Further Evidence on China’s B&R Impact on Host Countries’ Quality of Institutions

Abstract

As it has now been over eight years since the introduction of the Belt and Road initiative, analysts and scholars continue to debate the impact of the initiative on host countries. Recent academic studies reported a mostly positive relationship between China’s Belt and Road investments and host countries, including the governance of the latter. However, these studies mainly utilised data compiled by Chinese institutions, and questions have often been raised as to the quality of the data. This paper conceptually replicated one such investigation but utilised datasets compiled by non-China sources. The empirical methodology followed a novel sequential model selection approach, removing criticisms on the flexibility of generalised method of moments (GMM) panel estimations for researchers. Robust results were obtained: China’s Belt and Road investments had a significant positive impact on some measures of governance from the World Bank, and no significant negative effects were found.

1. Introduction

President Xi inaugurated the Silk Road Economic Belt in Kazakhstan on 7 September 2013 and the 21st Century Maritime Silk Road in Indonesia on 3 October 2013. The first initiative refers to the land corridor connecting China to Europe via Mongolia and Russia, Central and West Asia and the Indochina Peninsula. The second refers to the sea route connecting Southeast Asia, South Asia, Africa, the Middle East and Europe to China. The two initiatives are collectively known as the Belt and Road (B&R) [1]. This long-term infrastructure development policy plan aims to improve connectivity and cooperation among countries along the B&R routes. Huang [2] described B&R as the country’s greatest international economic ambition and provided four key points for understanding the initiative. First, the initiative’s purpose is to sustain China’s economic growth by developing new markets. Second, the intention is to exert more international influence and reshape global economic development policies away from those prescribed by the International Monetary Fund and the World Bank. Third, the initiative comprises more than infrastructure building, encompassing ‘policy dialogue, infrastructure connectivity, unimpeded trade, financial support and people-to-people exchange’. Fourth, it remains to be seen whether B&R will bring about the benefits touted. Chohan [3] investigated B&R through the lens of the surplus recycling of China’s industrial/manufacturing output and both human and financial capital. The official publication [4] elaborated on the B&R framework as consisting of six corridors and six means of communication across multiple countries and multiple ports. The corridors are the New Eurasian Land Bridge Economic Corridor, the China–Mongolia–Russia Economic Corridor, the China–Central Asia–West Asia Economic Corridor, the China–Indochina Peninsula Economic Corridor, the China–Pakistan Economic Corridor and the Bangladesh–China–India–Myanmar Economic Corridor. The communication modes are rail, roads, maritime transport, aviation, pipelines and aerospace integrated information networks. According to China’s official B&R web portal [1], the initiative has the support of over 100 countries and international organisations. On 17 March 2017, the United Nations Security Council adopted Resolution 2344, which in paragraph 34, pushed for the strengthening of regional economic cooperation through regional initiatives such as B&R and regional development projects.

To the critics of B&R, it is just another extension, albeit on a transcontinental scale, of China’s debt diplomacy. Brautigam [5] provided an insightful look at this alleged ‘debt-trap’, calling it a meme started by an Indian think tank and adopted by opponents of B&R. The author examined several ‘debt-trap’ cases regularly used to justify the meme, but found no evidence for it. In one instance, it was more a ‘lending trap for China’, as there were no tools for the country to secure repayments. Similarly, Singh [6] found no evidence supporting the ‘debt-trap’ narrative. Another criticism of B&R is the potential export of corruption along the corridors. The World Bank [7] highlighted this as a key risk area in infrastructure projects. This is understandable, as can be seen in the heatmap of Transparency International’s Corruption Perceptions Index (CPI) 2020, which is reproduced in Figure 1.

