𝑅𝑂𝐴_𝑖,−𝑡 is estimated by performing time series regressions.

After we obtain the normal returns in the effective post-issuing periods, we can continue to proceed to compute the post-issuing abnormal operating returns by the difference between the actual OROA and the estimated value in the effective post-issuing periods, which we present as ab_OROA in the equation. The formula to compute ab_OROA is:

𝑎𝑏_𝑂𝑅𝑂𝐴𝑖,+𝑡 = 𝑎𝑐𝑡𝑢𝑎𝑙 𝑂𝑅𝑂𝐴𝑖,+𝑡− 𝑛𝑜𝑟𝑚𝑎𝑙 𝑂𝑅𝑂𝐴𝑖,+𝑡 (6)

Finally, we sum up the values of all abnormal OROA in each effective post-issuing period as the cumulative post-issuing abnormal operating performance, which will be used as

measurement for firms` operating performance after issuing the new equity in the second step. This is used as the ultimate evaluating indicator of the factors` effect of dual class share structure on firms` operating performance. For convenience, we will present them as

cum_ab_OROA in the equation for short. By formula to compute cum_ab_OROA:

𝑐𝑢𝑚_𝑎𝑏_𝑂𝑅𝑂𝐴_𝑖,+𝑇 = ∑𝑇 𝑎𝑏_𝑂𝑅𝑂𝐴_𝑖,+𝑡

𝑡=1 (7)

𝑇 = +1, +2 𝑜𝑟 + 3

4.2 Step 2(A): Testing the single effect of factors on cumulative abnormal returns

Once we obtained the cumulative post-issuing abnormal operating performance for each new-issue event, we are able to test the dual-class effects of each factor on operating performance by running cross-sectional OLS regressions across the sample in the second step. However first of all, we would like to know the single effect of each factor on operating performance without inserting the interaction terms beforehand to help to observe how effects changed due to dual class share structure in step 2(B), the estimating equation for single effects is shown as below:

(21)

20 𝑐𝑢𝑚_𝑎𝑏_𝑂𝑅𝑂𝐴_𝑖 = 𝛼 + 𝛽₁𝑑𝑢𝑎𝑙_𝑖 + 𝛽₂ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ_𝑖 + 𝛽₃ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖 +

𝛽4𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦𝑖 + 𝛽5𝐼𝑃𝑂𝑖+ 𝛽6𝑓𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠𝑖 + 𝜀𝑖 (8)

Besides the four interested factors of technology, leverage level, issues type and economic level that we have described in section 2.2 Hypothesis, we also include the dummy variable of dual as the fifth factor to investigate its single effect which will be equal to 1 if the new issue is conducted through dual class share structure and 0 if otherwise. Similar as how we estimate the normal operating performance, the financial data of firms will be controlled for in the regressions, together with other data of firms` characteristics. The table below shows the elements included in the group of control variables:

Table 3. Firms` characteristics as control variables

Firm size = ln (total assets) Leverage ratio = total assets / total equity

Total assets Total liabilities

Sales revenues Costs of goods sold

Costs of selling, general and administration Depreciation and amortization

Employees Acquisitions

Operating return on assets in the previous year to the issuing year

(22)

21 The interpretation of the coefficients of the estimate for each factor variable (β1 – β5) is in

terms of how much amount of cumulative post-issuing operating performance will be

increased (decreased) if the firm issued the new equities under dual class share structure (β1),

if the firm is conducting high technology-oriented products (β2), if the firm bears a high

leverage level (β3), if the firm is regulated in advanced economies (β4) or if the firm issued

the new equities as initial public offers (β5).

