Impact of SOX on US Cross-listings : does analyst behaviour change?

(1)

Impact of SOX on US Cross-Listings:

Does Analyst Behaviour Change?

– Master Thesis

–

Submitted to

Dr. Jeroen Ligterink Finance Group

Faculty of Economics and Business University of Amsterdam Amsterdam Business School

by

Anna-Lena Westerfeld Ceintuurbaan 234 III

1072 GE Amsterdam

E-mail: anna-lena.westerfeld@web.de M.Sc. Business Economics - Finance

Student number: 10825673

(2)

Statement of Originality

This document is written by Anna-Lena Westerfeld 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)

Acknowledgements

I thank my thesis supervisor, Dr. Jeroen Ligterink for his regular support and valuable criticism throughout the whole process of writing my master thesis. I am especially thankful for the suggestion of some alternative measures, which were very insightful for this research.

(4)

This paper investigates the impact of the passage of the Sarbanes-Oxley Act (SOX) on a global sample of non-US firms cross-listed in the US. I use a difference-in-differences approach with a control group of non-US not cross-listed firms from the same countries, industries and with comparable firm size to the treatment group. I choose analyst measures to evaluate the effectiveness of the Act. These measures include analyst coverage, analyst forecast accuracy and the percentage of buy-recommendations. The main findings indicate an effective implementation of SOX for cross-listed firms relative to the control group. Especially the effect on forecast accuracy and the recommendation structure emerge immediately and with strong statistical and economic significance. Forecast accuracy increases far more for cross-listed firms subject to the Act compared to not cross-listed firms. The percentage of buy-recommendations decreases substantially more for cross-listed firms post SOX relative to the control group. Analyst coverage seems to respond with a delay to the passage of SOX and decreases more for cross-listed firms compared to not cross-listed firms. Moreover, all analyst measures show pronounced effects for companies from countries with civil law origin. Results are robust to several robustness checks.

(6)

1 Introduction

With the passage of the Sarbanes-Oxley Act (SOX or the Act) in 2002, the SEC (Securities and Exchange Commission) introduced a shift in its philosophy from solely regulating company disclosure to also aiming at substantially influencing corporate governance standards (Ribstein, 2003). SOX has been implemented as a response to various US fraud and accounting scandals like the Enron or Worldcom scandal. It refers to all SEC regulated companies including foreign companies (Piotroski and Srinivasan, 2008). The Act focusses on increased transparency, independent and objective accounting figures via enhanced financial reporting, fewer opportunities for conflicts of interests for auditors and more responsibilities for executives including penalties for non-compliance with the rules. Section 501 of SOX especially concentrates on potential conflicts of interest arising for investment banking analysts. Analyst recommendations can be biased by a conflict of interest if the investment bank employing the analyst also advises the company which the analyst evaluates.

This research aims at inspecting the effectiveness of the regulatory act and examines the differences in analysts’ reactions to non-US firms, which cross-list at an US stock exchange, before and after the implementation of SOX. The aim is to measure the effectiveness of Sarbanes-Oxley with respect to the degree of analyst following, accuracy of analyst forecasts as well as the percentage of buy-recommendations. If these characteristics change for cross-listed firms after the implementation of SOX, it is possible to identify the effect of regulation on cross-listings and how analysts evaluate it. Using different analyst measures, it is also possible to evaluate different aspects of the Act.

There is a large amount of literature which aims at evaluating the effect of SOX on US-firms. However, the methodology used is often not optimal to evaluate the effectiveness of SOX since no control groups are used. By contrast, the Act’s impact on foreign firms is rarely researched. As the passage of SOX serves as a natural experiment, a control group of not cross-listed firms, which is not subject to the new regulation, can simulate a counterfactual. Therefore, this research uses a difference-in-differences approach. Many studies use very short

(7)

time periods in which the full effect of the regulatory impact is might not be completely visible. Consequently, the sample period used in this study spans a relatively long period from January 2000 until December 2010. I use a global sample of all non-US firms cross-listed within this period. Moreover, this sample covers a high spectrum of institutional backgrounds controlling for different origins of law, which are important determinants of investor protection (La Porta et al., 1999) and seems to play a key role in research related to cross-listings.

The impact of SOX on cross-listed firms is not straightforward with resulting positive as well as negative possible developments. The regulatory changes related to SOX are criticized among academics and the business community, mainly with respect to the large costs for companies and the rules being too strict and thus deterring necessary risk-taking by the management. Moreover, the Act’s efficiency with respect to improving investor protection is questioned (Brown, 2006).

Analyst reports serve as a main source of information for investors. Analyst coverage has not been examined for cross-listed firms in connection with SOX so far. The only related empirical finding is that analyst coverage is generally higher for cross-listed firms (Baker et al., 2002 and Lang et al., 2003a). The development of forecast accuracy for cross-listed firms with respect to SOX has been researched by Arping and Sautner (2010). However, the authors only use a limited sample of EU-firms, which cannot capture effects of different institutional backgrounds such as origin of law and the wide range of investor protection backgrounds globally. The recommendation structure has been analysed for the introduction of regulation with similar aims as SOX, such as the Global Settlement and the NYSE and NASD rules. A decrease in optimism has been discovered (Clarke et al. 2011). However, this research does not inspect cross-listed firms and does not incorporate a control group to control for endogeneity quite rigorously.

The main findings indicate an effective implementation of SOX, especially with respect to forecast accuracy and the recommendation structure. Forecast accuracy increases by a substantial 27% more for cross-listed firms subject to SOX compared to not cross-listed firms. The percentage of buy-recommendations

(8)

decreases by up to 18 percentage points more for cross-listed firms compared to not cross-listed firms. Thus, optimism seems to decline to a more objective level. Analyst coverage seems to respond with a delay to the passage of SOX and decreases by up to 19% more for cross-listed compared to not cross-listed firms. This finding indicates that SOX might be able to substitute for the monitoring role of analysts.

All effects are even larger in magnitude for cross-listed firms from civil law countries. Since these companies have lower initial disclosure levels and transparency, there is larger room for improving forecast accuracy. Similarly, analyst coverage might be less necessary for civil law firms as soon as SOX takes over the monitoring role of analysts. Results are robust to several robustness checks with respect to sample variations and inclusion of multiple fixed effects. I further conduct plausibility checks with respect to the control group, which indicate that it is a valid counterfactual for the treatment group. A larger robustness check with an alternative cut-off for the post SOX period, starting in January 2005, indicates the delayed emergence of the drop in analyst coverage. In addition, it reveals a decline in the initial large increase in forecast accuracy as well as a decline in the initial decrease in optimism revealed in the lower percentage of buy-recommendations.

