Capital-Market Effects of Securities Regulation: The Role of Implementation and Enforcement *

Transcription

1 Capital-Market Effects of Securities Regulation: The Role of Implementation and Enforcement * Hans B. Christensen Booth School of Business, University of Chicago Luzi Hail The Wharton School, University of Pennsylvania Christian Leuz Booth School of Business, University of Chicago & NBER September 2010 (Preliminary Comments welcome) Abstract This paper examines capital market effects of changes in securities regulation. We focus on two key capital market directives in the European Union (EU) that tighten market abuse and transparency regulation and enforcement. All EU member states are required to adopt these two directives but their exact implementation into national law and the design of enforcement regimes differ across member states. We use this setting to highlight that implementation and enforcement of regulation play an important role for regulatory outcomes. To identify the effects on EU capital markets, our research design exploits the differential timing of when countries implement the two directives. We find that, on average, market liquidity increases and firms cost of capital decreases as EU member states improve and tighten market abuse and transparency regulation. The capital-market effects are stronger in countries with a stricter implementation of the directives, in countries with traditionally stricter securities regulation and in countries with a better track record of implementing regulation and government policies. These cross-sectional findings underscore that the success of regulation depends critically on how well the rules are implemented and enforced. JEL classification: Key Words: F30, G15, G18, G30, K22, M41 Capital market regulation, Enforcement, Disclosure, Law and finance, European Union, Liquidity, Cost of capital * We appreciate the helpful comments of Allan Bester, Zachary Kaplan and of workshop participants at the University of Chicago. We thank Philip Jacobs, Zachary Kaplan, Jeff Lam, Elton Lor and Michelle Waymire for their excellent research assistance.

2 1. Introduction The costs and benefits of securities regulation have been extensively debated in the literature. 1 While there is no clear consensus, what emerges from the debate is that whether or not securities regulation is beneficial to the economy depends crucially on how regulation is designed, implemented and enforced (see discussion in Section 2). The design of the rules generally receives much attention in the literature and in regulatory debates. However, how the rules are actually implemented, how the enforcement regime is designed (e.g., supervisory powers) and whether the rules are actually enforced is likely to play an important role as well. 2 To highlight the importance of implementation and enforcement for regulatory outcomes, this paper examines the capital-market effects of recent changes in EU securities regulation. We use the EU setting because it has several desirable features. First,EU directives apply to all member states, so in principle the regulatory act is the same and held constant across countries. But the exact implementation, i.e., the transposition of a directive into national law, and its enforcement, e.g., the actual supervision and the penalties for violations, are left to the EU member states. Second, the EU recently passed two directives that pertain to key elements of securities regulation: the Market Abuse Directive (MAD), which addresses insider trading and market manipulation, and the Transparency Directive (TPD), which addresses reporting requirements and their enforcement. In both cases, there were prior EU directives and national laws banning insider trading and stipulating extensive reporting requirements. Thus, the two directives can be largely viewed as harmonizing and tightening securities regulation, in particular by improving supervisory and enforcement regimes in the EU. To illustrate, the TPD requires 1 2 See e.g., Stigler (1964), Seligman (1983), Coffee (1984), Easterbrook and Fischel (1984), Mahoney (1995). For recent surveys of this debate, see Leuz and Wysocki (2008) and Zingales (2009). For example, Carvajal and Elliott (2007) point out that major shortcomings in the ability of securities regulators to effectively enforce compliance with existing rules is a recurring theme in IOSCO assessments. 1

3 member states to have a supervisory authority that, among other things, reviews firms financial statements on a regular basis. In doing so, the TPD makes few changes to reporting requirements for EU companies but it increases the extent to which financial reporting is monitored by a national supervisory agency (see Section 2 for more details). Thus, aside from shedding light on the benefits of transparency and insider trading regulation, the setting provides an opportunity to study implementation and enforcement issues. For instance, we can exploit cross-sectional variation with respect to how the two directives are implemented by the EU member states as well as examine how regulatory outcomes differ depending on countries prior regulation (e.g., the strength of existing securities regulation). We analyze capital-market effects around the two directives using two economic constructs, i.e., market liquidity and the cost of capital. Both constructs are closely linked to the goals of the EU capital market directives (e.g., Lamfalussy 2000; CRA, 2009) and, more generally, they are commonly used in justifications for securities regulation. Our design exploits that EU member states implemented the two directives at various points in time, which allows us to isolate the effects of regulatory changes from general time trends and other events in the EU capital markets. Specifically, we estimate quarterly panel regressions from 2001 to 2009, controlling for quarter-year, country and industry fixed effects. We measure market liquidity using the bid-ask spread and measure the cost of capital using the implied discount factor from a valuation model. 3 Both proxies have been used extensively for the underlying two constructs (see, e.g., survey by Leuz and Wysocki, 2008). We find that market liquidity increases when the TPD and the MAD come into force in EU member states and that the cost of capital decreases around the time the MAD comes into force. 3 We also conduct analyses using the percentage of zero-return days and dividend yields as alternative proxies. 2

