Cross-delisting : investor protection, ownership, culture and learning capabilities

(1)

UNIVERSITY OF AMSTERDAM, AMSTERDAM BUSINESS SCHOOL

Cross-delisting

Investor protection, Ownership, Culture and

Learning capabilities

Student: Frans Bedaux

Date: July 2015

Programme: MSc Business Economics, Finance track

Document: Master Thesis

(2)

1

Statement of Originality

This document is written by Student Frans Bedaux 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

In this paper the question is asked to why cross-listed firms delist from foreign exchanges. In the sample there are companies from United States, United Kingdom or Hong Kong equity markets over the period from 1995 through 2012. The two theories that are examined are the bonding theory and the learning hypothesis. For the bonding theory several multi period probit models are estimated to decide if investor protection, ownership and culture show the anticipated effect. For the learning hypothesis an event analysis is done to get an idea on what happens to the investment-to-price sensitivity. The bonding theory shows limited evidence of being a reason for cross-listing and the learning hypothesis is showing its first outcomes but has to be further researched to come to worthy conclusions.

(4)

3

1. Introduction

In the late 1990s the United States exchanges were the preferred exchanges for cross-listings. From the 2000s aggregated interest of firms in cross-listings started to decline. The globalization of the financial world and regulatory development (Doidge et al. 2009b: Karolyi, 2006) are seen as drivers for these changes in the markets. Cross-listings are listings by foreign companies, which are secondary listings for them. In other words a firm has a listing on its home country exchange and goes to a foreign country to list their stocks on that exchange as well. Over the years different reasons are found and tested by scientific researchers. One of the main reasons and early findings for cross-listing are funding (market liquidity, market segmentation) and reducing agency costs (Doidge et al. (2009a); Karolyi, 2012). These theories are well researched and due to the declining interest in cross-listings less popular in the recent years.

From this decline in cross-listings came a new topic for research called cross-delisting, which became interesting due to the new trend of foreign firms delisting from United States markets (Marosi and Massoud, 2008). This created an interesting new way to test the theories which were formed for the earlier mechanism cross-listing. The bonding theory is one of these theories, which will be the first theory tested in this study. This theory suggests that a foreign company can create benefits by bonding itself to an exchange outside of its own country (Stulz, 1999; Coffee 1999, 2002). This was shown by looking at United States exchange markets, where regulation requirements can cause for extra reporting and disclosure requirements. Through these extra requirements companies can be better judged by for instance analysts and through the SEC regulation laws investors can have better protection in case of investor expropriation, like self-dealing or fraudulent behavior. The second theory that will be tested is a more recent theory. The learning hypothesis suggests that managers within firms learn from their stock prices by evaluating investment decisions based on stock price movement. When a firm goes for a secondary listing the investment-to-price sensitivity is increased, which helps the managers to evaluate investment decisions by the stock price (Foucault and Frésard, 2012). These two theories are tested in order to get a better understanding of the determinants of firms to delist from foreign exchanges and to see if these theories apply to a more global dataset instead of only the United States. By answering these

(6)

5

testable implications the contribution of the paper is threefold. First, this study will extend to the current literature of reasons for firms to delist from the host markets. Second, to get a better understanding and broader picture of firm decision-making. The third contribution is to the field of managerial learning.

These contributions are explored by firstly using a multi period probit model that will determine if investor protection regulations in different countries, dispersed ownership and culture aspects have the effect that is consistent with the theories. Secondly, an event analysis is done to find out if the learning hypothesis can be explained by the investment-to-price sensitivity. For the sample 356 voluntary cross-delisted companies are found in the period of 1995 and 2012 for the United States, United Kingdom and Hong Kong markets. In this same period 930 cross-listed companies are used as control companies that do not leave the markets in this period.

This paper proceeds as follows. First a literature review is written about the relevant theories. Then the methodology and variables are shown and further explained. In section 4 the data is described and the relevant statistics are revealed. Section 5 contains the empirical results and a first comparison to other literature is made. In Section 6 the findings are brought together and discussed to the significance of the question asked. Then in section 7 a conclusion of the study is given.

(7)

6

2. Literature review

In the literature the theories and hypotheses behind the motives of a cross-listing have been researched a lot since the 1990’s. In this time cross-listing was a popular business phenomenon, with the United States as the main preferred market for foreign companies to go to for the extra listing as the main market. This trend started to change from the beginning of the 21st century when companies were delisting from the foreign market and staying active on their home country exchange. So in the more recent years companies are cross-delisting more and more and less are cross-listing on foreign exchanges. This brings a new topic into research: are the theories which are used for companies to cross-list still accurate and relevant when we talk about cross-delisting? In this section I will look into two main theories, namely bonding theory and learning hypothesis.

2.1 Bonding theory

One theory that is well researched for a reason why firms cross-list in foreign countries, is the bonding theory. This started with the work of Stulz (1999) and Coffee (1999, 2002) who stated that market segmentation hypothesis (Errunza and Losq, 1985) was not the only reason for cross-listing, but also the bonding theory was a motive. The market segmentation hypothesis said that due to the “separation” of global markets firms would be interested in listing abroad. Stulz and Coffee thought this was not the only reason for firms to cross-list and started working on investor protection and corporate governance reasons for cross-listing. This is how they came with the bonding theory as a reason for cross-listing.

The bonding theory tries to explain why firms decide to take on an extra listing next to their home country listing. Firms that use cross-listing on United States markets are obliged to adapt to United States laws and standards (United States GAAP), due to the regulation by the SEC. By these laws and standards corporate insiders have limited their opportunity to private benefits, which leads to a higher valuation of the firm (Stulz, 1999; Coffee, 1999, 2002; Reese and Weibach, 2002). So by investor protection the minority shareholder profits from the foreign listing. Another effect of the new requirements is that investors can easier assess firms and are less concerned about fraud and bankruptcy, since the compulsions for a listing in the

(8)

7

United States entail a lot of documentation and accounting standards, which decreases the changes of being picked off as a shareholder.

Reese and Weibach (2002) took an interesting approach by looking at equity offerings. They were interested in finding out if non United States firms choose to take a listing in the United States because of the protection for the minority shareholder, that comes with it. For looking into this matter they distinguished firms by law tradition of home countries and then built a case by looking at the frequency and location of equity offerings of United States cross-listed firms. By formulating hypotheses in which they put United States cross-listed firms from weak governance countries against United States cross-listed firms from strong governance countries. Firms from strong governance countries are more likely to offer equity in United States markets after the cross-listing in order to increase investor access, while firms from weak governance countries are using United States markets for bonding reasons so they can easier raise capital on their home markets.

2.1.1 Investor Protection

Over the past decades different perspectives and hypotheses were tested in order to put the bonding theory to the test. Investor protection is one of the main variables used by researchers in order to find support for the bonding theory. Each research discussed here is coming from a different insight but all have investor protection as their bonding theory reference point. For example, Lel and Miller (2008) tested investor protection for cross-listed firms by looking at CEO turnover of poorly performing cross-listed firms versus non cross-listed firms. There outcome shows that poorly performing cross-listed firms in the United States have a higher CEO turnover, due to investor protection in the United States, especially if these firms are located in weaker investor protection countries. In this way they demonstrate the cross-listing effect on corporate governance outcomes.