Although China’s 2020 CPI is mid-ranking, it is below all 38 member countries [8] of the Organisation for Economic Co-operation and Development (OECD), apart from Colombia and Mexico. Governance through effective, accountable and inclusive institutions is a key approach of the United Nations Development Programme (UNDP) to achieve the 2030 Agenda for Sustainable Development [9]. The B&R initiative is also part of the United Nations Environment Programme’s regional initiatives for increasing environmental sustainability through the Belt and Road Initiative International Green Development Coalition launched in Beijing (25–27 April 2019) [10]. A recent study [11], questioned whether China’s overseas investment in B&R countries had improved institutional quality (including control of corruption). The dataset for institutional quality was the World Governance Indicators (WGI) of the World Bank, covering six governance facets: (1) Voice and Accountability, (2) Political Stability and Absence of Violence/Terrorism, (3) Government Effectiveness, (4) Regulatory Quality, (5) Rule of Law and (6) Control of Corruption. Data for China’s overseas investment were extracted from the Statistical Bulletin of China’s Outward Foreign Direct Investment reported by China’s Ministry of Commerce (MOFCOM). Except for Voice and Accountability, the authors found positive influence on the other five institutional quality dimensions. In a report to members and committees of the United States Congress on tracking China’s overseas activities, Schwarzenberg [12] highlighted several issues with data from MOFCOM including only recording officially approved projects. Furthermore, the definition of construction activities versus investment projects has changed over the years making trend analyses difficult. In addition, the pervasive use of offshore financial centres for overseas investment by Chinese entities makes data collection/tracking challenging. There are nongovernmental organisations tracking China’s economic activities such as the China Global Investment Tracker (CGIT) published by the American Enterprise Institute (AEI) and the Heritage Foundation [13]. CGIT has its own issues, as only projects at USD 100 million or above are included and the recording occurs when a project is announced and not updated for changes, for example, cancellation, postponement, scale changes, etc. Still, as the author noted, data from databases such as CGIT is often used and quoted in global policy discourse. In the 2020 Global Go To Think Tank Index Report [14] of The Heritage Foundation is ranked sixth whilst AEI is ranked seventeenth. A ranking by the website AcademicInfluence.com (accessed on 11 November 2021) [15] has both the Heritage Foundation (second) and AEI (sixth) in its 2021 top ten most influential think tanks. Hence, the purpose of this paper is to investigate whether the conclusion of [11] holds true when data from CGIT (with different collection criteria) are used instead of data compiled by MOFCOM. In addition, as far as the authors are aware, studies in this area using the GMM linear dynamic panel model estimation method have yet to use a sequential model selection approach as shown in [16]. The quality of inference in GMM estimation is highly sensitive to the selections made by researchers from the numerous choices available [17]. Cheng and Bang [18] showed that studies need to be more explicit regarding the specifications and rationales for GMM model selection. A systematic selection approach adapted from [16,19] is used here.

2. Literature Review

The literature on foreign direct investment (FDI) and quality of institutions (QI) is part of the international business research on location-choice for FDI and the broader study on the determinants of FDI [20,21,22]. FDI is broadly defined as direct equity investments including greenfield, joint ventures and mergers and acquisitions overseas by firms, namely multinational enterprises (MNEs) [23]. For institutions in a country/location, North [24] defined these as ‘humanly devised constraints that structure political, economic and social interactions’, both formal and informal. Better institutions increase the potential return of such interactions either from increases in benefits/profits or reductions in transaction costs. Hence, the nature of host countries’ institutions will have an impact on FDI.

The eclectic or OLI (ownership, location and internalisation) paradigm is extensively used as the analytical framework for determinants of FDI and the foreign activities of MNEs [25,26,27]. A firm’s overseas activities (investment or trade) are ascertained by the interface of the three sets of interacting OLI variables that, alone, are the components of three subparadigms. The first subparadigm, the ownership (O) of specific advantages versus other foreign or domestic firms in a host country, would increase the likelihood of an FDI. The second subparadigm, the location (L) attractions of alternative countries/locations that allow an MNE to increase or use its O competitive advantage, would impact whether the firm undertakes an activity at home or overseas. The third subparadigm, the internalisation (I) benefits, allows a firm to assess how it would engage in an overseas activity given its O advantage in a selected L. The greater the I benefit, the higher the likelihood of a firm engaging in an FDI versus other nonequity arrangements such as trade, licensing or outsourcing. From this eclectic paradigm, countries/locations where institutions do enable firms to maximise their OLI advantages would be attractive destinations.

From firm decision-making literature, Francis et al. [23] expanded on the institutional isomorphism (firms becoming similar) theory from [28] to form a multilevel theoretical framework on FDI. DiMaggio and Powell [28] described three institutional environment pressures—coercive, mimetic and normative. Coercive pressure emanates from institutions as defined by [24]. Mimetic pressure comes from uncertainty which promotes imitation. Normative pressure stems from professional consensus. Francis et al. [23] explained the changing isomorphic pressure firms face at different levels (country, industry and firm) depending on the stages of FDI. At the initial entry stage, coercive pressure at the country level and mimetic pressure at the industry level have the most influence. Host country legal and regulatory factors such as profit repatriation and intellectual property protection would affect investment decisions. Meanwhile, firms tend to follow FDI decisions already made by industry peers, as they provide ‘evidence’ of the suitability of a particular country. For countries, the policy implication is clear. Having institutions closer to global expectation (reducing coercive) would be attractive to foreign firms, especially if the country is not yet a popular FDI destination (mitigating mimetic). For subsequent FDI entries, a firm already has the practical experience and, hence, its internal professional consensus becomes more important (reliance on normative).