4 The estimated effects are economically significant. The estimated spread reduction around the TPD (MAD) amounts to 21% (18%) of the pre-directive bid-ask spread. The estimated reduction in the implied cost of capital around the MAD is roughly 60 basis points (using the most stringent specification). Thus, our results suggest that improving key elements of securities regulation, in particular the enforcement regime, leads to substantial capital-market benefits. Furthermore, we document that the capital-market effects of the two directives are stronger in countries with traditionally stricter securities regulation and in countries with a history of higher regulatory quality. One explanation for these findings is that countries with a stronger securities regulation and with better track record of implementing regulation and government policies more generally are more willing and able to implement the new EU directives. The same forces that limited the strength of securities regulation and regulatory quality in the past (e.g., political pressures) are likely to be at work when implementing the two directives. In addition, there are likely important complementarities between existing and new regulation, especially when the latter as in our setting primarily improves the enforcement regime and hence the compliance with existing requirements. Consistent with these explanations, we also provide evidence using directive-specific measures that countries with a stricter implementation of the two directives experience larger capital-market effects. Together, the cross-sectional results underscore the role of implementation for regulatory outcomes. 4 Finally, we show that the introduction of International Financial Reporting Standards (IFRS), which the EU and many other countries around the world introduced during our sample period, has little effect on market liquidity and firms cost of capital. Instead, our evidence suggests that regulatory and 4 Our cross-sectional findings are not consistent with the notion that countries with weaker securities regulation are catching up with the others, which is an alternative hypothesis for cross-sectional effects. 3

5 enforcement changes in the EU, such as the MAD and the TPD, which are concurrent with the introduction of IFRS, are responsible for the observed improvements in EU capital markets. We conduct extensive sensitivity analyses and show that our results are robust to the introduction of firm-fixed effects as well as EU-specific quarter-year fixed effects. The latter demonstrates that our results can also be obtained from within-eu estimation. To gauge our identification strategy, we mechanically shift the implementation dates in time in our analyses. We find that the estimated treatment effects peak on or shortly after the true implementation dates. This finding mitigates the concern that our results simply reflect a within-eu trend of improving liquidity and cost of capital over the treatment period. We also conduct placebo analyses by randomly assigning implementation dates from 2001 to 2004, i.e., prior to the national implementation of the two directives. The placebo regressions rarely produce significant coefficients and show that our statistical tests have approximately the correct size and, in some cases, are even conservative. Our paper makes several contributions to the literature. First, we examine a novel setting to provide evidence that tightening securities regulation can have significant economic benefits in terms of market liquidity and cost of capital. Second, our findings support the notion that the success of regulation depends critically on how well regulation is implemented and enforced. We show that regulatory outcomes differ depending on the strength of prior regulation and countries ability to implement and enforce new securities regulation. These findings are consistent with the enforcement theory of regulation formulated in Djankov et al. (2003a) and its application to securities regulation in Shleifer (2005). Third, we add to the budding literature on securities law enforcement. As Bhattacharya (2006) points out, there is relatively little evidence on the role of enforcement. In an important 4

6 paper, Bhattacharya and Daouk (2002) provide evidence that the first enforcement of insider trading regulation lowers firms cost of capital. 5 Subsequent papers use the same dataset and provide evidence on additional capital-market effects associated with insider trading regulation (e.g., Bushman et al., 2005; Ackerman et al., 2008). Our analysis is not limited to insider trading regulation. Moreover, most of the evidence on securities law enforcement is based on ex-post measures, i.e., complaints, lawsuits, enforcement actions. The EU setting allows us to examine (ex-ante) capital-market effects associated with regulatory changes in the design of enforcement regimes and supervisory structures. Fourth, our study exploits that the same regulatory act applies to many countries at different times. It focuses our analysis on implementation and enforcement issues. It also helps the identification of the regulatory effects. Much of the early literature on securities regulation (e.g., Stigler, 1964) suffers from the lack of a proper control group. 6 Similar concerns apply to many studies examining the effects of the Sarbanes-Oxley Act of 2002 (see discussion in Leuz, 2007). Other recent studies on the effects of securities regulation are either cross-sectional (e.g., La Porta et al., 2006; Hail and Leuz, 2006) or analyze a single regulatory act (Bushee and Leuz, 2005; Greenstone et al., 2006), and they are not focused on implementation and enforcement issues as our paper. Finally, this study shows that the capital-market effects in the EU around IFRS adoption are largely driven by concurrent changes in securities regulation, rather than the change in the accounting standards per se. Daske et al. (2008) document improvements in market liquidity and 5 6 There is a large literature on the costs and benefits of insider trading. See, e.g., Bainbridge (2000) for an overview. An exception is Mahoney and Mei (2006). Their study highlights the lack of a control group as a major shortcoming of many studies assessing the 1933 Securities Act and the 1934 Securities Exchange Act. 5

