Do Anti-Bribery Laws Reduce the Cost of Equity? Evidence From the UK Bribery Act 2010

We examine the impact of the U.K. Bribery Act 2010 on the implied cost of equity. We find a significant reduction in the cost of equity amongst U.K. firms with high bribery exposure. We further show that the Bribery Act improves internal control system and increases stock liquidity of firms with high bribery exposure. Our results suggest that more stringent anti-bribery regulations are not always bad for the firm.


Introduction
Corruption is prevalent worldwide and is commonly cited as a significant deterrent to economic growth (Bardhan, 1997;Sensson, 2005;Shleifer and Vishny, 1993). As a common form of corruption, corporate bribery is increasingly becoming an important concern for policymaker as well as corporate stakeholders around the world (Karpoff, Lee, and Martin, 2017;Cheung, Rau, and Stouraitis, 2012). The World Bank Institute estimates that $1 trillion a year is paid in bribes (Rose-Ackerman, 2004). In a survey of corporate managers in 125 countries, D'Souza and Kaufmann (2013) find that more than 60% of participants believe that their competitors use bribes to secure a public contract. Similarly, the Dow Jones State of Anti-Corruption Survey in 2014 finds that a third of companies claim to have lost business to unethical competitors. Becker (1968) models corporate misbehaviors as an economic decision that involves trade-offs between benefits and costs. Paying bribes in foreign countries may potentially benefits firms by helping them expediting through the inefficient bureaucratic process (Huntington, 1968;Leff, 1964). Indeed, Cheung, Rau, and Stouraitis (2012) find that paying bribes brings an average firm $11 per each dollar of bribe paid. Karpoff, Lee, and Martin (2017) estimate the costs and benefits of bribery and find that, even after netting out expected financial and reputational costs, projects involving bribes remain profitable to firms. Zeume (2017) documents a decline in value of U.K. firms since the Bribery Act 2010 came into force due to reduced growth opportunities and concludes that paying bribes is a necessary cost of doing business.
Conversely, paying bribes can be costly to the firm. In addition to the risk of penalties and prosecutions (Murphy, Shrieves, & Tibbs, 2009), corporate bribery may endanger the firm's reputation and strain relations with stakeholders (D'Souza & Kaufmann, 2013;Serafeim, 2014). Focus on paying bribes to win business can also distract firms from investing in value-enhancing long-term projects (Birhanu, Gambardella, & Valentini, 2016).
Electronic copy available at: https://ssrn.com/abstract=3221984 legislation on bribery is among the strictest. 1 We argue that this significant change in the legal environments particularly affects firms that were already operating in bribe-prone countries before the passage of the Bribery Act, and examine the law's influence on their cost of equity in a difference-in-difference setting.
Controlling for decline in the expected growth rate, we find that the Bribery Act is associated with a 4.0% reduction in the implied cost of equity for firms with high bribery exposure. 2 Given that our model estimates the average cost of equity for these firms to be 12.5%, this is an economically substantial reduction, amounting to almost one-third of these firms' cost of equity. Our results suggest that the Bribery Act significantly reduces the risk of firms that operate in bribe-prone countries.
We perform several robustness checks to validate our main findings. We show that our results are not sensitive to how we classify firms as having high exposure to bribery, nor to several alternative earnings measures. We further show that our results hold when including loss observations in our sample and performing value-weighted instead of equally-weighted regressions (Easton 2007;Easton & Sommers, 2007). Because the U.K. contemporaneously changed its tax regime on foreign corporate income, we alleviate the concern that this may have driven our documented change in the cost of equity by directly controlling for each firm's exposure to tax on foreign income; we still find that the change in on cost of equity is only prevalent among firms with high exposure to foreign bribery. Further, we validate the parallel trend assumptions in our difference-in-difference design (Atanasov & Black, 2016;Roberts & Whited, 2012) by repeating our analysis using several artificial event years, and also performing our analysis on an entropy-balanced sample (Hainmueller, 2012;Hainmueller & Xu, 2013). Overall, these additional results support our main finding that the Bribery Act significantly reduces the cost of equity of firms with business operations in bribe-prone countries.
We present evidence that the Bribery Act affects cost of equity by improving corporate internal control (Ashbaugh-Skaife et al., 2009;Beneish, Billings, & Hodder, 2008) and stock liquidity (e.g., Lang et al., 2012;Pagano & Volpin, 2012). Using data from ASSET4's antibribery database, we find that the Bribery Act is associated with a significant improvement in corporate internal control to prevent bribery among firms that operate in bribe-prone countries.
Similarly, we find that the Bribery Act is associated with a significant decline in the bid-ask spread and Amihud's (2002) illiquidity measure, suggesting that the stocks of firms exposed to bribery have become more liquid after the Bribery Act passed. Overall, our results highlight the positive effects of the Bribery Act on a firm's cost of equity through improving the internal control system and stock liquidity.
This paper complements existing studies on the impact of foreign bribery prevention laws (e.g. Cheung et al., 2012;Karpoff et al., 2017;Zeume, 2017). In particular, Zeume (2017) finds that the U.K. Bribery Act is associated with reduced growth opportunities and a decline in firm value. Our results suggest that, despite the negative overall effect on firm value, the Bribery Act improves corporations' internal control system and stock liquidity. Consequently, equity shareholders demand a lower rate of return, which results in a lower estimated cost of equity.
Our study also contributes to the debate on the costs and benefits of anti-bribery laws.
Recent high-profile cases have demonstrated to policymakers that foreign bribery remains Electronic copy available at: https://ssrn.com/abstract=3221984 prevalent. 3 Proponents of anti-bribery laws argue that bribery increases the overall cost of business operations and undermines business confidence (Kennedy & Danielson, 2011). For instance, Christine Lagarde, as the Chairwomen of the International Monetary Fund, refers to bribery as "a corrosive force that eviscerates the vitality of business and stunts a country's economic potential". 4 In contrast, critics of these laws argue that they impose unnecessarily large compliance costs (e.g., Weissmann & Smith, 2010). Our results highlight the risk reduction benefit of anti-bribery laws. Despite the loss in business opportunities (Hines, 1995;Zeume, 2017) and the increased compliance costs (Aguilera & Vadera, 2008;Collins, Uhlenbruck, & Rodriguez, 2009;D'Souza & Kaufmann, 2013), strengthened internal control and enhanced corporate transparency can reduce adverse selection risk of these firms (Kolstad & Wiig, 2009;Pagano & Roell, 1996), resulting in the reduction of required return to shareholders.