Looking at the method and measures that were used by Lel and Miller (2008) they took 3 different country level measures for investor protection. First variable they used is from La Porta et al. (1997, 1998), which shows that generally speaking common law tradition countries show higher investor protection law then civil law countries. Second variable they looked at is the anti-director rights index, which has a revised version by Djankov et al.

(9)

8

(2008). This variable looks at different variables in order to determine how the corporate law in that particular country protects the minority shareholders. Third variable is the anti-self-dealing index, this is an index created to show the legal protections in a country for minority shareholders against self-dealing by corporate insiders. Self-dealing, also referred to as tunneling, is the problem of investor expropriation. This could be done by managers or controlling shareholders through diverting capital to themselves rather than sharing it with the (other) investors or managers. Examples of self-dealing are appropriation of corporate opportunities, excessive compensation, executive perquisites, transfer pricing and self-dealing transactions, like personal loans to insiders or directed equity issuance.

2.1.2 Sarbanes-Oxley Act and Exchange Act Rule 12h-6

The regulators in the United States market have made researchers test the cross-listing decision in different ways. In 2002 the Sarbanes-Oxley act was introduced to enhance standards of all United States listed companies. This was an outcome of the fraudulent cases with came out in the early 2000’s. For cross-listing firms this meant their costs for the foreign listing would go up, resulting in a deregistration of foreign firms on United States exchanges, especially small firms with low trading volume (Marosi and Massoud, 2008).

Another rule change by the SEC in 2007 called Exchange Act Rule 12h-6 made it easier for United States firms to deregister from the SEC. This was an important rule for non-United States companies, since most of their costs of the foreign listing were attributed to the reporting and disclosure requirements of the SEC. Before this rule was in place foreign companies could only deregister from the SEC if they looked through accounts of banks, brokers and other institutions in the whole world to count the number of United States resident’s holders of the equity. If this were more than 300 United States residents holding equity then the foreign company was not allowed to deregister from its obligations to the SEC. As you can imagine this was a difficult job to manage and this was not the only criteria the SEC had in place.

The Exchange Act rule 12h-6 started a new wave of researchers looking into the delisting decisions of cross-listed firms. One of these was Fernandes et al. (2010) who proclaimed that their results are in favor of the bonding theory. Since they find a negative market reaction

(10)

9

when Rule 12h-6 was announced for firms from a weak governance country, as oppose to firms who have already strong investor protection regulations in place. For firms with already a good investor protection in their home country there was no significant market reaction found. This suggests that shareholders of cross-listed firms pay attention to origin location of the firm when these firms delist.

Fernandes et al. (2010) find a weak relationship between delisting (market reaction) and compliance costs for being listed or financing needs of the firm. As methodology they did not use a standard event study, because of the cross-correlation of the error terms across firms. So instead they estimated a seemingly unrelated regression. They used proxies on a firm level in order to find the capital needs of the companies and seven different proxies for country level governance to look for disclosure requirements and investor protection of the country of origin of the company.

The country level proxies used not previously spoken of are disclosure requirements, disclosure, World Bank disclosure, disclosure of periodic filings, earnings management and efficiency of the legal system. In the methodology part more explanation will be given for variable choices.

In the same year Doidge et al. (2010) published their research on foreign firms leaving U.S. markets and they find support for the previously mentioned relationship as well. For firm characteristics that encourage delisting after the Sarbanes-Oxley act they show that low performance, a financing surplus and low growth expectations and low growth opportunities are the main drivers. They conclude with that the Sarbanes-Oxley act is not a major cause for companies to delist from United States exchanges.

To come to this conclusion Doidge et al. (2010) used a multi period logit regression, ordinary least squares (OLS) regression for returns of deregistering firms and another OLS regression to test the stock price reactions of firms around the announcement dates of the Sarbanes-Oxley act.

(11)

10

2.1.3 Culture

Another interesting added perspective was by Daugherty and Georgieva (2011) about cultures. They tested cross-delisting companies in the United States markets by using cultural factors and the Sarbanes-Oxley act. Cultural distance and the individualism index by Hofstede (1980; 2001) were added to their research. The motivation for this is that countries with more individualism are societies where the managers focus more on personal gain. So from a bonding theory perspective, countries which score high on individualism are expected to delist faster than companies from countries which are ranked lower on the individualism index. It was found that the propensity to delist is higher for companies which are ranked higher on the individualism index. They found this by looking at all non-United States companies which had a American depository receipts between 2000 and 2010 and put these into different logit regression analysis models and hazard models to test their hypothesis.

2.2 Learning Hypothesis

The second theory put to the test is the learning hypothesis. This hypothesis states that managers look at their stock price to evaluate corporate decisions. So following this hypothesis stock prices should contain information new to managers, that’s why the question of interest here is if the stock price contains information that could be useful to managers? Can managers learn from the stock price reaction to events or corporate announcements? Due to cross-listings it is shown that the stock price informativeness changes, therefore what can we expect if we talk about cross-delisting companies?

In the efficient market hypothesis it is stated that all information is shown in the stock prices in the market. So managers or companies who have inside information are exposed to external information shown by stock price movement. Therefore, the question of managerial learning, by using the external information shown in its stock price, is a relevant and broadly researched topic over different fields.

In corporate capital budgeting, as argued before, managers do not always make a decision in order to maximize the value of the company. These are the corporate governance problems which can be enhanced by a cross-listing (bonding theory). For managers that are looking for

(12)

11

value maximization, instead of self-interest, and have capital budgeting policies in place, it is shown by Durnev et al. (2004) that their stock prices are more informed. They showed this by looking into firm-specific return variation (Roll, 1988) within different industries and alignment of capital budgeting to maximize market value.

Cash savings is a part of corporate capital budgeting in which Frésard (2012) found evidence for managerial learning from the stock market. He searched for relation between sensitivity of cash savings to price and the amount of private information to stock prices. Specifically, he found that when there is more ‘new’ information for managers hidden in the stock price the cash savings are more sensitive to the price as well. His measure for stock price informativeness is firm-specific stock variation or price nonsynchronicity.

In the field of mergers and acquisitions managerial learning has been tested as well and there it is found that merging companies learn from the merger announcements’ market reaction for closing the deal (Lou, 2005). This result comes from different probit models used in combination with an economic model, which is demonstrated and proven in the article.

Aktas et al. (2011) find that CEO’s are looking at signals from investors in previous deals in order to decide on the bid premium they will offer for the next deal. In their study they looked if CEOs behavior is able to affect the cumulative abnormal return, by observing the acquisition activity from deal to deal. The learning hypothesis is the alternative explanation, which they worked with by following companies with two successful acquisition deals and tracking if the CEO behavior is influenced by the reaction of the investor.

