1. Corporate governance and stock liquidity:
Panel evidence from 2001 to 2013
Searat Ali*, Benjamin Liu, Jen Je Su
Griffith University, Australia
Accounting and Finance Association of Australia and New Zealand (AFAANZ)
4-5 July 2016
*Presenter

2. Research questions (RQs)
Does corporate governance quality (CGQ) determine stock liquidity?
RQ2
Does the impact of CGQ on stock liquidity depend on the choice of liquidity proxies or dimensions?
RQ3
How does CGQ affect stock liquidity?

3. Importance of stock liquidity
Investors want three things from the markets: liquidity, liquidity and liquidity (Handa & Schwartz, 1996).
In a perfect liquid market, any amount of a given security can promptly be converted to cash or vice versa at no cost. In a less than perfect world, a liquid market is one where the transaction costs associated with this conversion are minimized (Harris, 1990).
Liquidity
Illiquidity premium
(Amihud and Mendelson, 1986)
Cost of equity
(Lipson and Mortal, 2009)
Firm value
(Fang et al., 2009)
Determinants of stock liquidity
Stock characteristics
Market characteristics
External governance mechanisms
Internal governance mechanisms
Investor
Firm

4. Determinants of stock liquidity
Stock characteristics
stock price, return volatility, trading volume and number of trades
Market characteristics
market structure and competition
External governance mechanisms
cross country variations in regulatory and legal environments
Internal governance mechanisms
Limited studies
Our Focus
(Tinic, 1972; Benston and Hagerman, 1974; Branch and Freed, 1977)
(Bacidore and Sofianos, 2002; Brockman and Chung, 2003; Chung, 2006; Eleswarapu and Venkataraman, 2006)

5. Gap in the literature
Classical studies emphasize the role of internal CGQ in determining stock liquidity.
For example, Coffee (1991) argues that large investors support internal governance mechanisms because such mechanisms enhance stock liquidity that in turn makes exit less costly for them.
Chung, Elder, and Kim (2010) show that the CGQ improves stock liquidity in the US.
These findings, however, are limited to a short time-series (2001 to 2004) and coincide with the introduction of the Sarbanes-Oxley Act of 2002, which might have resulted in a spurious correlation between CGQ and stock liquidity.
Those authors use high frequency quote-based liquidity proxies capturing trading cost and price impact dimensions, and antitakeover provisions are one of the key components of their governance index.
This casts a doubt on the generalizability of the results from the US to other countries, where anti-takeover provisions and high frequency quote-based liquidity proxies are non-existent.

6. Gap in the literature
Other studies on CGQ and stock liquidity are conducted in emerging markets such as Malaysia (Foo and Zain, 2010), China (Lei, Lin, and Wei, 2013), and Thailand (Prommin, Jumreornvong, and Jiraporn, 2014).
Overall, these studies suffer from short-time series and limited liquidity proxies or dimensions.
For instance, Prommin, Jumreornvong, and Jiraporn (2014) document that better governance improves stock liquidity over time in Thailand. However, their finding does not survive in the cross-sectional setting. Since they select only 100 large firms during a short period of time (2006 to 2009), their findings may not be generalized to the wider economy.
Another important limitation of the prior studies on CGQ and stock liquidity is that they do not provide an empirical test of the channel through which CGQ affects stock liquidity.
Corporate governance provisions of large firms do not vary much. As can be seen in their summary statistics (Table 2), the governance index is 6 in the 25th percentile, 7 in the 50th percentile and 8 in the 75th percentile. This may be the reason that they have not found significant cross-sectional variation between governance quality and stock liquidity.