U.K. Bribery Act 2010
The Bribery Act was passed on March 25, 2009. This legislation is the first U.K. legislation that explicitly addresses bribery by corporations, and substantially increases the severity of penalties compared to previous anti-bribery legislations. The enactment of the Bribery Act has resulted in the U.K. moving from a country that failed to meet the OECD Anti-Bribery 3 A recent example is Goldman Sachs' involvement in alleged embezzlement from Malaysia's state-run investment fund. According to court documents, Goldman Sachs' employees paid bribes to state officials to secure large underwriting deals. See: https://www.nytimes.com/2018/11/01/business/goldman-sachs-malaysia-investmentfund.html. 4 See "Addressing corruption with clarity", Brookings Institution, Washington DC, September, 18, 2017. Available at: https://www.imf.org/en/News/Articles/2017/09/18/sp091817addressing-corruption-with-clarity.
Convention to be amongst the countries that have the most stringent anti-bribery legislation. 5 Specifically, the Bribery Act imposes unlimited fines and jail terms up to ten years for bribing and taking bribes. The legislation applies to individuals or companies that use bribes in the U.K. or elsewhere, and extends to corporations with U.K. operations, employing U.K. citizens, or providing services to any U.K. organization. Unlike previous U.K. anti-bribery laws and the U.S. Foreign Corrupt Practices Act (FCPA), the Bribery Act also covers facilitation payments that aim to induce the performance of routine and obligatory government tasks (Trautman & Altenbaumer, 2013). It also has a wider scope of application, including all forms of bribes to not only foreign public officials but also private operators.
The Bribery Act also imposes substantial fines if a corporation fails to prevent bribery because its internal control system is insufficient. Corporations must establish effective antibribery systems and controls, such as: (1) adequate procedures; (2) top (board)-level commitment; (3) risk assessment; (4) due diligence; (5) communication and training; and (6) monitoring and review. Risk-based due diligence must also extend to counterparties like contractors and suppliers. Overall, the Bribery Act has changed the basis for corporate criminal liability from personnel misconduct within the firm to the quality of the system governing the firm's activities (Mukwiri, 2015).