4.3 Step 2(B): Testing the dual-class effects of factors on cumulative abnormal returns

Based on the results of single effects from step 2 (A), we can further investigate if the dual class share structure matters to these single effects by adding interaction terms with dual into the estimating equation, which is shown as below, all other controlled variables remained:

𝑐𝑢𝑚_𝑎𝑏_𝑂𝑅𝑂𝐴𝑖 = 𝛼 + 𝛽1𝑑𝑢𝑎𝑙𝑖 + 𝛽2ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 + 𝛽3ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖 +

𝛽₄𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦_𝑖 + 𝛽₅𝐼𝑃𝑂_𝑖+ 𝛽₆𝑑𝑢𝑎𝑙_𝑖 ∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ_𝑖 + 𝛽₇ 𝑑𝑢𝑎𝑙_𝑖 ∗ ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖 + 𝛽₈𝑑𝑢𝑎𝑙_𝑖 ∗ 𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦_𝑖+ 𝛽₉ 𝑑𝑢𝑎𝑙_𝑖 ∗ 𝐼𝑃𝑂_𝑖 +

𝛽10𝑓𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠𝑖 + 𝜀𝑖 (9)

The interpretation of the coefficients of the estimates in the regression equation in step 2 (B) would somehow differ from that in step 2 (A). As in the case with interaction terms, the main objective is to test the dual-class effects of each factor on cumulative post-issuing abnormal operating returns. The summed value of the coefficients of both individual terms and corresponding interactions terms lead to the actual effects of each factor on operating

performance conditional on the issuance of new equities with the dual class shares structure. By comparing the results between step 2 (A) and step 2 (B), we can easily tell how the factors influences the effect of dual class share structure on operating performance and thus test the four hypotheses mentioned in section 2.2.

(23)

22

5. Empirical results and analysis

Following the steps described above in section 4 Methodology, we make a summary of the regression results in table 4, which is presented as below, in terms of each variable. Panel A shows the results for the single effect of each factor on the cumulative post-issuing abnormal operating performance without interaction terms, which corresponds to the test described in section 4.2 step 2 (A). Panel B displays the results for the dual-class effect of each factor on the cumulative post-issuing abnormal operating performance by including interaction terms. The interaction terms are defined as the joint effect of dual class shares structure and each factor, which corresponds to the test as described in section 4.3 step 2 (B). The columns of table 4 represent the three different effective post-issuing periods – the defined event windows - that we take into account for sensitivity analysis with T equals to +1, +2 and +3 for each column respectively.

5.1 F test for estimating normal operating returns

Table 4. F test for joint significance of variables estimating normal OROA

Significance level 0.1 0.05 0.01

% of total sample shows jointly significant results 97.39 96.29 92.61

In the first step of our methodology – 4.1 Estimating post-issuing abnormal operating performance, we basically estimate the normal operating returns by accounting data which are normally used to compute the operating income which presented on the Income

Statement, even though we include other relevant variables as well such as leverage ratio, liabilities, employees and acquisitions, it is still reasonable to doubt if these variables are jointly significantly significant to determine the value of the operating returns in the corresponding year based on only 15 observations. In order to test the significance of these variables, we employ F test both across sample and individually for each firm. What inspired

(24)

23 us according to the test results is that not only the sample as a whole shows jointly significant interpretability of the variables we employed, but also more than 90% of the total companies display the same robust interpretability individually. As shown in Table 5, if we restrict significance level at 1%, the estimated normal operating returns from 92.61% of the total sample firms can be jointly explained by the variables used in the estimating equation, if we relax the restriction of significance level to 5%, the number of firms exhibiting joint

significant results climbs to 96.29% and if the significance level is increased even more to 10%, we will find up to 97.39% of total firms show jointly significant interpretability of the variables we employed. Overall, the first step of our estimating methodology is proved to be robust and it makes sense to further conduct the second step based on the estimated data in this step.

5.2 Results overview

Generally speaking, as shown in panel A, factors of dual class shares structure, high technology and advanced economies display significant single effects on the operating performance after issuing the new equities. Both influences of dual class shares structure and high technology drive the operating performance to the downside respectively, however a positive impact of the dual-class issues on post-issuing operating performance has been discovered from the empirical results if the firm is operating in a developed economy. We next look at the results presented in panel B, the individual negative effect of the dual class shares structure on the post-issuing operating performance has disappeared after including the interaction terms. Similarly, the influencers of high technology and dual class shares structure together harm the operating performance after issuing the new equities, however the

developed economic environment was found to be favorable to the post-issuing operating performance and encourage the advantages of the dual class shares structure to be exacted out. In addition to the findings in panel A, a new relevant factor of firms’ leverage level was found to matter to the post-issuing operating performance if interacted together with the dual class shares structure. Similar to the technology factor, high leveraged firms under the dual class shares structure tend to deteriorate the operating performance after issuing the new equities.

