2.2 Literature Review
2.3.2 Variable Definitions
The major variables regarding the governance concept of interest and employed are World Bank’s Worldwide Governance Indicators (WGI) conducted by Kauffman et al (1999), drawing from 25 important sources and covering various aspects of governance such as International Country Risk Guide, Economic Freedom Index, and
Global Competitiveness Report. These indicators are compiled every year and cover
more than 245 countries, including most of the developed and emerging markets. The governance quality includes six indicators. Voice and Accountability (VA)
Political Stability and Absence of Violence (PS)
Government Effectiveness (GE)
Regulatory Quality (RQ)
Rule of Law (RL)
Control of Corruption (CC)
According to Kauffman et al (1999), the Voice and Accountability (VA) measures the extent of citizen rights such as election, freedom of expression, freedom of association, and a free media. Political Stability and Absence of Violence (PS) captures perceptions of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including politically-motivated violence and terrorism. Government Effectiveness (GE) measures the perceptions of the quality of public services, including the quality of policy formulation and a government’s commitment to policies’ implementation. Regulatory Quality (RQ) measures the ability of the government to formulate and implement sound policies and regulations that promote private sector development. Rule of Law (RL) captures the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, and the courts, as well as the likelihood of crime and violence. Control of Corruption (CC) measures the extent to which public power is exercised for private gain and corruption.
The measures are scored using the unobservable components model and range from -2.5 to 2.5, with a higher score representing a higher quality of governance in regards to that aspect of the governance infrastructure. Nevertheless, the WGI provides various aspects of governance indicators, and there could be a co-linearity problem in model estimation. Following Globerman and Shapiro (2002), we use the Principal Component Analysis (PCA) approach to extract the first component of the six indicators (VA, PS, GE, RQ, RL, CC) as the composite measure of governance quality (Gov).
To avoid model misspecification, we include some control variables based on the existing literature. We control country attributes such as GDP (GDP), while equity
market development is controlled by the ratio of stock market capitalization to GDP (Cap). In general, better stock market development often implies higher liquidity and more investment opportunities, which will attract more foreign investors by lowering the cost of financial intermediation (Levine and Zervos, 1996). On the one hand, a well developed equity market can attract institutional investors, reducing home bias.
Alternatively, domestic investors may ceteris paribus have less incentive to diversify their portfolios.
Mondria and Wu (2010) and Baele, et al. (2007) suggested that financial openness and trade openness can reduce home bias. We use two proxies to examine the marginal effects of trade and financial openness on cross-border investment. The sum of imports and exports scaled by GDP is used to proxy trade openness (Trade), while credit to the private sector as a share of GDP is employed to proxy financial depth (Fin). Moreover, Mann and Meade (2002) proposed that countries with a higher share of financial assets is seen as having a less diversified financial system and are not attractive for institutional investors. We use central government consumption as a share of GDP to proxy government size (Size), while the country size is proxied by the total population (Pop) in a particular country. Our measure of a country’s information penetration (IP) includes Internet users per 100 people, the number of secure Internet servers, and the number of mobile telephone subscribers. The
Internet
norm represents the number of Internet users per 100 people normalized by GDP per capita (in thousands of US dollars). The normalized procedure will be applied to other information channels in our estimation models. The data sources and definitions are summarized in Panel A of Table 2.2, while Panel B presents the descriptive statistics and correlations for variables used in this research.46
Table 2.2 Variables’ Definitions and Descriptive Statistics
Panel A
Variable Definition Source
Home Bias equity home bias measures the equity portfolio holdings’
deviation from CAPM CPIS
Log GDP logarithm of real GDP in 1990 US$ WDI
%Trade sum of exports and imports scaled by GDP WDI
%Cap market capitalization of listed companies scaled by GDP WDI
%Size central government consumption as a share of GDP WDI
Log Pop logarithm of total population WDI
%Fin credit to the private sector as a share of GDP WDI GOV first principal component of Governance Indices (VA, PS, GE,
RQ, RL, CC developed by Kaufmann et al. (1999) Kaufmann et al. (1999) Internet Internet users per 100 people normalized by GDP per capita WDI
Servers number of secure Internet servers normalized by GDP per
capita WDI
Mobiles number of mobile telephone subscribers normalized by GDP
per capita WDI
N = 344; correlations greater than 0.13 are significant at the 0.05 level; correlations greater than 0.17 are significant at the 0.01 level.
2.3.3 Model Specification
To test our argument that both governance quality and information penetration may have a negative effect on investment home bias, we construct an econometric model for home bias as a function of governance and information penetration and other control variables. The basic model we seek to address is whether governance quality affects international investment. We specify our basic model as follows:
0 1 2 3 4 5
6 7 8 ,
it it it it it it
it it it i t it
Home Bias GDP Size Cap Trade Pop
Fin Gov IP v
(2.4)
where
Home Bias denotes the home bias measure in the equity market, subscriber i
itPanel B
denotes the cross-section country, and t represents each time period (with t=2001, 2002,…,2009). The term β0 is a constant term while estimated coefficients βj
(j=1…8) capture the marginal effects of various independent variables. The term vi
captures unobservable country-specific effects, while δt denotes the time dummies.
Notation εit is the error term assumed to be independent and identically distributed and uncorrelated across countries and over time.
This study shows how cross-border investment is affected by the quality of governance environment. However, given the quality of governance, information diffusion also affects cross-border investment. As international investment bias can be affected by information penetration, there is no guarantee that governance quality decreases home bias. To address this problem, we analytically show how the change in home bias corresponding to a unit change in governance quality varies depending on the information penetration, as for instance, when IP equals Internetnorm. Since the interaction term attenuates the individual effect of information penetration
Internet
norm and governance quality Gov, omitting a significant interaction term will lead to a specification bias. Without an interaction term, the overall impact of a change in the flow of governance on home bias would be solely measured by β7. Therefore, we further specify our model by adding an interaction term as follows:0 1 2 3 4 5
6 7 8 9( ) ,
it it it it it it
it it it it it i t it
Home Bias GDP Size Cap Trade Pop
Fin Gov IP IP Gov v
(2.5)
In Equation (5), coefficient β9 captures the interaction effect of information penetration and governance quality on cross-border investment. With the interaction, the net marginal effect of the governance environment on home bias also depends on the level of information penetration. Therefore, the two variables Gov and IP×Gov modify the individual governance effect on home bias by β7+β9
IP. Moreover, ceteris paribus, the marginal effect of information diffusion is not constant, but varies with
governance quality. For example, with the interaction term, the marginal effect ofInternet
norm on home bias depends on the level of governance quality as:8 9
Our panel dataset consists of 44 countries and covers the period 2001-2009 (N=44 and t=9) in which the major developed and emerging markets are included.
To identify the governance effect on home bias, Model 1 is estimated by pooled OLS,
48
while Model 2 is estimated by the fixed effect after controlling other country characteristics such as government size and financial openness. For Model 3 to Model 5, we consider effects of information penetration on home bias using three explanatory variables to proxy information penetration (IP): Internet users per 100 people (Internetnorm), number of secure Internet servers (Serversnorm), and number of mobile telephone subscribers (Mobilenorm) normalized by GDP per capita. For Model 6 to Model 8, we examine the interaction of governance and information on home bias by including a set of the multiplicative interaction term of governance quality and information penetration in our model. To eliminate heterogeneity and potential misspecification bias, we use fixed effects panel data with the time dummies estimation approach. This procedure allows us to control time-invariant and country-specific unobservable effects, thus capturing the unobservable heterogeneity that causes the bias in the OLS regression.