For investment decision making by managers Chen et al. (2007) show that firms’ managers look at their stock price to evaluate corporate investment decisions. In their model they took two measures for private information in stock price, namely price nonsynchronicity and probability of informed trading. For both of these proxies strong positive correlations were found between the investment-to-price sensitivity and the quantity of private information in the stock price. Together with the first statement of Chen et al. (2007) this suggests that a higher investment-to-price sensitivity would be an advantage for the manager. Since with a higher investment-to-price sensitivity the manager will get a more severe reaction to his investment decision. Consequently, he can adjust his investment plans to the view of the whole market and incorporate the private information, which first was not at hand.

(13)

12

The increase in price informativeness of the stock is shown by Foucault and Gehring (2008) in their theoretical model by listing. In this model they build a new theory for cross-listing saying that it enhances the price informativeness of the stock and therefore improves the ability of the managers to evaluate corporate investment decisions by looking at their stock prices. In other words, companies receive more precise information about their growth opportunities. In the model they show that the cross-listing premium for firms in the United States is not from an governance improvement (bonding theory), but because of the managers ability to better exploit the growth opportunities after the cross-listing. A cross-listing enhances the option to collect private information from the stock price in two ways. Firstly, since the informed investor will now have more exchanges to trade on, this will give them a better possibility to exploit their private information about the stock price. Secondly, the stock will now be available to more foreign investors. Some investors have restrictions in investing in foreign stocks, so after the cross listing the private information of these foreign investors is also shown in its price.

Fernandes and Ferreira (2008) wanted to empirically test the effect of cross-listing on the information environment. They measured stock price informativeness by firm-specific stock return variation estimated from a two-factor international model with weekly returns. Lastly they tested changes in stock price informativeness around the cross-listing with an event study. Summing up they found that cross-listing gives a higher firm-specific stock return variation for developed countries and a lower firm-specific stock return variation for emerging markets. In countries with robust investor protection they find the largest increase of firm-specific stock return variation.

In the research of Foucault and Frésard (2012) they empirically test the learning hypothesis, which suggests that a motivation for cross-listing is the higher investment-to-price sensitivity. Since, this increases the ability of a manager to monitor the stock price of its firm in order to value his investment decisions. The amount of new information to managers will be shown more clearly in the stock price.

Going into the methodology they start by measuring investment-to-price sensitivity based on the model of Chen et al. (2007). They use a regression with the dependent variable being capital expenditures divided by lagged fixed assets (to account for investment) tested against

(14)

13

independent variables normalized stock price (book value of assets plus market value of equity minus book value of equity, divided by book assets), a dummy variable for if a firm is cross-listed, the interaction term of the previous two (which is one of the major interest variables in their study) and a set of controls. This is their basic regression from which they start and they use it in a lot of different ways. One of them being event time analysis where they replace the dummy variable cross-listing with dummy variables event time, which is centered around the cross-listing year. They use this last method in order to show that the before and after effect of a cross-listing is severe and the investment-to-price sensitivity indeed makes a jump from the cross-listing year and keeps this effect.

The learning hypothesis has not been researched in as a theory for delisting. If I consider it as a delisting reason it would state that a manager would delist its firm from a host market when the stock price informativeness has dropped down or is no longer in need. Therefore, the learning hypothesis suggests that after the delisting no significant change is seen in the investment-to-price sensitivity of the firm. I will go further into this in the hypothesis development.

2.3 Hypothesis development

In this paper I will test different hypotheses in different countries in order to proclaim certain effects of the bonding theory and the learning hypothesis.

2.3.1 Bonding theory

One of the main drivers of the bonding theory is the regulatory and disclosure system that is present in the country where the cross-listing happens. As earlier explained in the United States this regulatory and disclosure system is considered to be the strictest and therefore the most significant effect is expected to be found here. In this paper the focus will lay on three countries in order to test if this is indeed right. I will use the United States exchanges, the London exchanges and the Hong Kong exchange.

Although the bonding theory is tested mostly as a United States phenomenon, there are researches like Roosenboom and van Dijk (2009), who found support for the bonding theory

(15)

14

in the London market as well. While the previously discussed article of Lel and Miller (2008) found support for the bonding theory in the United States via CEO turnover and not in the United Kingdom. As well as Doidge et al. (2009a) who found support in the United States, but not in the United Kingdom. Also You et al. (2012) who found no support for the bonding theory in their study, which they explain by the fact that in their sample they see a lot of European cross-listing. These articles already give an interesting dilemma for further research.

For the bonding theory I will test the investor protection and cultural characteristics in order to determine if I find support for the bonding theory as a delisting reason. If a firm delists from the foreign exchange the bonding theory expects the company to have poor growth opportunities and a country with lower investor protection rules should be more eager to delist. This last statement is based on the fact that the agency problem, corporate insiders can extract benefits from the company from which minority shareholders can suffer, is more likely to happen in countries with low investor protection. So if there are no more opportunities for growth in the company and the potential of being cross-listed has become a cost instead of a benefit, then this is a plausible outcome.

As previously argued in this paper La Porta et al. (1997; 1998) show that generally speaking common law tradition countries show higher investor protection law then civil law countries. Based on this I can formulate the following hypothesis:

H1: Companies which have a non-common law tradition in their home country are expected to delist more often.

La Porta et al. (1998) worked on an anti-director rights index. This index was first created in 1998 and revised in 2008 and looks at different requirements in order to determine how the corporate law in that particular country protects the shareholders. A higher mark notes for a higher investor protection in that particular country. A more detailed explanation of the variable will be given later. Therefore I can formulate the following hypothesis:

H2: Companies which have a lower score on the anti-director rights index are expected to delist more often.

(16)

15

The anti-self-dealing index is a more recent index created by Djankov et al., (2008) in which it is argued that this index is a better tool for minority investor protection then the anti-director rights index. This is constructed to get a more accurate measure for investor protection, by looking at the self-dealing possibilities of a cross-l so partially the same reasoning applies as for the anti-director rights index. Therefore I can formulate the following hypothesis:

H3: Companies which have a lower score on the anti-self-dealing index are expected to delist more often.

The fourth hypothesis comes from Doidge et al. (2009a). They use the variable ownership, defined as closely held shares, to determine the spreading of the stock over investors. One can determine that following the bonding theory a company with a high percentage of closely held shares can be more focused on leaving the foreign markets when the benefits of this listing are over. Since more insiders held the shares and are more directly affected by this. Therefore I expect that high ownership relates with delisting. Formulating this in the next hypothesis gives:

H4: Companies with high ownership are expected to delist more often.

The last hypothesis for testing the bonding theory is from a cultural perspective. As written before Daugherty and Georgieva (2011) state that countries which score high on individualism index are expected to delist faster than companies from countries which are ranked lower on the individualism index. This corresponds with the bonding theory in the way that companies from higher individualism countries are expected to have a closer look at personal gain and benefits. So if the foreign exchange is no longer needed and the personal gain from delisting can outweigh the benefits of the foreign listing, then these companies are more likely to delist. Putting this into a hypothesis gives:

(17)

16

2.3.2 Learning hypothesis

Looking at the learning hypothesis partly the same characteristics of companies apply. The learning hypothesis and bonding theory are not mutually exclusive. For instance, if a firm would bond itself to a higher regulated exchange then this could cause for an incentive to investors in order to gain more information, which leads to make stock price more full of information (Fernandes and Ferreira, 2008).