(25)

24

5.3 Single effects of each factor on post-issuing operating performance

In this section, we would discuss the single effects of each factor on post-issuing operating performance, results of column (1) – (3) in Table 4 are relevant to the analysis in this part. Respectively, column (1) (2) (3) correspond to the effective post-issuing periods of one, two and three years.

Specifically, from Panel A, column (1), we could easily tell from the results that firms adopting dual class shares structure typically suffer statistically significant lower abnormal returns of operating activities than firms with non-dual class shares structure in the first year after issuing the new equities. On average, dual-class firms generate approximate 0.1 value of the abnormal OROA lower than the single-class firms. Such findings are basically consistent with the conclusions from many other existing literatures, such as Paul, Joy and Andrew (2009) and Michael and Eric (2008), even though they measure the post-issuing performance by firm value. Based on the analysis in their papers, the possible reasons leading this

phenomenon can be explained as that it is the control-enhancing mechanism rather than the family ownership that reduces a firm`s value. The negative impacts of the dual class shares structure have higher influencing power than its merits. However, the negative results from the factor of the dual class share structure disappears if we extend the length of the effective post-issuing period and even in the first year after the new issuance, the significance level of the negative result only shows to be at 10% level.

Similarly, firms operating in high technology industries typically bear lower post-issuing abnormal operating performance than non-technology-oriented firms for all the effective post-issuing periods with one-, two- and three-year. Numerically, high technology firms normally have around 0.08 value lower of the cumulative abnormal operating returns in the first year after issuing the new equities. The losses increase with the length of the effective post-issuing period, which are shown to be roughly 0.14 and 0.28 in the following second and third year respectively.

Unfortunately, we found no evidence to support the arguments that leverage level or issues type matters for the post-issuing abnormal operating returns as neither of the estimates for the coefficients of these two factors in panel A in any effective post-issuing period shows to have

(26)

25 statistically significant impacts. If we have to say, the only conclusion we can drive for these two factors from our results would be that when firms announce to issue new equities, we will not expect any average below post-issuing operating performance if the firms are facing high leverage ratios or the new-issues are going to be either IPO or secondary issues.

Even though both factors of the dual class shares structure and high technology are presented to be against to the post-issuing operating performance, one factor in our investigation is still shown to be favorable to improve the operating performance after the new issuing namely the economic level. The empirical results from the test for the single effects provide evidence to support firms to issue new equities in advanced economies as the post-issuing operating performance would be beneficial from the new issuing events. According to the empirical results, firms being regulated under environments with developed economic conditions enjoy significant higher post-issuing abnormal operating performance than firms being regulated under the developing economic environments. Based on the numbers, compared with the less developed economic environments, advanced economic environments improve the operating performance by approximate 0.1 value in the first year after issuing the new equities. Similar to the case of high technology, the losses increase with time as well, which are shown to be around 0.2 and 0.4 values in the following second and third year respectively.

Notwithstanding we find out negative effect of the dual class share structure on the post-issuing operating performance which is similar to Paul, Joy and Andrew (2009) and Michael and Eric (2008), the advantages of such a structure are still defended strongly and rationally in public and the questions of what factors actually beat these advantages up and how to trigger out the values of the structure to benefit society are of extremely importance and expected to be answered as well. As what we stated at the end of section 1 Introduction, the contributions of this paper is to further investigate what factors in nature drive the negative effect of the dual class shares structure and the aim of our research is to provide stock

exchange regulators and government policy makers the empirical evidence-supported criteria to restrict the new-issues with dual class shares structure in terms of their impacts on the post-issuing operating performance. In the next section 5.2, we will continue to present and

analyze the testing results of the dual-class effects following the methodology discussed in section 4.3 Step 2(B).