This thesis is organized as follows: Section 2 covers the related literature and develops the hypotheses. Section 3 describes the data collection and summary statistics. Section 4 explains the methodology and the control variables used. Section 5 describes the results and robustness checks, whereas section 6 concludes and gives an outlook for future research.

2 Literature Review

A cross-listing is the strategic decision of a company to list the shares trading at the home exchange secondarily at a stock exchange overseas (Karolyi, 2012). Cross-listings are an important source for cross-border capital flows, increasing

(9)

financial liberalisation and enhancing economic growth (Bekaert et al, 2001). There exist clear benefits and disadvantages for foreign firms cross-listing in the US. The following literature review shows that benefits of cross-listings can be large. Benefits and costs are presented in the following as well as indications of a possible shift in the profitability of a US cross-listing post SOX. Moreover, the Act’s implications on analyst research will be described in detail.

2.1 Characteristics of Cross-Listings

There is a range of empirical evidence on the benefits of cross-listings on an US- stock exchange. US-markets offer a larger investor base, stricter disclosure rules as well as more visibility and prestige compared to other cross-listing venues and especially compared to countries with weak investor protection. Moreover, a cross-listing in the US overcomes market segmentation, which might hinder capital raising due to regulatory barriers or information constraints in the firm’s home country (Roosenboom and Van Dijk, 2009 and Foerster and Karolyi, 1999).

When a foreign firm wants to cross-list at an US-exchange, it has to register with the US Securities and Exchange Commission (SEC), comply with US GAAP accounting as well as the rules of the stock exchange. Also, US shareholders are able to sue the cross-listed firm for fraudulent accounting statements (Reese and Weisbach, 2002). According to Reese and Weisbach (2002), companies from strong investor protection countries undertake a US cross-listing in order to get access to US capital markets and its large investor base, whereas firms from weaker investor protection backgrounds cross-list to bond with US regulatory standards and to be able to raise capital at lower cost.

Ammer et al. (2008) find that US investors nearly double their holdings in foreign companies as soon as these firms cross-list at NYSE (New York Stock Exchange) or NASDAQ. The increase in holdings by US investors is especially pronounced for firms from weak investor protection countries. Merton (1987) argues that investors choose only the stocks they are aware of since they are

(10)

constrained by incomplete information. This finding may partially explain the increase in holdings by US-investors after a cross-listing.

According to Karolyi (2012), a cross-listing overcomes the agency conflict of poor investor protection in the home country of the firm and leads to a “bonding” to stronger corporate governance in the country with the cross-listing. Berger et al. (2005) calls this mechanism an “exogenous improvement in investor protection”. Accordingly, a cross-listing is more highly valued for cross-listed firms from countries with poor investor protection because the US legal system protects minority shareholders from getting expropriated. Bonding can be split into legal and reputational bonding (Karolyi, 2012). Legal bonding refers to the quality of law and its enforcement whereas reputational bonding is associated with capital market intermediaries like securities analysts that serve as external monitors.

Cross-listings are especially beneficial for companies with large growth opportunities because they need to raise capital at low costs if internal funding or risk-free debt is not available. Lower funding costs can be more attractive for the controlling shareholders than pursuing expropriation of minority shareholders. Therefore, high growth firms choose to cross-list although private benefits of control will diminish for the management (Doidge et al., 2004).

Ball et al. (2013) find that foreign firms raise not only equity but also debt more often and at lower rates following an US cross-listing. Since they raise debt via the bond market more frequently, they are less dependent on bank loans. Lower cost of debt in combination with lower cost of equity can serve as evidence that no risk-shifting takes place at the disadvantage of debt holders when firms cross-list. As Ball et al. (2013) note, especially the high growth firms, which are typically cross-listing in the US, could incorporate higher potential for agency conflicts between debt and equity holders. This finding is important from a social welfare perspective since there seems to be no disadvantaged stakeholder visible.

Hail and Leuz (2009) discover a significant decrease in cost of equity between 70 and 120 basis points when a foreign firm cross-lists in the US between the years 1990-2005. This finding is robust when controlling for several risk proxies and firm-fixed effects. The cost of capital-effect is still present after

(11)

the implementation of SOX and larger for firms with weak investor protection backgrounds, consistent with the concept of bonding. The authors find no such effect for the London Stock Exchange (LSE).

In sum, higher valuations for cross-listed firms can result from lower discounting factors since cost of capital is lower, as well as from higher cash flows due to higher expected growth rates or potentially higher operational or financial performance (Doidge et al., 2009). Foreign companies cross-listed in the US have a 16.5% higher Tobin’s q in 1997 than firms from the same country without cross-listing (Doidge et al., 2004). The same premium reaches up to 37% for firms listing on a main stock exchange compared to OTC traded shares or private placements. Doidge et al. (2004) argue that it should be optimal for high growth firms to cross-list and especially for companies from countries with poor investor protection.

2.2 Consequences of SOX

On the one hand, the passage of SOX makes cross-listings less attractive due to higher costs with respect to stricter reporting standards. The costs relate to additional internal controls to meet the new compliance standards as to additional fees to be paid to outside auditors who have to check the company reports. According to an annual survey by Finance Executives International (FEI) in 2007, total costs make up an average fraction of 0.43% of revenues for US companies with average revenues of $6.8 billion and a market capitalization above $75 million. The survey reports that for 78% of the US-companies subject to SOX the costs of the regulatory changes are larger than its benefits. There are no such numbers available for cross-listed firms.

On the other hand, it might be even more beneficial to use cross-listings since SOX as a way to signal the will to bond to US regulation in times of higher uncertainty, especially following the financial crisis in the years 2007-2008. Iliev (2010) finds that SOX Section 404, which is the most costly part of the regulatory act, leads to more conservative earnings reporting that better reflects actual

(12)

profits but also to enormous costs resulting in declining firm values for smaller firms. Section 404 requires a management report as additional item in the annual report, which shall assure the quality of internal control processes and financial reporting procedures. An outside auditor has to validate the adequacy of these structures. Yet, the study considers only small firms with a public float of less than $75 million.

Lobo and Zhou (2006) confirm Iliev’s finding also for larger firms. They discover a tendency towards more conservatism in financial reporting after SOX with lower accruals numbers and earlier realised losses. Likewise, Cohen et al. (2004) discover a reduction of earnings management following SOX for US domestic firms which should make it easier to identify the firm’s actual corporate performance. These results imply an effective implication of the regulatory rules leading to more transparency and objective accounting figures.

