Institutions and Investment in the South and East Asia and Pacific Region : Evidence from Meta-Analysis

The ability of any country to attract inward FDI is considered to depend on the quality of the economic governance found within that state. This study seeks to explore the value of quality economic governance in attracting FDI in South and East Asia by meta-synthesising 771 estimates from 48 empirical studies published between 1980 and 2012. The authors find evidence that countries with good quality regulation and low levels of corruption are able to attract more FDI. They also find that countries with stronger legal systems are rewarded with higher levels of inward FDI. The results of the meta-regression suggest that investing in quality economic governance is rewarded in the global economy through high levels of inward FDI. Please notice: data available upon request (Published in Special Issue Meta-Analysis in Theory and Practice)


Introduction
Given the important role inward FDI can play in accelerating economic growth and transformation, developing countries are interested in attracting it. Amongst many other benefits such as creating employment and increasing technological development, inward FDI provides a more stable source of external financing than sources such as private debt and portfolio flows (Gastanaga et al., (1998); Globerman and Shapiro (2002a); Gani (2007)). Hence, countries in South and East Asia & Pacific region have liberalised their FDI regime and have pursued policies to attract FDI. They have also addressed various governance related issues to maximise such attraction. However, whether governance in these countries has achieved the purpose or not remains debatable.
Hence, the aim of this study is to contribute to evidence based policy making and to academic research on governance FDI relationship by providing meta synthesis of empirical evidence on various measures of governance and FDI, identifying factors causing heterogeneity in results, pointing to policy implications of our results and identifying potential avenues for future research within this field of study. In order to achieve the research aim, we raise the following questions: Is there a genuine effect of measures of governance on inward FDI? What is the directionality of such effect? We answer these questions by using all available empirical evidence obtained using systematic literature review from 1980 -2012 on effects of governance on inward FDI.
The definition of economic governance has evolved over the last few years. According to Kaufmann et al, (1999) Governance consists of the traditions and institutions by which authority in a country is exercised. This includes the process by which governments are selected, monitored and replaced; the capacity of the government to effectively formulate and implement sound policies; and the respect of citizens and the state for the institutions that govern economic and social interactions among them. Good, transparent and efficient governance in host countries ensures the safety of investments and thus attracts foreigners to invest. While there are many international and local authorities which give both subjective and objective information on governance, literature in the field of governance and inward FDI has used four main sources.
They are worldwide governance indicators provided by Kaufmann et al., (1996) under World Bank project, Freedom House measure of voice and accountability and political rights, Polity dataset and International Country Risk Guide (ICRG).
These different datasets on the quality of governance raise the issue of divergence in various measures of governance measured by these institutions. In order to synthesise governancegrowth effects, we delved deeper into the sub measures of each measure of governance to synthesise them based on the common sub measures. After observing the individual variables (representative sources) that have been used in measuring governance by these different data sources, we have classified governance into 7 measures based on World Wide governance measures. These seven measures are termed hereafter as voice and accountability, political stability, government effectiveness, regulation, law, corruption and aggregate governance.
The rest of the paper is organised as follows. Section 2 presents systematic review of literature with section 3 outlining the methodology used in the study. Section 4 presents results followed by concluding remarks in section 5.

Literature Review
While it is generally believed that good governance in a host country helps in attracting inward FDI, most of the empirical studies show that this is not the case. A systematic literature review of these empirical papers is presented here with a view to unearthing the issues within existing literature in terms of differences in their findings and the reasons causing such differences.