Looking at delisting and explaining this with the learning hypothesis in mind, I expect these companies to no longer need their investment-to-price sensitivity in order to evaluate their investments. As earlier shown by Doidge et al. (2010) the companies that delist have low growth opportunities, low performance, a financing surplus and low growth expectations. They used these characteristics for looking at the bonding theory and therefore in this research another measure is used which corresponds with the learning hypothesis and not the bonding theory. Going back to the learning hypothesis this would mean that if a firm delists the investment-to-price sensitivity will have no significant change after the delisting. The main reason for this expectation is the fact that the expected value gain that is seen in the stock price, due to the cross-listing, for the manager has decreased. Therefore I can formulate the following hypothesis:

(18)

17

3. Methodology

In order to test the hypothesis in the appropriate way I will first investigate possible multicollinearity by doing a correlation analysis. Multicollinearity is when two explanatory variables highly correlate. If this is the case it becomes hard to determine from which explanatory variable the effect comes on the dependent variable. This multicollinearity problem can shows up when two explanatory variables have a high correlation and it leads to an issue if the size of your sample is small. To account for this matter a correlation table is made and values are analyzed.

3.1.1 Model

In order to test the bonding theory I will make use of a multi-period probit model. The dependent variable will be set to zero when a cross-listed company is active in that year and to one when the cross-listed company delists in that particular year. After a company is delisted or before a company is listed, then their values will not be taken into account for that particular year. As argued by Shumway (2001) the multi period model is largely the same as the discrete-time hazard model, which could also be used here. But since the discrete-time hazard model does not work with a constant in the model, the multi period probit model is preferred. This brings me to the following formula for the model:

_ _, 1



The first variable common law stands for the law tradition in the domicile country of the firm. The second variable is the anti-director rights index revised (RADRI), this index looks at different variables in order to determine how the corporate law in that particular country

(19)

18

protects the shareholders. Third variable is anti-self-dealing index (ASDI), this is an index created to show the shareholder protection for the minority shareholder. Fourth variable is ownership (closely held shares), this essentially represents shares that are held by corporate insiders. Fifth variable is individualism, as earlier mentioned this is the index created by Hofstede (1980; 2001).

For control variables the several variables are put in the sample. Some variables are included to account for firm size and solvency, namely the market capitalization, the total liabilities and the logarithm of total assets of the firm. For investments and profitability the ratio capital expenditures to total assets and the Tobin’s q are added as controls. The Tobin’s q is a ratio with in the numerator equity market value plus liabilities market value and in the denominator equity book value plus liabilities book value. Another control used is GNI per capita, which stands for gross national income divided by the midyear population. Although, as shown later, due to high correlation with some of the explanatory variables it is left out in some regressions. This yearly data per country is downloaded from the THE WORLD BANK and is already put into United States dollars.

For testing the hypothesis dummies are created for separate groups (low or high score) of the explanatory index variables. For each variable of interest three dummies were created in order to test the hypothesis. These are put into the regressions in order to see if they come up with significant coefficients.

3.1.2 Explanatory Variables

The variables used in this research for the bonding theory will be explained here. First variable is common law, this will be a dummy variable which will show if a country has a common law tradition or not. If this variable is 1 then it is a company which is from a common law tradition country and if this is a zero the company is from a non-common law tradition country. For this variable I will use the list that is given in Spamann (2010).

The second variable anti-director rights index is a variable which can range from zero to six. It is a country level measure which is constructed by La Porta et al. (1998). They constructed this by looking at six requirements, for how the legal system in a country protects the

(20)

19

minority shareholder, and if a country meets its requirement they add one point to the index for that country and zero otherwise. The first requirement is if it is allowed in a particular country to mail a proxy vote of a shareholder to the firm. In some countries you can only vote in person, which can cause a problem for small investors with a global portfolio. Second requirement is if at the annual general shareholders meeting shareholders are not required to deposit their shares prior to this meeting. This is because in some countries shareholders are obliged to deposit their shares prior to this meeting in order to prevent them from selling their shares for some days around the annual general shareholders meeting. Third requirement is if it is allowed to accumulate voting for directors, by which minorities can have an influence in the board of directors, or is there an option in place for proportional representation on the board. Fourth is if there is a mechanism in place in a country for giving minority shareholders legal opportunities if they are oppressed by directors. With the exclusion of absolute fraud, which is illegal in all countries studied. Fifth requirement is if shareholders receive a preemptive right to buy new stocks when issued, this is a requirement in order to avoid shareholder dilution. Sixth and last requirement is how big of a percentage of the shares are needed in order to call for an extra shareholders meeting.

For the third variable, the anti-dealing index by Djankov et al. (2008), a hypothetical self-dealing transaction was setup between two firms. This case study was sent to 102 Lex Mundi law firms in 102 countries. They also received a questionnaire that was specifically constructed for this matter and the law firms were asked to answer everything with the legal requirements that were in to place in May 2003. This was supposed to be done on 10 different matters, which are shown in appendix A.1. Next to the report and the questionnaire the researchers asked for all the relevant laws, judicial precedent, statutes and regulatory opinions. They constructed the index together with the law firm by taking the average of ex ante and ex post private control of self-dealing, where ex ante is an index that is created by looking at the ex-ante disclosure and the approval of disinterest shareholders by country. For the ex-post private control of self-dealing the researchers and lawyers looked at the relevant information to determine the index for disclosure of periodic filings and ease of providing wrongdoing. The variable can reach from zero to one.

For the fourth variable ownership (closely held shares) I take the variable closely held shares from Datastream in percentages. So this variable can range from zero to 100. Closely held shares stands for shares held by insiders. In Datastream if a company has more than one class

(21)

20

of common stock, the closely held shares for all classes are added together. This incorporates shares held by cross holdings, corporations, holding company, government, employees, Individuals and Insiders.

The last variable individualism is an index created by Hofstede (1980; 2001). This index is created by setting individualism versus collectivism. If a country scores high on this index it means there is a high individualistic society present in this country. This indicates that persons in these countries are more focused on themselves and are expected to take care of their families. While if a country scores low in this index this means they are high in collectivism, which says that persons in a country take care of each other. Individuals look at themselves as we, instead of only looking at themselves as in the individualistic society.

3.2 Learning hypothesis

3.2.1 Model

For my research of the learning hypothesis I will look at the model for investment-to-price sensitivity used by Chen et al. (2007) and Foucault and Frésard (2012). The dependent variable will be capital expenditures divided by lagged fixed assets _, . For showing the relation between before and after the delisting event, I will use an event-time analysis, to look at the hypothesis. This method is also used by Foucault and Frésard (2012). I will look into two timeframes (-3, +3) and (-4, +5). A dummy is created for each year and firm, seen from the delisting event at t = 0 ( _, ). This give the following equation:

, , . . ,

, log , ,

Where are time fixed effects and country fixed effects. _, is the normalized stock price of a firm, this is calculated by taking the market value of equity minus the book value of the equity and plus the book value of the assets, divided by book value of assets. Variable are cash flows of the firm. In this research they are divided by total assets. And is the natural logarithm of total assets. I incorporate these because they have a clear relation with investment decisions, so therefore they are included to control.