Bribery and cost of equity
Recent empirical evidence suggests that bribery in foreign countries facilitates business expansion and, ultimately, firm growth. Cheung et al. (2012) find that each $1 paid in bribes 5 U.K. legislation on bribery can be traced back to the Public Bodies Corrupt Act 1889, which confined bribery to the public sector and criminalized the soliciting or receiving of a bribe by a public officer. The law was reformed by the Prevention of Corruption Act 1906, which expanded bribery to the private sector, and the Prevention of Corruption Act 1916, which further lightened the burden of proving corruption. Due to the complexity and the uncertainty arising from having several laws related to bribery, the U.K. failed to bring any foreign bribery case to court and risked being sanctioned by the OECD. increases a firm's market valuation by $11. Karpoff et al. (2017) find that, even when firms are caught bribing, the financial benefits from foreign bribery still more than offset any associated fines, legal expenses, and reputational losses. They also find that reputational losses associated with bribery are negligible.
Other studies contend that bribery prevention laws can potentially reduce a firm's competitiveness in countries where bribery is prevalent (Iriyama, Kishore, & Talukdar, 2016). Hines (1995) finds that the US Foreign Corrupt Practices Act is associated with a significant reduction in operations by U.S. firms in bribe-prone countries. Similarly, Zeume (2017) finds that the Bribery Act reduces the growth of U.K. firms operating in corruption-prone countries and argues that the Bribery Act has increased the cost of doing business overseas for U.K. firms.
The passage of the U.K. Bribery Act can result in a reduction in firm risk and, consequently, the cost of equity. The literature recognizes corporate misconduct, including bribery, as a business risk (e.g., Lyon & Maher, 2005;Murphy et al., 2009); consequently, a law that discourages managers from engaging in misconduct can potentially reduce a firm's business risk. Further, bribery can make firms more opaque: they may withhold financial information to divert funds to pay for bribes (Das, Nanda, & Xiao, 2015;Smith, 2016).
Information risk has been found to be priced in by equity holders (Ashbaugh-Skaife et al., 2009;Beneish et al., 2008), and can, thus, affect the cost of equity. By criminalizing failure to prevent bribery, the Bribery Act reduces managers' incentives to engage in bribery, and may thereby decrease a firm's risk and its cost of equity.

Hypothesis 1: The U.K. Bribery Act is associated with a reduction in cost of equity.
On the other hand, the Bribery Act may increase a firm's risk because it increases the propensity of and the costs associated with being caught. Compared to previous U.K.
Electronic copy available at: https://ssrn.com/abstract=3221984 legislation, the OECD Anti-Bribery Convention, and the U.S.'s Foreign Corrupt Practices Act, the Bribery Act applies to a significantly wider range of activities and imposes a much higher maximum fine. Therefore, the adverse financial consequences of being caught engaging in bribery have become more severe. Further, the Bribery Act potentially disadvantages U.K. firms compared to competitors not covered by the law, particularly local competitors in bribeprone countries (Zeume, 2017). Prohibition of paying bribes can also increase the likelihood of losing business opportunities to local competitors (De Jong, Tu, & Van Ees, 2012).

Hypothesis 2:
The U.K. Bribery Act is associated with an increase in cost of equity.

Sample
Our sample comprises 6,071 observations covering 1,022 firms listed on the London Stock Exchange (LSE) and the Alternative Investment Market (AIM) during the period of 2003-2015. 6 We obtain financial accounting and market data from Datastream/Worldscope. Following prior studies (e.g., Easton et al., 2002;O'Hanlon & Steele, 2000), we exclude observations with missing values, financial firms, and firms with negative book equity. We also follow the prior literature (e.g., Dechow, Hutton, & Sloan, 1999;O'Hanlon & Steele, 2000) and exclude observations with negative earnings, as these observations are unlikely to provide a meaningful anchor for valuation (Easton, 2007), and the importance of earnings to valuation of loss firms can differ to valuation of profitable firms (Collins, Pincus, and Xie, 1999).

