On the Pricing of Investable Securities and the Role of ... · PDF fileOn the Pricing of...
Transcript of On the Pricing of Investable Securities and the Role of ... · PDF fileOn the Pricing of...
On the Pricing of Investable Securities
and
the Role of Implicit Barriers
Abstract
Academics and practitioners implicitly assume that investable emerging market securities
are priced in the global context. However the removal of explicit barriers does not necessarily result in increased market integration if implicit barriers are important. To test this proposition, we use the conditional version of the Errunza and Losq (1985) model to estimate pricing of investable indices from twenty two emerging markets. Our results show that reduction in explicit barriers in conjunction with market liberalization does not lead to global pricing. The evidence suggests that across our sample of emerging markets the state and corporate governance together with the information environment plays a major role in globalization. • Key words: International Asset Pricing, Emerging Markets, Segmentation,
Liberalization. • JEL classification: G15, F30, G30.
2
1. Introduction
In the last two decades, developing countries have embarked on major programs aimed at
liberalizing their financial markets. Countries have proceeded toward this goal by
gradually lifting foreign investment restrictions and as a result, many of the stocks traded
domestically have become eligible for foreign investment. Eligibility per se does not
guarantee that institutional investors will be enticed into holding the security. What
matter is investability. For example CalPERS periodically evaluates foreign markets with
the goal of creating a list of permissible markets for their investment strategy.1 Therefore
this paper deals with investable securities.
Throughout the liberalization process stocks available to foreign investors have
traded alongside securities with ownership restrictions. Under these conditions, asset-
pricing theory suggests that assets available to both foreign and domestic investors should
be priced globally and local risks should not matter. On the other hand, non-investable
assets should command both the global and local risk premia.2 Yet, we still do not have
empirical evidence on the pricing of investable indices in emerging markets.3 Both
academics and practitioners assume that this subset of emerging market equities is priced
in the global context. Indeed, understanding whether these securities behave like a
separate asset class from the rest of the international securities is important in light of the
increasing focus of institutional investors and portfolio managers on global asset
allocation.
What characterizes investable indices? First, these indices are designed to reflect
the perspective of foreign investors as they take into consideration foreign investment
restrictions either at the national level and/or by the individual company's corporate
statute. Second, they span time periods of financial liberalization in emerging markets,
1 An article of the Los Angeles Times of March 15, 2004, CalPERS Clout Carries Overseas, reports on the importance of this list of investment-worthy emerging markets as the Ambassador of Philippines was going to plea the case for his country in front of the directors of the largest US pension fund. 2 See for example, Solnik (1974), Stulz (1981a,b), Errunza and Losq (1985), and Chaieb and Errunza (2007). 3 However, there is ample evidence that demonstrates the relevance of local risk factors for the pricing of market-wide EM indices, see for example, Errunza, Losq and Padmanabhan (1992), Harvey (1995), Bekaert and Harvey (1995, 1997), Carrieri, Errunza and Hogan (2007). Hou, Karolyi and Cho (2006) examine the importance of multi-factor risk models at the individual security level.
3
increasing foreign investment and introduction of a large number of cross-listings and
country funds. Third, there is strong evidence that increased availability of a country's
equity to foreigners is associated with an increase in the investor base (see Foerster and
Karolyi, 1999, and Kaniel, Li and Starks, 2005) and substantial changes in the
information environment which might attract foreign investors to emerging markets (see
Bae, Bailey and Mao, 2006).
While these facts would likely impact the pricing and degree of integration of the
investable indices, we lack clear evidence as to their pricing. Specifically, whether they
are priced globally or is the local factor still important. Indeed, there are reasons to
support the claim that investable indices might still be exposed to local factors. In some
markets, domestic assets are mainly held locally even though they could be traded
without restrictions by both domestic and foreign investors. While these indices include
securities that are accessible to foreign investors, they do not reflect the actual percentage
of stock effectively held by foreigners. As pointed out by Carrieri, Errunza and Hogan
(henceforth CEH (2007)), "the mere existence of barriers does not necessarily imply
market segmentation just as their removal does not necessarily result in increased market
integration."4
It is also the case that the investable indices account for the removal of legal
barriers but they ignore implicit barriers such as state of the local market, political risk,
availability of timely and quality information, investor protection, market regulation etc.
as suggested by Errunza (1977), Errunza and Losq (1987), Bekaert (1995) and Stulz
(2005). Country funds (CFs) may help to circumvent some of these barriers but they are
not perfect substitutes for the local securities.5 Therefore, market liberalization may not
result in global pricing of investable assets.
Investable portfolios have been the object of a few studies. By investigating the
cross-sectional relation between a stock's investability and its return volatility, Bae, Chan
and Ng (2004) show that the return volatility of highly investable emerging market
portfolios is increased due to greater exposure to world risk factors. Furthermore, Chari 4 As reported by Nishiotis (2004), indirect barriers can lead to segmentation even in the absence of strong capital flow restrictions. 5 Bekaert and Urias (1996) show that CFs do not span the corresponding IFC investable indices. Recently, Ammer et al. (2006) report that U.S. investors acquire investable shares directly in the firm's home market rather than securities like CFs that are traded in their domestic markets.
4
and Henry (2004) show that there is a difference in the world market betas of investable
and non-investable firms. Based on indirect evidence, these studies support the argument
that the investable indices are more integrated with the world than the non-investable
indices. However, the question remains whether these securities are still exposed to local
risk factors and the reasons for their lack of globalization.
If the investable securities are effectively integrated into the global market, then
theoretical asset pricing models suggest that only the global risk factors should be priced.
However if reducing explicit barriers in conjunction with market liberalization does not
lead to global pricing of investable securities, then the limits of globalization are likely
due to implicit barriers. Hence, this paper poses three key questions. First, are investable
indices fully integrated with the world market, i.e. are they globally or locally priced.
Second, if local factors are important, what is the extent of the departure from full
integration. And third, can we relate this departure to measures of implicit barriers.
We first estimate a conditional version of the Errunza and Losq (1985, henceforth
E-L) model for International Finance Corporation Investable (IFCI) indices for 22
emerging markets. We find evidence that exposure to country-specific risk factor is
rewarded. The price of the local risk factor is also statistically time-varying for many
markets. Since removing explicit barriers in conjunction with market liberalization does
not lead to global pricing of investable indices, we then investigate the role of implicit
barriers. We derive two measures of integration from our estimated model. We use the
Integration Index as in CEH (2007) and the ratio of Global to Total premium to capture
the extent of departure from full integration. We then relate these measures to a country's
state governance, corporate governance and information environment, after controlling
for factors that have been reported in the literature as significant drivers of market
integration. The evidence shows that implicit barriers matter as we find that our proxies
are significantly related to the degree of integration. The evidence is more robust using
the Integration Index and is confirmed with the ratio of Global to Total premium.
The rest of the paper is organized as follows. Section 2 discusses investability and
the available benchmarks. Section 3 presents the model and the empirical methodology.
Section 4 describes the data. Section 5 presents empirical results regarding global versus
5
local risk pricing and the integration measures. Section 6 investigates the impact of
implicit barriers on globalization. Conclusion follows.
2. Investability and Flows
Traditionally, market liberalization refers to removal of explicit barriers to foreign
portfolio investments such as inflow/outflow controls, restrictions on foreign exchange
transactions and taxes on investment proceeds. Markets that open their doors to foreign
investors through removal of these barriers are technically “eligible” for investments.
However, as Errunza and Losq (1985) first pointed out, investors are reluctant to invest in
some of these markets because, although technically open, they are perceived to be non-
investable. As a result, investment capital does not flow to such markets.
On the other hand, “investability” refers to the ability of foreign investors to
access markets and securities, i.e. the ease with which foreign institutional investors can
buy or sell securities and repatriate proceeds. It should include considerations of
openness (limits on foreign holdings), liquidity, size and float at the market and
individual security level. Since neither the locally available performance indicators nor
the initially available indices such as the IFC Global and the MSCI Emerging Market
Global were designed from this perspective, a number of so called “Investable” indices
were developed in 1990s by IFC, MSCI, and ING Barings.
The investable indices are thus designed to measure returns that foreign investors
would receive from investing in domestic stocks that are legally and practically available
for foreign investment. For example, the S&P/IFC first creates a variable called the
foreign investment limit (the degree open factor) with values ranging from zero to one.
Zero indicates that none of the stock is legally and practically investable; 1 indicates that
100% of the security's market capitalization is available for foreign ownership. S&P/IFC
determines stock's investability based on several criteria. It first determines whether the
market is open to foreign institutions with regards to the extent to which foreign
institutions can buy or sell shares on local exchanges and repatriate capital. It then
investigates whether there are any corporate by-laws, corporate charters, or industry
limitations on foreign ownership of the stock. It finally applies two further screening
6
criteria: size (at least $50 million in investable market cap) and liquidity (at least $20
million in annual trading).
In this paper, we use the IFC data.6 From the degree open factor of the IFC
database, we have detailed information on the amount of a company’s market
capitalization that a foreign entity may hold. As a result, for the IFC Investable indices,
we have information on the amount of total market capitalization that is available for
foreign ownership at the country level. At the same time, through the IFC Global indices,
we also have information on the total market capitalization of the country, including the
portion that is neither open nor practically available for foreign investment. The
combination of these two indices provides interesting information.7 We construct a
measure of investability from the ratio of market capitalization of the investable indices
(IFCI) over the global (total market) indices (IFCG).8 This measure lies between zero,
when no portion of the market capitalization is investable, and one, when the whole
market is investable. Figure 1 provides some information on the evolution of this
investability measure, MCI/MCG. We only show plots for those countries in our sample
with the earliest start date. The graphs show different patterns across countries and
regions. The ownership restrictions are lower for the Latin American countries.
Furthermore, the liberalization in Latin America occurred earlier than for the Asian
markets. In Argentina, most of the market cap had been available to foreign investors
since the official liberalization of the market in 1989. Brazil and then Mexico also
removed all of their ownership restrictions by 1990 and 1991 respectively. Over the
recent period, almost 100% of the MC of the Mexican market could have been fully
traded by foreigners. As for Chile, the country instituted higher ownership restrictions in
6 Discussions with S&P Index Services personnel suggest that at the end of July 2008 over 150 financial institutions were subscribing to their S&P/IFCI (or S&P/IFC) emerging markets database and estimate that over $65 billion were benchmarked to their indices. 7 MSCI used to provide non-free and free indices for some emerging markets. The non-free indices were discontinued on November 2001. Currently, MSCI/Barra only reports the free indices for the 25 emerging markets covered. Hence, all MSCI emerging market equity indices are fully investable by foreign investors and there is no information on the non-investable portion. In the data section we provide some more detailed information on both types of indices and compare the behavior of returns. 8 The ratio of the market capitalizations of an EM's IFCI and IFCG indices has also been used by Edison and Warnock (2003) and De Jong and De Roon (2005).
7
the early 1990s. By 1996, the MCI/MCG had increased dramatically from 25% to 100%.9
Korea also experienced a rapid increase in the measure in the middle of the Nineties,
reaching within two years almost full investability. On the other hand, India and Thailand
are still not fully investable by the end of 2006.
In figure 1 we also plot (against the right axis) a measure of transaction flows as a
percent of total market capitalization, flows/MCG. As numerator for this measure we use
the sum of purchases and sales between U.S. residents and counterparties located outside
the U.S. involving stocks of that specific country.10 Ideally we would like to add up
similar statistics for transactions between investors of all the major countries dealing with
stocks of the emerging markets in our sample. However, such data is not available so we
use the data of US residents as a proxy for the level of world-wide transactions involving
all countries. This measure provides a gauge of the interest of US investors for the stocks
of that particular country. An alternative measure could be provided by data on equity
holdings, but such data does not exist at this detailed level since the start of our sample
period. 11,12
As the plots show, the level of participation of US investors in the foreign stocks
has increased from the first part of the Nineties for all the countries. 13 The increase has at
times coincided with and often lagged the increase in the investability measure.14 This
pattern is evident in all cases but especially so for Brazil, Chile, Mexico and Korea.
However, with the exception of Mexico, the level of US transactions has leveled off in
9 Edison and Warnock (2003) argue that the jump should rather be registered in January 1992 when Chile implemented the DL 600 law that covers the foreign investments. Under DL 600, profits may be repatriated immediately, but none of the original capital may be repatriated for one year. However, the IFC included this law four years later. 10 The data is from Treasury International Capital System and it is collected from mandatory reports filed by banks, securities dealers, investors and other US entities dealing directly with foreign residents. 11 Treasury Capital International System also provides data on US holdings of foreign long-term securities. However these data are available on an annual basis only for 1994, 1997, 2001 and then with annual frequency from 2003. 12 With a limited dataset, Portes and Rey (2005) show that cross-border transaction flows in equity have informational pattern similar to equity holdings. 13 For the other countries in our sample, the official liberalization date basically coincides with the start of data availability for the investable indices, thus the corresponding investability measure does not show any interesting evolution. The level of investor’s transactions also shows a pattern similar to that of the countries we report. 14 Bacchetta and van Wincoop (2000) develop a model where liberalizations produce non-linearities in the inflows. Edison and Warnock (2008) show that inflows increase initially but then taper off and that only credible liberalizations generate sustainable increases in inflows.
8
the last few years, after increasing toward the end of the Nineties. Thus despite the
liberalization that has caused large jumps in the level of investability of most of the
countries by the middle of the last decade, the flows have not continued to increase
during the same time span.
This analysis is simply illustrative. Nonetheless, it validates our conjecture that
there are likely other factors driving investors’ participation beyond the removal of
explicit barriers.
3. The Asset Pricing Model. Theory and Empirical implementation We implement the IAPM of Errunza and Losq (1985) which accounts for barriers to
international investment. The model assumes a two-country world and two sets of
securities. All securities traded in the domestic market (e.g. the U.S.) are eligible for
investment by all investors. Securities traded in the foreign market (e.g. the emerging
market) are ineligible and can be held only by foreign investors. Thus, domestic investors
can invest only in domestic eligible stocks, while foreign investors can invest in their
local ineligible stocks as well as domestic stocks.