7 the cost of capital around IFRS adoption in the EU but caution that these effects could also stem for other regulatory changes in the EU. We confirm this conjecture. 2. Hypothesis Development and Institutional Setting In raising external financing, firms need to reassure outside investors. If outside investors have doubts that firms will return their money, they are unlikely to provide capital to firms in the first place (leading to low market liquidity) or, if they provide capital to firms, they are likely to demand a higher return on their capital (leading to a higher cost of capital for firms investments). As providing such reassurance is difficult and costly for firms, there is a long standing debate as to whether securities regulation is beneficial, for instance, by improving market liquidity and reducing firms cost of capital. The arguments in favor of regulation refer among other things to the existence of externalities, economy-wide cost savings, commitment problems and insufficient private penalties (see, e.g., Leuz and Wysocki, 2008, and Zingales, 2009, for recent overviews). However, these arguments often set aside or do not consider problems of how to implement and enforce securities regulation. 7 In contrast, Stigler (1971), Posner (1974), Peltzman (1976) and Becker (1983) highlight the difficulties of implementing and enforcing regulation in a way that is actually socially beneficial. They point out that regulators face serious information problems, are often incompetent or even corrupt, and can be captured in the regulatory process. Arguments against regulation (and in favor of private contracts) in turn rely heavily on courts and private litigation. However, courts and litigation can be quite imperfect as well (e.g., Johnson et al., 2002; Djankov et al., 2003b). 7 Shleifer (2005) argues that the same can be said for the (public interest) theory of regulation in general. 6

8 Against the backdrop of this debate, Djankov et al. (2003a) propose an enforcement theory of regulation. Their premise is that all strategies for implementing socially desirable policies (e.g., creating deep and functioning capital markets) are likely imperfect and optimal institutional design involves a tradeoff between imperfect alternatives. 8 Shleifer (2005) applies this theory to securities regulation. He argues that the inequality of weapons between corporate insiders and promoters on one side and (often unsophisticated) outside investors on the other side makes it unlikely that (pure) private litigation is an efficient solution in securities markets. In this situation, some regulation, e.g., rules that prescribe what firms have to disclose to investors, can be beneficial, because ex ante rules limit the discretion of courts and mitigate the inequality of weapons problem. 9 Thus, one hypothesis is that securities markets are an instance in which regulation is beneficial to the economy. Consistent with this conjecture, almost all economies have extensive securities regulation. Obviously, this observation alone is not sufficient to settle the case. However, empirical studies do not provide clear evidenceeither. 10 The mixed evidence in the literature may be related to a second hypothesis that follows from Djankov et al. (2003a). Securities regulation is likely to be more effective in richer countries with better institutions, more efficient bureaucracies, and a greater ability to implement and enforce such regulation. In countries with weak institutions and inefficient bureaucracies, the risk of abuse is much larger and hence securities regulation could actually be harmful (Shleifer, 2005; Bhattarcharya and Daouk, 2009). In addition, a country s track record with respect to implementing regulation in the past is likely revealing about the ability and political will of its government to put in place and enforce regulation that induces (curbs) policies that are deemed They characterize the problem as a tradeoff between two basic social costs: disorder and dictatorship. Based on prior research (e.g., Hay and Shleifer, 1998; La Porta et al., 2006), he argues that it might make sense to combine public rules with private enforcement through litigation. See also Jackson and Roe (2009). See, e.g., the recent studies by Glaeser et al. (2001), Bushee and Leuz (2005), Greenstone et al. (2006), Mahoney and Mei (2006). See also survey by Leuz and Wysocki (2008). 7

9 socially desirable (undesirable). In sum, the benefits of securities regulation (or any other regulation) depend crucially on its implementation and enforcement. These ideas form the basic conceptual underpinnings of our study. While our study cannot provide evidence that securities regulation or even a particular regime of securities regulation is indeed socially desirable, we can shed light on the aforementioned hypotheses and the underlying forces by analyzing capital-market effects around changes in securities regulation. Towards this end, our analysis exploits regulatory changes in EU capital markets for which implementation and enforcement issues are pertinent. The EU setting also has the desirable feature that a directive applies to all member states, so in principle the regulatory act is the same across countries. However, the exact implementation, i.e., the transposition of the directive into national law, and the enforcement, e.g., the supervision of and penalties for violations, are left to EU member states. Thus, cross-sectional variation in regulatory outcomes likely reflects differences in the implementation and enforcement of the directives. An alternative explanation for cross-sectional variation in regulatory outcomes is that countries with weaker securities regulation catch up with stronger countries as a result of the EU directives. As we discuss in more detail in Section 4.2, these two explanations have opposite predictions for the direction of the capital-market effects. We examine the Market Abuse Directive (MAD), which covers insider trading and market manipulation, and the Transparency Directive (TPD), which addresses reporting requirements and information regulation. These two directives are at the core of the EU s Financial Services Action Plan, which was established in 1999 with the goal to improve and integrate EU financial markets, and they address what are generally considered to be key elements of securities regulation. The MAD was passed by the EU legislature in January 2003 followed by several 8