(22)

21

4. Data and Descriptive statistics

In this section of the study the data sources will be discussed and later the sample will be shown with the relevant summary statistics.

In this research I will look at companies which are cross-listed on the United States exchanges, United Kingdom exchanges or Hong Kong exchange. Subsequently, I will look at the companies which will delist from these exchanges and keep their listing in their home country. Another important fact is to exclude companies which are merging or being bought, going bankrupt or are suspended by the local government regulatory agency. I need companies that are ‘voluntary’ delisting themselves from the relevant exchanges. This is the sample of firms needed to test my hypotheses. Over the counter (OTC) markets and companies which registered under Rule 144 (US) were left out of the sample, due to the smaller size of the companies and the less regulations needed in other to account for these markets. For the data from the United States exchanges I retrieved this from Wharton research data website and the Center for Research in Security Prices (CRSP). In CRSP data are cleaned for mergers, acquisitions, liquidations and suspensions by looking at the delisting code. All companies with a delisting code between 0 and 500 were deleted from the sample. Also the stocks that switch between the United States Exchanges were deleted, since this keeps the company to be cross-listed. All companies with total assets lower than $ 1 million were also removed from the dataset. In order to determine if a share was from a foreign company share code was added to the dataset so relevant delisting companies can be picked. In the end 204 companies delisted from United States exchanges from the period January 1995 to December 2012. What is left are the voluntary cross-delisting companies from the United States exchanges. As previous researchers already pointed out this is a time consuming and difficult process (Doidge et al., 2010), which results in different articles showing different numbers and restrictions for voluntary delisting companies.

For the United Kingdom exchanges and Hong Kong exchange Datastream is used. This was more of an effort to clean. Main part is done by hand using Datastream in combination with LexisNexis Academic in order to find delisting dates and clean for mergers, acquisitions, liquidations and suspensions. I had to leave out companies from Channel Island, Bermuda and Barbados, since there were no data on these countries and they are territories under the

(23)

22

jurisdiction and sovereignty of the United Kingdom. For companies from Zambia and Cyprus there were no data, so they were taking from the sample as well. The same as for companies listed on United States exchanges, all firms with total assets lower than $ 1 million were removed from the dataset. After this for United Kingdom 133 companies delisted in the period of January 1995 to December 2012, for the Hong Kong only 19 companies were left. For Hong Kong a difficulty was the listings that came from China, because the data from China did not show to be accurate a lot of the times.

Another important data filter used for all three countries is that a company that delists in the sample period needs to have at least one year in between the delisting year of the foreign listing and, if it is the case, the home country delisting. For comparison between the exchanges and countries all information is downloaded in United States dollars.

In Table 1 you see an overview of the number of companies that were cross-listed during the period, where delisted means that a company delisted and active means that a company was cross-listed in the sample period of 1995-2012. Both the delisted and active companies could be listed later than 1995. For the active companies this means they listed later then the year 1995, but they were at least still listed in 2012. When looking at the United States exchanges you see that the majority of the delisted and the active companies came from Canada, with 36% of the United States delisted companies and 38% of the United States active companies. Thereafter the United Kingdom had the most delisted (12%) and active (10%) companies. On the United Kingdom exchanges it looks like the portions are more split, with the United States having the largest numbers of 30 delisted (23%) versus 36 (19%) active companies. After this Germany comes up with the most delisted companies on UK exchanges, namely 17 (13%), while Ireland shows the next majority of the active companies with 35 (18%). Looking at the Hong Kong exchange the majority of active companies, namely 202 (83%), come from China. For the delisted companies Singapore has the largest portion, with 8 companies (42%), and China is the next country with 4 (21%) delisted companies.

As mentioned before the process of coming to a final list is a time consuming process and the list of voluntary delisting companies with which you end up can vary per research. In the study from Doidge et al. (2010) you see that they come up with a total of 141 deregistering firms from the period 2002-2008 for only in the United States. In this research 120 delisting

(24)

23

Table 1. Sample description.

This table show the number of companies that cross-delisted (Delisted) and cross-listed (Active) during the sample period of 1995-2012 per country of the home market. In total for the three countries together 356 cross-delisting companies and 930 cross-listed companies are shown.

United States United Kingdom Hong Kong

Delisted Active Delisted Active Delisted Active

Argentina 1 13 0 0 0 0 Australia 9 11 12 23 1 1 Austria 1 0 0 0 0 0 Belgium 0 3 3 0 0 0 Brazil 0 31 0 0 0 1 Canada 73 187 3 20 0 3 Chile 5 12 0 0 0 0 China 0 11 0 4 4 202 Colombia 0 4 0 0 0 0 Czech Republic 0 1 0 1 0 0 Denmark 3 1 0 0 0 0 Finland 0 2 0 2 0 0 France 5 9 9 3 0 0 Germany 7 8 17 5 0 0 Greece 0 3 0 0 0 0 Hong Kong 2 5 0 6 0 0 India 2 7 1 0 0 0 Indonesia 1 1 0 0 0 0 Ireland 3 2 4 35 0 0 Israel 11 37 2 4 0 0 Italy 2 6 15 0 0 0 Japan 12 16 3 7 0 0 Kenya 0 0 0 1 0 0 Luxembourg 1 1 2 1 0 0 Malaysia 1 1 1 1 1 0 Mexico 9 15 0 0 0 0 Netherlands 8 10 10 2 0 0 New Zealand 1 0 0 0 0 0 Nigeria 0 0 1 1 0 0 Norway 4 8 1 4 0 0 Peru 1 3 0 0 0 0 Philippines 2 1 0 0 0 0 Russia 0 3 0 2 0 0 Singapore 1 0 0 21 8 12 South Africa 3 7 6 4 0 0 South Korea 0 8 0 0 0 0 Spain 1 5 3 4 0 0 Sri Lanka 1 0 0 0 0 0 Sweden 6 1 7 0 0 0 Switzerland 2 6 0 1 0 0 Taiwan 0 6 0 0 0 0 Thailand 1 0 0 0 0 0 Turkey 0 1 0 0 0 0 United Kingdom 24 50 0 0 3 9 United States 0 0 30 36 2 15 Venezuela 1 0 0 1 0 0 Zimbabwe 0 0 3 2 0 0 Total 204 496 133 191 19 243

companies are shown in this period (2002-2008) for United States, United Kingdom and Hong Kong. A possible explanation for this is that in this research OTC markets and Rule 144a private placements were left out and another difference could be that deregistering firms

(25)

24

with the SEC show different numbers than actually delisting firms from exchanges (this research sample).

In Figure 1 the delisting companies are shown over the period of the sample. The peak of the delisting companies is shown in 1998 with 55 companies delisting from either United States, United Kingdom or Hong Kong Exchanges. Although the other peak in 2007 is not as high as in other studies, still it seems there is a significant effect of delisting companies in this year.

Figure 1. Number of delisting companies over time. This figure shows companies that voluntary cross-delisted over the

time period 1995 and 2012. The peak of delisting companies is shown in 1998 with 55 companies delisting from either United States, United Kingdom or Hong Kong exchanges.