Theoretical view on governance and inward FDI
Two main theoretical frameworks have been used in explaining the relationship between economic governance and inward FDI. Firstly, Dunning's OLI framework (1980) explains various reasons for which an MNC enters into a host country. According to Dunning (1980) an MNC will enter a host country when each of the ownership, location and organisation factors are met. In this context, economic governance can be seen as a location factor which might deter investments or serve as a helping hand for foreign investors depending on the form of investment and the industry into which these investments flow.
Secondly, North (1991) in his institutional theory posits that institutions in the form of political, economic and structural interactions are human-made constraints which aim to decrease the level of uncertainty and allow for firms and individuals to interact efficiently. While governance aims to facilitate investments, they effect transaction (ex: cost of protecting property rights) and transformation costs (ex: by effecting production interruptions) which in turn effects the profitability of such investments (Dahlstrom and Johnson, 2007). Both Dunning's and North's theories suggest that based on contextual factors, governance can have either positive or negative effects on FDI.
Political stability 1 measures the solidity of government to political shocks, terrorism and domestic violence which can eventually reduce the risk of doing business and deter investments.
Government effectiveness measures the quality of public services and the insulation of those services from political pressure. Through government effectiveness, government can exert discretionary power on economic activities by designing and implementing economic policies which can either deter or encourage investments Shapiro (2002a), Anghel, (2004)). Studies by Gastanaga et al., (1998), Arbatli (2011), Gordon et al., (2012) andJensen (2003) show mixed effects of government effectiveness on FDI under different models.
Regulation as one of the elements of governance indicators is the widest and diverse measure as it includes regulation related to aspects such as intellectual property rights, environment regulations, restrictive capital controls, accounting standards and corporate governance and tax and tariffs. Regulation captures the ability of a government in generating these policies and using them to promote private sector development. Through these policies regulation can affect FDI as they can either speed up or delay the investments alongside affecting the cost of investments.
There have been only three studies that have looked at the impact of Shapiro (2002a), Jadhav (2012), Gordon et al., (2012) which reported positive and significant, positive and insignificant and mixed effect respectively leaving a scope for both further research and conclusive results.
Law can affect investments through various legal institutions and property rights protection. This measure also includes the quality of contract enforcement, the police, the courts and the likelihood of crime. In a country where there are weak legal institutions and property rights protection, very few foreign investors would like to invest as it would put their investments at risk and vice versa. Positive and significant effect is shown by Anghel (2004), Gani (2007), Jadhav (2012) and Fan et al., (2009). While Globerman and Shapiro (2002a) have shown positive and insignificant effect of rule of law, Arbatli (2011) has shown negative and insignificant effect. Studies by Busse and Hefeker (2005) and Gordon et al., (2012) have reported mixed effects.
Corruption is viewed as one of the important measures of governance as it has an important bearing on investments. Corruption measures the extent to which public goods are misused or used for private purposes by individuals. However, corruption cannot be considered in isolation from other governance related factors as bad governance is closely associated with corruption.
Corruption is considered to affect foreign investments in two waysincrease in cost of investments leading to decrease in profitability of such investments and increase in uncertainty levels in host country. Some studies have also shown that corruption 'greases the wheels' of investments rather than 'sands the wheels of investment' (Globerman and Shapiro (2002a), Gastanaga et al., (1998), Hsiao andShen (2003) and Teksoz (2004)).
Finally, Globerman and Shapiro (2002b), Globerman and Shapiro (2004), Hur et al., (2007), Adeoye (2009), Wernick, Haar and Singh (2009), Ali et al.,(2010, Muhammad et al. (2011), Luca and Spatafora (2012), Ahlquist (2008), Goodspeed et al.,(2010), Gordon et al., (2012) have focused on the effect of aggregate governance on inward FDI. Overall governance includes various political, legal and institutional factors in a country that can have a bearing on investments. While governance is expected to show a positive effect on foreign investments by providing impartial, effective and efficient conditions to operate, there is no conclusive evidence on this.
Mixed results and seemingly contradictory arguments on the empirical relationship between measures of governance and inward FDI can be attributed to various measurements, conceptual and methodological differences in these studies (appendix 5). Given this situation, policy makers may be uncertain as to what kind of policy they should propose in order to create a favourable investment climate for foreign investors in terms of economic governance.
In order to address the above inconclusiveness, as outlined in the introduction section this study has the following research aims; firstly, to deal with the effect of measures of governance on inward FDI and secondly with respect to heterogeneity. With regards to the effect, the following two questions are raised: firstly, is there any genuine effect of each measure of governance (voice and accountability, political stability, government effectiveness, regulation, corruption and rule of law) on the inward FDI into South and East Asia & Pacific countries? Secondly, what is the directionality of such effect? With respect to differences in reported results the following questions will be answered. Why do governance-FDI studies report such divergent results? Is the heterogeneity due to the data generating process or is it due to differences in research design? An overall summary of this study is given in appendix 6.