10 implementing directives in December Its purpose is to ensure market integrity and equal treatment of market participants in EU securities markets by defining and prohibiting insider trading and market manipulation. Among other things, it establishes transparency standards requiring people who recommend investments to disclose their relevant interests. It also requires each member state to have a supervisory authority that is responsible for monitoring and dealing with insider trading and market manipulation and to give this authority the necessary supervisory and investigative powers. The MAD replaces Directive 89/592/EEC of 13 November 1989 which previously required EU member states to ban insider trading. Thus, while the MAD expands market abuse regulation in some areas, it is more appropriately viewed as harmonizing and improving the implementation and enforcement of market abuse regulation in the EU (e.g., Lamfalussy, 2000; CRA, 2009). The TPD was passed by the EU legislature in December 2004 and its implementing directive was enacted in March The directive requires issuers of traded securities to ensure appropriate transparency for investors by disclosing and disseminating periodic and ongoing regulated information. Regulated information comprises periodic financial reports, information on major holdings of voting rights and information disclosed pursuant to the MAD. However, prior EU directives, member state laws and exchange requirements already required annual and interim financial reports as well as other ongoing information. As such, the TPD does not necessarily expand existing disclosure requirements. In most areas, the TPD is more appropriately viewed as harmonizing and clarifying existing information regulation as well as improving its enforcement. Towards this end, the TPD stipulates that a supervisory authority in 11 Under the new Lamfalussy process, which was adopted in 2001 to make EU regulation on securities markets more flexible, the European Council and the EU parliament adopt a piece of legislation (directive), which at the first level establishes the key principles and guidelines on its implementation. The law then progresses to the second level, at which the European Securities Committee (ESC) and the Committee of European Securities Regulators (CESR) advise on technical details, leading to an implementing directive. 9

11 each member state assumes responsibility for supervising compliance with the provisions of the directive and that this authority examines the regulated and disclosed information. The TPD also stipulates that the authority is given appropriate enforcement tools. Thus, for both directives, there is extensive prior EU and national regulation. The two directives essentially address prior divergence in regulation, provide more detailed implementation guidance relative to prior directives, and stipulate appropriate supervisory and enforcement regimes. As such, the setting allows us to study the effects of tightening the implementation and enforcement of securities regulation. The stated goal of these directives in terms of market outcomes is to increase market confidence and, more specifically, to lower trading costs and firms cost of capital (e.g., CRA, 2009). Thus, our empirical analyses focus on capital market effects with respect to market liquidity and cost of capital. In addition to the MAD and the TPD, there are two other so-called Lamfalussy directives, namely, the Prospectus Directive (PD) and the Markets in Financial Instruments Directive (MiFID). Together, the four directives represent the key securities market directives under the FSAP. However, the PD is less relevant for our purposes as it pertains mainly to the market for securities offerings rather than secondary markets, for which we analyze market liquidity. The MiFID is more recent and was implemented in many member states only very recently and. Hence, it is too early to study its impact. Thus, for the most part, we focus on the MAD and the TPD. 12 There is one other piece of EU regulation that is potentially relevant in our context. IAS regulation (EC 1606/2002) requires the adoption of International Financial Reporting Standards (IFRS) in the EU by As the move to IFRS is often viewed as a major change in the rules 12 In our sensitivity analyses, we consider all four Lamfalussy Directives as one package and more generally discuss the issue that observed capital market changes may reflect broader regulatory changes in the EU. 10

12 that govern information regulation, we account for this change in our analysis. Interestingly, many other countries around the world also adopted IFRS around the same time. However, these countries did not necessarily make supporting changes in the information and enforcement regime as in the EU. Thus, the adoption of IFRS around the world (and the comparison to the EU) provides another opportunity to study the basic hypothesis of this paper that regulatory outcomes depend crucially on how regulation is implemented and enforced. 3. Research Design and Data 3.1. Empirical Model We test our hypotheses using a panel dataset of quarterly firm observations. The chosen data structure reflects a tradeoff between measuring our capital-market variables (e.g., liquidity) over some interval and capturing changes in these variables in a timely fashion, i.e., when the MAD and TPD come into force. A primary concern about our setting is that market-wide changes (e.g., due to macroeconomic shocks) and general time trends in the capital-market variables as well as other regulatory changes confound our analysis. Our identification strategy exploits that EU member states implement the same directives at different points in time. Thus, we can introduce an extensive fixed-effects structure to isolate the effect of the two EU directives on the capital-market variables. Specifically, we use the following model: CapEff = β 0 + β 1 MAD/TPD + β j Controls j + β i Fixed Effects i + ε (1) where CapEff stands for two different capital-market effects (i.e., bid-ask spreads and implied costs of capital), MAD or TPD are binary variables coded as 1 beginning in the quarter in which the corresponding directive comes into force in a given EU member state and 0 otherwise, Controls j denotes a set of firm-level and country-level control variables, and Fixed 11