Measuring firm exposure to bribery
Following prior literature (Chung et al, 2012;Karpoff et al., 2014;Zeume, 2017), we assume that a firm's propensity to engage in bribery is positively related to its exposure to bribe-prone countries. To measure firm-level exposure to bribery, we use Transparency International's Corruption Perceptions Index (CPI) to measure the prevalence of bribery in a country, and construct a weighted-average Segment CPI score for each firm based on its sales in different geographical segments. 7 Specifically, our firm-level measure for exposure to bribery is: where , is the CPI score for geographic segment s in year t. The CPI score ranges from 0 to 100, with a lower score indicating that a country is more prone to bribes. We then construct an indicator variable, Bribery Exposure, which equals 1 if Segment CPI is less than or equal to the 25 th percentile of the firms in our sample, and 0 otherwise. The firms with Bribery Exposure equal to 1 are likely to be more strongly affected by the Bribery Act.

Empirical model: The residual income valuation model
We follow Easton and Sommers (2007) by using the residual income valuation model based on current accounting and price data: where 0 = and 1 = ( − ′ )/ (1 + ′ ). (2000), 8 which allows us to simultaneously estimate the implied cost of equity ( ) and the expected growth rate in the residual income ( ′) from accounting and price information available at the end of period t. We use Equation 2 as the basis for our regression analysis. Specifically, the estimated coefficient for the intercept term ( 0 ) provides an estimate of the implied cost of capital, which is our main parameter of interest (Easton, 2007;Easton & Sommers, 2007).

This model is an adaptation of the model employed in O'Hanlon and Steele
While the expected growth rate is not this study's main focus, we avoid potential bias from arbitrary assumptions of this rate by estimating it simultaneously with the implied cost of equity. Further, using current realized earnings allows us to circumvent the problems associated with analyst earnings forecast data. For instance, Guay et al. (2011) identify a common delay between stock price changes and analysts updating their forecasts. Lys and Sohn (1990) find that analysts' short-term earnings forecasts contain only 66% of the information reflected in security prices before the forecast release date. Additionally, using only price and accounting variables allows us to include more firms in our sample.
We extend Equation 2 to allow both the implied cost of equity and growth rate to change after the Bribery Act was passed. First, we define a dummy variable Bribery Act, which equals 1 from 2010 to 2015 and 0 from 2003 to 2008, and include this variable in our model as follows: 8 Whereas Easton and Sommers (2007) use the financial year end's market price ( ) for the numerator of the right-hand-side variable ( − ), O'Hanlon and Steel (2000) use the average of a firm's market price at the end of the financial reporting month and six subsequent months. As our analysis compares the cost of equity before and after the Bribery Act was passed, we follow the former approach to avoid using information that is not available in the current period.
As the Bribery Act is likely to disproportionately affect firms that operate in bribeprone countries, we include an indicator variable Bribery Exposure, which equals 1 for firms with high exposure to bribe-prone countries, and 0 otherwise. The resulting difference-indifference specification is as follows: Our main coefficient of interest is 0 ′′ , which captures the change in implied cost of equity for firms with high bribery exposure after the Bribery Act passed, relative to the change for firms with low exposure to bribery. 9 (1) firms with low exposure to bribery before the Bribery Act passed ( = 0 ); (2) firms with low exposure to bribery after the Bribery Act passed ( = 0 + δ 0 ′ ); (3) firms with high exposure to bribery before the Bribery Act passed ( = 0 + δ 2 ); and (4) firms with high exposure to bribery after the Bribery Act passed ( = 0 + δ 0 ′ + δ 0 ′′ + δ 2 ). To measure the effect of the Bribery Act 2010 on firms with high exposure to bribery, we compare the difference between (1) the change in the implied cost of equity among high briberyexposure firms pre-and post-Bribery Act, ( 0 + δ 0 ′ + δ 0 ′′ + δ 2 ) − ( 0 + δ 2 ) = δ 0 ′ + δ 0 ′′ ; and (2) the change in the implied cost of equity among low bribery-exposure firms pre-and post-Bribery Act, ( 0 + δ 0 ′ ) − ( 0 ) = 0 ′ . The difference between these two groups is (δ 0 ′ + δ 0 ′′ ) − 0 ′ = δ 0 ′′ , which is the coefficient for Bribery Act * Bribery Exposure in Equation 4our main coefficient of interest. U.K. stocks. 10 In Column 2, which includes the Bribery Act indicator variable to allow both implied cost of equity and implied growth rate to change after the Bribery Act passed (Equation 3), we find that the average cost of equity is 11.3% in 2003-2008, and decreases by 1.7% to 9.6% in 2010-2015.