The expected return on a security i that can only be held by foreign investors is
given by:
( ) ( ) ( )( ) , , |i f i W u I i IE R R AMCov R R A A M Cov R R R= + + − e (1)
where is the expected return on the ith security in the Ith market that is accessible
only to its nationals, is the risk free rate, A(A
)( iRE
fR u) is the aggregate risk aversion
coefficient for all (Ith) market investors, RW(RI) is the return on the World (Ith) market
portfolio, M(MI) is the market value of the global (Ith) market portfolio, and eR is the
vector of returns on all securities that can be bought by all investors irrespective of their
nationality. Thus, the expected return on the ith security commands a global risk
premium and a super risk premium which is proportional to the conditional market risk.
The authors show that the eligible securities are priced as if the market were fully
integrated and command a world market risk premium. The ineligible securities
9
command a country-specific premium induced by the mildly segmented market structure.
However, if a subset of the EM securities (such as the IFC Investable index) is eligible in
the sense of being fully investable with no explicit barriers, the country-specific premium
should disappear. That is, these securities would command only a world market risk
premium, similar to US securities.15
Since we examine the pricing of the IFCI indices, we express equation (1) in
terms of the IFCI index by aggregating over the investable securities traded in the
emerging market. The expected excess return on the IFCI index is then given by:
( ) ( )DPIFCIIWIFCIWIFCI rrrrrE |var,cov)( λδ += (2)
where rIFCI is the excess return on the IFCI index, rW is the excess return of the world
market portfolio and rDP is the excess return on the diversification portfolio of the IFCI
index, i.e. the portfolio of eligible securities that is most highly correlated with the IFCI
index. 16
We estimate a conditional version of equation (2) where we allow prices and
quantities of risk to change through time as suggested in recent literature (see among
others Harvey, 1991, Dumas and Solnik, 1995, De Santis and Gerard, 1997, 1998). From
the E-L model, the following system of equations has to hold at any point in time,
15 In an earlier version of this paper, we also implemented the IAPM of Chaieb and Errunza (2007) for eight major EMs. That model accounts for barriers to international investment as well as purchasing power deviations and includes additional global and local risk premiums. However, its estimation is more difficult without commensurate additional benefits for our research question. The results are qualitatively similar to those reported here and available from the authors. 16 Note that the increasing convergence between IFCI and IFCG has led to the marginalization of the non-investable segment of most EMs and has contributed to the decision by both the S&P and MSCI to continue with only the investables indices going forward. Hence, we use IFCI as a proxy for the EM index. If we were to use the IFCG index as a proxy for the broader EM market, the pricing equation would be,
( ) ( )etIFCGtIFCIItWtIFCIWtIFCI rrrrrrE |,cov,cov)( ,,,,, λδ += which reduces to eq. (2) under the assumption
( , ,cov , | 0IFCI t IFCNI t er r r ≈) where rIFCNI is the excess return on the IFC non-investable index.
10
( ) ( ) ( )tDPtIFCIttItWtIFCIttWtIFCIt rrrrrE ,,11,,,11,,1 |var,cov −−−−− += λδ
( ) ( )tWtDPttWtDPt rrrE ,,11,,1 ,cov −−− = δ (3)
( ) ( )tWttWtWt rrE ,11,,1 var −−− = δ
where rIFCI is the excess return on the IFCI index, ,DP tr is the excess return on the IFCI’s
diversification portfolio, is the excess return on the world index, δtWr , W and λI are
respectively the world and the local price of risk. To keep the dimensionality of the
model reasonable, we test the model using one country at a time. Although such an
approach implies that power is lost since the procedure does not impose the equality of
global price of market risk across countries, it yields efficient estimates and permits
analysis of the contribution of each premium to the total premium.17
We further express [ ] )1)(var(|var ,,2
,,,1 tDPIFCItIFCItDPtIFCIt rrr ρ−=− where
is the correlation coefficient between the diversification portfolio and the IFCI
index return. Hence, for each country, we estimate the following system of equations,
tDPIFCI ,,2ρ
2
, ,, , 1 , , , 1 ,
, ,
(1 )IFCI DP t,IFCI t W t IFCI W t I t IFCI t IFCI t
IFCI t DP t
hr h h
h hδ λ− −= + − +ε
,
, , 1 , ,DP t W t DP W t DP tr hδ ε−= + (4)
tWtWtWtW hr ,,1,, εδ += −
where are the elements of , the 3x3 conditional covariance matrix of the assets in
the system. In particular,
tjh , tH
( )( )2,var 1IFCI IFCI DPr ρ− is parameterized as
( )2
, ,,
, ,
(1 )IFCI DP tIFCI t
IFCI t DP t
hh
h h− with , the time-varying covariance, and , ,IFCI DP th ,IFCI th ,DP th , the
time-varying variances.
17 An alternative approach would entail two steps estimation, similar to Bekaert and Harvey (1995) or CEH (2007). However, the two-step procedure does not allow us to analyze the simultaneous pricing of global and local risk, which is one of the goals of our paper.
11
Given that the model implies the price of global and conditional market risk must
be positive, we use a square function to model their dynamics as follows,18
1, −tWδ = (kW’ZW,t-1)2
1−Itλ = (kI’ZI,t-1)2
The dynamics of the conditional second moment Ht are specified following De
Santis and Gerard (1997) as,
1'
110 '*'*)'''(* −−− ++−−= tttt HbbaabbaaiiHH εε (7)
where i is a (3x1) vector of ones, a and b are (3x1) vectors of unknown parameters and *
denotes the Hadamard (element by element) matrix product.
We estimate the model by Maximum Likelihood assuming a normal conditional
density, and we use the quasi-maximum likelihood estimate (QMLE) of Bollerslev and
Wooldridge (1992). The estimation is performed using the BFGS (Shanno, 1985)
algorithm for updating the Hessian.
4. Data for the estimation of the asset pricing model
The estimation of the asset pricing model requires three groups of data: 1) Returns data
on the Morgan Stanley Capital International (MSCI) World index, MSCI Free indices,
and IFCI indices, 2) the eligible securities used to construct the diversification portfolios;
and 3) the instrumental variables including global and local variables.
1) Returns data on the MSCI and IFCI indices:
We include all emerging markets that have an investable index with returns data that
starts no later than 1994 to have enough observations and degrees of freedom for the
asset pricing estimation. We thus include Argentina, Brazil, Chile, China, Colombia,
Czech Republic, Hungary, India, Indonesia, Israel, Jordan, Korea, Malaysia, Mexico,
18 We also use an exponential form. Although we qualitatively obtain similar results, the point estimates with the exponential specification are extremely volatile, hence we only report results with the square specification.
12
Pakistan, Peru, Philippines, Poland, South Africa, Taiwan, Thailand, and Turkey, or 22
out of the 30 emerging markets with an S&P/IFCI index. Our sample also represents 22
out of the 25 MSCI global investable market indices. In November 2001, S&P/IFC
discontinued the IFCI indices of seven markets due to their small size or illiquidity.
These markets were Colombia, Venezuela, Pakistan, Jordan, Sri Lanka, the Slovak
Republic, and Zimbabwe. However, Colombia, Pakistan, and Jordan were still considered
investable by MSCI/Barra. Moreover, the returns on the S&P/IFCI indices for Israel
starts in 1997, while the starting date in MSCI/Barra is January 1993. Thus we use the
IFCI indices from the S&P/IFC Emerging Markets Database for all countries except
Colombia, Israel, Jordan, and Pakistan. For these countries, we use MSCI EM Free
indices.
All returns are monthly, dividend-inclusive, denominated in USD, and in excess
of the one-month Eurodollar deposit rate. Depending on the country, the sample period is
from January 1989 or later to December 2006. Table 1 summarizes the basic
characteristics of the IFCI and MSCI Free indices for those countries that are included in
this study. Although each vendor uses a different hierarchical process in constructing
their indices, their return behavior is very similar.19 Over time, each index has undergone
modifications to satisfy investor demand and for competitive reasons. Nonetheless, as
reported in Table 1, IFCI and MSCI Free indices continue to be highly correlated over the
period 1989-2006. The returns indices have greater than 96% correlation and the mean
differences, volatility differences, and tracking errors are small. Most of the countries
indicate a tracking error below 3%. The only countries where larger differences occur are
Argentina and China, where the correlation between the IFCI index and the MSCI EMF
index is respectively, 87% and 84%, and the tracking error is respectively, 9% and 6%.
For Argentina, much of the tracking error is due to the earlier part of the sample. As we
redo the comparison from February 1990, the tracking error drops from 9% to 2% and the
correlation increases from 87% to 99%.
Panel A of Table 2 provides some basic statistics on the composition of the IFCI
indices. The Table also reports some information on the market-wide indices (IFCG) that
19 See Bekaert et. al. (1997) for a detailed discussion and a comparison of the indices until 1996. Note that MSCI revised their indices in 2007. The new indices are called MSCI Global Investable Market Indices.
13
provide additional insights in combination with the investable portion. As of December
2006, at least half of the stocks in the IFCG indices are also included in the IFCI indices
for all countries. The number of stocks included in each IFCI index varies from 6 stocks
for Czech Republic to 242 stocks for Korea. These numbers range from 6 for the Czech
Republic global index to 411 for China global index. Over the same period, more than
half of the market capitalization of the IFCG indices is investable for all countries. The
market capitalization of the IFCG indices also indicates a substantial cross-sectional
difference in the size of the equity market for the emerging countries of our sample.
Pakistan’s IFCG index has a market capitalization of $10 billion and is the smallest,
while China’s IFCG index has a market cap of $511 billion and is the largest. Among the
countries with an S&P/IFC investable segment, Korea shows the largest investable
market cap with approximately $488 billion, which is 95% of the Korean IFCG index,
while Czech Republic with $15 billion investable market cap is the smallest. Nonetheless
the Czech Republic investable market capitalization as of December 2006 makes 100%
of its IFCG index.
Panel A of Table 2 also reports summary return statistics of the investable indices.
There is substantial cross-sectional variation in the average returns of the IFC investable
indices. For some countries, the returns are negative due mainly to the financial crisis
experienced over the sample period. The investable indices exhibit high volatility and
substantial deviations from normality. The Bera-Jarque test of normality rejects the
hypothesis of normality in all the countries, except India, at the 95% confidence level. In
addition, there is significant autocorrelation in the return series for half of the sample of
emerging markets, as indicated by the statistics. There is also significant
autocorrelation in the squared return series for most of the emerging countries. Overall,
the return behavior is similar to that of IFCG indices reported in past studies (see e.g.
Harvey, 1995; Bekaert and Harvey, 1995, 1997).
12)(zQ
2) Eligible set and construction of the diversification portfolios:
The eligible set includes the MSCI World index, 38 global industries, 16 closed-end
country funds (CFs) and 65 cross-listings whether these are direct placements, American
14
Depository Receipts (ADRs), or Global Depository Receipts (GDRs). 20 The stocks cross-
listed outside the US are either listed in the UK or Germany. To preserve the degrees of
freedom in the regression, we only include for each country the first incepted country
fund and the five earliest cross-listings when available.21 In general we observe that
countries from Latin America started cross-listing in mature markets earlier than
countries in Asia. Some of the countries of our sample, e.g. Israel, Mexico, Chile, Brazil,
and more recently China have a large number of cross-listings. Appendix A provides a
detailed list of the eligible set and more information on the data sources. The monthly
returns (adjusted for dividends) for CFs are obtained from the Center for Research in
Security Prices (CRSP) database. The return data on ADRs are collected from CRSP,
while return data on GDRs are compiled from Datastream.
To build the diversification portfolios, we follow CEH (2007). We first regress the
return of the country investable index on the returns of the 38 global industries along with
the MSCI World index. Using a stepwise regression procedure with forward and
backward threshold criteria, we obtain the diversification portfolio of global securities,
RG. We then regress the return of the country investable index on RG, CF, and the cross-
listings. We allow the weights assigned to previous securities to vary upon the
availability of new overseas listings as in CEH (2007). The fitted value from this
regression is the return on the diversification portfolio RDP that we use in the estimation
of system (4).
Panel B of Table 2 contains pairwise correlations between the world index,
country j investable index, and the diversification portfolio of country j. As expected, the
correlations between the diversification portfolios and the world index are higher than the
20 Data on the end of month total return on the 38 global industries are collected from Datastream that uses the FTSE industry classification. For a detailed description, see "FTSE Global Classification System", available at http://www.ftse.com. 21 Of the available securities, we only use the listings that remain active over the entire sample, with a minimum of three years of returns data, and that are relatively liquid, i.e. without many zero returns. If a company delisted pre-December 2006, or that it has many zero returns we exclude it and use the next earliest listing. However, we include all country funds regardless of whether they open-ended or liquidated pre-December 2006. We do so because country funds preceded cross-listings in almost all cases except for Philippines and South Africa, and the first country fund or cross-listing has a stronger effect in spanning the investable index than the subsequent funds or listings. In addition, since some countries have at most one cross-listing, including the country fund (whether active over the entire period or not) helps to further span the investable segment.
15
correlations between the country investable index and the world index. The return
correlation between the investable index of a country and the world market index ranges
from 0.11 for Jordan and Pakistan to 0.5 for Israel. The return correlation between the
diversification portfolio of the investable index and the world market index ranges from
0.25 for Pakistan to 0.8 for Poland.