13 Effects i represents country, industry and a set of quarter-year fixed effects. As the key variables of interest vary only at the country level, we assess statistical significance using standard errors that are clustered by country. We include a benchmark sample comprising firm-quarter observations from non-eu countries, which are unaffected by the introduction of MAD and TPD. The inclusion of non-eu benchmark firms helps us to identify worldwide changes and general trends in market liquidity and cost of capital (and to estimate the coefficients for the control variables). Our extensive fixed-effects structure captures any unobserved (time-invariant) heterogeneity across countries and industries. It also includes quarter-year fixed effects, which estimate a flexible quarterly time trend and control for common shocks over time. We also consider an alternative model that includes separate quarter-year fixed effects for EU and non-eu countries. This specification essentially amounts to within-eu estimation and uses only variation that is not common across EU member states in a given quarter to identify the capital-market effects of the directives. However, the introduction of EU-specific, year-quarter fixed effects is very demanding and may lead us to dismiss a true treatment effect, particularly if there is clustering of the implementation dates across countries, if the dates are measured with noise or if the directives have a more gradual rather than a sharp effect. 13 Our set of control variables also accounts for mandatory adoption of IFRS in the EU and many other countries around the world. We include this control because the move to IFRS is often viewed as a major improvement in the transparency and comparability of financial reporting. 14 As IFRS were introduced in the EU during our sample period, it could potentially As a compromise, we also analyze a model that includes a quadratic (and hence less flexible) time trend in the EU. The results for this specification (not tabulated) are very similar to those reported in the text. See, e.g., Hail et al. (2010) for a summary and evaluation of the IFRS debate. 12

14 confound the effects of the MAD and TPD. We define a binary indicator variable, IFRS, taking on the value of 1 beginning in the quarter in which IFRS becomes mandatory for a firm and 0 otherwise. Thus, the construction of the IFRS variable is firm-specific and depends on a firm s fiscal year end and the date when IFRS becomes mandatory, which for most EU countries was fiscal years ending on or after December 31, Our sample period starts in the first quarter of 2001, i.e., before the EU adopted the MAD and the TPD, and hence well in advance of the first country-specific entry-into-force dates for the MAD (October 2004) and TPD (January 2007). The sample period ends in the second quarter of 2009, which is the most recent quarter for which we have the necessary data. We include all the firm-quarter observations from EU and non-eu countries for which we have the necessary data to compute the capital-market and control variables to estimate our basic regression in Eq. (1). The maximum sample comprises 560,868 firm-quarter observations from 25 EU countries and 30 non-eu countries. 16 Table 1 provides information on the sample composition by EU country Construction of the Variables In this section, we describe the coding of our two primary test variables, i.e., the Market Abuse Directive (MAD) and the Transparency Directive (TPD) indicator variables, and the construction of our dependent and independent variables in the regression analysis For example, when a Spanish firm has a fiscal-year end of April 30, the IFRS indicator is set to 1 beginning in the second quarter of Our sample comprises only 25 EU member states as we do not have data for Malta and Romania. To minimize the issue that firms are subject to a mix of regulations, the sample excludes US GAAP adopters and firms that are cross-listed on exchanges or in the over-the-counter markets in the U.S. In addition, we require the dependent variable be available for at least eight quarters for a firm to be included and, for benchmark countries, that there are at least twenty firms for a given country. 13

15 After an EU directive is enacted, each member state must take the necessary measures to comply with the directive before a deadline specified in the directive. In the case of MAD and TPD, compliance required amending national law(s) in all member states. We select the dates where the national law(s) that implements the respective directives come into force in each member state (the Entry-Into-Force date). Our primary sources for the Entry-Into-Force dates are the European Commission for the MAD and the international law firm Linklaters LLP for the TPD. We validate the dates from these primary sources by comparing them to the dates on which the EU member states inform the European Commission that they comply with the MAD and the TPD. For countries for which there are discrepancies, we contact the national securities regulator to ensure that we use the most accurate dates and make adjustments where necessary. Table 1 provides the final dates used in our analyses. The MAD Dates vary from April 2004 (Lithuania) to January 2007 (Bulgaria) and the TPD Dates vary from January 2007 (Bulgaria, Germany, United Kingdom) to August 2009 (Italy). Thus, there is material variation in the dates on which the two directives became effective across member states. The MAD (TPD) indicator, used in the analysis, takes on the value '1' beginning in the quarter of the country-specific MAD Date (TPD Date) and is '0' before. In studying the economic consequences of MAD and TPD implementation, we focus on market liquidity and the cost of capital. Reducing insider trading as well as enhancing disclosures should reduce information asymmetries between investors and, hence, increase market liquidity (e.g., Diamond and Verrecchia, 1991; Verrecchia, 2001). More transparent reporting should also lower non-diversifiable estimation risk, which in turn reduces the cost of capital (e.g., Coles et al., 1995; Lambert et al., 2007). Thus, proxies for market liquidity, information asymmetry, and the cost of capital should reflect, among other things, major changes 14