Methodology
The review methodology used in this thesis i.e the methods used for searching studies, study selection, critical evaluation and data extraction is informed by three sources. First, Cambell and Cochrane Collaboration guidelines on systematic reviews in health care and social policy; second, Centre for Reviews and Dissemination (CRD, 2009) of the University of York; third, Evidence for Policy and Practice Information and Co-ordinating Centre (EPPI-Centre) of the Institute of Education. Data analysis is informed by Doucouliagos et al., (2010), Doucouliagos andUlubasoglu (2008) and Stanley and Doucouliagos (2012). Reporting guidelines are informed by Stanley et al., (2013).
We started by establishing a pre-established search criteria to identify all studies in the English language on measures of dependent variable (FDI) and independent variable (governance). This is done in two stages: the first stage involves identifying databases for published and unpublished studies. The second stage involves specifying key words, searching databases and storing results. Bureau of economic research and IMF e-library were used. In addition to these databases, two search engines namely Google scholar and web of knowledge provided by University of Greenwich were utilised. In addition to the above, manual search was performed in order to identify grey literature using two approachessnowball approach and random search of studies in 5 journals. Under the snowball approach we have started with the reference list of studies identified through systematic review and proceeded to find new studies. These exhaustive searches were carried out to identify all possible studies on measures of governance and inward FDI.

Figure 1: Summary of methodology used in the study
With a pre-defined list of key words for measures of governance and inward FDI (appendix 1), 'title', 'abstract', 'text' and 'keyword' were searched in the above databases. The time period of the study was January 1980 -December 2012. A total number of 4996 studies were retrieved which have analysed the relationship between measures of governance and inward FDI. From this, 150 and 109 duplicate studies were removed using automatic and manual duplicate searches respectively. This left a total of 4728 unique studies for further screening. Figure 1 summarises the methodology used in this study.
The relevance of each study was checked based on whether the study estimates or analyses the relationship between measures of governance and inward FDI? While the earlier study is coded as 'E', later ones are coded are 'T'. If a study estimates and analyses the relationship then it is coded as 'TE'. Studies which do not satisfy any of these criteria are not included in meta-analysis. 131 studies were selected from the initial screening stage and these were considered for the critical evaluation stage. This was done using PIOS (Population-Independent variable-Outcome variable-Study design) criteria (appendix 2). While 94, 62, 68 and 94 studies have satisfied population, independent variable, outcome variable and study design respectively, only 40 studies have satisfied all four criteria (appendix 3). Another 8 studies were added to this number by hand searching, making a total of 48 studies for meta-analysis. Our exclusive search for studies on South Korea did not result in any records.
The following data were obtained from 48 studies. Firstly, bibliographical information such as name of the first author and University, year of publication of study and type of study (whether it is a published or unpublished study). Secondly, study characteristics such as kind of data used, information on dependent and independent variables such as their functional form and their data sources, and estimation methods. Thirdly, outcome related information such as estimated parameters, t values, standard errors, P value, Z value, F value for linear, non linear and squared terms was obtained.
The general form of econometric model used in the primary empirical studies with linear terms only (equation 1) and that with linear, non-linear and squared terms (equation 2) is shown below: Y it = α 0 + α 1 X it + γF it + ε it equation (1) Y it = α 0 + α 1 X it + α 2 X it · K it + α 3 X 2 it + γF it + ε it equation (2) In above equations, The effect size is measured using partial correlation to allow for meaningful comparison across different models. Various estimates of α 1 are converted into partial correlations using the formula r = [t/√ (t 2 + dof). Where, t stands for t -statistics of the multiple regression coefficient, dof stands for the degrees of freedom of the respective t -statistic.

Modelling simple and meta-regression analysis
The following equation is used for simple meta-regression analysis for estimating the overall effect after correcting for publication bias 1 : (3) The following equation is used for multiple meta-regression analysis for estimating the overall effect after correcting for publication bias: r ij =β 0 + β 1 SE 2 ij + + β 2 X ij + ε ij equation (4) The following equation is used for multiple meta-regression analysis with study and journal specific moderator variables. r ij =β 0 + β 1 SE 2 ij + β 2 X ij + β 3 Z j + ε ij equation (5)   It is worth highlighting at this point that while some studies have defined r on a scale of 0-1 from low to high governance, others have used it as 0-1 high to low governance. In order to aggregate estimates, we have rescaled all estimates as 0-1 low to high governance 2 . This was done by inversing and multiplying both coefficients and standard errors of estimates defined on the opposite scale (i.e. 0-1 high -low governance) by -1.