(27)

26 Table 5.

The influencing power of factors on the relationship between dual class structure and operating performance: estimates for technology, leverage, IPO and economic level

This table looks at the single effects (column 1 – 3) and the dual-class effects (column 4 – 6) of the

investigated factors (dual class shares structure, technology, leverage, IPO and economic level) on the post-issuing operating performance. In addition to the variables included in column 1 – 3, column 4 – 6 add 4 more interaction terms to test the joint effect of dual class shares structure and each factor. The magnitude of the effects in different effective post-issuing periods can be compared across the rows. Column 1 and 4 look at effects within one-year post-issuing period (T = +1), column 2 and 5 look at effects within two-year post-issuing period (T = +2) and column 3 and 6 look at effects within three-year post-issuing period (T = +3). The regression use cross-sectional data of 2989 firms. Robust standard errors are reported in

parentheses. All the investigated factors of dual class shares structure, technology, leverage, IPO and economic level are dummy variables that take the value 1 if they firm is in the category. For definitions of high technology firms and high leveraged firms see Appendix I and II. *, ** and *** indicate significance at 10%, 5% and 1% respectively.

Dependent variable: cum_ab_OROA+T

Panel A: no interaction terms Panel B: With interaction terms

Event Window 𝑇 = +1 (1) 𝑇 = +2 (2) 𝑇 = +3 (3) 𝑇 = +1 (4) 𝑇 = +2 (5) 𝑇 = +3 (6) Factors 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 -0.2679** -0.4331 -0.6418 -0.2761** -0.3792** -0.4838** (0.12) (0.17) (0.26) (0.12) (0.17) (0.24) 𝑑𝑢𝑎𝑙𝑖 -0.1015* 0.0787 0.3666 -0.1155 0.0246 0.1932 (0.06) (0.17) (0.35) (0.12) (0.22) (0.40)

(28)

27 ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 -0.0837* -0.1371** -0.2766*** -0.0822* -0.1170* -0.2331** (0.04) (0.07) (0.11) (0.05) (0.07) (0.11) ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖 -0.0550 -0.0255 0.0142 -0.0637 0.0061 0.1063 (0.08) (0.11) (0.17) (0.09) (0.12) (0.18) 𝑖𝑝𝑜𝑖 0.0468 0.0526 0.0498 0.0356 0.0417 0.0462 (0.08) (0.13) (0.14) (0.08) (0.14) (0.15) 𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦𝑖 0.1009* 0.2330** 0.4162*** 0.1078* 0.1628* 0.2226* (0.05) (0.09) (0.15) (0.06) (0.10) (0.12) 𝑑𝑢𝑎𝑙𝑖∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ𝑖 -0.0425 -0.3283* -0.6474* (0.10) (0.19) (0.36) 𝑑𝑢𝑎𝑙𝑖 ∗ ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒𝑖 0.1029 -0.4530* -1.2871** (0.12) (0.29) (0.58) 𝑑𝑢𝑎𝑙𝑖∗ 𝑖𝑝𝑜𝑖 0.1528 0.1399 0.0386 (0.15) (0.25) (0.38) 𝑑𝑢𝑎𝑙𝑖 ∗ 𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦𝑖 -0.0279 0.3077* 0.8621** (0.11) (0.22) (0.44) Firm size (ln(total asset)) -0.0054 -0.0023 0.0055 -0.0049 -0.0023 0.0044 (0.01) (0.02) (0.03) (0.01) (0.02) (0.03) Leverage 0.0077*** 0.0077** 0.0068** 0.0077*** 0.0081** 0.0078*** (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Total liabilities 0.0000 0.0001 0.0001 0.0001 0.0001 0.0001 (0.0000) (0.00) (0.00) (0.00) (0.00) (0.00)

(29)

28 Sales revenues -0.0001** -0.0002** -0.0002** -0.0001** -0.0002** -0.0002**

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Costs of goods sold 0.0001** 0.0001** 0.0001** 0.0001** 0.0001** 0.0002**