16 in securities regulation, particularly, with respect to insider trading, transparency and corporate reporting, assuming the rules are properly implemented and enforced. We use two primary proxies for market liquidity and the cost of capital. 17 Our first proxy is the Bid-Ask Spread, which is commonly used in empirical studies to measure market liquidity and information asymmetry (e.g., Welker, 1995; Healy et al., 1999; Leuz and Verrecchia, 2000; Lang et al., 2009). We obtain the closing bid and ask prices for each day and compute the daily quoted spread as the difference between the two prices divided by the mid-point. To obtain firmquarter observations, we compute the median daily spread over the quarter for a given firm. Our second proxy is the implied cost of capital. The basic idea of the various models used in the literature is to substitute price and earnings forecasts into an accounting-based valuation equation and to back out the cost of capital as the internal rate of return that equates current stock price with the expected sequence of future abnormal earnings (e.g., Claus and Thomas, 2001; Gebhardt et al., 2001; Hail and Leuz, 2006). Conceptually, the models are consistent with discounted dividend valuation. Since the estimation of the implied cost of capital neither requires a long time series of data nor market integration, it is particularly suited for crosscountry settings for which these conditions are often not given (e.g., Hail and Leuz, 2006; Pástor et al., 2008; Lee et al., 2009). To avoid problems with analyst forecast bias and staleness of analyst forecasts (e.g., Easton and Sommers, 2007), and to extend our sample to firms without analyst following and hence countries with less extensive analyst coverage, we do not use analyst estimates of future earnings. Instead, we follow Hou et al. (2009) and use the predicted values from a pooled cross-sectional regression of future (realized) earnings on a set of contemporaneous firm characteristics. We plug these forecasts into the 12-year version of the 17 In the sensitivity analyses, we analyze two additional proxies for liquidity and cost of capital, i.e., the percentage of zero-return days and the dividend yield. 15

17 Gebhardt et al. (2001) valuation model and solve the model for the internal rate of return, r GLS, that equates a firm s intrinsic value with its market value of outstanding equity at the end of each calendar quarter. 18 This cost of capital estimate is assigned to the quarter of the pricing date. We delete extreme implied cost of capital estimates if they exceed 50 percent and if they would lead to negative risk premiums, i.e., they are below the local risk-free rate. In the liquidity regressions, we control for firm size using the market value of equity, share turnover and return variability (Chordia, Roll, and Subrahmanyam, 2000; Leuz and Verrecchia, 2000). We follow the prior literature and estimate the liquidity models in a log-linear form, i.e., we use the natural logarithm of the continuous variables. We lag the above control variables by one quarter. For the cost-of-capital specifications, we follow Hail and Leuz (2009) and control for the risk-free rate, firm size measured by total assets, financial leverage, and return variability. We measure total assets and leverage as of the last fiscal year-end, and return variability over the last calendar year before the pricing date used in the cost of capital imputation. To estimate our models, we obtain financial data from Worldscope, returns, bid-ask spreads, price, market value, risk-free interest rates and trading volume data from Datastream. Except for variables with natural lower and upper bounds, we truncate all variables at the first and 99 th percentile. Table 2 provides descriptive statistics on the dependent and independent variables as well as further details on the variable measurement. 18 More specifically, we require each firm-quarter observation to have positive one-, two-, and three-year-ahead earnings forecasts. These forecasts are the predicted values of a pooled cross-sectional regression of future realized earnings on the market value of the firm, total assets, dividend payments, current earnings, operating accruals, and a dividend payment as well as a loss indicator variable (see Hou et al., 2009, for details). To allow for differences in accounting practices across countries, we include country-fixed effects in the model. We estimate this regression for each forecast horizon (i.e., t+1, t+2, and t+3) and year using up to ten years of previous data. The predicted values of annual earnings are strictly out-of-sample. That is, we multiply the coefficient estimates of the pooled cross-sectional regressions with the yearly realizations of the independent variables that occur after the estimation period, but before the pricing date (i.e., the end of the calendar quarter). For details on the additional input parameters and the estimation procedure of the Gebhardt et al. (2001) approach, see also the appendix of Hail and Leuz (2006). 16

18 4. Results 4.1. Capital-Market Effects of Tighter Securities Regulation in the EU We begin our analysis by examining the effects on firms stock market liquidity and cost of capital following the implementation of MAD and TPD in the EU member states. As described in Section 2, MAD and TPD are an attempt to harmonize and tighten EU securities regulation, particularly, with respect to the implementation and enforcement of key existing regulations (e.g., firms reporting requirements). We use cross-sectional, time-series panel regressions, which benchmark EU firms after the entry into force of MAD and TPD against their own history before the introduction of the two directives and against a global sample of non-eu firms that are not subject to the new directives. Table 3 presents results from OLS regressions for the average capital-market effects of MAD and TPD on liquidity and the cost of capital. For each dependent variable and directive, we report coefficient estimates from estimating Eq. (1) with two different fixed-effects structures: (i) industry, country, year-quarter fixed effects, and (ii) industry, country, year-quarter fixed effects but allowing the year-quarter fixed effects to vary between EU and non-eu countries. These models essentially represent opposite ends of a spectrum. The first model estimates a very flexible trend for liquidity and the cost of capital but assumes that countries face similar economic shocks over time. The second model allows the trend to differ between the EU and the rest of the world. However, as discussed in Section 3, introducing EU-specific yearquarter fixed effects is very demanding and risks throwing out a treatment effect even when it exists. In providing results from both models, we gauge the sensitivity of our results to the fixed-effects structure. 17