7. Theoretical linkage Internal corporate governance
Information disclosure
Information asymmetry
Stock liquidity
Corporate governance is assumed to affect stock liquidity through the channel of information transparency.
In particular, better CGQ imposes more monitoring on managers and, therefore, prevents opportunistic managers from concealing and distorting information.
Thus, better corporate governance improves the informational transparency of a firm and mitigates information asymmetry, between insiders (e.g., managers) and outsiders (e.g., investors), as well as among outsiders.
When information asymmetry is less severe, traders face fewer adverse selection problems (Glosten and Milgrom, 1985); hence, they provide more liquidity to stocks of well governed firms.
Give examples on how Board quality and audit quality monitor the managerial behavior

8. Contributions
The importance of the given issue and the limitations in the empirical literature motivate us to provide comprehensive and robust evidence on the governance–liquidity nexus
To the best of our knowledge, no prior researcher has examined such a relationship
in Australia
using an internal CGQ,
a larger cross-section (1,207 unique firms)
a longer time-series data (2001 to 2013),
and three main dimensions of stock liquidity (trading cost, price impact and immediacy)
that are calculated by using both high and low frequency data in one paper.
In addition, we are the first to show empirically how CGQ improves stock liquidity in Australia. We study disclosure as a channel that helps explain the results.
Give examples on how Board quality and audit quality monitor the managerial behavior

9. Contributions (1/5) Australian Context
Trading mechanism
order-driven
Corporate governance
less stringent corporate governance
weak market for control
high ownership concentration
Australia possesses a distinctive regulatory and institutional framework that casts a doubt on the generalizability of results from the US to the Australian market(see e.g., Monem, 2013; Méndez, Pathan, and García, 2015).

10. Contributions (2/5) Channel
Corporate governance >> disclosure quality
(Beekes and Brown, 2006; Beekes et al., 2015)
Disclosure quality >> stock liquidity
(Chang et al., 2008),
Corporate governance >> Disclosure >> stock liquidity??

11. Contributions (3/5) Liquidity proxy Frequency High frequency
Low frequency
Dimensions
Trading cost
Price impact
Immediacy
Data type
Quote-based
Volume-based
Price-based
Is there any differential effect of CGQ on various dimensions of liquidity?
Liquidity proxy
Prior literature on corporate governance and stock liquidity either uses high frequency or low frequency proxies of stock liquidity. We extend literature by incorporating both high frequency (i.e., time-weighted quoted spread) and low frequency (e.g., Amihud illiquidity estimate, liquidity ratio and turnover) stock liquidity proxies in a single study.
In addition to this, prior studies on corporate governance and stock liquidity include one or two dimensions of stock liquidity. We extend the literature by incorporating liquidity proxies that capture three dimensions of the liquidity namely trading cost, price impact of trade and immediacy in a single study. It helps us to know if corporate governance has a differential effect on various dimensions of stock liquidity.
Finally, based on the type of data, Lesmond (2005) classifies liquidity proxies into three categories: price-based, volume-based and quote-based. Each liquidity proxy has its weaknesses and strengths. Prior literature on CGQ and stock liquidity does not cover all. However, our liquidity proxies in this paper cover all the three categories.

12. Contributions (4/5) Sampling Sample period and size Major events
Large cross section 1217 firms
Long time series 2001 to 2013
Major events
ASX CG reforms (2003)
GFC (2008)
Generalizability
Entire SIRCA database
All non-financial firms
Generalize to wider economy
Time series and cross sectional investigation separately.
Sample period
Most of the studies on corporate governance and liquidity suffer from either small cross section or short time series. However, our study covers a large panel dataset i.e. large cross-section (1,217 unique firms) and long-time series (2001 to 2013).
Investigating the governance–liquidity linkage in Australia in such a period is important because of two major events. The first, the introduction of the Australian Securities Exchange corporate governance reforms (hereafter ASX CG reforms) in 2003, enables access to data both pre- and post-CG reforms; the second, the GFC in 2008.
This is the first study which uses the entire universe of Australian governance data from SIRCA database. The sample firms come from all non-financial industries, ages, and are heterogeneous in size and profitability. Therefore, the extended dataset allows us not only to generalize results to wider economy (small, medium and large firms) but also helps us to investigate the relationship between CGQ and stock liquidity in both time series(within firm) and cross sectional (between firms) setting separately.