3) Global and local instrumental variables:
For reasons of comparability, we follow previous research in selecting the data on the
global and local instrumental variables [see Ferson and Harvey (1993), Bekaert and
Harvey (1995, 1997), Dumas and Solnik (1995), De Santis and Gerard (1998) and
Carrieri, Errunza and Majerbi (2006) among others]. The global instruments include the
change in the US term premium measured by the yield difference between the 10-year T-
bond and the 3-month T-bill, the world dividend yield in excess of the one-month Euro-
dollar interest rate, and the US default premium measured by the yield difference
between Moody's Baa and Aaa rated bonds. The local instruments include the lagged
local equity market return, the local dividend yield in excess of the one-month Euro-
dollar interest rate, and the change in bilateral exchange rates $/FCj where j is the
currency of country j. Since these instrumental variables have been widely used in other
studies, we omit a detailed description of their properties. Panels C and D of Table 2
show some basic statistics as well as the pairwise correlations among the instruments.
Notice that the correlations among the information variables are small.
5. Empirical Results from the Asset Pricing Model
This section reports the results based on the empirical asset pricing model described in
Section 3 and discusses the findings on the degree of integration.
5.1. Global versus local risk pricing
Panel A of Table 3 contains the results of the joint hypothesis tests from the country-by-
country estimation of the multivariate system (4). For each country we report robust
Wald tests for the significance and time-variation in the prices of world market risk and
conditional market risk. A number of interesting findings emerge from Panel A. First, the
local risk factor (conditional market risk) is priced and time-varying for most of our
16
indices. Specifically, the price of local market risk is significant in 17 out of 22 countries
and is significantly time varying in 14 cases. In addition, the price of world market risk is
significant in all cases, while it is only significantly time-varying in 6 cases. The average
estimate across countries of the price of world market risk of 2.8 is economically
significant and is consistent with previous studies. The results overall indicate that for
most of the countries both global and local risk are significant pricing factors, although
the time-variation is not always confirmed.22
Panel B of Table 3 reports some diagnostics tests on the estimated residuals.
There is evidence that GARCH effects have been removed and the non-normality in the
data is reduced although not eliminated. Furthermore, there is no more serial correlation
in the squared standardized residuals. We also report the Engle-Ng test for asymmetry.
The Engle-Ng tests indicate that, with the exception of five cases, there is no evidence of
negative asymmetry in the residuals while there is evidence on the presence of positive
asymmetry only in four cases. Hence there is no consistent evidence of asymmetric
response of the conditional second moments to past innovations. We also report the
pseudo R-squares (R²) computed from our model.23
The analysis provides clear evidence that the local risk is still a relevant factor in
explaining time-variation of the investable return indices. These countries therefore
display characteristics that substantially differentiate them from the rest of the securities
in a global portfolio.
5.2 Degree of Integration across Investable Emerging Markets
To capture the extent of globalization we use two measures of integration that are
directly drawn from the theoretical model.
The first one is the E-L integration index (II) defined as the ratio between the
variance of the unspanned part of the investable index and its total variance. The index
lies between 0 and 1 and is defined as,
22 In the case of Mexico and Peru the estimation generates extreme values for the premia that are due to the imposition of the non-linear specification for the prices of risk. 23 For each asset, the pseudo R-squared is the ratio between the explained sum of squares and the total sum of squares. Due to the cross-equation restrictions, there is no guarantee that the pseudo R-squared are positive for all assets.
17
[ ][ ] )(
)1)((1
|1
,1
2,,,1
tIFCIt
tDPIFCItIFCIt
IFCI
DPIFCI
rVarrVar
rVarrrVar
II−
− −−=−=
ρ
Note that this index is derived from the conditional second moments estimated in the
model (4).
The second measure is the ratio of Global to Total premia (GT) of the index
returns and it is designed to capture the relative importance of global and local risks.
Specifically, the total premium is constructed as the sum of the global and the local
premia, thus
, 1 1 , ,
, 1 1 , , , 1 1 , ,
[ cov ( , )][ cov ( , )] [ var (
W t t IFCI t W t
W t t IFCI t W t I t t IFCI t DP t
abs r rGT
abs r r abs r r )]δ
δ λ− −
− − − −
=+
The sum of the two premia is computed from absolute values of the estimates of model
(4) and thus by definition GT lies between 0 and 1.
If the investable index is perfectly spanned by the eligible set, the DP is perfectly
correlated with the index, the II will be equal to 1 and markets are fully integrated. In this
case, the conditional local market risk will be 0 and thus the GT will be 1, exactly like the
integration index. In the other extreme case, when the correlation with the diversification
portfolio is 0, the conditional market risk is equivalent to the unconditional risk and the II
will be 0. In this case the local risk will be very large and thus the GT ratio will also tend
toward 0. However, in between the two extremes, the two measures are not exactly the
same since while the II places emphasis only on local market risk and on substitute
assets, the GT ratio also depends on the covariance with the world factor and prices of
risk. From the point of view of estimation, the II depends only on time-varying second
moments, while the GT also depends on prices of risk that have been shown to be highly
time-varying in previous research. This will affect the dynamics and characteristics of the
estimates of the two measures. 24
Table 4 reports summary statistics. Panel A presents the Integration Index while
Panel B contains the Global to Total ratio. In both panels, a low value indicates that the
24 Recently, Pukthuanthong-Le and Roll (2009) propose another measure of global integration, the R-squared from the regression of a country’s index returns on common global factors. As explained in the paper, our II measure and theirs would yield similar results. Please refer to their paper for a discussion of other proposed measures of integration in the literature.
18
spanning of the eligible set is limited (in the case of the II) or that the contribution of
global risk is not very large (for the GT ratio). The evidence shows that the extent of
globalization is not uniform within this large sample of Emerging Markets. Although, as
expected, the estimate of the degree of integration for each country is not the same across
the two measures, it is interesting to note that the average of both the Integration Index
and the Global to Total ratio is very similar, just above 0.5. Thus, taken together, these
countries are still far from being fully integrated, as it is also suggested by the evidence in
Table 4. The cross-sectional variation is more pronounced for the GT measure, with a
standard deviation for the pool of 0.314 versus 0.237 for the II, confirming that prices of
risk contribute to higher variation in the measure.
Based on the Integration Index, the least integrated country is Jordan with 0.04
while Israel is the most integrated at 0.820. From the Global to Total ratio, the lowest is
Turkey followed by Poland and Jordan while Israel is the sixth highest after Mexico,
Peru, Korea, Brazil, and Argentina. In addition, the ranking of the countries is quite
uniform across the two measures. For example, except for Peru, all the countries that
show a very low level for the II, also rank low based on the GT. All the countries below
the mean of the integration index, have one or no cross-listing, while all the top five
countries have five (or more). Except for Chile and India, all the countries at the bottom
of the GT distribution have one or no cross-listing, and also in this case, all the top-five
countries have five cross-listings. Thus we find a link between the two distributions based
on the number of cross-listings of the diversification portfolios included in estimation.
We can also draw interesting insights from the sub-period analysis. Contrary to
our a priori expectations, for both measures we do not observe a general increase across
subperiods. In particular, for most of the Asian countries, we observe a slowing of the
progress toward globalization, with a decrease in one or both measures. Nonetheless,
some countries such as South Korea and India, have experienced large increases in the
integration index measure. Remarkably, both measures indicate a decrease for Colombia
and Pakistan. Indeed these countries have been dropped from the sample of investable
markets of the S&P/IFC in 2001 based on small size and low liquidity. Thus our
estimates of the degree of integration confirm a reversal in the process similar to what is
documented for some countries by CEH (2007).
19
6. Implicit Barriers and Integration
As reported in the previous section, reduction in explicit barriers in conjunction with
market liberalization has not resulted in global pricing of investable indices. Moreover,
our integration measures indicate a stalling and in some cases a reversal of the process.
What are then the causes, other than explicit barriers, that might represent hindrance to
the integration process?
The E-L model provides a framework since it assumes that a subset of securities
can only be held locally as a result of a prohibitive tax. This tax on foreign assets can
arise from high information and monitoring costs to foreign investors.25 To the extent that
substitute assets that trade in the eligible markets (such as cross-listings and CFs) do not
alleviate these costs, a large fraction of the total premium would be explained by the local
risk premium and the market would be less integrated.26
Thus we expect that information asymmetry can explain the limits to financial
integration through different channels. It can be the case that information and monitoring
costs discourage foreign investors. Alternatively, high ownership by corporate insiders as
outcome of poor investor protection yields to poor disclosure and high information
asymmetry.27 Moreover, better information disclosure in an economy helps investors’
recognition and improves risk sharing and thus should be empirically related to
differences in the degree of integration. Thus, we would expect the informational,
institutional and governance environment to play a major role in the globalization
process.
25 Along the same vein Merton (1987) emphasizes the role of incomplete information and investor recognition on risk sharing and asset pricing. 26 The interpretation of the prohibitive tax as extremely high information and monitoring cost assumes away the role of intermediation in mitigating information disadvantage to foreign investors as also argued in Leuz et al. (2008) and especially makes a strong assumption about information endowments. 27 There is sizeable evidence on the role of information asymmetry in international equity markets. Many papers show that local investors have an information advantage relative to foreign investors (see for example, Kang and Stulz, 1997, Portes and Rey, 2005, Choe, Kho, and Stulz, 2005, and Dvorak, 2005), while a few papers find that foreign investors are better informed than local investors (see Seasholes, 2004). The literature points out the empirical importance of the information asymmetry in explaining international equity flows (see Brennan and Cao, 1997, and Portes, Rey and Oh, 2001), institutional investors holdings across countries (Ahearne, Griever and Warnock, 2004), and home bias (Bae, Stulz, and Tan, 2008). There is also theoretical and empirical evidence that information asymmetry is priced in international equity markets based on market liquidity and adverse selection (e.g. Chan, Menkveld, and Yong, 2008).
20
Past literature has touched on some of these issues. For example, Errunza (1977)
emphasizes the importance of implicit barriers including state of the local market,
political risk, availability of timely and quality information and market regulation for
investments in emerging markets. Bekaert (1995) empirically relates some implicit
barriers to market integration. Nishiotis (2004) provides evidence that some type of
indirect barriers, such as liquidity, credit ratings and inflation can explain the premium
and discounts of EM closed-end funds. Recently, Stulz (2005) has suggested that the
limits to financial globalization are likely due to state and corporate governance.
Specifically, he identifies the twin agency problems related to expropriation by the state
and by corporate insiders at the expense of outside investors as the primary hindrance to
financial globalization.28
Hence, this section offers extensive evidence in assessing the impact of implicit
barriers on the globalization of emerging markets. We use our two measures as dependent
variables and relate them to a number of implicit barriers.
6.1 Analysis of Implicit Barriers
We focus our analysis on three broad determinants of implicit barriers, those that
are due to the institutional environment, those that depend on corporate governance and
those related to the quality of information available to investors.29 Table 5 contains some
summary statistics for the explanatory variables. They are presented by country as
averages over the sample years. More detailed explanation of all the variables and their
sources can be found in Appendix B.
To capture the relevance of the institutional environment we use the ratings
provided by the Political Risk Services’ International Country Risk Guide (ICRG)
political risk index. These ratings are a composite of a number of elements such as
government stability, investment climate, corruption, law and order tradition,
bureaucratic quality (see e.g. Erb, Harvey, and Viskanta (1996) for detailed description of 28 For more insights on the differences in corporate governance and ownership across the world, see La Porta, Lopez-de-Silanes and Shleifer (1999) and Classens, Djankov and Lang (2000). 29 CalPERS’ list of permissible markets is based on evaluation of a number of country and market factors that are considered as important determinants for its investment strategy. These are three country factors (Political Stability, Transparency and Productive Labor Practices) and four market factors (Market Liquidity and Volatility, Market Regulations/Legal System/Investor Protection, Capital Market Openness and Settlement Proficiency/Transaction Costs).
21
the ICRG political index). All these aspects are very crucial for investors concerned about
the transparency and fairness of the political and legal institutions of a country. A high
number for this variable (POL) implies that the country scores very high with respect to
these elements, and thus indicates low political risk. For our sample of emerging markets,
the average rating is 68.
We collect two measures of investors protection. We consider the revised anti-
director rights index and the anti-self-dealing index of Djankov et al. (2008). The anti-
director index (A-DIR) varies between 0 and 6, with a higher score for those countries
that show better protection of minority shareholders.30 We interpret this variable as an
indicator of the weakness of the corporate law of a country. The average of the scores in
our sample of countries is 3.5. The anti-self-dealing index (ASD) focuses on private
enforcement mechanisms such as disclosure, approval and litigation that would protect
outsiders in the case of self-dealing transactions of insiders. This index ranges from 0 to
1. Djankov et al. (2008) report that the world average for the index is 0.44, while the
average of our pool of countries is 0.49.
Stulz (2005) also shows that the twin agency problems help explain ownership
concentration across the world. This is because "As the twin agency problems worsen,
greater ownership concentration becomes more efficient and corporate insiders must co-
invest more with other investors."31 To further address issues related to corporate
governance we collect data on ownership concentration. From Worlscope data we
construct the "closely held shares" variable (C-HELD) that is a value-weighted average
of the shares held by insiders in each country. The average fraction of closely held shares
over the period for our countries is 56 %. As a comparison, this fraction was 15.68 % for
the U.S. in 2002 (Stulz, 2005). Taiwan and Korea have the lowest value weighted
ownership concentration at respectively 29% and 35%, while Czech Republic, Pakistan,
and Turkey have the highest values at around 70%. We also collect the information
reported in La Porta et al. (2006) on the average ownership of the three largest
30 The index covers six areas, indicating if proxy by mail is allowed, shares are not blocked before a shareholder meeting, cumulative voting for directors is allowed, oppressed minorities are protected, preemptive rights at new equity issuances, and the right to call a special shareholder meeting. 31 Leuz, Lins, and Warnock (2008) and Kho, Stulz, and Warnock (2008) find that US investors invest less in poorly governed firms, that is, firms with large block ownership by insiders. In addition, Ferreira and Matos (2008) find that institutional investors hold fewer shares of firms that are closely held.
22
shareholders for the ten largest companies (OWC). For our countries, the average
ownership concentration is 0.49. Also in this case, Korea and Taiwan have the lowest
ownership concentration while Colombia, Mexico and Turkey show the highest.