19 Panel A reports the coefficient estimates and t-statistics using bid-ask spreads as the dependent variable. As is common for these models and given our extensive fixed effects structure, the explanatory power of the model is high with an R 2 of 84 percent. The firm-specific control variables are highly significant and exhibit the expected signs. That is, large firms and firms with a high trading volume have lower bid-ask spreads while firms with more volatile returns have higher spreads. In terms of our test variables, the coefficient on TPD is negative and statistically significant regardless of the fixed effects structure. The estimated effects are also economically significant. For instance, a coefficient of suggests that, on average, bid-ask spreads decrease by 21 percent, which is equal to a 54-basis point reduction relative to a pre-directive median of 256 basis points. 19 The spread effects of MAD are also negative. The effect is statistically significant in the first model and close to conventional levels for the second model (two-sided p-value of 10.6%). That is, for the MAD, the introduction of EU-specific quarter-year effects attenuates the coefficient substantially but the estimated effect is still economically significant. In contrast, the introduction of mandatory IFRS reporting does not seem to affect the bid-ask spread and including this variable does not alter the estimated capitalmarket effects for the MAD or the TPD. 20 Panel B of Table 3 reports results for the cost of capital as the dependent variable. Consistent with the expectation that cost of capital estimate are noisier, the explanatory power of the model is substantially smaller with an R 2 of 39 percent. But the control variables behave as expected and, except for return variability, are highly significant. Cost of equity capital is higher for firms in countries with higher risk-free interest rates, for firms with higher financial leverage, We compute the spread reduction and its level after the implementation of the TPD as [e ln(0.0256) ] = (or 202 basis points). The insignificant IFRS coefficient is consistent with the idea that without supporting changes in the institutional environment (e.g., stricter enforcement) a change in accounting standards alone, which is what our regressions isolate, is unlikely to improve financial reporting (e.g., Ball, 2006; Hail, Leuz, and Wysocki, 2010). 18

20 and for smaller (and presumably less established) firms. For TPD we do not observe an effect on the cost of capital. The coefficients are positive but close to zero and statistically insignificant. One potential explanation for the results is that firms cost of capital anticipates the effects of TPD prior to its entry-into-force date. That is, if investors lower the discount rate that they apply to future cash flows in expectation of the TPD, this effect is already present in our pre-tpd cost of capital estimates, which in turn attenuates the estimated TPD coefficient. In contrast, the coefficient on MAD is significantly negative in both models suggesting that tightening insider trading regulation reduces firms costs of capital. The magnitude of the estimate differs depending on the fixed effects structure. For the more conservative model, the estimated reduction in cost of capital amounts to 61 basis points. A key concern about our identification strategy is that all EU countries implement the directives after a given date and that the implementation dates spread over just a few years. Thus, the indicators for the directives could pick up within-eu trends in liquidity and the cost of capital that are unrelated to the directives. To assess this concern and the importance of the implementation dates, we mechanically shift the actual implementation dates back and forth in time. If the true implementation dates matter, the treatment effect should be attenuated as we shift away from the true dates. We find that the (absolute) magnitude of the estimated treatment effect peak on or shortly after the true implementation dates. This finding mitigates the concern that our results simply reflect within-eu trends with respect to liquidity and cost of capital that are unrelated to the directives. In addition, we conduct a series of placebo tests for which we randomly assign implementation dates to the EU member states in the period before the first member state implements one of the directives, i.e., from 2001 to the second quarter of For

21 replications, these placebo regressions estimate coefficients for the random treatments that are on average much smaller than the coefficients for the actual treatments. In fact, the coefficients for the random treatments are always smaller than the actual estimate of the treatment effect in all 250 regressions (for both spreads and the cost of capital). These tests highlight that our results are unlikely to be the result of chance. Finally, we can use the placebo tests to gauge the size of the statistical tests and the choice to draw inferences based on standard errors that are clustered at the country level. The placebo tests show that our tests have approximately the correct size and, in some cases, are even conservative. For example, using randomly assigned dates and 250 replications, we obtain only 11 (4.4%) spread regressions with coefficients for the random treatments that are statistically significant at 5% level or better Role of Prior Regulation and Differing Implementation of MAD and TPD The previous analyses estimate the average capital-market effects from the two securities regulation directives and do not provide evidence on the cross-sectional differences across member states. However, as discussed in Section 2, it is unlikely that the effects are the same that the two directives are uniformly and consistently implemented in all member states. Thus, in this section, we examine whether and how the capital-market effects of the MAD and the TPD vary with prior regulatory quality and prior securities regulation. As discussed in Section 2, it is likely that countries with a prior track record of higher regulatory quality and stronger public enforcement of securities regulation implement the two new directives in a better and stricter way. Furthermore, it is plausible that the two new directives complement existing 20