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Costs of selling, general & administration 0.0001* 0.0001* 0.0001 0.0001* 0.0001* 0.0001 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Employees -0.0006 -0.0007 -0.0012 -0.0006 -0.0009 -0.0015 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Depreciation & amortization 0.0001 0.0001 -1.6100 0.0001 0.0001 0.0000 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Acquisitions -7.0100 -0.0000 -0.0000 -6.9900 -0.0000 -0.0000 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) OROA-1 0.9824* 0.8850 -0.1348 0.9788* 0.8857 -0.1268 (0.55) (0.93) (1.52) (0.56) (0.93) (1.53) OROA 1.9862** 2.8691** 4.3020** 1.9902** 2.8643** 4.2820** (0.90) (1.27) (1.75) (0.90) (1.28) (1.76) Total assets 5.7200 7.4100 0.0000 5.6300 5.4300 0.0000 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) N 2989 2989 2989 2989 2989 2989 𝑅 − 𝑠𝑞𝑢𝑎𝑟𝑒𝑑 0.0901 0.0692 0.0473 0.0902 0.0704 0.0509

(30)

29

5.4 Dual-class effects of each factor on post-issuing operating performance

Following the analysis of the single effects in the last section, we will carry on investigating the dual-class effects by looking at panel B in Table 4, in which the interaction terms between the dummy variable of the dual class shares structure and each respective variable of the investigated factor have been added into the estimating equation presented in section 4.3 Step 2(B). If we start the analysis by looking at column (4) with the effective post-issuing period of one year, we could hardly find out any significant results from these interaction terms. However, if we extend the effective post-issuing period to two- or three year, as shown in the column (5) and (6) respectively, some of these variables with interaction terms start to show up their influencing power in the results. Anyhow, the negative single effect of the dual class shares structure on the post-issuing operating performance disappeared after including the interaction terms, which can be explained as that the bad influencing power has been split between the interaction terms – the external influencers of the dual class shares structure. Compared with other firms, the dual class shares structure itself does not influence the operating performance of firms after issuing the new equities. We will discuss each circumstance in details in the following analysis.

Let us start the discussions with the joint influencing power of high technology and the dual class shares structure. In section 5.1, we have found out evidence of the downside influencing power of technology factor to firms` post-issuing operating performance individually in panel A, similar negative effect of this factor but together with the dual class shares structure on the operating performance after issuing the new equities has been discovered as well in panel B. Numerically, as shown in column (5) and (6), we can easily indicate from the estimates of the coefficients that, on average, if firms are operating in high technology industries, they

typically suffer around 0.3 and 0.6 value of the post-issuing abnormal operating performance lower than non-high technology firms respectively if we consider the effective post-issuing period for two- and three-year. Clearly, the magnitude of the effects increases with time as well – the length of the effective post-issuing periods. Surprisingly, the empirical results from our regression analysis are completely opposite to our Hypothesis I but supportive to Alon Brav, Wei Jiang, and Lucian Bebchuk`s argument that activist shareholders` interventions in nature help to improve operating performance of firms lagging behind peers consistently in

(31)

30 the long run. What these results in nature imply to the stock exchange regulators and

government policy makers is that no potential benefits can be expected if granting high technology firms the access to issuing dual-class new equities. In other words, technology-oriented firms tend to hurt the potential benefits which normally could be obtained from the dual class shares structure, the argument that dual class share structure potentially protect the long-run technology-related research programs from myopia investors is broken down in our findings.

Even though we failed to find out any correlation between leverage level and firms` post-issuing operating performance in section 5.1, we discovered a negative connection between them if the high leveraged firms are operating under the dual class shares structure in this section. Such a result is similar to the findings of the influencing power of technology factor but in line with our Hypothesis II and supportive to the argument that high leveraged firms issue new stocks potentially aiming at raising capital by the new funds rather than improving their operating performance and the dual class shares structure with less-costly centralized control rights help to realize the firms` insider controllers` aim of the decisions control and deteriorate the operating performance after issuing the new equities. Numerically, if the firm has a high leverage ratio, they tend to suffer 0.5 and 1.3 value of post-issuing cumulative abnormal operating performance lower than low leveraged firms in the second and third year respectively just after issuing the new equities, both of the significance level and the effect magnitudes increase with the length of the effective post-issuing periods. These findings offer the recommendations to the stock exchanges regulators and the government’s policy makers that more restricting criteria are needed to be set up for the high leveraged firms when they are planning to issue new equities through a form of the dual class shares structure.

