Presentation on theme: "Stock Market Liberalization and the Information Environment by Kee Hong Bae, Warren Bailey, Connie X Mao Discussed by: Campbell R. Harvey Duke University."— Presentation transcript:
Stock Market Liberalization and the Information Environment by Kee Hong Bae, Warren Bailey, Connie X Mao Discussed by: Campbell R. Harvey Duke University and NBER
Introduction Purpose of paper: Study association between equity market openness and information environment for emerging market firms Findings: Information measures increase with openness.
Proxies for Openness Proxies for Openness: Discrete Events: –Official liberalization announcement –Cross-listing or country fund in U.S./U.K. –Breakpoint in U.S. portfolio flow to country Gradual Measures: –Fraction of stocks available to foreign investors –Gross U.S. portfolio flow to country
Proxies for Information Proxies for Information: Firm specific volatility Earnings related measures: –(1) Number of analysts (resources) –(2) Absolute time series forecast error (naïve forecast precision) –(3) Absolute analyst forecast error (analyst precision) –(4) Analyst information advantage=|(3)-(2)| (analyst valued added) –(5) Forecast dispersion (uncertainty)
Hypotheses Openness increases H1: Number of analysts increases H2: Firm specific volatility increases H3: Reported earnings volatility increases H4: Analyst information advantage improves H5a: Earnings forecast dispersion decreases H5b: Earnings forecast dispersion increases
What we know Bekaert, Harvey and Lumsdaine (JFE 2002) “Dating Integration of World Equity Markets”
What we know Increased turnover and value traded imply: Better information environment
What we know Bekaert and Harvey (JF 2000) “Foreign Speculators”
What we know Cross-sectional standard deviation increases: Implying less cross-correlation and More idiosyncratic volatility Hence, contribution of the really focuses on the accounting based measures of information.
Suggestions 1.Econometric Example regression would have firm specific volatility on LHS and other stuff on RHS. No correction for autocorrelation and heteroskedasticity – and the usual techniques do not apply. Need to correct for firm or country-specific autocorrelation and heteroskedacity. Correction detailed in Bekaert, Harvey and Lundblad (JDE 2001)
Suggestions 2. Inference Regression evidence details changes in key variables after liberalization. Even with the correct BHL (2001) standard errors, there are issues with small sample properties Our experience suggests that it is best to do a Monte Carlo experiment with random liberalization dates to get the empirical cutoff for the t-ratios.
Suggestions 3. Specification Regression uses dummy variables, one year before and one year after. Existing evidence suggests one year after is not enough time to measure impact of liberalization.
Suggestions 4. Omitted variables Regression uses fixed and time effects. This should take care of omitted variables. However, it is often more economically interesting to specify a set of control variables.
Suggestions 4. Omitted variables Bhattacharya, Daouk, Welker (AR 2003) They dig deeper into the accounting number: ‘earnings aggressiveness’, ‘earning smoothing’, ‘overall opacity’ and ‘loss avoidance’
Suggestions 5. Heterogeneity The impact of liberalization differs across countries. However, the effect is forced to be the same across countries in all but the Korean example in Table 8.
Suggestions 5. Heterogeneity Break liberalization dummy into two pieces: one for countries with above average value of characteristic and the other for below average values.
Suggestions 5. Heterogeneity Candidate characteristics: Financial development indicators –Private credit/GDP –Market size Quality of institutions –Judicial efficiency –Speed of process –ICRG institutional quality Risks present in country –Conflict –Economic environment
Suggestions 6. Endogeneity What comes first liberalization or information environment? Econometrically, difficult to deal with. Lead-lag tests in the paper indicate some bi-directionality Usual technique is instrumental variables. However, one needs to come up with an instrument that predicts liberalization but is uncorrelated with the information environment.
Conclusions I believe the results
Off-line suggestions Update BHL “Growth volatility” reference. Actually, we find that economic volatility never increases after liberalization and most of the time significantly decreases Cite Bekaert (WBER 1995) for the first use of the investibility measure. The Harvey website is now referred to as Bekaert, G. and Harvey, C.R. “A Chronology of Important Economic, Financial and Political Events in Emerging Markets” Footnote #17 says no EIV because variable on LHS and footnote #18 says you used lagged dependent variable in some of the analysis
Off-line suggestions Lagged dependent correction is problematic for many reasons, best to go with BHL Some results seem at odds with BCN (JFE 2003), in particular, their Table 5. Overall, nice work!