DISCUSSION OF RESULTS
We present and analyse results of simple meta-regression analysis (SMRA) and multiple metaregression analysis in this section. Before that, funnel plots and graphs of chronological order of estimates are presented. These graphs are used in order to offer a clear picture of the state of empirical knowledge in governance FDI studies. Aggregate Governance and FDI graphs above. Funnel plot is used to trace the relationship between the effect size which is measured using partial correlation (shown on X axis) and its precision measured as inverse of standard error (shown on Y axis). While high precision estimates are generally few and are compactly distributed at the top of the funnel, low precision estimates are at the bottom of the funnel and are widely distributed. One possible reason for the wide dispersion of estimates (which is the case in most of the graphs) is publication bias 2 (Doucouliagos and Ulubasoglu, 2008). In each of the above graphs, the centre of the plot represents the estimated true underlying effect of respective measure on growth. In contrast to graphs of political stability, the other graphs show wide dispersion of governance-inward FDI values around the central value.
We have tested for publication bias using Funnel Asymmetric Test (FAT) and Precision Effect Test (PET) (appendix 2.9). Despite the presence of publication bias, PET results suggests that there is genuine effect of each measure of governance on FDI along with aggregate governance. However, they are not robust in case of corruption and aggregate governance.

Aggregate Governance and FDI
The graph above shows the chronological order of estimates of measures of governance on inward FDI. X-axis shows end year of sample period and Y axis shows partial correlation.
Chronological ordering of graphs offers an insight into evolution of effect sizes and highlights the trends. With the exception of voice and accountability and political stability graphs, we see a downward trend in the estimates 3 . Downward trend has an important economic interpretation as it indicates that governance over a period of time has a declining effect on inward FDI as opposed to initial years of investment. As an alternative explanation, the downward trend can also be due to the fact that the econometric techniques have got better at controlling econometric problems and therefore smaller estimates are found.  indicating that a higher measure of each measure leads to more FDI. For instance, tighter regulations are associated with more FDI. In the case of corruption, results should be read inversely (due to rescaling) i.e. more corruption leads to less FDI. A positive effect of aggregate governance in the last column indicates that better governance is good for FDI.

SIMPLE META-REGRESSION ANALYSIS
Except for corruption, all the estimates are significant and unreliable as the R 2 value of each of these measures is very low (R 2 value ranges from 0.002 for aggregate governance to 0.33 for regulation). In addition to lower R 2 values, another shortcoming with this method of estimation is that the unweighted method treats all estimates equally with equal weight. Therefore studies with a large number of estimates can have an undue influence on the statistical assessment.
Therefore these results can be biased and misleading. Hence, following Stanley and Doucouliagos (2012), we ran the above models using the weighted least squares method where estimates are weighed by precision. we calculate precision as inverse of standard error as it is proven to be the optimal way of calculating weights from a statistical point of view.
When estimates are weighted by precision it is noted that, the size and significance of all measures has changed. A change in the size and significance of estimates indicates that undue influence by estimates is possibly removed. In terms of the effect, positive effect of regulation for instance indicates that more of regulation is good for FDI, whereas in the case of corruption, positive effect indicates that more corruption is still bad for FDI.