The flip-side of more conservative financial reporting might be private costs for the management. If there is a controlling shareholder or the management who pursues private benefits, this should be harder post SOX (Hostak, 2013). A cross-delisting could be a solution for the management. Negative stock price reactions in the home stock markets following such a cross-delisting indicate that a cross-cross-delisting is not perceived in terms of saving costs for investors but rather by the investors’ assumption of agency issues and management pursuing private benefits.

Overall, there is a dominance of net costs related to the passage of SOX with respect to delisting trends for in the US cross-listed firms. Marosi and Massoud (2008) have discovered an increasing trend of cross-delistings since SOX despite substantial deregistering costs with the SEC. In addition, there are several difficulties due to the delisting process from an US-exchange for foreign firms since they are subject to SEC disclosure until they have less than 300 US investors. Firms from poor corporate governance countries (measured by the La Porta antidirector-index including origin of law) especially tend to pursue delistings, which contrasts the bonding hypothesis that especially these firms should benefit from bonding to the strict US-regulation. The pronounced voluntary cross-delistings by firms from poor corporate governance countries is

(13)

rather in line with an avoidance-theory meaning that companies want to escape stricter regulation, which result in higher costs.

Nevertheless, Piotroski and Srinivasan (2008) find that the preferences of large firms in the choice between a cross-listing at an US or UK stock exchange in the years 1995-2006 does not change after the implementation of SOX. The authors suggest that US-exchanges might even have become more attractive due to a pronounced bonding mechanism. The marginal likelihood for large emerging market companies increased to list at NASDAQ instead of LSE’s Main Market. However, small firms rather tend to list at an UK-exchange due to the higher relative costs of the accompanied stricter reporting standards with respect to SOX in the US. Doidge et al. (2009) use another methodology to analyse the number of new listings and delistings for US (main exchanges and OTC) as well as London (main market and AIM) cross-listings. They project pre SOX numbers of cross-listings based on firm characteristics via a logistic regression on the after SOX period. Their findings suggest that the decline in the number of listings at both venues is due to changing firm characteristics and not due to changes in benefits of the listing venues due to SOX. The authors find no support that a firm cross-listing in the US in the 1990s would no longer do so after the passage of the Act. More precisely, there are just not that many large, fast growing companies with low debt, which fit the typical profile of foreign firms cross-listing in the US. In fact, the authors discover listing surpluses in the US when taking firm characteristics into account.

Overall, these mixed findings at least indicate that bonding is not strong enough to keep some sorts of firms listed in the US but they can potentially indicate that SOX is effective in constraining corporate governance practices more strongly. Bianconi et al. (2013) assume that SOX has contributed to a segmentation of markets with the incentive for firms with lower valuations heading to other cross-listing destinations like Hong Kong. This migration development would not be in the interest of the Act to incentivise better corporate governance for a larger number of firms.

There are also some contradicting findings regarding the valuation premium for listed firms since SOX. Doidge et al. (2009) find that the

(14)

cross-listing premium is present in every year from 1990 to 2005, however with varying magnitude. Litvak (2008) however, discovers a declining premium on Tobin’s q for in the US cross-listed firms in 2002, when SOX has been implemented, which stays significantly below the pre-SOX level until 2005. However, there seems to be an overreaction since a third of the decline recovers in 2003.

Litvak (2007) further finds that only companies from poor governance countries with high growth might benefit from the Act. The author matches a cross-listed firm with a not cross-listed firm from the same country on the propensity to list including firm characteristics. She compares US cross-listed firms via ADR Level II and III with Level I and IV not subject to SOX as well as with not cross-listed firms. ADR Level II and III firms, being subject to SOX, experience negative stock price reactions for events indicating the likelihood of the Act’s passage compared to not cross-listed firms. Level I and IV react similarly, but with a weaker response (Litvak 2007). The author finds the strongest reaction for small firms and firms which already had a high level of disclosure before SOX. Consequently, the passage of SOX can be interpreted mainly as increased net costs not contributing to a stronger bonding in this case.

These empirical findings might not reveal if SOX results in net costs or benefits in general but they stress that firm characteristics such as the legal background seem to play a large role for firms cross-listed in the US.

2.3 Origin of Law

Extensive research by La Porta et al. (1998, 1999 and 2000) shows that legal families shape a country’s legal rules, which in turn influence the development of financial markets. Most importantly, French civil law and Socialist countries provide weaker investor protection than English common law countries. German and Scandinavian civil law countries are somehow in between with very strong law enforcement and rather strong creditor protection but weak shareholder protection. La Porta et al. (1999) stress that Socialist law countries aim at

(15)

maintaining the state’s power and do not intend to protect property or minority interests.

The authors find that the differences in laws and its enforcement can explain where companies obtain their funding. Only shareholders and creditors protected by law get their claims back and are willing to finance firms. When investor protection is strong, private benefits for company insiders diminish and refinancing costs decrease (La Porta et al., 2000). According to La Porta et al. (1998), strong investor protection is associated with effective corporate governance mechanisms, broad financial markets, efficient allocation of capital and dispersed share ownership, thus, no pyramids and large cross-ownerships.

High ownership concentrations are more common in countries with weak protection of minority shareholders and poor law enforcement. Control stakes can be a way to compensate weak legal investor protection and gain control over firm assets, according to Shleifer and Vishny (1997). High ownership concentration typically appears in form of family ownership and is common in Western Europe, but also in South and East Asia, the Middle East, as well as Latin America and Africa (Burkart et al., 2003). In legal regimes with weak investor protection like in many emerging market countries, both, ownership and management are in the hands of the family. In intermediate investor protection countries like Western Europe the ownership refers mainly to the family but management is delegated to a professional. Strongest investor protection can be aligned with a widely held professionally managed company, which is associated with Anglo-Saxon countries (Burkart et al., 2003).

Similarly, private control benefits are associated with less developed capital markets and more concentrated share ownership (Dyck and Zingales, 2004). Nenova (2003) finds, that private benefits of control are especially pronounced in civil law countries like Brazil, Chile, Mexico, France, Italy and South Korea. They are valued up to a half of the companies’ market capitalizations. By contrast, control benefits in English common law and Scandinavian law countries are only valued at average 4.5% of the market capitalization since investor protection and law enforcement are stricter. The author also mentions, that expropriation of minority shareholders can decrease

(16)

by up to 37% when moving from a weak to strong investor protection background. Exactly this movement can be performed by a cross-listing in the US when bonding to stricter regulatory standards. Legal protection of minority shareholders by country laws and its enforcement explain 75% of the control benefits across countries. Corporate charter characteristics explain only a small part in the variation.