To capture the transparency and the quality of information in global financial
markets we collect data related to analysts following and construct two variables. The
first variable, AN-F, is the mean number of analysts following each firm listed in I/B/E/S
in a specific year, while the second measure, AN-D, is the diffusion of analysts, or the
proportion of firms covered by I/B/E/S over the number of firms listed in the country in a
given year.32 For both of these variables, a high number indicates that there is more
private information that is divulged in the economy through the analyst channel. As
Hong, Lim and Stein (2000) argue, more analyst coverage improves the transparency of a
firm. Thus a higher number at the aggregate level should be linked to higher integration.
For our group of countries, the analysts-per-firm variable has a mean of 4.91 while the
corresponding number is 29% for the analysts-diffusion variable.
Accounting standards are another important channel for dissemination of
information in financial markets. Healy and Palepu (2001) show that disclosure helps to
reduce information asymmetry between the firm and its investors, as well as among
investors. Thus we also include a measure of the disclosure practices in a country (ACC).
This measure is an aggregation of the practices observed in the annual reports for the
sample of domestic firms collected by the Center for International Financial Analysis and
Research (CIFAR). The score attributed by CIFAR is based on the analysis of different
categories up to a maximum value of 100. The average among the countries of our
sample is 67.
Table 6 reports the correlations among the variables. All the independent
variables that proxy for the implicit barriers show correlations of the expected sign with
the integration measures except for the anti-self-dealing index that shows a negative
correlation with the Global to Total ratio. Some of the variables proxying for the same
type of information show high correlations with each other. It is the case that the two 32 Both these variables have been used in other papers, most notably by Bushman et al. (2005). That paper also discussed some of the limitations of such data and the related assumptions. Some of those concerns are more limited in our case since we focus on a period after 1990 when I/B/E/S substantially extended its coverage. The corresponding averages of our variables are 2.6 and 21% for the period 1987-2001 which could be taken as indication that the I/B/E/S coverage has extended significantly in later years.
23
measures of ownership concentration are highly correlated (0.49), similar to the two
indices of investors protection (0.40) and the two analysts variables (0.57). The pairwise
correlations between some of our implicit variables are of the opposite sign; as is the case
for the political risk ratings with both indices of investor protection, the accounting
variable with both measures of ownership concentration, the anti-director index with the
analyst diffusion (AN-D) variable and the ownership concentration with the average
analyst per firm. We also observe high correlations between the value traded to GDP with
the two variables measuring ownership concentration.
6.2 Main Results
Given the annual frequency of most of our independent variables, we time-aggregate the
integration measures for each country and then pool our cross-section and time series for
panel estimation. All regressions are performed with time-fixed effects. We use Panel
Corrected Standard Errors as suggested by Beck and Katz (1995, 1996) and correct for
cross-equation correlation and different error variance in each cross-section.
We start by separately analyzing the importance of our proxies. We run
preliminary univariate regressions of some of the independent variable shown in table 5
on our measures of integration. As a measure of ownership concentration we use the
closely held variable (C-HELD). The other concentration variable (OWC) has no cross-
sectional variation and it is also likely to underestimate the degree of ownership
concentration as it is constructed from the largest firms. To capture the extent of investor
protection we use the anti-director right index (A-DIR) since the correlation is of the right
sign with both dependent variables. As proxy for the information transparency of a
country, we rely on the average number of analyst per firm (AN-F) because the total
number of listed firm that we use to construct the analyst diffusion variable has very large
variation when compared to other data sources. The results of our regressions are in table
7. Panel A reports the results with the Integration Index as the dependent variable while
panel B contains those using the Global to Total Premia ratio. We find that these proxies
for implicit barriers have the expected sign with both integration measures. In addition,
24
they are highly significant in all instances, with the exception of the accounting index that
is significant only for the integration index.33
We then proceed by separately analyzing the impact of the political environment,
the corporate governance, and the information environment. Since country characteristics
have been linked in the past to different integration measures, we want to make sure that
our implicit barriers provide additional information in explaining what drives the
globalization process. Thus in our analysis we include widely used proxies. We add the
trade to GDP, as a measure of economic openness, the market capitalization to GDP to
control for financial development and the value traded to GDP to account for the level of
liquidity in financial markets.34 Summary statistics on these variables are also included in
Table 5.
Models 1 (a through c) in Table 8 contain the results. We find that the political
environment is important. Both the Integration Index and the Global to Total ratio show
a positive relation with the political risk variable, meaning that countries with a higher
risk rating, i.e. smaller political risk, are more integrated. However, the coefficient is only
significant with the Global to Total ratio in this first round of estimation. To capture the
relevance of the corporate environment we use the closely held variable and the revised
anti-director index. The closely held variable has a negative and significant coefficient, as
indication that countries with concentrated insider ownership are more exposed to local
factors and less integrated with the world market. Broadly speaking, a decrease in the
fraction of a (large) firm owned by insiders by 10% increases the integration index by 2%
and the Global to Total ratio by 5% (Model 1b). The coefficient for the anti-director
rights index, the other corporate governance proxy, is positive though marginally
significant (Model 1b) indicating that countries with a higher protection for minority
shareholders also have relatively higher level of integration. Thus, when insider
ownership is high and minority shareholders are less protected, the integration level is
relatively lower, suggesting that the impact of globalization is partly explained by the
33 We also run univariate regression with the alternative proxies of implicit barrier reported in Table 5. It is the case that the ownership concentration (OWC) and the anti-self-dealing index (ASD) are not significant for the Global to Total measure, while the analyst diffusion is of the right sign and significant for both measures. 34 These variables have been previously linked to integration and liberalization (see for example CEH ,2007 and Bekaert et al. 2008).
25
degree of investor protection and insider ownership. Analyst per firm and the accounting
variable are capturing the information environment in model 1c. It is worth pointing out
that the intersection of these variables creates the pool with the smallest number of
observations. The coefficient on analyst activity is positive and highly significant with
both measures of integration. The analyst activity is also economically important as an
increase in analyst coverage by 10 would increase the integration index by 0.2 and the
Global to Total ratio by 0.4. The accounting coefficient in both regressions is positive as
we expect that countries with better accounting practices are more integrated but, as with
the univariate analysis, we find significance only for the regression with the integration
index. Nonetheless, the economic magnitude of the accounting measure is similar across
the two measures of integration. A 10% increase in the disclosure index leads to a 0.4
increase in the two measures. Overall it is the case that a more transparent information
environment with better analyst coverage and better accounting practices is linked to
higher levels of integration.35 In all estimations, the country characteristics have the
expected sign in the case of the liquidity proxy, however the sign and significance of the
other two variables, namely the market capitalization to GDP ratio and the trade to GDP
ratio, varies across the two regressions.
Models 2 and 3 in Table 8 extend our analysis by looking at the combined effects
of the proxies for implicit barriers. Since both models include the information
environment variables, the pool for estimation are the smallest in our analysis to this
point. Model 2 only considers the implicit barriers proxies while model 3 also includes
the control variables. The evidence indicates that both measures of integration are
significantly higher when the institutions in the country are sound, the corporations are
widely held and analyst coverage is higher. The political risk variable, the closely held
variable and the average number of analyst per firm variable are all very significant. The
accounting variable has coefficients of different sign for the two integration measures but
they are not significant. This could be indication that the index is a poor proxy for the
quality of information. Alternatively the finding could be due to the statistical
characteristics of this variable that has no time variation and the smallest cross-sectional
35 Bae, Ozoguz and Tan (2008) show that greater investibility reduces the delay with which individual stock prices respond to the global and local market information.
26
dispersion among all other explanatory variables. While the coefficient on the anti-
director index is positive and significant for the integration index, it is negative and
significant with the Global to Total ratio. We thus perform a check (results are not shown
but available upon request). We eliminate Mexico from the pool since it appears as an
outlier in the statistics of table 4 with a Global to Total ratio of 1 while the Integration
Index shows a mean of 0.66. Table 5 also shows that with a score of 3.0 Mexico is below
the median for the anti-director variable. In the Global to Total ratio regression without
Mexico, the sign of the coefficient is now positive, although not significant, while it is
positive and significant if we use a larger pool, dropping the information environment
variables. All these results are robust to the inclusion of the control variables as indicated
in model 3 of Table 8. For the Integration Index, the adjusted R2 increase from 26.1% in
model 1c to 36.7% in model 3 that includes all proxies for implicit barriers. For the
Global to Total ratio, the increase is from 14.8% to 23.3%.
6.3 Extensions and Robustness Checks
Table 9 includes a second set of estimations. For brevity, we tabulate only the main
variables although all our regressions are run with the inclusion of the controls.
There is the concern that some of our proxies for implicit barriers might be
strongly linked to the liberalization process. For example, as countries liberalize, the
domestic institutions are also modernized and some of the proxies might actually be
picking up the effect of the changes in liberalization. We thus include in our model a
variable that captures the institutional changes. We use the intensity of capital controls as
a proxy for explicit barrier. This measure is equal to one minus the fraction of market
capitalization of the investable indices over the country's total market capitalization.36
When this measure is zero, the market capitalization of the investable indices is equal to
that of the market-wide indices, indicating the lack of institutional barriers to foreign
investment. We extend this measure that was first used by Edison and Warnock (2003)
by accounting for additional changes in investability for our sample countries.37 Table 8
36 This fraction is shown against the left axis in Figure 1. 37 Specifically in 2001, Colombia, Jordan and Pakistan were moved from the S&P/IFCI to the S&P/IFCG due to their small size or illiquidity, which made them not investable or too expensive to hold in a truly
27
model 4 reports the results of the two regressions. The explicit barrier variable is negative
and significant at any statistical level for the Integration Index and for the Global to Total
ratio. Thus our integration measures reflect the decrease in the explicit barrier.
Nonetheless, in both regressions the proxies for implicit barriers preserve their sign and
significance, except for the accounting variable that looses significance.38
High transaction costs are another potential obstacle for investing in emerging
markets. A measure of transaction costs is notoriously difficult to compute. No consensus
exists on what would be the best proxy and there is an additional challenge for emerging
markets where data availability is always a problem. We proxy transaction costs by the
zero-return measure that is used for Emerging Markets in Lesmond (2005) and Bekaert,
Harvey and Lundblad (2007).39 This measure is computed from the proportion of zero
daily returns observed in a year and a higher fraction of zero returns is indication of
higher level of transaction costs. Although an imperfect measure, Bekaert et al. (2007)
shows that it might capture features of transaction costs that are not related to other
measures of market liquidity. Model 5 in table 9 reports the results from the two
regressions. We should note that adding this variable to our base regressions further
shrinks the pool of observations. The zero-return measure is negative, indicating that
markets with lower transaction costs have higher level of integration; however the
coefficient is only significant for the Integration Index. We also run another check using
data on the costs of transaction and settlement collected by Wilshire associate for the
CalPERS report on Emerging Markets accessibility. For this variable, a higher number
indicates lower transaction/settlement costs. Results (not reported) indicate that countries
that are more integrated are those with lower transactions costs. Thus the direction of the
relationship is also confirmed with this proxy. In these regressions, power becomes a
problem as we have only 69 observations and we find that significance is affected not
only for this additional variable but also for some of the implicit barrier proxies of our
investable benchmark. Thus our averages for the ICC variable in table 5 include a number of observations with a reversion to 1 for these three countries. 38 We also perform the same check with a revised version of the explicit barrier proxy excluding for Colombia, Jordan and Pakistan those observations corresponding to the reversal of liberalization from 2001 and for Israel those preceding the liberalization until 1997. As a result, we only use returns from IFC investable data, reducing the number of observations included in the pool. The inference on the implicit barrier proxy remains qualitatively unchanged. 39 We thank Christian Lundblad for kindly providing us with his data.
28
base model. Nonetheless, the direction of the sign of all our proxies remains consistent
throughout.
While all the proxies have consistent sign and comparable significance level
between the two dependent variables, the coefficients on the anti-director index in our
base model show a different impact on the two dependent variables. We thus further
refine the information on shareholders protection and investigate whether the differences
in a country’s legal origin matter. We construct a dummy from the classification of La
Porta et al. (1998) for countries of civil law origin and common law origin and interact
the anti-director index with this dummy. The dummy is equal to one if the country
belongs to a civil law system. Model 6 of table 9 includes both main and interaction
coefficients, in addition to the previous variables for implicit barriers, thus it basically
estimates separate slope coefficients for the anti-director index depending on the legal
origin of a country. In these instances, the minority shareholder protection variable has
the same relationship with both our measures of integration. We find that the main
coefficient for the civil law origin is negative while the interaction term is positive and
significant, indicating that the effect of better shareholder protection are higher in civil
law countries. There is thus evidence of relatively higher integration in civil law origin
countries that have better protection of minority shareholders. This result also confirms
what has been found in previous research about the diminishing returns to shareholders
protection (see for example Bushman et al., 2005) since its impact is heightened among
civil law countries.
We conduct a final robustness check. In the regressions for both dependent
variables, the time fixed effects show more negative signs in the earlier years. We then
re-estimate our expanded regression of model (3) with a trend instead of time fixed
effects. The trend coefficient is positive but not significant in the case of the Integration
Index and marginally significant for the Global to Total ratio. The conclusions on the
importance of the implicit barriers do not change qualitatively.
In summary, our analysis indicates that the limits to globalization are related to
country-specific implicit barriers. Although our results are stronger and always consistent
with respect to the Integration Index, a finding that we attribute to the statistical
properties of the series, we find further confirmation using the Global to Total ratio. In
29
particular, we find that proxies related to state and corporate governance together with the
information environment play a strong role in explaining the lack of integration and
limits to globalization across our sample of emerging markets.
