How Does Corporate Governance Affect Firm Behavior? Panel Data

How Does Corporate Governance
Affect Firm Behavior?
Panel Data versus Shock-Based Methods
by Bernard Black, Woochan Kim, and Julia Nasev
Comments by
Stijn Claessens, Federal Reserve Board,
CEPR and University of Amsterdam
The 5th INTERNATIONAL CONFERENCE ON
CORPORATE GOVERNANCE IN EMERGING MARKETS
HHL Center for Corporate Governance
Leipzig, September 25-26, 2013
Disclaimer!
This presentation represents my own
views and not necessarily those of the
Federal Reserve Board of Governors or
its staff.
General motivation of the study
• Whether CG practices (CGPs) and rules (CGRs)
affect firms subject to debate, empirical question
– E.g., do CEO/Chairman split, RPT, P4S, IDs matter?
– Yes: GIM (2003), Bebchuk et al. (2004), others; But…
• Results can (greatly) vary by econometric used..
– Much work suffers from endogeneity => “associations”
– (Too) few analyses focus on causality, using “events”
• Study looks at robustness of results to methods
– Impact of various econometric designs. Finds:
– Effects are not (so) robust, even for event!!
What the paper exactly does
• Study detailed effects of new CG rules (CGR)
– Korea required in 1999 large firms to have at least 50%
outside directors, an audit committee and an outside
director nominating committee (with ≥ 50% ODs)
– Q: What are effects of these CGRs on reporting quality,
accruals, investments, growth (know market value went
up)? What is causal and what are associations?
• Study associations at event with outcomes using:
– Multiple techniques: classic, simple + combined causal
• Conclusion and Lessons
– Results fall away when applying causal methods
– Classic and simple causal methods not suffice to infer!
How it extends the current literature
• “Clean” event, with sharp size cutoff
– Large vs. small firms
• Good index of CGPs (“KCGI”)
– Detailed CG (sub-)indicators, long period
• Much more careful econometrically, using
numerous tests, of three types:
– Simple: Pooled OLS, firm Fixed & Random E
– Simpler Causal: DiD, regression discontinuity
(RD), IV, year-by-year DiD
– Combined Causal: combining DiD, IV and
year-by-year DiD with RD
My praise and relevance of study
• Very careful work, hard to disagree with any econometrics!
– Very rigorous, robust regression results
– Clear exogenous event (hard to reason with it)
• Relevant: relationship CGRs↔outcomes is of much interest
– Like to know best type of CGRs to encourage best type of CGPs
– While many studies on impact of CGPs and CGR, findings do vary
• Some (thus) question effects of (specific) CGPs and CGRs
– Some question appropriateness of specific CGRs, e.g.:
• Debate on excessive regulation, post Enron/SOX
• Or, today, on too little CGR for financial industries?
– EMs specifically question whether global model best for them
– Generally, questions on appropriate CGPs/CGRs given country, etc.
• In parts as CGPs interact with CGRs, other institutions
Three Sets of Comments
1. What should be on the LHS?
•
•
Are these right measures? Are disclosure,
accruals, investment, growth etc. best/right?
Why is value not enough given reforms?
Does matter for causality tests, and channels
2. What thus to do given what you are after?
•
Related tests to find firm reforms that lead to
better minority rights and valuation ↑
3. What more can be done?
• Hard to suggest much more, but nevertheless.
• Some minor comments
1. What should be on the LHS?
• What did specific reforms/CGRs aim at?
– Arguably: “Protecting minority rights,” through
less expropriation, less RPTs, etc. Value ↑
• Reforms are (indirect) ways to this goal
– Goal is hard, in Korea, given the many forms
through which value gets tunneled out of firms
• Is thus value not the best perspective, as a
measure for minority shareholders?
• And value seems to rise with these CGRs!
– Clean event test in Black and Kim (2012)
1. What should be on the LHS?
• If valuation is a good summary, given
objective of specific reforms adopted, why
and what more do we need to analyze?
• To find channels. But… many actions of
management, controlling owners to test
• Since CGRs are esp. meant for minority
rights, some matter more and some less?
– Disclosure, inv., accruals: less directly so?
– RPT, tunneling: more so. And BKJP say: Yes
2.What thus to do,
given what you are after?
1. Redo the valuation test of Black and Kim
–
Then tests whether various econometric tests lead to
different conclusions for valuation effects
2. Study more actions, expropr., RPTs, etc.
– Since these are closer related to adopted CGRs
– Could include ownership changes, M&As
– As well as qualitative data, e.g., news report on
expropriation, minority rights violations
– Again test whether various econometric tests lead to
different conclusions
2.What thus to do,
given what you are after?
3. Study interactions of CGPs with firms
– Acknowledging again econometric testing challenges
• H0: “CGRs matter differently for certain firms as
they react differently to (lack of) minority rights”
• Expect, given firm differences, variations:
– E.g., effects of RPTs should vary between closely-held
and wide-held firms or ownership concentration.
– Or disclosure to vary with stock exchange
listing/enforcement
– Also specific sub-indexes tell more on effects of event
3. What more can be done?
1. Try “general to specific” (Hendry)
2. Do simulations to “test” econometrics
– Study if true model is “X is F(CGRs)” and
then whether econometrics will confirm it
3. Take endogeneity explicitly into account!
– Methodology → Build and test model of CGP
and their effects, given CGP‘ costs, covering
various outcomes (value, inv., performance,
profitability, minority rights, social, etc.)
– Allows system of variables/data to be used
3. What more can be done?
• Minor comments
– Like to see the details of how KCGI made
• Do not refer to other paper
– Raw stats of some variables look funny
• BTM (Tobin’s Q?) up to 35000
– Outliers may be an issue
• e.g. abs(AA) can be 44
– Tell more on # of obs.
• Do not know what is observations large vs. small