13. Contributions (5/5) Governance proxy Horwath report Comprehensive
Well-recognized
Limitations
Simplification
Extensions
To capture internal governance quality we follow the Horwath report, which is comprehensive and well-recognized in the research community.
This report pays special attention to the aspects that have been viewed as important in CG best practice codes in Australia and elsewhere (IFSA Australia, 1999; ASX CG reforms, 2003; OECD Report, 1999; US Blue Ribbon Committee, 1999; UK Hampel Committee, 1998; Ramsay Report, 2001).
Prior studies that use the Horwath report as a measure of CGQ either are cross-sectional or have linked CGQ to corporate activities other than stock liquidity, such as firm performance (Linden and Matolcsy, 2004), information disclosure (Beekes and Brown, 2006; Beekes et al., 2014), and corporate social responsibility (Chan et al., 2014).
Given some limitations of the Horwath report (see Section 3.2.2), we construct a new CG index that makes several contributions.
First, our CG index simplifies the Horwath report by excluding the subjective criteria and by using the equally weighted scoring methodology.
Second, our CG index extends the Horwath report to both cross-sectional (including small- and mid-cap firms) and time series (including 2008 to 2013).

14. Hypothesis development
Hypothesis 1. : Better CGQ improves the stock liquidity.
Hypothesis 2. : The relation between CGQ and stock liquidity may vary based on the liquidity dimensions (tightness, price impact and immediacy).
Hypothesis 3. : Board quality, i.e., independent directors, independent chairman and board diligence, improves stock liquidity.
Hypothesis 4. : The presence and quality of board sub-committees (audit, nomination and remuneration) improves stock liquidity.

15. Data and methodology Sample and data
The initial sample of 13,500 firm-years consists of all the Australian listed firms whose corporate governance data are available in the SIRCA during the period from 2001 to 2013.
We exclude financial firms because of their unique financial characteristics (e.g., composition of financial statements; substantially higher debt ratios) and operating environment.
We also exclude firm-year observation with missing data on any one of the following : (1) governance data; (2) financial data; and (3) liquidity data.
The final sample comprises 10,179 observations on 1,207 non-financial firms across all size groups (small, medium, and large).
The financial information about the sample firms, obtained from the Morningstar DatAnalysis Premium database. The data for the calculation of corporate governance index and stock liquidity such as stock price and trading volume is obtained from SIRCA.

16. Data and methodology Measures of corporate governance (CGQ)
Horwath report
Focus: Quality of a firm’s internal structures and processes.
Composition: 1) board structure 2) audit committee 3) nomination committee 4) remuneration committee 5) external auditor independence 6) code of conducts and other policy disclosures
Limitations
Only top 250 firms
only from
Methodology is confidential
Subjective categories (5 and 6)
No category scores
Construction of CG index
We address these issues by collecting an extended CG dataset across both cross section (large, mid-cap and small firms) and time series ( ) on the objective Horwath categories.
These categories are based on 17 criteria.
To construct the CG index, we use an equally weighted scoring methodology that has been used in extant corporate governance research (see e.g., Gompers, Ishii, and Metrick, 2003).
These individual values are then aggregated to construct a composite CG index which ranges from 0 to 17 where 0 indicates the ‘worst’ governance and 17 indicates the ‘best’ governance.

17. Data and methodology Measures of stock liquidity
Dimension 1: Trading cost
Time-weighted quoted spread (TWQS)
Zero return measure (ZERO)
Dimension 2: Price impact of trade
Amihud illiquidity estimate (ILLIQ)
Liquidity ratio (AMIVEST)
Dimension 3: Immediacy
Stock turnover (STO)
Turnover-adjusted zero daily volumes (LM)
Other immediacy proxies
number of trades (TRADES), number of levels (LEVELS), trading volume (VOLUME)

18. Data and methodology Control variables Firm size
Number of shares outstanding times share price at the end of a financial year.
Leverage
Total liabilities over the book value of total assets
Return volatility
Standard deviation of daily stock returns
Asset tangibility
Ratio of net property, plant and equipment to total assets
Share price
Reciprocal of share price
Firm age
Natural logarithm of the number of years since the firm's listing
Growth opportunities
Market value to book value ratio
Year effect
Industry effect

19. Potential endogeneity
Data and methodology
Research Methodology
Baseline
Pooled OLS
Fixed effect
Between Estimator
Potential endogeneity
Lagged independent variables
Instrumental variable approach
Dynamic panel estimation