7. Conclusion
As portfolio management becomes increasingly focused on global asset allocation, it is of
primary importance to understand whether specific risks differentiate EM countries as a
separate asset class. We thus analyze investable indices, a subset of EM assets that takes
into account technical and practical foreign investment restrictions and as a result
represent the segment of choice for institutional investors.
We investigate the pricing behavior of investable portfolios represented by the
IFCI using the E-L model that accounts for market segmentation. This model allows us to
uncover whether local risk is relevant alongside global risk. We estimate a conditional
version of the model for the investable indices of 22 emerging markets over a period
characterized by increasing financial liberalization. Our results can be summarized as
follows. In spite of decreasing restrictions on foreign investment at the institutional level,
there is strong evidence that local factor - the conditional market risk - is still relevant in
pricing the returns of the investable indices. Indeed, the returns on investable indices are
determined by a combination of domestic and global factors with the local risk premium
still contributing significantly in economic terms to the total premium as indicated by the
global to total ratio of premia. Although the extent of globalization in our sample is not
uniform, the relative importance of global risk has significantly increased in the more
recent period.
The relevance of the local factor in the conditional pricing of investable indices
suggests that an important source of segmentation for emerging markets is related to
implicit barriers. Our results further show that state and corporate governance, together
with the information environment, play a significant role in the globalization of emerging
markets. Hence, regulatory reform efforts should include improvement in governance,
transparency and institutional environment, as they can complement market liberalization
policies and further integrate emerging markets.
30
References Ahearne, A., W. Griever, and F. Warnock, 2004. “Information costs and home bias: an analysis of US holdings of foreign equities,” Journal of International Economics, 62, 313-336. Ammer, J., S. Holland, D. Smith and F. Warnock, 2006. “Look at me now: what attracts U.S. shareholders?” Working paper. Bacchetta, P. and E. van Wincoop, 2000. “Capital flows to Emerging Markets: liberalization overshooting and volatility” in S. Edwards (ed.), Capital Flows and the Emerging Economies, NBER and University of Chicago Press. Bae, K.H., W. Bailey and C. X. Mao, 2006. “Stock market liberalization and the informa-tion environment,” Journal of International Money and Finance 25, 404-428. Bae, K.H., K. Chan and A. Ng, 2004. “Investability and return volatility,” Journal of Fi-nancial Economics, 71, 239-263. Bae, K.H., R., Stulz, and H. Tan, 2008. “Do local analysts know more? A cross-country study of the performance of local analysts and foreign analysts,” Journal of Financial Economics forthcoming. Bae, K.H., A. Ozoguz, and H. Tan, 2008. “Do foreigners facilitate information transmission?” Working Paper. Beck, N. and J. Katz, 1995. “What to do (and not to do) with Time-Series Cross-Section data,” American Political Science Review 89, 634 647. Beck, N. and J. Katz, 1996. “Nuisance versus Substance: Specifying and Estimating Time-Series Cross-Section Models,” Political Analysis 1-36. Bekaert, G., 1995. “Market integration and investment barriers in emerging equity mar-kets,” The World Bank Economic Review, 1, 75-107. Bekaert, G. and C. Harvey, 1995. “Time-varying world market integration,” Journal of Finance, 50, 403-444. Bekaert, G. and C. Harvey, 1997. “Emerging equity market volatility,” Journal of Financial Economics, 43, 29-77. Bekaert, G. and C. Harvey, 2002. “Chronology of important financial, economic and political events in emerging markets,” http://www.duke.edu/~charvey/chronology.htm Bekaert, G., C. Harvey and C. Lundblad, 2007. “Liquidity and Expected Returns: Lessons from Emerging Markets,” Review of Financial Studies, 20, 2007, 1783-1831.
31
Bekaert, G., C. Harvey, C. Lundblad, and S. Siegel, 2008. “What segments equity markets?” Working Paper. Bekaert, G. and M. Urias, 1996. “Diversification, integration and emerging market closed-end funds,” Journal of Finance, 51, 835-869. Bollerslev, T. and J. Wooldridge, 1992. “Quasi-maximum likelihood estimation and inference in dynamic models with time-varying covariance,” Economic Reviews, 11, 143-172. Brennan, M. and H. Cao, 1997. “International Portfolio Equity Flows,” Journal of Finance, 52, 1851-1880. Bushman, R., J. Piotroski and A. Smith, 2004. “What determines corporate transparency?” Journal of Accounting Research 42, 207-251. Bushman, R., J. Piotroski and A. Smith, 2005. “Insider trading restrictions and analysts’ incentives to follow firms.” Journal of Finance, 60, 35-66. Carrieri, F., V. Errunza and K. Hogan, 2007. “Characterizing world market integration through time,” Journal of Financial Quantitative Analysis, 42, 915-940. Carrieri, F., V. Errunza and B. Majerbi, 2006. “Does emerging markets exchange risk affect global equity prices?” Journal of Financial Quantitative Analysis, 41, 511-540. Chaieb, I. and V. Errunza, 2007. “International asset pricing under segmentation and PPP deviations,” Journal of Financial Economics, 86, 543-578. Chan, K., A. Menkveld and Z. Yang, 2008. “Information asymmetry and asset prices: Evidence from the China foreign share discount,” Journal of Finance, 63, 159-196. Chari, A. and P. B. Henry, 2004. “Risk sharing and asset prices: Evidence from a natural experiment,” Journal of Finance, 59, 1295-1324. Choe, H., B.C., Kho and R., Stulz, 2005. “Do domestic investors have an edge? The trading experience of foreign investors in Korea,” Review of Financial Studies 18, 795-829. Claessens, S., S. Djankov and L.H. Lang, 2000. “The separation of ownership and control in East Asian corporations,” Journal of Financial Economics, 58, 81-112. De Jong, F. and F.A. De Roon, 2005. “Time-varying market integration and expected returns in emerging markets,” Journal of Financial Economics, 78, 583-613. De Santis, G. and B. Gerard, 1997. “International asset pricing and portfolio
32
diversification with time-varying risk,” Journal of Finance, 52, 1881-1912. De Santis, G. and B. Gerard, 1998. “How big is the premium for currency risk?” Journal of Financial Economics, 49, 375-412. Djankov, S., R. La Porta, F. Lopex-de-Silanes and A. Shleifer, 2008. “The law and economics of self-dealing,” Journal of Financial Economics, 88, 430-465. Dumas, B. and B. Solnik, 1995. “The world price of foreign exchange rate risk,” Journal of Finance, 50, 445-479. Dvorak, T., 2005. “Do domestic investors have an information advantage? Evidence from Indonesia,” Journal of Finance 60, 817-839. Edison, H.J. and F.E. Warnock, 2003. “A simple measure of the intensity of capital controls,” Journal of Empirical Finance, 10, 81-103. Edison, H.J. and F.E. Warnock, 2008. “Cross-border listings, capital controls, and equity flows to emerging markets”, Journal of International Money and Finance, 27, 1013-1027. Engle, R. and V. K. Ng, 1993. “Measuring and testing the impact of news on volatility”, Journal of Finance, 5, 1749-1778. Erb, C. Harvey and Viskanta, 1996. “Political risk, Economic risk, and Financial risk,” Financial Analysts Journal, 52, 29-46. Errunza, V. and E. Losq, 1985. “International asset pricing under mild segmentation: Theory and test”, Journal of Finance, 40, 105-124. Errunza, V., 1977. “Gains from portfolio diversification into less developed countries’ securities”, Journal of International Business Studies, 83-99. Errunza, V., E. Losq and P. Padmanabhan, 1992. “Tests of integration, mild segmentation and segmentation hypotheses”, Journal of Banking and Finance, 16, 949-972. Ferreira, M., and P. Matos, 2008. “The colors of investors’ money: The role of institutional investors around the world,” Journal of Financial Economics 88, 499-533. Ferson, W. E., and C. R. Harvey, 1993. “The risk and predictability of international equity returns”, Review of Financial Studies, 6, 527-566. Foerster, S. and A. Karolyi, 1999. “The effects of market segmentation and investor recognition on asset prices: Evidence from foreign stocks listing in the U.S.,” Journal of Finance, 54, 981-1014.
33
Healy, P. and K. Palepu, 2001. “Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature,” Journal of Accounting and Economics 31, 405-440. Harvey, C. R., 1991. “The world price of covariance risk”, Journal of Finance, 46, 111-157. Harvey, C. R., 1995. “Predictable risk and returns in emerging markets,” Review of Financial Studies, 8, 773-816. Standard&Poor’s, 2006. “S&P Emerging Markets Indices.”, Standard&Poor’s, New York. Hong, H., T. Lim and J. Stein, 2000. “Bad News Travels Slowly: Size, Analyst Coverage, And The Profitability Of Momentum Strategies,” Journal of Finance, 2000, 55, 265-295. Hou, K., A. Karolyi and B.C. Cho, 2006. “What fundamental factors drive global stock returns?” Working Paper. Kang, J.K. and R. Stulz, 1997. “Why Is There a Home Bias? An Analysis of Foreign Portfolio Equity Ownership in Japan,” Journal of Financial Economics, 46, 3-28. Kaniel, R., D. Li and L. Starks, 2005. “Investor Visibility Events: Cross-Country Evidence”, Working Paper, University of Texas at Austin. Karolyi, A., 2004. “The role of ADRs in the development of emerging markets,” Review of Economics and Statistics 86, 670-690. Kho, B., R. Stulz, and F. Warnock, 2006. “Financial globalization, governance, and the evolution of the home bias,” NBER Working Paper 12389. La Porta, R., F. Lopez-de-Silanes and A. Shleifer, 1999. “Corporate ownership around the world”, Journal of Finance, 54, 471-517. La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. Vishny, 1998. “Law and Finance”, Journal of Political Economy, 106, 1113-1155. La Porta, R., F. Lopez-de-Silanes and A. Shleifer, 2006. “What works in securities laws?” Journal of Finance, 61, 1-32. Lesmond, D. 2005. “Liquidity of emerging markets,” Journal of Financial Economics, 77, 411-452. Leuz, C., K. Lins, K. and F. Warnock, 2008. “Do foreigners invest less in poorly governed firms?” Review of Financial Studies forthcoming.
34
Merton, R., 1987. “A Simple Model of Capital Market Equilibrium with Incomplete Information,” Journal of Finance, 42, 483-510. Nishiotis, G., 2004. “Do indirect investment barriers contribute to capital market seg-mentation?” Journal of Finance and Quantitative Analysis, 39, 613-30. Portes, R. and H. Rey, 2005. “The determinants of cross-border equity flows”, Journal of International Economics, 65, 269-296. Portes, R., H. Rey and Oh, 2001. “Information and capital flows: the determinants of transactions in financial assets,” European Economic Review 783-796. Pukthuanthong-Le, K., and R., Roll, 2008. Global Market Integration: An Alternative Measure and Its Application. Journal of Financial Economics, forthcoming Seasholes, M., 2004. “Re-examining information asymmetries in emerging stock markets,” working paper, Haas School of Business. Shanno, David F. 1985, “On Broyden-Fletcher-Goldfarb-Shanno Method,” Journal of Optimization Theory and Applications, 46(1), 87-94. Solnik, B., 1974. “An equilibrium model of the international capital market”, Journal of Economic Theory, 8, 500-524. Stulz, R., 1981a. “A model of international asset pricing”, Journal of Financial Economics, 9, 383-406. Stulz, R., 1981b. “On the effects of barriers to international investment”, Journal of Finance, 36, 923-934. Stulz, R., 2005. “The limits of financial globalization”, Journal of Finance, 60, 1595-1638.
35
Country Country Funds
ARGENTINA ARGENTINE INVESTMENT (1991/10-2001/06)
BRAZIL BRAZIL FUND (1988/03-2006/06)
CHILE CHILE FUND (1989/09)
CHINA CHINA FUND (1992/07)
COLOMBIA N.A.
CZECH REPUBLIC N.A.
HUNGARY N.A.
INDIA INDIA FUND (1994/02)
INDONESIA INDONESIA FUND (1990/03)
ISRAEL FIRST ISRAEL FUND (1992/10)
JORDAN N.A.
KOREA KOREA FUND (1984/08)
YPF S.A.(USA, 1993/07), BBVA BANCO FRANCES S.A.(USA, 1993/11), TELEFONICA DE ARGENTINA S.A.(USA, 1994/03), TRANSPORTADORA DE GAS DEL SUR, S.A.(USA, 1994/11), METROGAS S.A.(USA, 1994/11)
BANCO GANADERO (USA, Jan-95 to Mar-01)
N.A.
SINOPEC SHANGHAI PETROCHEMICALS A (USA, 1993/07), HUANENG POWER 'H' (USA, 1994/10), GUANGSHEN RAILWAY (USA, 1996/05), CHINA EASTERN AIRLINES (USA, 1997/02), CHINA SOUTHERN AIRLINES (USA, 1997/07)
PT INDONESIAN SATELLITE CORPORATION TBK (USA, 1994/10), PT TELEKOMUNIKASI INDONESIA TBK (USA, 1995/11)
TISZAI VEGYI KOM (Germany, 1996/11), MAGYAR TELECOM (USA, 1997/11)
CESC LTD (UK, 1996/08), STATE BANK OF INDIA (Germany, 1997/02), MAHANAGAR TEL.NIGAM(UK, 1998/02), INFOSYS TECHNOLOGIES LIMITED (USA, 1999/03), SIFY LTD. (USA, 1999/10)
Appendix A - Securities for the Diversification Portfolios
TEVA PHARMACEUTICAL INDS LTD (USA, 1982/02), ELRON ELECTRONIC INDUSTRIES LTD (USA, 1982/11), OPTROTECH LTD (USA, 1984/08), RADA ELECTRONIC INDUSTRIES LTD (USA, 1985/06), GALAGRAPH LTD (USA, 1987/04)N.A.