Main results: the U.K. Bribery Act 2010 and the implied cost of equity
[ Table 2 about here] Column 3 presents our difference-in-difference estimation (Equation 4). Controlling for changes in the expected growth rate, we find that the coefficient for Bribery Act * Bribery Exposure is negative and significant, indicating that the Bribery Act is associated with a reduction in the cost of equity among high bribery-exposure firms relative to firms with low bribery exposure. The coefficient estimates in Column 3 suggest that the Bribery Act has reduced the cost of equity for high bribery-exposure firms by 4.0% (1.1% + 2.9%), whereas the reduction is 1.1% for low bribery-exposure firms. Since Column 3 estimates that the average cost of equity for high bribery-exposure firms is 12.5% (11.1%+1.4%), this reduction in the cost of equity is economically substantial, amounting to almost one-third of the average cost of equity of high bribery-exposure firms.
Column 4 includes year and firm fixed effects to eliminate the influence of any unobserved heterogeneity that is time invariant across firms and any market-wide variation across time periods. We find that the coefficient for Bribery Act * Bribery Exposure remains statistically significant (p < 0.05). We also find that including firm and year fixed effects results in a larger estimated effect on the implied cost of equity (3.6%, compared to 2.9% in Column 3). 10 The average return on the FTSE 100 Index was 8.4% during 2003-2015 (or 11.5% if 2008 is excluded).
Overall, our main results indicate that the Bribery Act has significantly lowered the cost of equity for firms exposed to bribe-prone countries. Our evidence suggests that bribery regulations reduce the risk for equity holders, leading them to demand a lower rate of return for their investment. Table 3 presents the results from various robustness checks. Panel A shows that our main results are not sensitive to how we classify firms as having high exposure to bribery. In our baseline results, the indicator variable Bribery Exposure is equal to 1 when a firm's weightedaverage Segment CPI score is equal to or below the 25 th percentile, and 0 otherwise. In Columns 1-4, we change this threshold to equal to or below the 10 th , 20 th , 30 th , and 40 th percentiles, respectively. We find that the coefficient for Bribery Act * Bribery Exposure is statistically significant when firms with high bribery exposure are classified as those with a weightedaverage Segment CPI score equal to or below the 10 th or 20 th percentile (Columns 1 and 2), but becomes weaker as we relax this threshold in Columns 3 and 4. These results support our conjecture that the Bribery Act has more strongly affected firms with higher exposure to bribeprone countries. In Column 5, the Bribery Exposure indicator is replaced with the negative value of the weighted-average Segment CPI scores; our main finding continues to hold. 11 In Column 6, we use FTSE4Good bribery exposure data as an alternative data source to the CPI, and continue to find consistent results. 12 Overall, these results provide further evidence that firms with high bribery exposure have experienced a significant decrease in their cost of equity since the Bribery Act passed. 11 As a lower Segment CPI score suggests that a country is less transparent and more bribe prone, we use the negative value so that the coefficient is in the same direction as in our main results. 12 FTSE4Good bribery exposure data evaluates firms based on their operating industries, countries, and the extent to which the firms are involved in public contracts. We thank FTSE for providing us with the data.