Third again, same as the estimating results for the variable of IPO dummy in the test for single effects in the analysis of last section, we do not find out any significant evidence to support our Hypothesis III: Dual class share structure has a negative effect on operating performance if the new equity is issued as IPO. That is to say, either the new issues are conducted as IPO or secondary issues will not violate the impacts of the dual class shares structure on the operating performance after issuing the new equities and time is irrelevant. Overall, it does not make sense to especially consider about the issues type when firms request to issues dual-class equities in terms of the post-issuing operating performance.

(32)

31 Favorably, in accordance to our Hypothesis IV, factor of advanced economic level tends to have statistically significant positive effect on the post-issuing operating performance. Numerically, if firms are operating in an advanced economy, the cumulative post-issuing abnormal operating returns would be proximately 0.15, 0.2 and 0.27 of value higher than firms operating in developing countries respectively by considering the effective post-issuing periods of one-year, two-year and three-year. The potential reasons behind this findings might be the better corporate governance and comprehensive regulations in the advanced countries, which ensure the effectiveness of the dual class shares structure on improving the post-issuing performance of firms` operating activities.

To sum up the empirical results based on our database and methodology, dual class shares structure itself has non-value, neither positive nor negative, to firms` post-issuing operating performance. However, other factors external to this structure have the potential power to violate its neutral effect, which are usually done through the manipulation of the dual class shares structure by the votes` controllers – the management team. To specify, insider

controllers of high technology firms tend to extract private benefits by making use of the dual class share structure, which is considered to be the true motivation behind the adoption of the dual class shares structure that will be covered by the beneficial-looking long term research program. Similarly, high leveraged firms are more likely to issue dual class shares equities to finance their debt, the stressful financial states actually drive the new-issues. However, to some extent, good economic environment such as developed countries contribute to help good insider controllers to implement their knowledge-embedded and long-run visionary strategies smoothly, which will ultimately bring its values to all shareholders. Besides, no matter the new dual-class issues are conducted as IPO or secondary issues, the post-issuing operating performance will not be affected.

(33)

32

5.5 F test for interaction terms and factors in the dual-class effects

Table 6: F test for column 4-6 in table 5 (10% significance level)

Tested variables The effective post-issuing periods

T=+1 T=+2 T=+3

dual + high_tech + dual*high_tech Y Y Y

dual + high_leverage + dual*high_leverage N N Y

dual + ipo + dual*ipo N N N

dual + advanced economy + dual*advanced economy N Y Y

In Table 6, the joint significance of dual class shares structure and each factor are presented based on different effective post-issuing periods. In the column on the most left, the variables which are being tested jointly in the estimation equation are displayed. The right three

columns represents the effective post-issuing period of one-, two- and three-year respectively, which are indicated by T. In the table, Y indicates that the variables on the left are shown to be jointly significant to explain the dependent variable and vice versa, N indicates a jointly insignificant result. After checking the F Distribution Table, the criteria for significance results would be at 10% level with F value larger than 2.15. From Table 6, we can see that in the first year after issuing the new equities, only variables of high technology, dual class shares structure and their interaction term together show significant explanatory power. In the second year, in addition to the technology factor, economic level of a country starts to show significant predictability together with dual class shares structure and their interaction term as well. In the third year, only IPO factor together with the dual class shares structure and their interaction term show jointly insignificant explanatory power to the dependent variable.