2.4 The Role of Research Analysts

The focus of this study is on one potential channel, namely the reactions of analysts to SOX. By focusing on analyst behaviour I am able to examine the Act’s effectiveness in interaction with an important intermediary in financial markets. Karolyi (2006) describes the function of analysts as reputational intermediaries, monitoring companies also when company-internal monitoring mechanisms fail. He pronounces that future research needs to focus on informational intermediaries to better understand what the consequences of changes in disclosure requirements on listed shares are. According to Karolyi, research should also analyse how analyst activity changes and how the quality of their reports alters when the environment of capital markets changes.

Investors use analyst reports as their primary source to receive information about firms (Marcus and Wallace, 1991). Walther (1997) even discovers that investors put more weight on analyst forecasts for firms with higher analyst coverage. This resembles an amplifying effect of analysts becoming more important the more analysts follow a firm.

According to a prediction of a model by Chemmanur and Fulghieri (2006), a cross-listing at a high-reputation stock exchange with strict disclosure requirements should lead to an increase in the production of available information about the firm as well as higher analyst coverage. The higher listing standards and disclosure requirements at a high-reputation exchange such as the NYSE lead to higher costs, which should be balanced by greater benefits such as higher analyst coverage. Empirical results by Lang et al. (2003a) confirm this

(17)

model prediction and report that in the US cross-listed firms have greater analyst coverage than not cross-listed firms. This increase occurs around the cross-listing and should therefore not be due to other reasons. There are 2.44 more analysts covering the average cross-listed firm compared to a matched sample of only domestically listed firms. Emerging markets companies with civil law legal background cross-listed in the US have even 4.1 more analysts than a not cross-listed firm.

Likewise, Chang et al. (2000) find that common law countries have on average lower analyst coverage. A possible explanation could be a higher demand for analyst coverage in countries where investors find an opaque informational situation. If minority shareholder protection is low, investors as well as analysts are not provided with available information. A higher antidirector-index by La Porta et al. (1997), which tells how easily investors can pursue their rights against management, suggests a lower forecast error (Chang et al., 2000). The authors find lower forecast error when analyst coverage is high and the legal origin is common law.

A positive analyst reaction to cross-listings in the US occurring after the implementation of SOX, mirrored by increased analyst coverage, could highlight the positive effects of this regulation to the “bonding” mechanism of cross-listings. Stricter disclosure requirements lead to stronger protection of minority shareholders. Extracting private benefits becomes more costly and declines for cross-listed firms after the passage of SOX (Karolyi, 2012).

Yet, stricter disclosure requirements could also substitute the need for analysts as monitors and analyst coverage could decrease after SOX. This substitution idea derives from a weakly positive correlation between analyst following and investor protection measures (Berger et al., 2005). An analyst can via her intermediation of disclosure serve as a monitor who makes it more difficult for company insiders to extract private benefits. As soon as SOX takes over at least part of the monitoring role of analysts, their role is weakened and investment banks might employ less of them. In turn, demand and supply for analysts will adjust downwards.

(18)

Additionally, the conflict of interest inherent in analyst recommendations, further described in section 2.6, could lead to lower analyst coverage. Until the implementation of the regulatory changes in 2002, analyst research was funded by investment banking business. Since this is prohibited by now, opportunities for analysts to acquire information could be more difficult and even a cut in research jobs could occur (Begley et al., 2009). However, since external monitoring by analysts should be higher valued for firms from weak investor protection countries, there could still be an increase in analyst coverage for these countries post SOX even if there should be a decline for firms from countries with strong investor protection.

2.5 Cross-Listings and Stock Price Informativeness

Analysts’ efforts in providing market analysis and issuing forecasts as well as recommendations enhance informational efficiency by detecting mispricing (Green, 2006). Likewise, greater analyst coverage is associated with an improved and faster incorporation of common information into prices, holding firm size constant (Brennan et al, 1993).

Foucault and Gehrig (2008) find that cross-listings have the potential to lead to better investment decisions since cross-listed firms incorporate higher stock price informativeness. When considering informationally segmented markets, informed traders can trade on their information in the domestic market without affecting the foreign market immediately. This applies as long as trading is more or less evenly distributed between the two markets. Prices in the two different markets can thus differ slightly in the very short run. If informed traders trade more aggressively and can benefit from such price discrepancies due to segmented markets, more informed trading will take place. As a result, stock prices informativeness measured by stock price volatility can increase. However, liquidity trading becomes more costly because informed traders’ profits equal liquidity traders’ losses. The authors note that informational segmentation

(19)

can also occur in the form that foreign investors are not as good informed about a firm as domestic ones. This information asymmetry can be overcome via a cross-listing as long as it leads to substantially more available information to foreign investors.

In an informationally integrated market, however, the benefits from informed trading will decrease. This relation serves as a potential explanation for less firms cross-listing in Europe since European financial markets have become more integrated (Pagano et al., 2002). In the context of the model by Foucault and Gehrig (2008), the authors suggest that an increase in publicly available information, such as due to SOX, can potentially decrease trading benefits of informed trading with private information and lead to a decline in stock price informativeness.

This idea is confirmed empirically by Fernandes and Ferreira (2008). They use idiosyncratic stock price variation as a proxy for the rate of private information incorporated in the stock price via trading. They find that cross-listed companies from emerging markets experience a decline in idiosyncratic stock price volatility after the cross-listing, which can be explained due to increased analyst coverage dragging down the positive effects of higher disclosure requirements. The difference in available information before and after the cross-listing could be so large that informed trading benefits diminish. The opposite is true for companies from developed markets and strong investor protection backgrounds: They experience increased stock price volatility after the cross-listing in the US.

Fernandes and Ferreira explain that more available private information leads to more informed trading since private information is cheaper. Their sample covers the period 1980-2003, thus not including the period subject to SOX. They conclude that increased stock price informativeness can only be confirmed for developed market firms, at least when measured with stock price volatility. More analysts covering a cross-listed emerging market firm make it less beneficial to acquire private firm-specific information. Thus, analyst coverage increases the availability of public information but deters other investors through a crowding-out mechanism from collecting firm-specific

(20)

information and idiosyncratic return volatility declines (Fernandes and Ferreira, 2008).

In conclusion, more information is not necessarily better from every perspective. However, if initial analyst coverage is low, a cross-listing has the dominant effect of increased disclosure and improved price informativeness. If analyst coverage increases further, a cross-listing can have the opposite effect when reaching a certain threshold (Fernandes and Ferreira, 2008).