Robustness checks
[ Table 3 about here] Panel B shows that our findings hold when using alternative proxies for earnings.
Column 1 uses I/B/E/S Actual EPS, which excludes one-time items and, consequently, may be more representative of a firm's sustainable earnings (Easton, 2007). Column 2 follows Easton et al. (2002) by using I/B/E/S forecast EPS instead of reported earnings. 13 The results in both columns are consistent with our main results.
In Panel C, we test the sensitivity of our findings to including loss observations. While our main tests follow prior studies by excluding observations with negative earnings from the sample (e.g., Dechow et al., 1999;O'Hanlon & Steele, 2000), only considering healthier firms may lead to biased inferences (Easton, 2007). For brevity, we only include the coefficient estimates for Bribery Act * Bribery Exposure. In Column 1, we find that the coefficient for Bribery Act * Bribery Exposure is negative but not statistically significant. This is potentially due to the large proportion of observations with negative earnings in this sample (32%), which suggests that this estimate for cost of equity may not be reliable (Easton, 2007). As our sample includes firms listed on the FTSE or the AIM, Columns 2 and 3 present separate results for observations from these two stock markets. We find that the coefficient for Bribery Act * Bribery Exposure is negative and significant among FTSE firms, which notably have the smallest proportion of loss observations (16%). The overall results indicate that a large proportion of loss observations significantly influences the cost of equity estimation. Following Easton (2007) and Easton and Sommers (2007), we also estimate value-weighted regressions 13 The residual income valuation model in Easton et al. (2002) is where +1 is the next period's earnings forecasted at t, 0 is the implied growth rate, and 1 is the difference between the cost of equity and the implied growth rate. To directly estimate the cost of equity, we algebraically transform this model to where the intercept term ( 0 ) is the cost of equity.
to reduce the influence of negative earnings, which tend to be from smaller firms in the sample.
We find that the value-weighted coefficients for Bribery Act * Bribery Exposure are negative and statistically significant across all three columns.
Panel D alleviates the concern that our results are confounded by the contemporaneous passage of the Finance Act 2009, whereby the U.K.'s tax regime transitioned from a worldwide tax system to a territorial tax system. After 2009, U.K. multinationals ceased to pay dividend tax on earnings from foreign countries with lower corporation tax rates than the U.K.'s (Liu, 2018). 14 Because the effect of this tax reform is likely to manifest in financial year end 2010, we directly control for the tax exposure of U.K. firms on their foreign income. Specifically, we construct two measures for tax exposure: Tax Exposure 1 is defined as the sum product of foreign sales and the foreign corporation tax rate scaled by total sales: Tax Exposure 2 is defined as the sum product of foreign sales ratio and a dummy variable that equals 1 when the foreign corporation tax rate is below the U.K.'s corporation tax rate: 15 In Columns 1 and 2 of Panel D, we include the two foreign tax exposure variables and their interaction with Bribery Act as additional control variables in the model. We find in both columns that Bribery Act * Tax Exposure does not enter the model significantly, implying that 14 Prior to the 2009 reform, U.K. companies paid the difference between the U.K. corporation tax rate and the applicable foreign corporation tax rate on any income repatriated into the U.K. Income from countries whose tax rate was higher than the U.K.'s incurred no additional tax charge. 15 We use 2010 tax rates obtained from the OECD Statutory Corporate Income Tax Rates, available at https://stats.oecd.org/index.aspx?DataSetCode=Table_II1.
the variation in U.K. firms' tax exposure on foreign income does not explain the change in their cost of equity after the Bribery Act passed. By contrast, we continue to find that the coefficient for Bribery Act * Bribery Exposure is negative and statistically significant. These results suggest that the reduction in cost of equity is more likely to have resulted from the Bribery Act than the Finance Act 2009.