(34)

33

6. Robustness

6.1 Robustness check without control variables of firms` characteristics

In our main equation of the regression (6), which is described in section 4.3 step 2(B), firms` characteristics are used as control variables to test the influencing power of our investigated factors on the relationship between the dual class share structure and the operating

performance. In this section, in order to check the robustness of the results presented in Table 5, we would like to exclude these control variables of firms` characteristics and redo the testing process illustrated in section 4 Methodology. Equations (7) and (8) explain the new estimation model for the single effects and the dual-class effects respectively,

𝑐𝑢𝑚_𝑎𝑏_𝑂𝑅𝑂𝐴_𝑖 = 𝛼 + 𝛽₁𝑑𝑢𝑎𝑙_𝑖 + 𝛽₂ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ_𝑖 + 𝛽₃ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖+

𝛽₄𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦_𝑖 + 𝛽₅𝐼𝑃𝑂_𝑖 + 𝜀_𝑖 (10)

𝑐𝑢𝑚_𝑎𝑏_𝑂𝑅𝑂𝐴_𝑖 = 𝛼 + 𝛽₁ 𝑑𝑢𝑎𝑙_𝑖 + 𝛽₂ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ_𝑖+ 𝛽₃ ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖 + 𝛽₄ 𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦_𝑖+ 𝛽₅𝐼𝑃𝑂_𝑖 + 𝛽₆ 𝑑𝑢𝑎𝑙_𝑖∗ ℎ𝑖𝑔ℎ_𝑡𝑒𝑐ℎ_𝑖+ 𝛽₇ 𝑑𝑢𝑎𝑙_𝑖 ∗

ℎ𝑖𝑔ℎ_𝑙𝑒𝑣𝑒𝑟𝑎𝑔𝑒_𝑖+ 𝛽₈ 𝑑𝑢𝑎𝑙_𝑖 ∗ 𝑎𝑑𝑣𝑎𝑛𝑐𝑒𝑑 𝑒𝑐𝑜𝑛𝑜𝑚𝑦_𝑖+ 𝛽₉ 𝑑𝑢𝑎𝑙_𝑖 ∗ 𝐼𝑃𝑂_𝑖 + 𝜀_𝑖 (11)

Table 7 displays the new results with only the investigated factors remained. Compared with the results presented in Table 5, the values of the R-squared decreased sharply. This is reasonable since some of the control variables that are subtracted from the old estimation equation (8) and (9) have significant explanatory power such as leverage ratio, sales revenues, costs of goods sold, actual OROA and etc. Excluding these variables would definitely lower the explanatory power of all the independent variables.

Furthermore, for the single effects in column (1) – (3), we can find that the factor of dual class shares structure still shows negative impact on the post-issuing operating performance only in the first year after issuing the new equities. Both factors of high technology and

(35)

34 advanced economy display significant influencing power of positive and negative effects respectively. The only difference from the main results presented in section 5 would be that the magnitudes of these effects increase slightly and the significance levels decrease in a way, especially for the factor of economic level. In addition, for the dual-class effects in column (4) – (6), most of the results are in line with our main results except for that some estimates of the coefficients in the one- and two-year effective post-issuing period turn out to be

statistically insignificant, e.g. advanced economic level and interaction term between high technology and dual.

Such relatively robust results indicate that our conclusions are convincing and reliable. Even though small differences existed between the main results in table 5 and the robustness check results in table 7, the directions of the effects (positive or negative) and the factors that are showing significant influencing power are all consistent.

6.2 Robustness check by taking average value of firms` characteristics

In our main methodology, we use the value of firms` characteristics in the issuing year only irrespective of the lengths of the effective post-issuing periods. Some would doubt that the firms` characteristics actually vary overtime and the value used in the cross-sectional regression should be consistent with the length of each effective post-issuing period. Therefore in this section, we would like to take the average value of firms` characteristics within each effective post-issuing period to redo the testing process described in section 4.2 Step (A) and 4.3 Step (B). Table 8 below presents the new results. Compared with Table 5 with our main results, except for that the estimate for the coefficient of the interaction term between the dual class share structure and high technology in column (5) – the second post-issuing period becomes statistically insignificant, nothing else changed significantly. These results are in line with our main results and therefore the conclusions appear to be robust and convincing.

What factors determine the effect of dual class share structure on firms` operating performance? : the influence of technology, leverage, issues type and economic level