KOREA ELECTRIC POWER CORP. (USA, 1994/10), POSCO (USA, 1994/10), SK TELECOM CO. LTD. (USA, 1996/06) , KT CORPORATION (USA, 1999/04), MIRAE CORP. (USA, 1999/11)
ARACRUZ CELULOSE (USA, 1992/05), COMPANHIA BRASILEIRA DE DISTRIBUICAO (USA, 1997/06), COMP. PARANAENSE DE ENERGIA-COPEL (USA, 1997/07), COMPANHIA SIDERURGICA NACIONAL (USA, 1997/11), PETROLEO BRASILEIRO S.A. (USA, 2000/08)
COMPANIA DE TELECOMUNICACIONES DE CHILE (USA, 1990/07), COMPANIA CERVECERIAS UNIDAS S.A. (USA, 1992/10), MADECO S.A. (USA, 1993/05), SOC. QUIMICA Y MINERA DE CHILE, S.A. (USA, 1993/09), ENERSIS S.A. (USA, 1993/10)
Cross-listings
36
Appendix A continued
Country Country Funds
MALAYSIA MALAYSIA FUND (1987/05)
MEXICO MEXICO FUND (1981/06)
PAKISTAN PAKISTAN INVESTMENT FUND (1993/12-2001/06)
PERU N.A.
PHILIPPINES FIRST PHILIPPINES (1989/11-2003/06)
POLAND N.A.
SOUTH AFRICA SOUTHERN AFRICA FUND (1994/02-2004/11)
TAIWAN TAIWAN FUND (1986/12)
THAILAND THAI FUND (1988/02)
TURKEY TURKISH INVESTMENT FUND (1989/12)
Listings are from the four primary depository banks, Citibank, JP Morgan, the Bank of New York Mellon, and Deutsche Bank. Direct and ADR listings are then complented from the US major exchanges; NYSE, AMEX and NASDAQ, OTCBB and pink sheets. GDR listings are complemented from major world market exchanges, Datastream, as well as the 1998 listings available from Sergei Sarkissian website, http://web.management.mcgill.ca/Sergei.Sarkissian/. Given that the effective dates provided by the depository banks are the dates of the last change in listing, we made sure to have the date of initial listing. All the information was cross-checked and supplemented with the listed company’s website and LEXIS/NEXIS. For a full description on the procedure to obtain the ADRs listing please refer to Karolyi (2004).
MACRONIX INTERNATIONAL COMPANY LIMITED (USA, 1996/05 ) TAIWAN SEMICONDUCTOR MANUFACTURING CO. (USA, 1997/10 ) ADVANCED SEMICONDUCTOR ENGINEERING INC. (USA, 2000/09) UNITED MICROELECTRONICS CORPORATION (USA, 2000/09) SILICONWARE PRECISION IND., CO. LTD. (USA, 2000/09)
COMPANIA DE MINAS BUENAVENTURA (USA, 1996/05)
PHILIPPINES LONG DISTANCE TELEPHONE (USA, 1970/01), PSI TECHNOLOGIES (USA, 2000/03)
MOSTOSTAL EXPORT (USA, 1999/10-2001/07)
HIGHVELD STEEL AND VANADIUM (1981/10), SASOL (1982/04) ANGLOGOLD ASHANTI (Germany, 1988/09), BARLOWORLDZERT (Germany, 1988/09), ABSA GROUP (Germany, 1996/01)
THAI AIRWAYS INTL. (Germany, 1997/07), TT&T (Germany, 1998/01), INTERNET THAILAND (Germany, 2002/01)
TOFAS OTOMOBIL ADR (UK, 1994/12), TURKCELL ILETISIM HIZMETLERI (USA, 2000/07)
Cross-listings
N.A.
PETALING TIN (UK, 1986/03)
TELEFONOS DE MEXICO SA DE CV - SERIES A (USA, 1976/01), TUBOS DE ACERO DE MEXICO S.A. (USA, 1976/01), VITRO S.A. DE C.V. (USA, 1991/11), EMPRESAS ICA S.A. DE C.V. (USA, 1992/04), GRUPO RADIO CENTRO S.A. DE C.V. (1993/07)
37
Variable Sources
Political risk POL
International Country Risk Guide
Closely Held CHELD
WorldScope
Ownership concentration OWC
La Porta et al. (2006)
Anti-directors rights index A-DIR
Djankov et al. (2008)
Anti-self-dealing index ASD
Djankov et al. (2008)
Analyst coverage AN-F
I/B/E/S
Analyst diffusion AN-D
I/B/E/S and EMDB of S&P
Accounting standards ACC
Bushman, Piotroski, and Smith (2004). International accounting and auditing trends, Center for International Financial Analysis and Research (CIFAR).
Trade to GDP TR/GDP
World Bank Development Indicators.
Mcap to GDP MC/GDP
$&P/IFC emerging market and World Bank
Value traded to GDP VT/GDP
Standard and Poor's/International Finance Corporation's Emerging Stock Markets Factbook & World Bank Development Indicators.
Intensity of Capital Controls ICC
Standard and Poor's/International Finance Corporation's Emerging Stock Markets Factbook
Zero returns Z-RET
Kindly provided by Christian Lundblad as used in Bekaert, Harvey and Lundblad (2007)
Appendix B - Variable definition
Description
ICC = (1-Investability) where investability is defined as the ratio of the market capilization of the IFCI index over the market capitalization of the IFCG index. Frequency: monthly.
The sum of exports and imports of goods and services measured as a share of gross domestic product. Frequency: Annual.
Equity market capitalization divided by gross domestic product. Frequency: Annual.
The ratio of equity market value traded to GDP. Frequency: Annual.
Value weighted average fraction of firm stock market capitalization held by insiders i.e. corporate officers, directors, immediate family members, by individual shareholder holdings representing more than 5%, by other coporations (except shares held in fiduciary capacity by financial institutions), and by pension/ benefit plans and trusts. Frequency: annual.
Political risk ratings based on the sum of 12 weighted variables covering both political and social attributes. The index has 100 points, with higher scores indicating lower risk. Frequency: annual.
Mean number of analysts providing a forecast for a specific firm in a given calendar year. Frequency: annual.
Average percentage of common shares owned by the top three shareholders in the ten largest non-financial, privately-owned domestic firms in a given country
Proportion of firms with analyst coverage in a given calendar year, or number of firms included in IBES/number of listed companies. Frequency: annual.
Index created by examining and rating companies' 1995 annual reports on their inclusion or omission of 90 items. These items fall into seven categories (general information, income statements, balance sheets, funds flow statement, accounting standard, stock data, and special items). A minimum of three companies in each country were studied. The companies represent a cross section of various industry groups.
Aggregate index of shareholder rights. The index ranges from 0 to 6 and it is formed by summing: (1) vote by mail; (2) shares not blocked or deposited; (3) cumulative voting; (4) oppressed minority; (5) pre-emptive rights; and (6) capital.
The proportion of zero daily returns observed over the relevant year for each equity market, used as measure of transaction cost. Frequency: annual.
Average of ex-ante and ex-post private control of self-dealing. The index ranges from 0 to 1. It measures approval by disinterested shareholders, ex-ante disclosure, disclosure in periodic filings and ease of proving wrongdoing.
38
Country
Start date of the overlapping sample
Average return difference
Volatility difference Tracking error Correlation
(IFCI-MSCI EMF) (IFCI-MSCI EMF) (IFCI-MSCI EMF) (IFCI-MSCI EMF)
ARGENTINA Jan-89 0.1% 3.3% 9.1% 0.87BRAZIL Jan-89 0.1% 1.2% 3.0% 0.99CHILE Jan-89 0.0% 0.0% 1.6% 0.98CHINA Jan-93 0.5% 0.3% 6.1% 0.84CZECH REPUBLIC Jan-95 -0.2% 0.1% 1.7% 0.97HUNGARY Jan-95 -0.2% 0.5% 1.8% 0.98INDIA Dec-92 -0.1% -0.1% 1.8% 0.98INDONESIA Oct-90 0.1% -0.1% 2.2% 0.99ISRAEL Jan-97 0.1% -0.7% 3.7% 0.88KOREA Feb-92 -0.1% 0.3% 1.6% 0.99MALAYSIA Jan-89 -0.1% 0.4% 3.6% 0.92MEXICO Jan-89 -0.1% -0.1% 1.8% 0.98PERU Jan-93 0.1% -0.6% 2.3% 0.97PHILIPPINES Jan-89 -0.2% 0.5% 3.7% 0.93POLAND Jan-93 0.2% 0.0% 2.0% 0.99SOUTH AFRICA Jan-93 0.2% 0.1% 1.5% 0.98TAIWAN Feb-91 0.0% 0.2% 1.2% 0.99THAILAND Jan-89 0.0% -0.4% 2.2% 0.98TURKEY Sep-89 -0.2% 0.2% 4.6% 0.96
Average 0.0% 0.3% 2.9% 0.96
Table 1- Comparison of IFC and MSCI Emerging Market Investable indices
The IFC Investable emerging markets equity indices are from the S&P/IFC Emerging Markets Database. The MSCI emerging markets Free (EMF) equity indices are from Datastream. Returns are monthly percentage, denominated in USD. The period is from January 1989 or later to December 2006. For each country, the table presents the starting dates (this is the latest from the two benchmarks IFCI or MSCI EMF), the return difference between IFCI and MSCI EMF, the volatility difference, the tracking error (defined as the standard deviation of the difference between the index returns), and the correlation between IFCI and MSCI EMF.
39
Table 2: Summary statistics for assets excess returns and information variablesPanel A presents basic statistics of the investable indices. The emerging markets investable equity indices returns are proxied by IFC investable indices (IFCI) from the S&P/IFC Emerging Markets Database for all countries except Colombia, Jordan, Pakistan. The investable return series for these three countries has been discontinued by S&P/IFC since November 2001 due to their small size or illiquidity. For these countries, we use MSCI EM Free indices. For Isreal, we also use MSCI EM Free as the returns data starts on January 1993, while it starts on January 1997 for the Israelian IFCI. The world market portfolio (WMP) return is the U.S. dollar return on the MSCI value-weighted world market portfolio. Returns are monthly percentage, denominated in USD and in excess of the one-month Eurodollar deposit rate. The period is from January 1989 or later to December 2006. For each country, the table presents the starting dates for the return data, the number of firms in the IFCG index (NG) and IFCI index (NI) as of December 2006, the market cap. of the IFCG index (MCG) and the market cap. of the IFCI index (MCI) in millions of U.S. dollars as of December 2006. B-J is the Bera-Jarque test for normality based on excess skewness and kurtosis. Q is the Ljung-Box test for autocorrelation of order 12 for the excess returns and the excess returns squared. **and *denote statistical significance at the 1% and 5% levels respectively. Panel B presents for each country, the correlation between the Diversification Portfolio, the IFCI index, and the World Market Portfolio. The diversification portfolio is constructed as described in Section 3. Panel C presents the basic statistics for the global information variables. The global instruments include a constant, the world dividend yield in excess of the one-month Euro-dollar interest rate (XWDY), the change in US term premium (∆USTP), and the US default premium (USDP). All variables are in percent per month, lagged one month with respect to the returns series.Panel D reports the basic statistics for the local information variables. The local instruments include a constant, the lagged emerging market excess returns (LagRet), the local dividend yield in excess of the one-month Euro-dollar interest rate (XLDY), the change in bilateral exchange rate (∆FX). All variables are in percent per month, lagged one month with respect to the return series.