Artificial event years and covariate balancing
The validity of our difference-in-difference design hinges on the parallel-trend assumption (Atanasov & Black, 2016;Roberts & Whited, 2012): specifically, absent the Bribery Act, the cost of equity for all our sample firms should move in the same way regardless of their exposure to bribe-prone countries. In this section, we perform additional sensitivity tests to check the validity of this assumption.
[ Table 4 about here] We first rule out the possibility that our results are driven by any pre-existing trend (Atanasov & Black, 2016) Table 4. As expected, the coefficients for Bribery Act * Bribery Exposure are not statistically significant when we use these artificial event years. These results suggest that our findings are unlikely to be due to any pre-existing trend.
We also combine our difference-in-difference design with covariate balancing to ensure that the observations in our treatment and control groups are similar along several dimensions of covariates. Specifically, we use entropy balancing (Hainmueller, 2012;Hainmueller & Xu, 2013) to impose a weighting scheme on our observations, such that the means and variances of all covariates are exactly identical across treatment and control groups. Entropy balancing has several advantages over other matching methods, such as propensity score matching or nearest-neighbor covariate matching. First, entropy balancing can be used to obtain a higher degree of balance for a large set of covariates, because it directly searches for a set of observation weights that give the treatment and control groups the same means and variances (Hainmueller, 2012). Second, the matching scheme in entropy balancing allows the weight of observations to continuously vary across different units (Hainmueller & Xu, 2013). Therefore, unlike nearest-neighbor or radius matching schemes, we do not need to keep or drop observations from the matched sample based on any arbitrary threshold. Consequently, we can retain a larger proportion of observations in the matched sample. Finally, the weights resulting from entropy balancing can flexibly be used with any standard estimators, including the difference-in-difference estimator.
We use entropy balancing to match our observations along several key covariates that can influence a firm's cost of equity: market capitalization, book-to-market ratio, cash flow from operations, stock return, sales growth, and leverage (see, e.g., Ashbaugh-Skaife et al., 2009;Lang et al., 2012). 16 The estimation using the entropy-matched sample is presented in Column 4 of Table 4. Consistent with our baseline results, the coefficient for Bribery Act * Bribery Exposure is negative and statistically significant. This balanced-sample result provides further confidence that our main findings are unlikely to be driven by differences between firms with high and low exposure to bribery.

Potential mechanisms
We find that the passage of the Bribery Act is associated with a significant reduction in cost of equity among firms with high exposure to bribe-prone countries. We postulate that this reduction is due to firms becoming less risky to equity holders, such that they demand a lower rate of return (Lang, Lins, & Maffett, 2012). This section provides evidence of potential mechanisms through which the Bribery Act reduces firm risk.

Strengthened internal control system
One potential mechanism through which the Bribery Act decreases the cost of equity is by improving corporate internal control system. The Bribery Act requires firms to implement an internal control system for bribery and corruption prevention and imposes significant personal and legal liability on managers if this system fails to prevent bribery (Bargeron, Lehn, & Zutter, 2010;Litvak, 2008). This should encourage firms to improve their internal control system to reduce the probability of negative outcomes associated with paying bribes (Birhanu et al., 2016;D'Souza & Kaufmann, 2013;Serafeim, 2014). Prior literature has evidenced the association between a firm's internal control system and its cost of equity (Ashbaugh-Skaife et al., 2009;Beneish et al., 2008).
[ Table 5 about here] In Table 5, we construct a variable Anti-Bribery Score from ASSET4's anti-bribery database. 17 The variable Anti-Bribery Score ranges from 0-6, with a higher score indicating that the firm has a stronger internal control system. Column 1 employs the difference-indifference estimation with Anti-Bribery Score as the dependent variable. We find that the coefficient for Bribery Act * Bribery Exposure is positive and significant, suggesting that, among firms with high exposure to bribery, the Bribery Act is associated with a significant improvement in corporate internal control to prevent bribery. In Column 2, we include several 17 ASSET4's anti-bribery data comprise six indicators, namely whether the company: (1) mentions a public commitment to anti-bribery/corruption at the senior management or board level; (2) strengthens anti-bribery/corruption in its code of conduct; (3) adopts an internal management tool for bribery/corruption such as hotlines or whistle-blowing systems; (4) has a policy to cope with bribery/corruption in business transactions; 5) communicates on briberyrelevant issues with employees in organizational processes; and 6) conducts employee training on anti-bribery/corruption. firm characteristics that can influence a firm's internal control system: market capitalization, book-to-market ratio, leverage, working capital ratio, and ownership concentration (see e.g., Doyle, Ge, & McVay, 2007); our results continue to hold. Overall, these findings indicate that the Bribery Act is indeed associated with improved internal control among high briberyexposure firms.