Furthermore, I expected more delisting companies after the year 2000, since markets became more global and you would have anticipated more delisting companies. Likely this shows the fact that since the year 2000 the number of new cross-listings dropped significantly, as shown by other studies (Doidge et al., 2009; Youet al., 2012).

In Table 2, Table 3 and Table 4 descriptive statistics are shown for the most important variables used in the previously mentioned multi period probit model. In the United States (Table 2) the anti-self-dealing index shows hardly any difference between active and delisted companies. In the United Kingdom (Table 3) this difference is a bit larger, while in Hong Kong (Table 4) there is a large difference. Although I have to watch out for any conclusions,

25 20 23 55 25 21 14 ₁₃ 17 8 15 12 43 12 20 5 8 5 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Number of delisting firms

(26)

25

since the delisted sample in Hong Kong are only 19 firms and the sample form Hong Kong in total is largely dominated by China.

Table 2. Descriptive statistics for multi period probit sample United States.

This table shows the summary statistics for the United States exchanges. It is shown for cross-delisting companies (Delist) and cross-listed (Active) companies separately. The data Total Assets, Market Capitalization, TobinQ, Capital Expenditures divided by Total Assets and GNI per capita are all winsorized at 0.01 and 0.99. Firm years are from 1995 to 2012. Variable definitions are in appendix Table G.1.

US Delist

Mean Std Dev Min Max Firm years

ASDI 0.59 0.22 0.17 1.00 2753 RADRI 3.97 0.69 2.00 5.00 2753 Individualism 69.51 19.23 14.00 90.00 2753 Closely held 30.90 24.12 0.00 99.80 801 Total Assets 35100000 77600000 1000176 790000000 942 Market Cap 13600000 26100000 7453 181000000 930 TobinQ 1.40 0.75 0.66 8.63 929 Capex/TA 0.05 0.04 0.00 0.29 927

GNI per Capita 31446 14937 380 98880 2747

US Active

ASDI 0.58 0.22 0.17 0.96 6636 RADRI 3.93 0.95 1.00 5.00 6636 Individualism 61.64 23.84 13.00 90.00 6636 Closely held 26.73 28.95 0.00 100.00 2578 Total Assets 80500000 193000000 1000968 1030000000 4057 Market Cap 23600000 34800000 3150 181000000 3574 TobinQ 1.66 1.61 0.21 31.66 3556 Capex/TA 0.07 0.06 0.00 0.66 3988

GNI per Capita 24198 16173 380 98880 6566

For the United States (Table 2) and United Kingdom (Table 3) the means of closely held shares vary around 30%, while the Hong Kong (Table 4) active market shows a far larger number with 53.99%. This is probably due to the domination of the Chinese companies that are cross-listed on the Hong Kong exchange. Looking at the TobinQ ratio all the averages are at approximately 0.5 point above one, meaning that the market value of the firms are one and a half time the asset value of the firms. The available firm years for testing are shown in the last column of each table. United Kingdom (Table 3) shows the highest number of years for delisting companies. Although the United States (Table 2) have the largest amount of delisting companies, due to the variable closely held shares the United Kingdom comes up with the most firm years. This is probably due to the fact that this variable is extracted from Datastream instead of CRSP.

(27)

26

Table 3. Descriptive statistics for multi period probit sample United Kingdom.

This table shows the summary statistics for the United Kingdom exchanges. It is shown for cross-delisting companies (Delist) and cross-listed (Active) companies separately. The data Total Assets, Market Capitalization, TobinQ, Capital Expenditures divided by Total Assets and GNI per capita are all winsorized at 0.01 and 0.99. Firm years are from 1995 to 2012. Variable definitions are in appendix Table G.1.

UK Delist

ASDI 0.49 0.19 0.2 0.95 1825

RADRI 3.31 0.86 2 5 1825

Individualism 74.99 13.7 26 91 1777

Closely held 33.29 26.48 0 100 808

Total Assets 36900000 121000000 1004761 1.03E+09 860

Market Cap 9950889 16100000 2300 131000000 840

TobinQ 1.79 1.42 0 14.19 838

Capex/TA 0.05 0.05 0 0.53 831

GNI per Capita 33588 14372 170 98880 1824

UK Active

ASDI 0.69 0.21 0.09 1 2464

RADRI 3.99 0.99 1 5 2464

Individualism 63.92 26.49 12 91 2431

Closely held 30.58 27.18 0 100 1160

Total Assets 87100000 201000000 1007668 1.03E+09 1219

Market Cap 30400000 47500000 2300 181000000 1184

TobinQ 1.47 0.82 0 9.37 1184

Capex/TA 0.05 0.04 0 0.32 1212

GNI per Capita 30137 14097 170 98880 2463

Looking at the gross national income per capita you see some diversion in the various tables. The biggest diversion shows in Table 4 where the active companies display an average GNI per Capita of 7653. As mentioned earlier this result could occur due to the China domination in the sample for Hong Kong.

The Table B.1 in the appendix shows the values of the variable of interest when the dummy variable is one. ASDIQ1, ASDIQ2 and ASDIQ3 are dummies for the variable anti-self-dealing index (ASDI). RADRIQ1, RADRIQ2 and RADRIQ3 are dummies for the revised anti-director rights index (RADRI). HCHQ1, HCHQ2 and HCHQ3 are dummies for closely held shares (Closely held). HIHIQ1, HIHIQ2 and HIHIQ3 are dummies for the individualism Hofstede index. In this table you can see the cut off points for which values the dummies are made. Also the years for when the dummies show a value of one are displayed.

(28)

27

Table 4. Descriptive statistics for multi period probit sample Hong Kong.

This table shows the summary statistics for the Hong Kong exchange. It is shown for cross-delisting companies (Delist) and cross-listed (Active) companies separately. The data Total Assets, Market Capitalization, TobinQ, Capital Expenditures divided by Total Assets and GNI per capita are all winsorized at 0.01 and 0.99. Firm years are from 1995 to 2012. Variable definitions are in appendix Table G.1.

HK Delist

ASDI 0.89 0.13 0.65 1 278

RADRI 3.94 1.59 1 5 278

Individualism 45.36 33.6 20 91 278

Closely held 32.72 29.54 0 99 141

Total Assets 120000000 279000000 1018167 1.03E+09 144

Market Cap 28300000 48100000 84618 181000000 144

TobinQ 1.72 1.78 0.57 14.19 144

Capex/TA 0.04 0.04 0 0.19 144

GNI per Capita 26896 15240 530 59760 278

HK Active

ASDI 0.77 0.08 0.27 1 3080

RADRI 1.55 1.24 1 5 3080

Individualism 29.47 23.72 20 91 3080

Closely held 53.99 30.1 0 100 1161

Total Assets 59700000 167000000 1004752 1.03E+09 1287

Market Cap 22300000 42800000 52598 181000000 1127

TobinQ 1.76 1.5 0.35 16.64 1127

Capex/TA 0.07 0.07 0 0.46 1278

GNI per Capita 7653 13110 530 59760 3080

For the event analysis Table 5 shows descriptive statistics of the relevant firms characteristics. Here you can clearly see that the ratio capital expenditures over lagged property, plant and equipment (dependent variable) shows differences on the means and standard deviations for the United States, United Kingdom and Hong Kong. For the normalized stock price the United States and Hong Kong listed companies show almost the same characteristics, while the companies that were listed in the United Kingdom show a higher mean with higher standard deviation.