22 securities regulation by improving the enforcement regime. In addition, we examine directly how the capital-market effects differ depending on how countries implement the new directives. We begin with an analysis of cross-sectional differences in the capital-market effects of the MAD and the TPD depending on prior regulation. We separately estimate the model in Eq. (1) for two distinct groups of countries, and then compare the coefficients on the MAD or TPD variables across sub-samples using a Chow test with standard errors clustered by country. We use two institutional variables to partition the data: (i) Regulatory Quality is an index taken from Kaufmann et al. (2009) that measures the government s ability to formulate and implement sound policies and regulations that permit and promote private sector development. Higher index values indicate better regulatory quality. (ii) Public Enforcement is an index capturing the market supervision by a securities regulator and its investigative powers and sanctions, such as its rule-making power or the ability to hold corporate management criminally liable for filing misleading information. This index is taken from La Porta et al. (2006), and it takes on higher values for stricter enforcement. We split the treatment and the benchmark sample by the EU median for regulatory quality measured in 2003 and public enforcement measured in This allows us to form two approximately equal sized groups of EU countries (see Table 1 for the partitions), and benchmark them to countries with similar institutional characteristics. 22 We report the results from these regressions in Table 4. Panel A presents coefficient estimates and t-statistics of the spread regressions. The results are fairly consistent across the two partitioning variables and for the two directives. The We assign a value of zero to countries with missing data on public enforcement in La Porta et al. (2006). The results are not sensitive to this research design choice and the inferences remain unchanged if we delete these countries from the analysis instead. Note that, as our identification strategy relies on different implementation dates for the directives and we use quarter-year fixed effects, partitioning the sample is demanding as it reduces the variation in adoption dates. 21

23 implementation of the MAD and the TPD seems to reduce bid-ask spreads only in countries with high prior regulatory quality and strong public enforcement in the past. That is, the coefficient on MAD and TPD is negative and statistically significant from zero for the group with high prior regulation. The estimated effect is also significantly different from the TPD and MAD coefficient for the group with low prior regulation. For them, we do not find any change in liquidity following the entry-into-force dates of the MAD or the TPD. The coefficients on the control variables are as expected and similar across sub-samples. Again, the introduction of mandatory IFRS reporting by itself is not associated with a significant impact on liquidity. In Panel B, we report results for the implied cost of capital. Similar to the average effects presented in Table 3, the period following the implementation of TPD is not associated with a lower cost of capital. The introduction of MAD, however, is related to a lower cost of capital, but as for the spread, this effect is only significant in countries with strong prior regulation for both of our regulatory proxies. However, because the coefficients on MAD in the low quality group are also negative, the difference across samples is not statistically significant. We note that the coefficient on IFRS has opposing signs in the sub-samples and, for the TPD, is even significantly negative. Closer inspection reveals that the MAD, which came into force over a similar time period as mandatory IFRS reporting, likely acts as a correlated omitted variable in this specification. Once we control for the implementation of the MAD in the regression, the coefficient on IFRS again loses its statistical significance. We come back to this issue and the joint effects of EU regulatory changes in more detail in Section 4.3. Overall, our liquidity and, to a lesser degree, cost of capital results using a country s prior regulation as a split variable to tease out cross-sectional effects support the notion that implementation and enforcement are important factors for economic outcomes of new securities regulation. 22

24 To examine this hypothesis more directly, we construct variables that indicate how member states have implemented the MAD and the TPD. EU member states have some discretion when translating new EU directives into national law. This discretion can play out in various ways, for example, in the way rules are implemented nationally, supervisory regimes are set up or in the way rules are monitored and enforced once violations occur. 23 We analyze the existence of these effects by estimating the following variation of our model: EconCon = β 0 + β 1 MAD/TPD Implementation Strong + β 2 MAD/TPD Implementation Weak + β j Controls j + β i Fixed Effects i + ε, (2) where Implementation Strong and Implementation Weak stand for binary indicator variables that split the post-mad or post-tpd observations into two non-overlapping groups, one for countries with substantial changes around the implementation of the directives and with stricter implementation, and one for the remaining EU member states with weaker implementation. We use three implementation-specific variables to partition the coefficients: (i) TPD_Shift indicates a substantial change in the enforcement activity of the local supervisory authority during our sample period. We gather this information from a survey that we sent to the enforcement institution as well as to the technical department of the local PricewaterhouseCoopers affiliates in each EU member state. 24 (ii) TPD_Full represents the sub-set of EU countries that fully comply with all the enforcement principles of CESR Standard No. 1 by the end of (iii) For related studies in the insider trading literature, see also Bhattacharya and Daouk (2002) and Ackerman et al. (2008). We coded TPD_Shift as 1 if the local enforcement authority indicated that it implemented a comment and review process for the first time, and the audit firm replied that, according to their own assessment, a significant shift in the intensity of enforcement did occur during the sample period. In cases of disagreement between the two and if we could not resolve the issue either by other answers to our survey questions or by going back to the respondents, we coded TPD_Shift as zero. CESR Standard No. 1 outlines 21 principles on how the EU wants to achieve harmonization of the enforcement and oversight mechanisms of the provision of financial information by its market participants (e.g., on the requirement to set up an independent supervisory authority by each member state, on the varying 23