Similarly, Bailey et al. (2006) find that reactions of volume and absolute return to earnings announcements increase when a firm cross-lists, and especially if it is a developed country firm. However, they state that a cross-listing should lead to more precise information and lower costs for acquiring this information about a firm and thus, lower disagreement and less volatile reactions to earnings announcements. As Kim and Verrecchia (1991, 1994) show, the volume and volatility reaction to an earnings announcement should increase in information asymmetry among investors. This implies that lower information asymmetry for cross-listed firms should actually decrease volume and volatility reactions and not increase it. Disagreement among investors is due to different expectations about the future stock price and thus, information asymmetry, which causes trading volume. A less volatile reaction would reveal an increased quality of information and less uncertainty among investors. This would also result in more liquidity trading accompanied by lower cost of acquiring information, more precise information incorporated in the price and higher analyst following. These effects should be pronounced for emerging market countries since they have lowest disclosure standards (Bailey et al., 2006). Explanations for the increased volume reaction are rather speculative and include the potential higher uncertainty a firm is facing when expanding overseas and the unknown behavior of certain stakeholders like controlling shareholders after the cross-listing.

Still, it is not clear how information asymmetry and reactions to earnings announcements relate to forecast accuracy. As Lang et al. (2003a) find, analyst forecast accuracy increases by 1.15% for listed instead of 0.4% for not cross-listed firms in the same period when a firm cross-lists in the US. The authors

(21)

conclude that a cross-listing in the US provides analysts with more reliable and complete information. When more disclosure informs better about future earnings, analysts can make more precise estimates. This increase is more pronounced for firms from civil law emerging markets. As Chang et al. (2000) note, earnings management and smoothed income make it easier for managers to meet earnings forecasts by analysts, which can bias the obtained results, especially if earnings management is more pronounced in poor governance countries.

By contrast, as Lang et al. (2003b) show, increased information disclosure can also lead to more dispersed analyst forecasts and therefore also to less accurate forecasts. This finding stresses again the disagreement component in the provision of additional information. More available information has a larger potential to be interpreted differently by several analysts depending on the additionally disclosed information being private or common information. In case that analysts share the same forecasting model and common information, but have different private information, increased firm-provided disclosure and thus informativeness will lead them to put less weight on their private information. Therefore, dispersion among analyst forecasts will decrease. On the other hand, if analysts have the same common and private information but weight them differently, more disclosure can lead to more dispersed forecasts (Lang and Lundholm, 1996).

Arping and Sautner (2010) find that forecast error and forecast dispersion decrease for EU firms listing at an US-exchange after SOX compared to a control group without such a cross-listing. The authors refer to the two measures as opaqueness factors and find a decrease in them also for the control group after SOX but a significantly larger effect for the treatment group of cross-listed firms. Since SOX especially aims at increased transparency and enhanced reliability of financial records, analyst forecasts should become more precise after the implementation of the Act.

Correspondingly, Begley et al. (2009) find an improvement in the quality of common and private information after SOX (and related regulation). However, the increase is maintained only for one year. After that, the levels of both,

(22)

common and private information decline to pre-SOX levels. The temporary improvement and following decline in the information environment can best be seen for companies with already high prior information quality. By contrast, firms with the lowest information quality experience an increase in common information quality. A potential explanation could be that the management refuses to disclose additional voluntary information after the passage of SOX, since the CEO has to certify the correctness of financial statements and penalties for misleading information increased substantially.

Firms with high prior information quality have the largest scope to reduce disclosure. Consequently, analysts might have less access to information, especially to private information. If analysts are seen as proxy for informed investors, the information asymmetry might decline between informed and uninformed investors, as Begley et al. (2009) note. Finally, stock prices could become less informative, information asymmetry should decrease and stock prices might become less volatile. These effects should have different implications for forecast accuracy. Yet, the initial level of a firm’s information quality and thus, the legal background of the firm should play a significant role with respect to the final effect of SOX on forecast accuracy.

2.6 Analyst Conflict of Interest

SOX does not only include guidelines regarding enhanced disclosure of financial figures and the responsibilities of the management but also regulation about conflicts of interest stemming from relations to auditors and securities analysts. Section 501 concentrates especially on potential conflicts of interest for investment banking analysts.

The conflict of interest arises when the investment bank employing the analyst acts as an underwriter for the same company covered by the analyst. Agrawal and Chen (2008) find that affiliated analyst recommendation levels are more optimistic. Recommendations are positively related to the magnitude of the

(23)

conflict of interest the analyst faces. The study controls for several analyst characteristics as well as firm size and examines the period 1994-2003. The conflict can be quantified as the proportion of revenues from investment banking and brokerage business to total revenues and thus the importance of these businesses (Agrawal and Chen, 2008). However, trading patterns suggest that the market recognizes the conflict and discounts the recommendations. Therefore, no systematic misleading of investors seems to exist.

Motives for overly optimistic recommendations can be the attraction of new business or rewards for past underwriting business, according to Agrawal and Chen. With respect to brokerage houses, analyst compensation is linked to trading commissions and thus, trading volume. Due to short selling constraints (bans or costs), recommendations tend to be biased towards buy.

Michaely and Womack (1999) find that firms with buy-recommendations issued by affiliated analysts acting as lead underwriters have a worse long-run post-recommendation performance compared firms with unaffiliated buy- recommendations. The same analysts issue buy-recommendations with superior post-recommendation performance when they are in an unaffiliated situation. Thus, the driving force is not ability but a conflict of interest. The authors examine stock performance for IPO firms after the quiet period during which analysts make no recommendations or forecasts, and relate this finding to a significant evidence of bias in analyst recommendations with affiliation. In addition, lead-underwriters, analysts subject to a potential conflict of interest, publish 50% more buy-recommendations after an IPO than other analysts. Furthermore, securities analysts tend to start following those firms for which they can give a buy-recommendation (McNichols and O’Brien, 1997). Thus, analysts tend to self-select firms they cover on being able to evaluate them favourably.

However, Cowen et al. (2006) find that full-service investment banking firms issue less optimistic recommendations and more accurate forecasts than pure brokerage firms. Less optimistic underwriter forecasts and recommendations can in part be explained by reputation concerns of bulge bracket investment banks. There appear to be additional factors contributing to

(24)

biased forecasts and recommendations besides the underwriter-analyst conflict of interest. In the case of brokerage firms, the linkage of analysts’ salaries to trading commissions seems to introduce a bias towards more optimistic reports. Overall, these findings suggest that regulation introduced by the Global Settlement might not completely tackle the source of analyst optimism (Cowen et al., 2006).