20. Data and methodology Research methodology
Baseline models and estimation methods
To test H1 (i.e., better CGQ improves stock liquidity), and H2 (i.e., the relationship between CGQ and stock liquidity may vary based on three liquidity dimensions), we formulate the following regression equation:
First, we employ pooled ordinary least squares (OLS). The standard errors are clustered by firm to control for heteroskedasticity and within-firm correlation in the residuals (Petersen, 2009). We also include the year and industry effects to capture the variation over time and across industries, respectively.
Second, we employ the fixed effect (FE) method to control for unobserved heterogeneity, due to time-unvarying omitted variables that differ across firms but are constant over time. FE method focuses on over time (causal) changes in the variables.
Third, we employ a between estimators (BE) method that is based on the group means of the variables, therefore capturing only the cross-sectional variation.

21. Data and methodology Research methodology Endogeneity
One may raise concern about endogeneity between CGQ and stock liquidity. It is possible that CGQ and stock liquidity are endogenously determined: not only CGQ may affect stock liquidity, but also stock liquidity may trigger changes in CGQ simultaneously (Li, Chen, and French, 2012; Edmans, Fang, and Zur, 2013).
This endogeneity bias is not addressed in the study of Chung et al. (2010); however, we use three alternative model specifications to address this potential endogeneity concern.
The first alternative replaces the contemporaneous values of CGQ and the other control variables with one or two years lagged values. Eq. (9) is re-written as Eq. (11) where j=1 or 2 to show this effect of lagged independent variables on stock liquidity.
Regressions based on contemporaneous variables are susceptible to endogeneity bias due to reverse causality, whereas regressions based on lagged values of independent variables help to control for reverse causality, and thus tend to be less susceptible to endogeneity effects.

22. Data and methodology Research methodology Endogeneity
The second alternative specification uses a two-stage least squares (2SLS) approach to further address the reverse causality issue.
Based on the literature, we consider two IVs (Jiraporn, Kim, and Kim, 2011; Liu, Wei, and Xie, 2014; Prommin, Jumreornvong, and Jiraporn, 2014).
First, we use a dummy variable that is equal to 1 for the years after 2003 and 0 otherwise (the CG reform after 2003)
Second, we use the average CGQ of all the firms in firm i’s industry (excluding firm i’s score).
The use of CG reforms as an exogenous variable is based on the assumption that in the post reforms period (2004 to 2013), firm-level CGQ should be higher, suggesting a strong correlation between firm-level CGQ and CG reforms. However, we find no reason to believe that CG reforms have a direct linkage with firm-level stock liquidity.
The intuition behind using industry-average CGQ as an exogenous variable is that although firm-level liquidity may affect firm-level CGQ, it is less likely to affect industry-level CGQ. Further to this, managers may influence governance choices of their own firm, but should have little or no influence on the governance choices of other firms.

23. Data and methodology Research methodology Endogeneity
The third alternative specification includes a lagged liquidity into main regression and estimates the augmented regression using the Arellano and Bover (1995) and Blundell and Bond (1998) dynamic two-step system GMM.
In GMM, first-differenced variables are used as instruments for the equations in levels and the estimates are robust to unobserved heterogeneity, reverse causality, and dynamic endogeneity (if any).
Compared to the 2SLS method, dynamic GMM has at least two benefits. First, internally generated instruments. Second, it explicitly models the dynamic nature of the relationship.
The consistency of GMM estimation depends on two important conditions. The first is the serial independence of the residuals. The second condition is the validity of instruments, which is tested through Hansen J-statistics of over-identifying restrictions.

24. Descriptive statistics

25. Descriptive statistics

26. Correlation and T-test

27. CGQ and stock liquidity (H1 and H2) Baseline results: Pooled OLS regression
Time-series and cross sectional variation
CGQ improves stock liquidity.
Unaffected by the liquidity dimensions (trading cost, price impact, and trading frequency)
unaffected by the use of high frequency (TWQS) or low frequency (e.g., ILLIQ) liquidity proxies.