For the cash flows over total assets the means and standard deviations are roughly the same, although looking at the minimum and maximum values you see there are indeed some differences. Firm years here are different from the previous table, because in the event analysis time can vary outside the sample period of 1995 till 2012. Also in order to create the dependent variable earlier data were needed. So firm years in the event analysis counts from 1989 till 2014.

(29)

28

Table 5. Descriptive statistics for event analysis.

This table show a summary of the characteristics used in the event analysis. All data used for capital expenditures, property, plant and equipment, capital expenditures divided by lagged property, plant and equipment, normalized stock prize (NormP) and cash flow divided by total assets are winsorized. Firm years are from 1989 till 2014. Variable definitions are in appendix Table G.1.

US Delist

Capex 1133321 1852436 0 7648694 649 PPE 6404635 9590241 8449 38600000 650 Cap/PPE-1 0.21 0.45 0.00 10.84 668 NormP 1.32 0.58 0.66 6.30 642 CF/TA 0.07 0.07 -0.21 0.50 622 UK Delist

Capex 478813 846014 0 7648694 824 PPE 2686722 4894865 9118 35800000 862 Cap/PPE-1 0.32 1.11 0.00 18.77 847 NormP 1.62 1.07 0.62 10.82 836 CF/TA 0.07 0.08 -0.27 0.40 619 HK Delist

Capex 643364 918018 0 4742890 103

PPE 3795281 4592135 26273 17900000 103

Cap/PPE-1 0.25 0.75 0.00 7.17 99

NormP 1.30 0.54 0.69 3.92 103

(30)

29

5. Empirical results

In this part the methodology will be put to the test and results will be shown and interpreted. First a correlation table will be analyzed and after that the outcome of the regressions will be discussed, the relation to the hypotheses and the first comparison to the existing literature is made.

The correlation table is shown in appendix Table C.1. Here all the correlations are shown between the variables. The variables which show a higher correlation that 0.8 or lower than -0.8 are highlighted in red. Analyzing this no multi collinearity problem is expected, due to the fact that the n in the sample is large enough to deal with this. For the regressions the United States and United Kingdom are also looked at individually. Hong Kong is not shown separately because of the small amount of delisted firms (19). However, the sample of Hong Kong is taken into account when the total sample is used.

5.1.1 Investor protection: Common law

For testing the first hypothesis three multi period probit models are estimated as seen in Table 6. In model (1) the total sample, meaning United States, United Kingdom and Hong Kong with all index variables and closely held shares is estimated. Model (2) and (3) are for the United States and United Kingdom separately. As this is a probit model the magnitude of the coefficients cannot be interpreted from the table, but the sign of the coefficient can be interpreted. Translating this to this research a cross-listed company is more or less likely to delist following the sign of the dummy variable common law.

Hypothesis one states that companies with a non-common law tradition in their home country are expected to delist more often. This gives the prediction that the coefficient of the dummy variable common law should have a negative sign in the estimated model. Table 6 shows a significant common law dummy for the total sample (1) at a significance level of 0.05, the United States (2) at a significance level of 0.01 and United Kingdom (3) at a significance

(31)

30

level of 0.05. Analyzing the coefficients a negative sign is shown for the total sample and the United States. This is in line with the hypothesis, meaning that if a company comes from a country with a common law tradition it is less likely to delist from the foreign exchange. However, for the United Kingdom a positive sign is found. This indicates that cross-listed companies, with a common law country tradition, on United Kingdom exchanges are more likely to delist. The other variables of interest RADRI, ASDI, Closely held and individualism show significance in these first models but are discussed later. The control variables GNI per Capita, Total Liabilities and Capex/Total Assets, TobinQ and Market Cap show significance in one or two of the models, while log(Total Assets) is not showing any significant coefficients.

For the bonding theory this is a relevant outcome. As cited earlier the United States is seen as the main country for proclaims on the bonding theory, since in the United States the most regulated and prestige exchanges are based, as discussed in the literature review. In this result you see an opposite outcome of the hypothesis in the United Kingdom. The effect here suggests that cross-listed companies form a common law tradition are more likely to delist. This is rejecting the hypothesis. To make a claim towards the bonding theory keep in mind that this is one outcome and all hypotheses should be considered all together in order to make a statement about the outcome of the study. For the United States it suggests that cross-listing companies from higher investor protection countries are less eager to delist, due to the fact that the insiders are less eager to leave foreign markets for compliance and regulatory issues. They have higher investor protections laws in their own country, leaving less benefits for insiders from common law countries. Comparing this result to others gives different outcomes. First Roosenboom and van Dijk (2009) found evidence for the bonding theory in the United States and United Kingdom by using a civil law dummy. This is different than what is found here, it might be due to the fact that they looked at cross-listing instead of cross-delisting. However, for the United States it is in line with their results. Witmer (2005) only looked at United States data and found the same significant relation between common law and cross-delisting as in this study.

(32)

31

5.1.2 Investor Protection: Anti-director rights index

For testing the second hypothesis nine probit models are estimated. The second hypothesis states that companies which have a lower score on the anti-director rights index are expected to delist more often. In order to test the companies with a lower score on the anti-director rights index three dummy variables (RADRIQ1, RADRIQ2 and RADRIQ3) are created for the anti-director rights index. Values for the different cutoff points are shown in appendix Table B.1. Dummies are created to show the clear distinction between a company which has a low value on the index compared to the other values of the index. This is different for interpretation of the index variable itself since this would show the relation between delisting and the index and not for a specific group. For each of the dummies separate models are estimated and the significant one (RADRIQ1) is shown in Table 6 (model (4)-(6)), the other two dummies (RADRIQ2, RADRIQ3) are displayed in appendix Table D.1 (model (1)-(6)). This means that the only significant effect was found using the dummy variable with the smallest sample of lowest scores on the revised anti-director rights index.

In Table 6 for the total sample (1) and United States sample (2) variable RADRI shows a significant positive coefficient at a 0.05 level. If this sign of the coefficient is interpreted than I would say that a higher score on the revised anti-director rights index gives a more likely change of a cross-listed company to delist. So a company from a country with higher investor protection is more likely to delist. When the RADRI is replaced with its dummy (model (4)-(6)) it shows a negative significant coefficient at a 0.05 level for the total sample (4) and no significance for the United States and United Kingdom. This is an interesting outcome and actually the opposite as from what was expected for the sign. If the coefficient is interpreted of the dummy variable than I would say that a lower score on the revised anti-director rights index gives a less likely change of a cross-listed company to delist. So a company from a country with low investor protection is less likely to delist. These results are not in line with the hypothesis. As discussed and empirically tested by Djankov et al. (2008) the anti-dealing index is a better measure for investor protection, especially when looking at self-dealing by corporate insiders. This however fails to explain the opposite effect that has arisen in these results. Also Daugherty and Georgieva (2011) find a significant negative coefficient which is the opposite of what is found in this study. The control variables GNI per Capita, Total Liabilities and Capex/Total Assets, TobinQ and Market Cap show significance in one or two of the models, while log(Total Assets) is not showing any significant coefficients.