MULTIPLE META-REGRESSION ANALYSIS
It can be noted that in spite of weighting these estimates, R 2 values are still low indicating that the above models are weak in explaining the effect of governance on FDI. Hence similar to unweighted results these results can be misleading. One possible reason for a low R 2 value is due to the possible presence of heterogeneity. The expected value of governance FDI estimates will often depend on many other factors such as study, author and journal related. As these factors are unaccounted for, it is possible that both simple unweighted and weighted measures may capture the real effects of governance on FDI. Hence, we include the following moderator variables in order to validate simple meta-regression results. While some of the variables are included out of intuition (author specific variables) others are included as they are proved to have a significant effect by earlier meta studies (Doucouliagos and Ulubasoglu, 2008 This potential bias is removed by running MMRA using cluster analysis where each study is treated as a cluster. Results of cluster analysis are used to validate the results obtained by the weighted method. Before we analyse the results, it is worth noting the following five points. First of all it is important to comment on the good overall fit of the models. With an adjusted R 2 value ranging from 0. 07 for government effectiveness to 0.94 for political stability, these models have done a reasonable job explaining the heterogeneity in governance FDI literature (Stanley and Docouliagos, 2012). As compared to R 2 values of simple meta-regression results, the explanatory power of these models has increased after inclusion of moderator variables. Hence, these estimates are more reliable as compared to simple meta-regression estimates.
Secondly, we could not test for endogeneity due to the limited number of estimates (in most cases it was less than 10). Therefore, the effects reported can be due to the possible presence of causality. Thirdly, in terms of the statistical significance, all estimates are statistically significant. In the fourth instance, robustness of all these results is confirmed by cluster analysis. In the fifth instance, with more than 140 estimates and an adjusted R 2 value of more levels of this measure in these countries is associated with high levels of FDI into them.
These results reflect the tendency of MNC's to not to invest in countries where people are given voice to express their views and interests on government policies and processes.
Secondly, the overall effect of political stability on inward FDI is found to be positive and significant, which are in line with the findings reported by Anghel (2004), Baek and Qian (2011) and Busse et al., (2011). Therefore in general political stability does matter for foreign investors and it can be assumed that they like to invest in countries with high levels of stability. These results also suggest that foreign investors would not like to see frequent changes in the leadership and that they prefer long term government.
Thirdly, government effectiveness has positive and significant effect on FDI. A positive effect of government effectiveness indicates that higher levels of government effectiveness are correlated with higher levels of FDI. This contrasts the view that foreign investors are not happy with the cumbersome rules and tight procedures that effect the process and productiveness of investments (Khamfula, 2007;Gastanaga et al., 1998 andArbatli, 2011).
However, it is worth noting that with the lowest number of observations and a lower R 2 value, results for this measure are not strong enough. The lack of government effectiveness data may have caused biggest challenge in this area of research. Hence, further research is advised in this field of study before any strong conclusions can be made.
In the fourth instance, while on the one hand, effective and efficient policies along with incentives can attract foreign investments (Globerman and Shapiro, 2002a), on the other hand burdensome regulations can negate such investments (Jadhav, 2012). MMRA results on regulatory quality suggest that tighter regulations or regulations enforced in friendly manner are preferred by foreign investors as it has a positive and statistically significant impact on FDI. Therefore my results contrast the view that reducing the regulatory burden and making regulations easier for foreign investors would attract more FDI (Globerman and Shapiro, 2002b).
In the fifth instance, my results on rule of law contrast Arbatli (2011)'s view that a strong and impartial legal system is not preferred by foreign investors as the rule of law has a negative and statistically significant effect on inward FDI. As one would expect stronger laws to facilitate and protect investments, negative effect of law contradicts this view (Anghel, 2004;Gani, 2007;Jadhav, 2012;Fan et al., 2009). This shows a need for host country governments to develop their legal systems further and incline them in favour of foreign investors. Similar to the government effectiveness measure, despite a higher R 2 value, we have limited number of observations for this measure and hence these results must be interpreted carefully.
In the sixth instance, a positive sign of corruption indicates that the higher the corruption, lower is inward FDI. This suggests that foreign investors view corruption as an extra cost of operation rather than viewing it as helping hand. My results are not in line with the literature arguing that corruption is good for foreign investors (Gastanaga et al., 1998;Globerman and Shapiro, 2002a;Teksoz, 2004;Voyer and Beamish, 2004;Khamfula, 2007;Mathur and Singh, 2013). Negative effect inform us that investors prefer not to invest in countries with high corruption or where there is a lack of anti-enforcement laws. Results on corruption confirm the view that corruption sands the wheels of investment rather than greasing them. In the case of study related factors, whether a particular study has been published or not in an academic journal matters as it is statistically significant and have reported higher effects in the case of political stability as compared to estimates from unpublished studies. For instance, published studies on an average have reported a value of 0.33 as opposed to an overall effect of 0.26. Except in case of law, estimates using yearly data on FDI show a negative effect with reference to those using non-yearly data. This could presumably be because governance takes time to show its impact on FDI. There is also evidence to suggest that estimation techniques matter for governance FDI relationship. Models estimated using OLS and Probit techniques proved to be statistically significant compared to estimates estimated using other methods.
Governance and FDI data sources also mattered.
Under real world factors, as expected, country composition of the sample did matter as there were few regional specific effects. For instance, models including China in their list of sample countries have reported an average effect of -0.81 which is lower than those which did not include China. Similarly, inclusion of South Korea mattered as reported results are higher (i.e. 0.67) in case of corruption as opposed to an overall effect of 0.28. Thus we infer that governance FDI association did alter with inclusion or exclusion of any particular region.
These results are consistent with the notion that there can be many country specific factors that can have an important bearing on how governance works. It is interesting for future research to explore the reasons behind such differential impacts.
In the case of author related aspects, with the exception of political stability, law, corruption This has been a common problem with several other meta-analysis studies and thus highlights the need for more extensive research in this field with interaction and non-linear terms.
Thirdly, the quality of results in this study is as good as the quality of studies included for Whether results on the effect of governance on inward FDI would significantly differ if it 6 There were about 15 different types of interaction terms ranging from a minimum of 1 to a maximum of 11 observations. 7 There were only 2 different non-linear terms with less than 12 observations.
were possible to carry out research at regional level or by sector is uncertain (Globerman and Shapiro, 2002b