The Global Settlement tries to mitigate the conflict of analysts working for an investment bank that acts as underwriter for the same firm which is covered by the analyst. Analyst bonuses were tied to the amount of new underwriting business analysts were able to attract. Some large investment banks got sanctioned and are now required to fund research through trading commissions instead of fees from underwriting services. As shown in Cowen et al. (2006) this shift in funding might pronounce an already strong source of optimism.

The Global Settlement (GS) is much related to SOX Section 501 as well to Regulation Fair Disclosure (Reg. FD). Reg. FD was introduced 23rd_{October, 2000.}

It prohibits selective information to preferred investors or analysts and requires communication to all parties at the same time. This reform has the potential to restrict information flow to analysts which can reduce the stock price informativeness as found in earlier mentioned studies. Following, in May 2002, regulatory rules by NASD (National Association of Securities Dealers) and NYSE were introduced and accepted by the SEC on 29th_{July, 2002. They aim at heavily}

reducing communication between both departments and making analyst compensation independent of investment banking business. The GS took place on 23rd_{April 2003 between the SEC, NYSE, NASD, the New York Attorney’s}

General Office and 10 large investment banks, which had to pay $1.4 billion fines for providing investors with misleading information (Begley et al, 2009).

Finally, SOX Section 501 demands analysts to disclose conflicts of interest and lengthen the quiet period after equity issues when affiliated analysts are not allowed to issue recommendations. Overall, SOX strengthened parts of the NASD and NYSE rules. Very importantly, in the US cross-listed foreign issuers are not subject to Reg. FD (Piotroski and Srinivasan, 2008). Since cross-listed firms are only subject to GS and SOX, the regulatory changes concentrate from summer

(25)

2002 to spring 2003. This is a more precise time frame to construct a cut-off between prior and post-regulation as compared to US firms, for which the regulatory changes started in 2000. It is very difficult to assign developments during the time of these regulatory changes to a single one of them. Therefore, this paper subsumes all three regulatory changes under the term SOX when referring to hypothesis 3 on analyst recommendations. All of the regulations aim primarily at disentangling the ties of the research analysts with the investment banking department.

Clarke et al. (2011) examine the reaction of analysts to the Global Settlement, NASD Rule 2711 and NYSE Rule 472. They find a decrease in strong-buy-recommendations after the passage of these rules for affiliated (21% to 13%), unaffiliated (24% to 17%) as well as independent (28% to 23%) analysts. Affiliated analysts are less likely to issue strong-buy-recommendations in the post regulation period compared to independent analysts. However, as independently classified analysts also decrease their strong-buy- recommendations. The authors assume that they might potentially also be subject to a conflict of interest or intend to not become subject to regulation in the future by reflecting a less optimistic recommendation structure.

Downgrades are interpreted as being less informative for all analyst types by the market after the passage of the regulation since their according announcement returns decreased compared to prior GS. However, affiliated analyst recommendation changes are still seen as being more informative as those by independent ones. If recommendation changes were more informative, incorporating strategic information, the market would show a greater reaction and respond with trading. At least, upgrades by unaffiliated analysts show an increased announcement return. The shift in recommendations might thus be less optimistic but not necessarily of higher quality. The study by Clarke et al. has no control group and thus, cannot control for underlying unrelated trends affecting their results. Furthermore, their results apply to US domestic and not to cross-listed firms.

Kadan et al. (2008) confirm decreased fractions of buy-recommendations and also find increased numbers of hold- and sell-recommendations. They also

(26)

note that many analysts switched from a five-category rating system to three categories. Post-regulation buy-recommendations appear to be more informative and hold and sell recommendations less informative when comparing announcement returns pre and post regulation. Overall informativeness has declined after the passage of the regulation. The study only looks at a short period from 2000-2004 and considers only US-firms. Since some SOX reforms only became binding in 2005 or 2006, I suspect this time period to be too short.

This research does not distinguish between affiliated and unaffiliated analysts, since it is not clear if only current underwriting relations lead to conflicts of interests. Simply the intention to attract firms for new coverage can create conflicts and biased recommendations. Furthermore, existing research has not found a precise distinction in pattern between different groups of analysts, maybe because they do not differ systematically.

2.7 Hypotheses

When analyst coverage increases for cross-listings in general, it might increase even more for the period after the implementation of SOX since this regulation might strengthen the bonding characteristics of US-markets. At least, this should be the case for civil law firms where bonding plays a pronounced role. Bushman et al. (2004) find a positive correlation between analyst following and investor protection as well as disclosure quality. Both should increase due to the passage of SOX. Evidence for improvement in investor protection can be drawn from positive announcement returns for stock prices of foreign private issuers for SOX-related events. This return is even higher for firms from countries with weak investor protection (Berger et al., 2005). Consequently, an increase in analyst coverage post-SOX can be expected.

On the other hand, it is also possible to find an increase in analyst following for weak investor protection countries but a decrease for firms from strong investor protection backgrounds. Especially, countries with strong investor protection might experience SOX as a substitute of analyst monitoring

(27)

(Berger et al., 2005). Therefore, including a dummy for the investor protection / origin of law is crucial. As contradiction to Berger et al., Lang and Lundholm (1996) find that increased firm-provided disclosure in general does not substitute for analyst services. Overall, it is an empirical question to find out the trend of analyst coverage post SOX. Nevertheless, I suspect analyst coverage to increase at least for civil law firms.

Hypothesis 1:

Analyst coverage for cross-listed firms increases since the

implementation of SOX compared to not cross-listed firms.

If analyst coverage increases, the information asymmetry between shareholders and the management declines (Kind and Schläpfer, 2011). Consequently, also analysts should be provided with more information. Yet, it is not clear if this additional information is common or private or if there might even be some firms with initial high disclosure getting more cautious about disclosure and providing the requested amount of information but potentially less than pre SOX in case they voluntarily published additional information before.

Backed by the evidence from Foucault and Gehrig, an increase in information in the market, in which the cross-listing takes place and which is pronounced by SOX should lead to informational integration of the domestic and the foreign market. Increased information available to the public reduces the room for private information of informed traders and hence, the profitability of their trades. As a result, informed trading could decrease in the US, but potentially not in the domestic market of the cross-listing firm. Therefore, stock price informativeness could decrease for in the US cross-listed firms. This can have an influence on forecast accuracy depending on which exact information is used by analysts to form their forecasts.