40
Panel A: Distributional Statistics of the IFC investable indices
Start date NG MCG NI MCI Mean Std. Dev. B-J Q(z)12 p-value Q(z2)12 p-value
ARGENTINA Jan-89 18 20145 14 19476 1.18 18.11 1659.8** 19.97 0.07 75.05 0.00
BRAZIL Jan-89 123 316627 115 302505 1.12 17.72 827.5** 22.90 0.03 41.07 0.00
CHILE Jan-89 56 57673 50 55588 1.14 7.07 27.83** 16.84 0.16 15.92 0.19
CHINA Jan-93 411 511238 195 363133 -0.04 9.66 44.17** 28.84 0.00 118.83 0.00
COLOMBIA Jan-93 17 16558 na na 0.78 8.39 19.55** 21.74 0.04 96.25 0.00
CZECH REPUBLIC Jan-94 6 15531 6 15531 0.45 8.07 532.95** 20.97 0.05 12.15 0.43
HUNGARY Jan-93 11 27938 8 27699 0.94 9.58 350.29** 16.62 0.16 15.85 0.20
INDIA Dec-92 192 305838 182 210157 0.37 7.16 2.76 18.84 0.09 31.26 0.00
INDONESIA Oct-90 42 37823 42 37823 -0.23 12.79 88.48** 33.51 0.00 213.56 0.00
ISRAEL Jan-93 70 65396 69 65273 0.15 6.56 29.07** 13.96 0.30 84.05 0.00
JORDAN Jan-89 na na na na 0.29 5.15 28.57** 37.92 0.00 17.58 0.13KOREA Feb-92 305 511238 242 487752 0.21 10.68 166.4** 8.19 0.77 141.06 0.00MALAYSIA Jan-89 113 75997 113 71436 0.11 9.33 152.85** 46.98 0.00 189.49 0.00MEXICO Jan-89 55 160441 52 160088 1.25 9.34 203.95** 26.58 0.01 24.18 0.02PAKISTAN Jan-93 48 10227 na na 0.28 9.83 132.38** 11.62 0.48 55.80 0.00PERU Jan-93 30 17310 19 15883 0.86 7.25 130.63** 24.59 0.02 20.83 0.05PHILIPPINES Jan-89 29 24899 20 15791 -0.17 10.08 34.19** 22.16 0.04 26.87 0.01POLAND Jan-94 53 53229 53 53229 0.18 9.94 169.68** 21.37 0.05 46.80 0.00SOUTH AFRICA Jan-93 146 256044 144 255907 0.74 7.13 293.38** 10.22 0.60 49.86 0.00TAIWAN Feb-91 176 400894 176 391923 0.04 8.80 65.92** 19.83 0.07 21.19 0.05THAILAND Jan-89 85 65376 71 38245 -0.06 11.37 24.70** 36.37 0.00 119.39 0.00TURKEY Sep-89 60 42917 55 42135 0.53 17.05 10.06** 7.88 0.79 9.73 0.64WMP Jan-89 0.29 4.06 17.73** 7.56 0.82 19.17 0.08
41
Panel B: Pairwise Correlations for Assets Returns
mean
IFCI and
WMP
DP and
IFCI
DP and
WMP
ARGENTINA 0.06 0.13 0.57 0.35BRAZIL 0.05 0.39 0.58 0.68CHILE 0.01 0.37 0.78 0.44CHINA 0.00 0.35 0.85 0.41COLOMBIA 0.02 0.21 0.46 0.65CZECH REPUBLIC 0.01 0.36 0.53 0.72HUNGARY 0.03 0.47 0.71 0.69INDIA 0.01 0.29 0.80 0.47INDONESIA 0.02 0.34 0.82 0.50ISRAEL 0.01 0.54 0.90 0.63JORDAN 0.01 0.11 0.13 0.82KOREA 0.02 0.46 0.80 0.66MALAYSIA 0.01 0.40 0.77 0.58MEXICO 0.03 0.48 0.80 0.64PAKISTAN 0.02 0.11 0.62 0.24PERU 0.04 0.28 0.51 0.45PHILIPPINES 0.01 0.37 0.80 0.52POLAND -0.01 0.47 0.53 0.83SOUTH AFRICA 0.01 0.51 0.85 0.67TAIWAN 0.00 0.41 0.83 0.55THAILAND 0.01 0.44 0.78 0.58TURKEY -0.02 0.32 0.74 0.54
correlations Counrty pairwise correlations between returns of IFCI, Diversification Portfolios and World Market Portfolio
0.0
0.2
0.4
0.6
0.8
1.0
ARG
ENTI
NA
BRAZ
IL
CHIL
ECH
INA
COLO
MBI
A
CZEC
H RE
PUBL
ICHU
NGAR
YIN
DIA
INDO
NESI
AIS
RAEL
JO
RDAN
KORE
AM
ALAY
SIA
MEX
ICO
PAKI
STAN
PERU
PHIL
IPPI
NES
POLA
ND
SOUT
H AF
RICA
TAIW
AN
THAI
LAND
TURK
EY
Correlation between IFCI and WMP Correlation between DP and IFCI
Correlation between DP and WMP
42
Panel C: Global information variables
Mean Std. Dev.XWDY -0.23 0.19 1.00 0.03 0.23∆USTP -0.01 0.23 1.00 0.18USDP 0.84 0.20 1.00
Panel D: Local information variables
Mean Std. Dev. Mean Std. Dev.With LagRet With ∆FX
ARGENTINA -0.19 0.21 -0.196 0.248 -3.60 18.24BRAZIL -0.07 0.29 -0.198 0.235 -7.47 13.11CHILE -0.05 0.19 -0.032 0.055 -0.36 2.16CHINA -0.31 0.22 0.068 0.003 -0.35 3.28COLOMBIA -0.05 0.21 0.110 0.053 -0.90 2.40CZECH REPUBLIC -0.08 0.27 0.179 0.108 0.17 4.15HUNGARY -0.17 0.16 0.099 0.292 -0.59 2.94INDIA -0.26 0.19 0.144 0.131 -0.51 2.16INDONESIA -0.25 0.26 0.110 0.022 -0.77 8.20ISRAEL -0.31 0.22 0.010 0.050 -0.46 2.07JORDAN -0.15 0.21 0.250 0.302 -0.19 0.99KOREA -0.26 0.21 0.064 0.028 -0.14 3.79MALAYSIA -0.19 0.22 0.022 0.011 -0.14 2.59MEXICO -0.24 0.18 -0.016 0.076 -0.73 3.99PAKISTAN 0.05 0.35 0.038 0.172 -0.55 1.78PERU -0.23 0.26 0.071 0.245 -4.06 15.47PHILIPPINES -0.29 0.20 0.089 0.102 -0.39 2.70POLAND -0.30 0.22 0.088 0.169 -1.88 7.72SOUTH AFRICA -0.20 0.28 0.058 0.162 -0.52 3.68TAIWAN -0.30 0.22 0.060 0.133 -0.02 2.41THAILAND -0.19 0.19 -0.003 -0.022 -0.17 3.32TURKEY -0.14 0.18 -0.056 0.090 -3.10 5.28 0.414
0.2710.3740.4180.583
0.5130.148
0.2000.349
0.3970.0940.5850.523
Pairwise Correlations
0.117
0.4200.619
Pairwise Correlations
0.108With LagRet
Pairwise Correlations
0.3960.2560.234
XLDY ∆FX
0.2260.427
43
Table 3: Hypothesis testing of the model
The estimated model is: r IFCI,t = δW,t-1 cov (rIFCIt,rWt) + λIt-1 var (rIFCIt|rDPt) + εIt
rDP,t = δW,t-1 cov (rDPt,rWt) + εDP,t
rW,t = δW,t-1 var (rW,t) + εW,t
where rIFCI,t is the country investable index excess returns, rDP,t is the diversification portfolio excess returns, rW,t is the world index excess returns, δW is the price of world covariance risk, λI is the prices of local risk and εt| ϑt-1 ~ N (0, Ht). The time-varying prices are estimated with a different set of conditioning information. Price specifications are given by: δW,t-1 = ( κW' Zw,t-1 )
2
where ZW is a set of information variables which includes a constant, the U.S. default spread, the U.S. term structure spread and the world dividend yield in excess of the risk free rate, λI,t-1 = ( κi' Zi,t-1 )
2
where ZI is a set which includes a constant, the lagged local equity return, the local dividend yield and the change in the local exchange rate.Ht is the time-varying conditional covariance parameterized as: Ht = H0 * (ιι' - aa' - bb') + aa' * Σt-1 + bb' * Ht-1 ,where * denotes the Hadamard product, a and b are (3 x 1) vector of constants, ι is (3 x 1 ) unit vector, and Σt-1 is the matrix of cross error terms, εt-1ε't-1. Country equity investable indices are from IFC and MSCI and the world equity index is from MSCI. The risk free rate is the one-month Eurodollar rate from Datastream. All returns are denominated in USD. Sample is from January 1989 or later to December 2006. The model is estimated by Quasi-Maximum Likelihood. P-values for robust Wald test for the hypothesis are reported under each country. B-J is the Bera-Jarque test for normality based on excess skewness and kurtosis. Q is the Ljung-Box test for autocorrelation of order 12 for the eresiduals and the residuals squared. EN-AN and EN-AP are respectively the Engle-Ng (1993) negative size bias and positive size bias test on the squared residuals.
44
Panel A. Specification tests
Null hypothesis
d.f. p-value d.f. p-value d.f. p-value d.f. p-valueARGENTINA 4 0.000 3 0.000 4 0.987 3 0.957BRAZIL 4 0.000 3 0.117 4 0.000 3 0.000CHILE 4 0.000 3 0.216 4 0.002 3 0.037CHINA 4 0.000 3 0.103 4 0.000 3 0.000COLOMBIA 4 0.000 3 0.307 4 0.011 3 0.021CZECH REPUBLIC 4 0.000 3 0.050 4 0.000 3 0.000HUNGARY 4 0.000 3 0.164 4 0.574 3 0.407INDIA 4 0.000 3 0.009 4 0.097 3 0.103INDONESIA 4 0.000 3 0.080 4 0.000 3 0.000ISRAEL 4 0.000 3 0.306 4 0.007 3 0.004JORDAN 4 0.000 3 0.232 4 0.034 3 0.018KOREA 4 0.000 3 0.460 4 0.006 3 0.004MALAYSIA 4 0.000 3 0.028 4 0.000 3 0.000MEXICO 4 0.000 3 0.161 4 0.970 3 0.998PAKISTAN 4 0.004 3 0.583 4 0.000 3 0.001PERU 4 0.000 3 0.802 4 0.998 3 1.000PHILIPPINES 4 0.051 3 0.285 4 0.010 3 0.034POLAND 4 0.000 3 0.445 4 0.000 3 0.145SOUTH AFRICA 4 0.000 3 0.368 4 0.000 3 0.275TAIWAN 4 0.000 3 0.483 4 0.001 3 0.173THAILAND 4 0.000 3 0.027 4 0.000 3 0.000TURKEY 4 0.000 3 0.009 4 0.000 3 0.020
for insignificant world market risk
for constant world market risk
for insignificant local market risk
for constant local market risk
45
Panel B. Diagnostics for the residual
B-J p-value Q(z)12 p-value Q(z2)12 p-value EN-AN p-value EN-AP p-value R2(%)ARGENTINA 202.080 0.000 6.801 0.871 7.996 0.785 1.470 0.072 2.656 0.004 1.22%BRAZIL 99.806 0.000 10.434 0.578 12.419 0.413 0.951 0.171 0.048 0.481 5.61%CHILE 5.947 0.051 14.677 0.260 7.838 0.798 1.682 0.047 1.208 0.114 5.27%CHINA 3.042 0.219 19.635 0.074 8.352 0.757 1.077 0.142 1.352 0.089 -2.21%COLOMBIA 2.271 0.321 8.292 0.762 19.829 0.070 -0.232 0.408 -0.950 0.172 5.55%CZECH REPUBLIC 13.771 0.001 13.141 0.359 19.975 0.068 -0.505 0.307 -0.323 0.374 3.53%HUNGARY 150.840 0.000 9.318 0.676 11.031 0.526 0.063 0.475 0.374 0.354 0.90%INDIA 17.765 0.000 8.106 0.777 7.005 0.857 -0.522 0.301 -1.507 0.067 1.30%INDONESIA 75.798 0.000 14.030 0.299 9.358 0.672 -1.796 0.037 -1.913 0.029 -8.57%ISRAEL 11.388 0.003 14.598 0.264 6.356 0.897 -0.544 0.294 -1.304 0.097 1.18%JORDAN 6.958 0.031 20.946 0.051 5.349 0.945 1.975 0.025 1.070 0.143 2.78%KOREA 6.626 0.036 6.946 0.861 27.108 0.007 -3.536 0.000 -0.240 0.405 7.04%MALAYSIA 75.506 0.000 18.588 0.099 8.942 0.708 -1.836 0.034 -1.793 0.037 6.30%MEXICO 279.410 0.000 20.107 0.065 6.384 0.896 -0.548 0.292 -0.129 0.449 0.98%PAKISTAN 19.295 0.000 10.345 0.586 8.685 0.730 -0.987 0.163 -0.421 0.337 4.53%PERU 20.258 0.000 23.154 0.026 12.434 0.411 0.500 0.309 0.798 0.213 -0.18%PHILIPPINES 14.644 0.001 9.658 0.646 8.158 0.773 -2.717 0.004 -2.246 0.013 0.58%POLAND 18.645 0.000 11.696 0.470 14.848 0.250 -0.362 0.359 1.363 0.088 1.69%SOUTH AFRICA 19.069 0.000 6.776 0.872 24.042 0.020 0.746 0.228 -0.621 0.268 0.07%TAIWAN 19.658 0.000 11.661 0.473 13.179 0.356 0.203 0.420 0.396 0.346 0.01%THAILAND 12.596 0.002 23.277 0.025 20.729 0.054 0.615 0.270 -0.837 0.202 2.30%TURKEY 1.866 0.393 6.906 0.864 14.035 0.298 -0.605 0.273 -0.681 0.248 0.02%
46
Mean Before 1995 After 2001 Max. Min. Std. Dev. Obs.
ARGENTINA 0.550 0.336 0.636 0.803 0.065 0.226 18BRAZIL 0.537 0.197 0.839 0.900 0.097 0.315 18CHILE 0.660 0.577 0.724 0.787 0.494 0.081 18CHINA 0.737 0.763 0.670 0.879 0.528 0.111 14COLOMBIA 0.202 0.273 0.161 0.266 0.156 0.034 14CZECH REPUBLIC 0.273 0.206 0.349 0.455 0.119 0.099 13HUNGARY 0.575 0.372 0.694 0.805 0.186 0.213 14INDIA 0.664 0.429 0.772 0.802 0.383 0.161 14INDONESIA 0.661 0.518 0.746 0.798 0.423 0.111 17ISRAEL 0.820 0.825 0.830 0.885 0.776 0.034 14JORDAN 0.044 0.040 0.043 0.065 0.014 0.014 18KOREA 0.598 0.387 0.681 0.852 0.277 0.166 15MALAYSIA 0.630 0.626 0.592 0.781 0.472 0.080 18MEXICO 0.661 0.654 0.635 0.762 0.590 0.055 18PAKISTAN 0.401 0.370 0.293 0.586 0.171 0.120 14PERU 0.274 0.202 0.304 0.372 0.137 0.081 14PHILIPPINES 0.589 0.493 0.518 0.865 0.181 0.227 18POLAND 0.332 0.232 0.415 0.452 0.155 0.100 13SOUTH AFRICA 0.752 0.733 0.761 0.772 0.721 0.016 14TAIWAN 0.671 0.639 0.709 0.872 0.594 0.078 16THAILAND 0.593 0.492 0.635 0.774 0.250 0.150 18TURKEY 0.628 0.458 0.758 0.846 0.431 0.152 18
Country pool 0.543 0.446 0.580 0.237 348
Panel A. Integration Index
Panel B. Global to Total ratio
Table 4 - Summary statistics for the estimated Integration Indices and Global to Total ratio
For each IFCI index, the table presents in Panel A summary statistics of the integration index estimated from the model in table 3. The estimated monthly integration indices are averaged to obtain yearly values. Panel B presents summary statistics of the global to total ratio (GT). The global premium is the world market premium, the local premium is the conditional market premium. These premiums are estimated from the model in table 3. The Global to Total ratio is then computed as the absolute value of the global premium devided by the sum of absolute values of global and local premiums. Hence by construction the Global to Total ratio lies between 0 and 1. The estimated monthly GT ratios are then averaged to obtain yearly GT ratios. The mean, subperiod means, maximum, minimum and standard deviation are reported for each measure over the period 1989-2006. The table also reports mean and standard deviation of the pool of observations. Last column of Panel B reports the correlation between the Integration Index and GT for each country in our sample.