(33)

Table 6. Multi period probit model for testing H1, H2 and H3.

This table presents several multi period probit regressions that examines the impact of three different investor protection measures on delisting. Common law is a dummy variable that is one if a firm’s home country has a common law tradition, RADRI is the revised anti director rights index, ASDI is the anti-self- dealing index, Closely held is a percentage of closely held shares by corporate insiders and individualism is the individualism Hofstede index. In model (1), (4) and (7) the whole sample is taken into account, containing United States, United Kingdom and Hong Kong companies. The other variable definitions are in appendix Table G.1. The standard errors are shown in parenthesis and are adjusted for clustering on firm level. Statistical significance levels for t-tests are shown by *** p<0.01, ** p<0.05, * p<0.1.

(1) (2) (3) (4) (5) (6) (7) (8) (9) Total US UK Total US UK Total US UK

Common law -0.477** -1.189*** 0.902** -0.568*** -1.145*** 0.649 0.697 -0.826 4.413*** (0.196) (0.272) (0.379) (0.210) (0.274) (0.432) (0.466) (0.600) (0.714) RADRI 0.0780** 0.168** -0.0943 -0.0560 0.228** -0.263** (0.0370) (0.0808) (0.0665) (0.0609) (0.0950) (0.105) RADRIQ1 -0.295** -0.172 -0.115 (0.118) (0.208) (0.195) ASDI -0.453 0.884** -4.083*** -0.337 1.136*** -3.894*** (0.291) (0.436) (0.756) (0.304) (0.414) (0.838) ASDIQ3 1.193*** 0.0185 4.946*** (0.377) (0.438) (0.562) Closely held 0.00758*** 0.00943*** 0.00666** 0.00781*** 0.00867*** 0.00737*** 0.00708*** 0.00971*** 0.00494 (0.00180) (0.00305) (0.00281) (0.00178) (0.00310) (0.00282) (0.00185) (0.00290) (0.00304) Individualism 0.0221*** 0.0195*** 0.0103 0.0234*** 0.0180*** 0.0166** 0.0154*** 0.0184*** 0.0169** (0.00383) (0.00523) (0.00780) (0.00422) (0.00536) (0.00849) (0.00417) (0.00677) (0.00689)

GNI per Capita 1.85e-09 1.37e-05*** -2.15e-05*** -1.10e-06 1.35e-05*** -2.20e-05*** -1.78e-06 1.36e-05*** -2.12e-05***

(3.09e-06) (4.66e-06) (5.12e-06) (3.30e-06) (4.82e-06) (5.11e-06) (2.95e-06) (4.47e-06) (5.28e-06)

Total Liabilities -5.13e-10 -8.14e-10** -6.54e-10 -4.75e-10 -8.55e-10** -6.98e-10 -4.20e-10 -8.86e-10** 2.19e-10

(3.54e-10) (4.07e-10) (8.49e-10) (3.45e-10) (4.18e-10) (8.48e-10) (3.44e-10) (4.21e-10) (7.02e-10)

log(Total Assets) -0.0266 -0.0550 0.0569 -0.0269 -0.0463 0.0450 -0.0344 -0.0422 0.0189 (0.0425) (0.0619) (0.0764) (0.0420) (0.0644) (0.0793) (0.0423) (0.0610) (0.0715) Capex/Total Assets -2.919*** -3.849** -1.814 -2.948*** -3.846** -1.824 -2.896*** -3.699** -1.178 (0.976) (1.710) (1.684) (0.966) (1.660) (1.722) (0.995) (1.745) (1.684) TobinQ -0.0544 -0.303*** 0.246*** -0.0574 -0.296*** 0.245*** -0.0515 -0.283*** 0.229*** (0.0495) (0.105) (0.0781) (0.0506) (0.107) (0.0786) (0.0481) (0.104) (0.0724)

Market Cap -4.00e-09* -8.87e-10 -1.41e-08*** -4.05e-09* -8.64e-10 -1.35e-08** -3.75e-09 -1.01e-09 -1.31e-08***

(34)

33

Table 6 – Continued

(1) (2) (3) (4) (5) (6) (7) (8) (9) Total US UK Total US UK Total US UK

Constant -2.666*** -3.243** -0.597 -2.396*** -2.736** -1.159 -2.908*** -3.308** -6.290***

(0.724) (1.285) (1.614) (0.741) (1.213) (1.548) (0.716) (1.293) (1.222)

Observations 4,993 2,919 1,081 4,993 2,919 1,081 4,993 2,919 1,081

(35)

34

5.1.3 Investor protection: Anti-self-dealing index

For hypothesis three nine probit models are estimated. The third hypothesis states that companies which have a lower score on the anti-self-dealing index are expected to delist more often. In order to test for companies with a lower score on the anti-self-dealing index three dummy variables (ASDIQ1, ASDIQ2 and ASDIQ3) are created for the anti-self-dealing index. Values for the different cutoff points are shown in appendix Table B.1. In the previous section it was already explained why the dummies were created and not only the index variable itself is tested. Following this for each of the dummies separate regressions are estimated and the significant dummy shows in Table 6 (ASDIQ3). The other dummies are shown in appendix Table D.1. This means that the only significant effect on the total sample was found using the dummy variable with the largest sample of lowest scores on the anti-self-dealing index.

In Table 6 for the coefficient of variable ASDI significance is shown at a 0.05 level for United States (2) and for a 0.01 significance level for United Kingdom (3). The sign for United States is positive and for United Kingdom is negative. Interpreting this means that in the United States cross-listed companies which come from countries with a higher score on the anti-self-dealing index are more likely to delist. Looking at the United Kingdom the opposite effect is shown indicating that cross-listed companies with a lower score for the anti-self-dealing index are more likely to delist. When the variable ASDI is replaced with the dummy variable ((7)-(9)) the coefficients show significance at a 0.01 level for the total sample (7) and for the United Kingdom (9), both with a positive coefficient. Looking at the appendix Table D.1 in model (12) variable ASDIQ2 shows also a positive significant coefficient for United Kingdom. These results indicate that, for where the coefficient is significant, a cross-listed company with a low score on the anti-self-dealing index is more likely to delist than the cross-listed companies with the higher scores. Going to the hypothesis it would be hard to give a strong conclusion. The variable ASDIQ3 gives a result in favor of the hypothesis. However the variable ASDI shows different signs for United Kingdom and United States, which makes an interpretation in favor of the hypothesis difficult. Comparing this to other studies Doidge et al. (2010), who only worked with United States data, found no significance for the coefficient for anti-self-dealing index in any of their tests. No studies looked at different dummies for different groups within the index. For the control variables GNI per Capita, Total Liabilities and Capex/Total Assets, TobinQ and Market Cap show significance