Nevertheless, Arping and Sautner (2010) find decreased forecast error after the passage of SOX. The results of this study could be even more pronounced compared to their findings since this sample contains a larger variety of countries including a wider range of weak investor protection origins with

(28)

potentially more severe deficits in corporate governance quality. EU-15 firms used in the sample by Arping and Sautner are also more homogeneous due to a more harmonized regulatory framework and the adoption of IFRS accounting standards in the European Union. Consequently, if a firm from weak investor protection background cross-lists in the US with enforced strong investor protection since SOX, I expect a stronger improvement of analyst forecast accuracy for civil law countries.

Hypothesis 2:

Analyst forecast accuracy increases especially for firms from civil

law countries after the implementation of SOX relative to not cross-listed firms.

If SOX and related regulation are effective, the percentage of buy-recommendations should decrease. Clarke et al. (2011) already find a decrease in optimism revealed in fewer buy-recommendations for US domestic firms after GS and related regulation. Since SOX is enhancing related regulation, a pronounced effect should be the result.

Yet, another potential scenario is that buy-recommendations even increase due to an increased bonding effect since the implementation of SOX. Hence, if analyst coverage increases after SOX, this could indicate and increased interest in the group of cross-listed firms, voluntarily applying to the high standards of corporate governance in the US. Higher analyst coverage and the reported more conservative accounting could not only make those firms appear better but indicate indeed higher value leading to more buy-recommendations, which truthfully represent good companies. Recommendations would become more realistic and less rosy. Consequently, resulting analyst recommendations would become more valuable.

This reasoning incorporates a self-selection bias stemming from the fact, that only those firms being able to appear as a valuable company on paper will choose to apply to SOX via the cross-listing. If the selection process includes the recognition of some firms being bad, the number of cross-delistings can also

(29)

increase. More delistings of bad firms can in turn increase the fraction of buy-recommendations again.

All in all, the latter relations should be less dominant than the decline in the conflict of interest and thus, lead to a decrease of the percentage of buy-recommendations in total. Since SOX aims at increased disclosure concerning this conflict of interest and there is evidence for a decreased bias in recommendations for US-firms, I suspect a lower percentage of buy-recommendations for cross-listed firms after SOX as well.

Hypothesis 3:

The percentage of buy-recommendations decreases more for

cross-listed firms after the implementation of SOX than for not cross-cross-listed firms.

3 Data

A cross-listing can be pursued by an ordinary listing process at a foreign stock exchange in combination with additional capital issuance. Alternatively, shares can be listed on the overseas exchange via depository receipts and in case of the US via American depository receipts (ADR). For both procedures, regulatory requirements for the cross-listed firms are approximately the same (Reese and Weisbach, 2002). An ADR represents the stock of a foreign company issued in the US. The custodian bank makes the ADR available and holds the foreign shares denominated in the foreign currency but issues the ADR in US dollars (Bianconi et al., 2013). As Marosi and Massoud (2008) describe, the custodian bank is responsible of taking care of the transfer payments, receiving dividends in local currency and paying out dividends in US dollars to US investors. There are several levels of ADRs: Rule 144a private placements (Level IV), Level I ADRs and Pink Sheets which trade on OTC markets, whereas Level II and III ADRs trade on regulated stock exchanges (Karolyi, 2012). This research considers Level II and III ADRs since only these are subject to the strict regulatory framework in the US and only here, concepts like bonding can be assumed.

(30)

I use data from NASDAQ, NYSE and AMEX to download ADRs as well as direct cross-listings at these exchanges. To cross-check, I include data on ADRs from large depository banks. The websites of the Bank of New York and Citibank provide data on ADRs (also used in Reese and Weisbach, 2002 and Hostak et al., 2013). I do a cross-check on all samples with exchange listing data from CRSP. I delete all companies from tax havens such as Bermuda, the Bahamas, the Cayman Islands and the British Indian Ocean Territory because they usually have the primary listing in the US and are only foreign firms for tax reasons (Doidge et al., 2009).

I include firms that are listed on an US exchange during the period 2000-2010. I do not restrict my sample to firms being listed over the complete sample period because I assume that a firm listing before the passage of SOX in 2002 and not cross-delisting afterwards for a couple of years should be quite comparable to a firm that chooses to list after 2002. However, I delete those companies from my sample that delist within the sample period due to voluntary reasons not related to mergers or liquidations but to decisions made by the company to deregister from the SEC, to go private or delist from an exchange but start being traded OTC (CRSP delisting codes 520, 570 and 573). This is the case for 716 observations. One can only speculate why these firms delist and if the decision is related to the passage of SOX, but the reasons mentioned seem to be most likely related to the Act. As Arping and Sautner (2010) argue, firms that are less transparent and want to avoid stricter regulation are more likely to delist around SOX events. Not excluding them from the sample would spuriously appear as a transparency enhancing effect of SOX only due to a change in the sample selection over time.

I exclude all firms with a firm size of less than $100 million in total assets as done in Doidge et al. (2009). This step decreases the sample size at this point in the process by 1,437 companies. Smaller firms might react differently to the introduction of SOX than larger firms, mainly because small firms did not have to comply with Section 404 of the Act by the end of the years 2007 or even 2009 (Arping, and Sautner, 2010). Studying the effects on small firms should pronounce the fixed costs effect of SOX being a worthwhile research approach but

(31)

is beyond the scope of this thesis. However, the range of total assets after excluding small firms is still very large. Data on firm specific identification and accounting information can be downloaded from the CRSP/Compustat merged database. I obtain a treatment group with 651 cross-listed companies and 3,385 company-year observations. Further details about the sample are described in the descriptive statistics.

Furthermore, I need to construct a control group. I match the cross-listed firms with a set of not cross-listed firms based on country, industry and firm size, as done in Litvak (2007) as well as Arping and Sautner (2010). I collect a control sample from Compustat for non-US firms without cross-listing in the US. The data retrieved is denominated in the domestic currency. I convert the variables to USD with historical foreign exchange rates retrieved from Compustat Global Legacy. I exclude all firms with total assets below $100 million like in the treatment group. Further, I delete all firms from countries which are not represented in my treatment group of cross-listed firms to eliminate any misleading effects from a different composition of countries. I also check that the control group does not include industries not covered in the treatment group. I use the SIC industry classification available in Compustat, which divides the companies into 11 sectors and 83 industries (2 digit SIC codes). The industries not represented in the treatment group and therefore dropped from the control group are mainly from the retail trade, services and public administration segments. Finally, there are in a total 6,087 firms in the control group and 24,995 firm-year observations. Overall, 56 industries and 37 countries are represented in the whole panel.

The origin of law is added to the treatment as well as the control sample. Legal classifications are obtained from tables from La Porta et al. (1998, 1999) and can be sorted into five groups: English common law, French, German and Scandinavian civil law as well as Socialist law.