47
Mean Before 1995 After 2001 Max. Min. Std. Dev. Obs. Correlation with Integration Index
ARGENTINA 0.770 0.592 0.821 0.997 0.094 0.293 18 0.729BRAZIL 0.796 0.541 0.930 0.998 0.231 0.255 18 0.543CHILE 0.316 0.119 0.474 0.584 0.007 0.175 18 0.823CHINA 0.559 0.599 0.742 0.891 0.115 0.258 14 -0.184COLOMBIA 0.307 0.687 0.203 0.674 0.019 0.214 14 0.892CZECH REPUBLIC 0.329 0.591 0.299 0.591 0.089 0.187 13 0.137HUNGARY 0.662 0.596 0.655 0.894 0.347 0.169 14 0.011INDIA 0.317 0.265 0.421 0.746 0.009 0.228 14 0.484INDONESIA 0.596 0.640 0.620 0.972 0.132 0.252 17 0.175ISRAEL 0.736 0.903 0.752 0.956 0.353 0.166 14 0.275JORDAN 0.271 0.112 0.209 0.821 0.020 0.242 18 -0.181KOREA 0.846 0.802 0.835 0.991 0.586 0.121 15 0.128MALAYSIA 0.634 0.665 0.738 0.959 0.226 0.251 18 -0.053MEXICO 1.000 1.000 1.000 1.000 1.000 0.000 18 -0.025PAKISTAN 0.407 0.528 0.359 0.739 0.042 0.204 14 0.325PERU 1.000 1.000 1.000 1.000 1.000 0.000 14 --PHILIPPINES 0.572 0.559 0.484 0.954 0.249 0.185 18 0.267POLAND 0.167 0.277 0.242 0.430 0.019 0.145 13 0.358SOUTH AFRICA 0.432 0.445 0.571 0.781 0.235 0.169 14 0.125TAIWAN 0.708 0.717 0.623 0.906 0.321 0.155 16 0.431THAILAND 0.470 0.493 0.430 0.910 0.150 0.228 18 0.132TURKEY 0.126 0.054 0.259 0.446 0.019 0.122 18 0.564
Country Pool 0.554 0.554 0.576 0.314 348
48
POL C-HELD OWC A-DIR ASD AN-F AN-D ACC MC/GDP TR/GDP VT/GDP ICC Z-RET CIVIL
ARGENTINA 0.71 0.67 0.53 2 0.34 6.25 0.41 0.68 0.36 0.25 0.04 0.05 0.28 1BRAZIL 0.66 0.56 0.57 5 0.27 4.75 0.23 0.56 0.36 0.18 0.15 0.24 0.49 1CHILE 0.77 0.62 0.45 4 0.63 3.74 0.26 0.78 0.97 0.52 0.10 0.18 0.36 1CHINA 0.67 0.66 1 0.76 5.47 0.19 0.32 0.44 0.33 0.73 0.14COLOMBIA 0.54 0.53 0.63 3 0.57 2.21 0.10 0.58 0.20 0.29 0.02 0.37 0.47 1CZECH REPUBLIC 0.79 0.72 4 0.33 5.77 0.17 0.23 1.02 0.12 0.42HUNGARY 0.80 0.46 2 0.18 6.91 0.54 0.23 1.02 0.15 0.28INDIA 0.61 0.52 0.40 5 0.58 5.33 0.04 0.61 0.42 0.21 0.46 0.67 0.23 0INDONESIA 0.54 0.66 0.58 4 0.65 6.27 0.37 0.27 0.54 0.11 0.31 0.35 1ISRAEL 0.63 0.50 0.51 4 0.73 2.77 0.04 0.74 0.59 0.57 0.28 0.01 0.20 0JORDAN 0.71 0.62 0.52 1 0.16 1.00 0.30 1.08 0.87 0.37 0.63 0.51 1KOREA 0.77 0.35 0.23 5 0.47 4.89 0.36 0.68 0.51 0.60 1.08 0.47 0.16 1MALAYSIA 0.73 0.53 0.54 5 0.95 8.39 0.35 0.79 1.71 1.78 0.75 0.18 0.29 0MEXICO 0.70 0.54 0.64 3 0.17 8.61 0.49 0.71 0.28 0.55 0.09 0.08 0.26 1PAKISTAN 0.51 0.71 0.37 4 0.41 2.00 0.05 0.73 0.19 0.31 0.44 0.49 0.28 0PERU 0.61 0.58 0.56 4 0.45 1.23 0.03 0.28 0.28 0.04 0.27 1PHILIPPINES 0.66 0.65 0.57 4 0.22 7.51 0.32 0.64 0.55 0.86 0.14 0.50 0.36 1POLAND 0.78 0.63 2 0.29 5.17 0.33 0.17 0.49 0.06 0.29SOUTH AFRICA 0.70 0.46 0.52 5 0.81 4.46 0.41 0.79 1.79 0.45 0.54 0.18 0.24 0TAIWAN 0.78 0.29 0.18 3 0.56 4.08 0.56 0.58 0.59 0.84 2.46 0.61 1THAILAND 0.69 0.45 0.47 4 0.81 6.27 0.57 0.66 0.55 0.98 0.39 0.56 0.26 0TURKEY 0.59 0.69 0.59 3 0.43 0.58 0.22 0.32 0.32 0.04 0.35 1
Country PoolAverage 0.68 0.56 0.49 3.5 0.49 4.91 0.29 0.67 0.54 0.61 0.38 0.34 0.30 0.67Standard Deviation 0.10 0.15 0.13 1.2 0.23 2.78 0.22 0.08 0.51 0.39 0.58 0.35 0.17 0.49
Table 5 - Summary statistics for the independent variables
The table presents averages of the variables for each country and for the pool. The period is from 1989 to 2006. Not all variables are available in every period for every country. The definition of the variables in in Appendix B.
49
POL C-HELD OWC A-DIR ASD AN-F AN-D ACC MC/GDP TR/GDP VT/GDP ICC Z-RET
C-HELD -0.33 1.00
OWC -0.37 0.49 1.00
A-DIR -0.17 -0.22 -0.14 1.00
ASD -0.12 -0.20 -0.16 0.40 1.00
AN-F 0.42 -0.10 0.13 0.16 0.07 1.00
AN-D 0.59 -0.49 -0.21 -0.10 0.00 0.57 1.00
ACC 0.20 0.04 0.03 0.28 0.46 0.29 0.15 1.00
MC/GDP 0.25 -0.29 -0.06 0.34 0.52 0.16 0.27 0.67 1.00
TR/GDP 0.49 -0.18 -0.12 0.06 0.22 0.48 0.48 0.39 0.54 1.00
VT/GDP 0.34 -0.62 -0.77 0.08 0.25 -0.05 0.38 -0.17 0.28 0.30 1.00
ICC -0.32 0.02 -0.29 -0.23 0.03 -0.24 -0.12 -0.39 -0.05 0.08 0.21 1.00
Z-RET -0.19 0.33 0.55 -0.12 -0.46 -0.33 -0.06 -0.53 -0.04 -0.02 -0.50 0.13 1.00
II 0.09 -0.32 -0.19 0.35 0.49 0.40 0.23 0.39 0.21 -0.02 0.25 -0.34 -0.70
GT 0.12 -0.45 -0.05 0.13 -0.07 0.29 0.26 0.11 -0.05 0.00 0.15 -0.26 -0.33
Table 6 - Cross-correlations of variables - by country
The table presents correlations of the variables computed from the averages of each country. The period is from 1989 to 2006. The definition of the variables in in Appendix B.
50
POL 0.232** 0.390***(0.114) (0.144)
C-HELD -0.271*** -0.517***(0.070) (0.107)
A-DIR 0.060*** 0.027***(0.006) (0.008)
AN-F 0.028*** 0.016***(0.004) (0.006)
ACC 0.615*** 0.162(0.145) (0.136)
Nobs 348 331 348 293 245 348 331 348 293 245
Period FE yes yes yes yes yes yes yes yes yes yesAdj. R2 2.9% -0.01% 12.1% 12.6% 15.1% 3.9% 3.9% 3.6% 2.4% 1.50%
*** 1% significance level** 5% significance level* 10% significance level
dependent variable = II dependent variable = GT
Table 7 - Effects of implicit barriers - Univariate analysis
The table reports the estimated parameters from a pooled regression of our integration measures on proxies for implicit barriers. We run unbalanced regression as not all the explanatory variables are available for all the cross-sectional units. Standard errors in parentheses are Panel Corrected Standard Errors based on Beck and Katz (1998). The sample period is from 1989 to 2006. Estimates of the time effects are not reported. Definition of the variables is in Appendix B.
51
Model (1a) Model (1b) Model (1c) Model (2) Model (3) Model (1a) Model (1b) Model (1c) Model (2) Model (3)
POL 0.189 0.712*** 0.834*** 0.436** 0.691*** 0.872***(0.146) (0.122) (0.143) (0.177) (0.160) (0.197)
C-HELD -0.187** -0.164** -0.165*** -0.524*** -0.208*** -0.201*(0.076) (0.064) (0.061) (0.139) (0.092) (0.110)
A-DIR 0.023 0.028*** 0.035** 0.020* -0.071*** -0.038**(0.015) (0.014) (0.017) (0.012) (0.015) (0.018)
AN-F 0.024*** 0.013*** 0.019*** 0.040*** 0.023*** 0.032***(0.004) (0.004) (0.004) (0.008) (0.006) (0.008)
ACC 0.439*** 0.251 0.368** 0.47 -0.247 0.393(0.166) (0.160) (0.180) (0.302) (0.234) (0.284)
TR/GDP -0.087** -0.036 -0.109*** -0.146*** -0.066 -0.004 -0.086 -0.113**(0.034) (0.023) (0.030) (0.027) (0.044) (0.038) (0.053) (0.055)
MC/GDP 0.078*** 0.060* 0.068*** 0.027 -0.046* -0.067** -0.147*** -0.123***(0.022) (0.032) (0.022) (0.025) (0.028) (0.031) (0.042) (0.047)
VT/GDP 0.058*** 0.019 0.050*** 0.0005 0.070*** 0.009 0.106*** 0.041(0.014) (0.014) (0.018) (0.017) (0.019) (0.027) (0.028) (0.034)
Nobs 347 322 213 207 206 347 322 213 207 206Period FE Yes Yes Yes Yes Yes Yes Yes Yes Yes YesAdj. R2 7.2% 6.9% 26.1% 30.9% 36.7% 5.2% 9.2% 14.8% 20.2% 23.3%
*** 1% significance level** 5% significance level* 10% significance level
Table 8 - Effects of implicit barriers - Multivariate analysis
dependent variable = GTdependent variable = II
The table reports the estimated parameters from pooled regressions of our integration measures on proxies for implicit barriers and other country characteristics. We run unbalanced regression as not all the explanatory variables are available for all the cross-sectional units. Standard errors in parenthesis are Panel Corrected Standard Errors based on Beck and Katz (1998). The sample period is from 1989 to 2006. Estimates of the time effects are not reported. Definition of the variables is in Appendix B.
52
Model (4) Model (5) Model (6) Model (4) Model (5) Model (6)
POL 0.602*** 0.736*** 1.098*** 0.434** 0.976*** 0.546**(0.144) (0.160) (0.197) (0.208) (0.213) (0.238)
C-HELD -0.172*** -0.117* -0.139** -0.213* -0.197* -0.197*(0.065) (0.064) (0.066) (0.117) (0.118) (0.108)
A-DIR 0.042*** 0.043*** -0.068** -0.026 -0.069*** -0.130**(0.015) (0.015) (0.031) (0.016) (0.022) (0.063)
AN-F 0.015*** 0.017*** 0.021*** 0.026*** 0.024*** 0.034***(0.004) (0.005) (0.005) (0.008) (0.008) (0.008)
ACC 0.158 0.261 0.047 -0.005 -0.052 0.635*(0.188) (0.234) (0.188) (0.345) (0.329) (0.321)
ICC -0.173*** -0.326***(0.050) (0.098)
Z-RET -0.373*** -0.218(0.135) (0.207)
CIVIL -0.514*** -0.448(0.166) (0.309)
CIVIL*A-DIR 0.098*** 0.137*(0.035) (0.072)
Nobs 206 191 206 206 191 206Period FE Yes Yes Yes Yes Yes YesAdj. R2 42.1% 43.1% 40.5% 31.5% 24.7% 27.6%
Table 9 - Extensions
dependent variable = II dependent variable = GT
The table reports the estimated parameters from a pooled regression of our integration measures on proxies for implicit barriers and other country characteristics. We run unbalanced regression as not all the explanatory variables are available for all the cross-sectional units. Standard errors in parenthesis are Panel Corrected Standard Errors based on Beck and Katz (1998). The sample period is from 1989 to 2006. Estimates of the time effects and coefficients on the control variables are not reported. Definition of the variables is in Appendix B.
53
Figure 1
Argentina Brazil
Chile Mexico
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
5
10
15
20
MCI/MCG flows/MCG
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
2
4
6
8
10
MCI/MCG flows/MCG
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
0.0
0.2
0.4
0.6
0.8
1.0
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MC flows/MCG
54
Figure 1
India Indonesia
Korea Malaysia
0
0.2
0.4
0.6
0.8
1
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
0.0
0.2
0.4
0.6
0.8
1.0
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
55
Figure 1Philippines Taiwan
Thailand
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
2
4
6
8
10
MCI/MCG flows/MCG
0
0.2
0.4
0.6
0.8
1
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
0
1
2
3
4
5
MCI/MCG flows/MCG
56