Tests – Discrete Openness Events 2. Table 3 – Difference in Means Panel A, Panel B Findings support: H1: Number of analysts increases H2: firm specific volatility increases H3: naïve forecast error increases H4: analyst advantage increases H5b: dispersion increases Note: Portfolio Flow Breakpoints significant both before & after
Tests – Discrete Openness Events 3. Table 4 – Regression Tests Panel A (firm-specific volatility): Openness firm-specific volatility increases. Panel B (earnings related measures): Openness measures increase. Note: De-liberalization has no effect (different from Table 3) Critique: - Why does ‘before’ & ‘after’ dummy only include obs 1 year before & after event? Why not all years before & after event? Is it to avoid other shocks unrelated to liberalization event? Include crisis dummy for Mexico 1994, Asia ? -Are std errors adjusted for heteroskedasticity and contemporaneous cross- correlation?
Summary of before & after analysis Tables 3 & 4 Ranking of importance of liberalization events: 1) Portfolio flow breakpoints, cross-listing and country fund events 2) Official liberalization announcements Critique: Compare H2 with Table 5 in BCN(2003): BCN find idiosyncratic firm volatility is not linearly related to firm investibility. Differences: BCN strip out more systematic factors – investibility, country, industry, size. BMM only accounts for local market index (and global market index but not reported). BCN uses firm level time series investibility (more comparable to Table 5 BBM). BBM use country level before & after dummies.
Tests – Gradual measures of openness Table 5 – Investibility, Portfolio flows Panel A: (Investibility), find support for : H3: absolute time series forecast error increases H4: analyst advantage increases H5b: forecast dispersion increases H2: firm volatility - insignificant, agrees with BCN (2003), Panel B: (Portfolio Flows), find support for: H1: Number of analysts increases H3: Absolute time series forecast error increases H4: analyst advantage increases H5b: Forecast dispersion increases Overall note: lagged investibility important for firm volatility. Contemporaneous investibility important for earnings related variables takes time for attention to spread from specific earnings events to general coverage. Critique: Use firm-country-year to increase sample size & adjust std errors for contemporaneous correlation across firms within each country?
Local vs foreign analysts Table 6 Findings: Number of foreign analysts increases by more than local analysts with openness Foreign analyst advantage increases relative to local Forecast dispersion increases more for foreign analysts Forecast error increases more for foreign analysts (mixed evidence) BMM concludes: “openness increases amount of skilled foreign resources applied to local market”. Critique: - what’s defn of ‘foreign’ - foreign firm or analyst of foreign origin? Alternative explanation: foreign firms attract experienced skilful local analysts with high pay. Local firms hire new analysts with less skill. Results reflect movement of skilful analysts from local to foreign firms rather than increased resources. (partly based on anecdotal evidence from Malaysia)
Old vs New analysts Table 7 Findings: Old analyst advantage don’t change after openness More forecasts per earnings event after openness BMM conclude: openness analysts work harder Critique: Related issue to previous Table 6. Does ‘old’ analyst refer to firms or to individual analysts? My guess is it refers to firms reporting to IBES. Individual analysts could be different while firm name remains the same.
Case study: Korea Table 8 – Firm level analysis Panel A (Investibility): - Consistent with country level analysis - H2: openness leads to increased firm volatility (different from BCN(2003)) - Interactive terms (chaebol dummy*openness): strongly negative Panel B (Portfolio Flows): - Results similar to Panel A. BMM conclude: Interaction among openness, information and corporate governance. Firms with poor governance attract more attention (more analysts) but experience less improvement in info environment with openness. Critique: Sign on Market Cap coefficient for firm specific volatility is opposite to sign in Table 5 for country level analysis. Why? Are std errors adjusted for contemporaneous correlation across firms for each year? Failure to do this could bias t-statistics upwards
Openness information ? Table 9 – VAR Causality Test Findings: Portfolio Flow Firm specific volatility Portfolio flow Number of analysts Investibility Forecast error Investibility Forecast dispersion Other pairs show insignificant results BMM conclude: (weak) evidence in support of openness leading improved information environment