I collect data on analyst coverage, analyst forecast accuracy and recommendations from IBES. I proxy analyst coverage with the average number of analysts forecasts made per company per year. As O’Brien and Bhushan (1990) note, analyst coverage tends to increase during the year unrelated to the level of

(32)

initial analyst coverage. Therefore, it is useful to measure analyst coverage in the same month of the year during the whole sample period. There seems to be a pattern visible that analyst coverage is at its annual peak in the eleventh month of the fiscal year as the authors note, and stays constant until the date of the earnings announcement at the fiscal year’s end. When a firm has its fiscal year end e.g. in December, I collect data on analyst coverage for November.

The data needed to calculate analyst forecast accuracy include the actually obtained EPS value, the “actual”, and the median analyst forecast for the estimation period. I choose the median instead of the mean forecast, to get measures less influenced by outliers. Since I measure analyst coverage in the eleventh month of the fiscal year, I decide to measure forecast accuracy at the same point in time. When more analysts provide their estimates, a lower forecast error should be more likely (Lang and Lundholm, 1996). Consequently, calculating the forecast accuracy at different points during the year would make them less comparable over the years. Reporting currencies for median estimates and actual values are often not in the same currency. Therefore, I convert all numbers to USD based on daily exchange rates retrieved from the IBES currency section. Furthermore, I delete all actual values equal to zero since the forecast error cannot be calculated otherwise.

Analyst recommendations can be obtained from the detail section in IBES. The different investment banks and research companies issuing the recommendations differ in their classifications from sell to buy. IBES classifies the different recommendation structures into one system with 5 different levels from strong buy (1) to sell (5) to make them comparable. Slight differences might be lost when subsuming all recommendations into one classification scheme. However, I rely on the IBES classification for this research. I further download summary statistics including the percentage of buy, hold and sell recommendations as well as median recommendation and its standard deviation. I choose to calculate the recommendation distribution again for the eleventh month of the fiscal year. A very detailed summary of the data collection process including company identifiers used, all variables downloaded and the exact names of the database sections can be found in Appendix g).

(33)

3.1 Summary Statistics

The vast majority (68.9%) of the foreign firms are cross-listed at NYSE, followed by NASDAQ (25.8%) and AMEX (5.3%). As depicted in detail in table 1, the country distribution differs substantially between the treatment and the control group: Cross-listed firms are to more than a third (36.4%) represented by Canadian firms. Great Britain, Israel, the Netherlands, Ireland and Mexico also contribute to large fractions (4-9%) of the sample. Not cross-listed firms are highly represented by Japan (26.51%) as well as Australia, China, France, Great Britain, Taiwan and Germany (4-8%). The differences in the country distributions are also reflected in the distribution across the legal origins between common and civil law but especially within the group of civil law countries, which are represented in table 2. These differences have to be kept in mind during the whole analysis. The difference in country distributions will be taken into account by country fixed effects whereas I include different dummies for the legal origins.

Cross-listed firms come to nearly two thirds from common law countries. The largest proportion within civil law countries stems from French civil law (22.75%), followed by German, Socialist and Scandinavian origins. By contrast, the control group is strongly represented by civil law countries. Nearly 70% come from the less investor protective countries. The largest fraction within this group has German legal origin (37.37%), followed by French, Socialist and Scandinavian origins. Common law countries make up a fraction of 30.44% only. Concluding, civil law in the treatment group consists predominantly of French origin and civil law in the control group has mainly German origin. These differences can make a difference since German civil law protects investors significantly better than French civil law (La Porta et al., 1999).

(34)

Table 1: Country of Incorporation Distribution

This table shows summary statistics for cross-listed and not cross-listed firms with respect to the firms’ country of incorporation. The sample period is 2000-2010. Cross-listed firms are all non-US firms listed at an US-exchange during this period. All firms need to have more than $100million total assets over the whole sample period to be included. The data is retrieved from Compustat Global, Compustat NorthAmerica or the CRSP/Compustat merged database respectively.

Country

Firm-years Percentage Firm-years Percentage Firm-years Percentage

Argentina 59 1.74% 79 0.32% 138 0.49% Australia 24 0.71% 1,279 5.12% 1,303 4.59% Belgium 14 0.41% 352 1.41% 366 1.29% Brazil 65 1.92% 304 1.22% 369 1.30% Canada 1,232 36.40% 759 3.04% 1,991 7.02% Switzerland 77 2.27% 651 2.60% 728 2.57% Chile 110 3.25% 72 0.29% 182 0.64% China 80 2.36% 1,660 6.64% 1,740 6.13% Columbia 4 0.12% 6 0.02% 10 0.04% Germany 48 1.42% 1,159 4.64% 1,207 4.25% Denmark 16 0.47% 325 1.30% 341 1.20% Spain 38 1.12% 496 1.98% 534 1.88% Finland 13 0.38% 449 1.80% 462 1.63% France 81 2.39% 1,441 5.77% 1,522 5.36% Great Britain 308 9.10% 2,106 8.43% 2,414 8.51% Greece 13 0.38% 373 1.49% 386 1.36% Hong Kong 39 1.15% 239 0.96% 278 0.98% Indonesia 11 0.32% 268 1.07% 279 0.98% India 49 1.45% 916 3.66% 965 3.40% Ireland 160 4.73% 101 0.40% 261 0.92% Israel 246 7.27% 21 0.08% 267 0.94% Italy 39 1.15% 736 2.94% 775 2.73% Japan 108 3.19% 6,626 26.51% 6,734 23.73% Korea, Republic of 56 1.65% 735 2.94% 791 2.79% Mexico 142 4.19% 157 0.63% 299 1.05% Netherlands 142 4.19% 336 1.34% 478 1.68% New Zealand 3 0.09% 185 0.74% 188 0.66% Peru 13 0.38% 32 0.13% 45 0.16% Philippines 12 0.35% 165 0.66% 177 0.62% Portugal 9 0.27% 151 0.60% 160 0.56% Russia 26 0.77% 43 0.17% 69 0.24% Singapore 36 1.06% 440 1.76% 476 1.68% Sweden 12 0.35% 504 2.02% 516 1.82% Turkey 9 0.27% 268 1.07% 277 0.98% Taiwan 40 1.18% 1,105 4.42% 1,145 4.03% Venezuela 9 0.27% 3 0.01% 12 0.04% South Africa 42 1.24% 453 1.81% 495 1.74% Total 3,385 100% 24,995 100% 28,380 100%

Impact of SOX on US Cross-listings : does analyst behaviour change?