28. CGQ and stock liquidity (H1 and H2) Economic significance

29. CGQ and stock liquidity (H1 and H2) Baseline results: Fixed Effects and Between Effects
Time-series variation and time in varying omitted variable bias
The existing literature on CGQ and stock liquidity (Chung et al., 2010; Prommin et al., 2014) employ fixed effect regression in the short time-series (n=4).
However, FE may not be suitable for relatively short time series (Baltagi, 2005: p 13).
Cross-sectional variation
These results are in contrast to Prommin et al. (2014) as they do not find significant linkage between CGQ and stock liquidity in a cross-sectional setting.
We argue that their sample is based on large 100 Thai firms and most of the large firms may have similar corporate governance and similar firm characteristics and thus may not produce cross-sectional variation in stock liquidity.

30. CGQ and stock liquidity (H1 and H2) Endogeneity bias: Lagged Independent variables
These results not only provide an additional support to our principal results but also suggest that CGQ has an ability to predict stock liquidity–a high level in CGQ “current year” leads to a greater stock liquidity “next year”
Why endogeneity is an issue?

31. CGQ and stock liquidity (H1 and H2) Endogeneity bias: 2SLS and GMM
Instrument selection?

32. Governance categories and Stock liquidity (H3 and H4)
All the governance categories are significantly related to the three dimensions (trading cost, price impact and immediacy) of stock liquidity.
Large coefficient of board quality than those of the other three categories
This is consistent with the notion that board quality (proportion of independent directors, CEO duality and board meetings) is directly related to management control and thus prevent them to distort information.
Overall, the evidence suggests the relationship between CGQ and stock liquidity is not driven by few governance categories.

33. Individual governance variables and Stock liquidity (H3 and H4)

34. How does CGQ affects stock liquidity
Since we argue that CGQ could improve stock liquidity by improving firm disclosure, we examine whether (1) CGQ is associated with a higher disclosure, and (2) a higher level of disclosure improves stock liquidity.
We use document count as a proxy of disclosure level. The document count is the natural logarithm of the total of all documents (TD), or those classified by the ASX as price sensitive (PS) or as non-price sensitive (NPS).
OC index?????

35. How does CGQ affects stock liquidity
We use a split sample regression to check if CGQ has a stronger effect on stock liquidity for the firms that disclose more information.
We split the sample into firms with high disclosure and firms with low disclosure using TD
We classify a firm in a low disclosure category if TD is in the 25th percentile (Q1) and in a high disclosure category if TD is in the 75th percentile (Q4)
OC index?????

36. How does CGQ affects stock liquidity
A stream of literature (e.g., Edmans, Fang, and Zur, 2013) provides evidence on the reverse effect of stock liquidity on governance.
Although such literature measures governance through ownership variables (blockholders) that differ from our governance proxy (board and sub-committees), we use the three-stage least squares (3SLS) method to eliminate the endogeneity problem from simultaneity bias (if any) among CGQ, disclosure and stock liquidity.
We endogenize CGQ and disclosure, given existing literature (Monem, 2013; Haß, Vergauwe, and Zhang, 2014; Beekes, Brown, and Zhang, 2015) on their determinants.
OC index?????

37. Additional analysis: CGQ, ownership and stock liquidity
Ownership concentration may negatively related to stock liquidity.
Large shareholders have an access to private, value relevant information and may trade on this information to extract the private benefits and thus increase adverse selection costs and reduce stock liquidity (Heflin and Shaw, 2000).
High ownership concentration is inversely related to trading volume and continuity of order flow because large shareholders trade with lower frequency that leads to wider spreads and lower depth (Kothare, 1997)
Since ownership concentration permits close monitoring of a firm’s management, it may reduce the demand for alternative monitoring mechanisms (e.g., proportion of independent directors and separation of CEO from board chair)
OC index= top 20 shareholdings + substantial shareholders + CEO shareholdings + directors with substantial shareholdings
OC index?????

38. Conclusion
By using a use large cross sectional data (1,207 firms) over the period from 2001 to 2013 we find better governance improves stock liquidity in Australia.
These findings are robust to
different dimensions of stock liquidity,
Trading cost
Price impact
Immediacy
alternative estimation methods
pooled OLS (within and across)
Fixed effects (within firm)
Between estimators (across firms)
different sample specifications
GFC period
balanced data
size effect
endogeneity bias.
Lagged independent variables
2SLS
GMM
Governance categories and individual governance variables
Disclosure as a channel
Ownership concentration

39. Thank you