Increase in stock liquidity
As an enhanced internal control system can improve corporate transparency by mitigating management's ability to distort financial information or engage in concealed transactions (Dass, Nanda, & Xiao, 2016;Leuz, Nanda, and Wysocki, 2003), the Bribery Act could reduce the information asymmetry between the firm and outsider shareholders (Beneish et al., 2008).
This would, in turn, result in better stock liquidity by reducing the need for private information gathering and also reducing adverse selection in trading of the firm's stock (Chung, Elder, and Kim, 2010;Diamond and Verrecchia, 1991;Francis et al., 2007;Holmstrom and Tirole, 1993).
As better liquidity reduces the cost of equity (e.g. Lang, Lins, and Maffett, 2012;Lin, Singh, and Yu, 2009), the improvement in stock liquidity could be a potential mechanism through which the Bribery Act reduces cost of equity.
[ Table 6 about here] Table 6 examines the effect of the Bribery Act on two key proxies for stock liquidity: (1) the bid-ask spread (Easley & O'Hara, 1987;Glosten & Milgrom, 1985;Kyle, 1985); and (2) Amihud's (2002) illiquidity measure. Column 1 shows that the bid-ask spread significantly decreases after the Bribery Act passed. In Column 2, we follow prior literature (e.g., Amiram, Owens, & Rozenbaum, 2016;Daske et al., 2008) by including additional control variables: market capitalization, return volatility, trading volume, and stock turnover; our results continue to hold. Columns 3 and 4 present consistent evidence when using Amihud's (2002) illiquidity measure. Overall, these findings offer evidence that the stock liquidity of high bribery-exposure firms increased after the Bribery Act passed.

Conclusions
This paper examines the effect of the Bribery Act on cost of equity. We find that the implied cost of equity of firms with high exposure to bribe-prone countries has significantly decreased since the Bribery Act passed. Among the same firms, we also find that the Bribery Act is associated with improvement to the corporate internal control system and also stock liquidity, suggesting that these are potential mechanisms through which the Bribery Act reduces firm risk.
With the costs and benefits of anti-bribery laws still being debated by policymakers and academics, our results contribute by showing the positive impact of one such law. Despite the concern that these laws may impede business expansion in some foreign markets and increase compliance costs, we highlight significant benefits in terms of improved internal control and stock liquidity. Our findings suggest that anti-bribery laws can benefit equity investors.
communicates on bribery-relevant issues with employees in organizational processes; and 6) conducts employee training on anti-bribery/corruption. Notes: Segment CPI score is a sum product of Transparency International's Corruption Perception Index score and the firm's sales in different geographical segments, scaled by total sales. Anti-Bribery Score is an index constructed using six indicator to anti-bribery/corruption provisions provided by ASSET4. Other variables are from Datastream/Worldscope. All variables are defined in Appendix A.    (Hainmueller, 2012;Hainmueller and Xu, 2013), in which 1,348 high bribery exposure observations are matched with 4,907 low bribery exposure observations. Standard errors (in parentheses) are clustered at the firm level. *, **, and *** denote statistical significance at 10%, 5%, and 1% respectively. Notes: The dependent variable is Anti-Bribery Score, an index constructed using six indicators related to anti-bribery/corruption provisions provided by ASSET 4. ASSET4's anti-bribery data comprise six indicators, namely whether the company: (1) mentions a public commitment to anti-bribery/corruption at the senior management or board level; (2) strengthens anti-bribery/corruption in its code of conduct; (3) adopts an internal management tool for bribery/corruption such as hotlines or whistle-blowing systems; (4) has a policy to cope with bribery/corruption in business transactions; 5) communicates on bribery-relevant issues with employees in organizational processes; and 6) conducts employee training on anti-bribery/corruption. Other variables are defined in Appendix A. Robust standard errors clustered at the firm level are presented in parentheses. *, **, and *** denote statistical significance at 10%, 5%, and 1% respectively. Notes: The dependent variables are Ln Bid-Ask Spread (Column 1-2), defined as the logarithm of daily closing ask price less the closing bid price scaled by the midpoint of the closing ask and bid prices, and Ln Illiquidity, defined as the logarithm of the average of daily ratio of absolute stock return to GBP trading volume multiplied by 1,000 (Amihud, 2002). Other variables are defined in Appendix A. We exclude observations in 2014-2015 due to the exemption of Stamp Duty Reserve Tax for trading on AIM and LSE high-growth stocks. Robust standard errors clustered at the firm level are presented in parentheses. ***, **, and * denote statistical significance at the 1, 5, and 10% levels respectively.