There is extensive empirical scholarship on quality of institutions (QI) being a key determinant of FDI. In a study of 164 countries from 1996 to 2006 (both developed and developing), Buchanan et al. [29] found a 1 to 1.69 standard deviation ratio relationship between QI and FDI. This accords with an earlier study by [30] that found institutions to be a significant determinant, separately from GDP per capita. Daude and Stein [31] and Ali et al. [32] showed the importance of QI on FDI, with results consistent for different model specifications and techniques. With China being a significant FDI recipient, Fan et al. [33] investigated whether China’s large FDI inflows, given its weaker QI, invalidate the positive relationship between QI and FDI. The authors concluded that, controlling for factors such as economic growth track record and demography, China’s level of FDI per capita is at comparable levels with other host countries with the familiar QI-FDI relationship.

In line with development in the global institutional framework [34,35], more recent and, for this paper, more relevant literature is investigating whether FDI leads to an improvement in QI. This is especially relevant given the widespread FDI outflows from China to developing and less developed countries, even before B&R. Fukumi and Nishijima [36] investigated 19 countries across Latin America and the Caribbean using data from the International Country Risk Guide (ICRG) [37], Freedom in the World [38] and the World Bank’s Database of Political Institutions 2000 [39]. The authors used simultaneous equations to avoid endogeneity biases and regression results point towards a bidirectional relationship between QI and FDI and indicate a virtuous cycle of FDI improving QI, which in turn attracts further investment. Similarly, the results of [40] investigating property rights also suggest the positive influence of FDI on QI. Foreign investment brings better manufacturing and production technology together with better social and institutional norms. The authors utilised a panel of 70 developing countries from 1981 to 2005 and argued that not only are lower tax rates important for foreign investors, but also better QI. This incentivises governments to provide better institutions as they compete for foreign investments. Economic data were sourced from UNCTAD and the World Bank whilst QI data were from ICRG. For a robustness check, the paper tested both FDI flows and FDI stock as an explanatory variable. The results were consistent across the two variables. The authors indicate that current FDI stock affects the quality of property rights protection, whereas FDI flows could be more a reaction to improving property rights protection. Instead of cross-country panel data, Dang [41] used within-country data from 60 provinces in Vietnam with the QI measured using a survey of private businesses’ qualitative assessment of provincial economic governance (Vietnam Provincial Competitiveness Index). Foreign investment was instrumented using the distance of a provincial capital to the main economic centres of Hanoi, Da Nang or Ho Chi Minh City. The results using two-step GMM confirm that FDI leads to improvement in QI and a size effect (larger disbursements of FDI deliver higher QI). Using similar methodology, Long et al. [42] also found a similar relationship of ‘FDI-induced institutional improvement’ for host regions in China. The authors examined two aspects of QI for Chinese domestic firms, namely tax/fee burdens and quality of legal protection. The paper utilised data from a 2005 survey conducted by the World Bank and the National Bureau of Statistics of China (NBSC) of 12,400 firms in 120 cities across all Chinese provinces except Tibet, the Chinese Private Enterprise Survey in 2006 and 2008 and NBSC’s industrial survey in 2000. Two instrumental variables selected for FDI were provincial highway density in 1937 (reflective of current infrastructure but not current local QI) and the number of World Heritage Sites in each province (correlation with foreign visitors/potential investors, but not local QI). Positive correlation is reported between higher levels of FDI (four years’ lag) and reduced tax/fee and higher quality rule of law.

A dissenting paper is [43]; using bilateral FDI data for 134 countries (developed and developing) from 1990–2009, the author found no significant positive effect (reduction in QI gap) of both developed and developing bilateral FDI into developing countries. On an aggregate level, developing countries’ FDI into a developing host country had a negative effect (increase in QI gap between aggregate QI and host country QI). The use of the QI gap could be problematic, as an increase in gap could be due to the host country’s QI improvement occurring at a slower rate versus the aggregate QI rate of investing countries. For 19 developing Asian countries from 2002–2015 (data from The Freedom House, The Heritage Foundation, ILO, UNESCO, World Development Indicators [WDI], World Economic Forum and WGI), Huynh et al. [44] investigated the three-way linkages between FDI, the shadow economy and QI. The authors found a bidirectional positive relationship between FDI and QI. The bidirectional relationship between QI and the shadow economy and between the shadow economy and FDI were both negative. Zhang et al. [45] investigated the spillover effects of China’s B&R investments into 36 developing countries from 2003 to 2017 using panel vector autoregressive modelling (with impulse response function, variance decomposition analysis and Granger causality test). The dataset used was a QI index of the six WGI measures created using the entropy-weighted method and data from the Human Development Index (UNDP). Spillovers were positive with the most impact on QI occurring in the fourth year after a Chinese investment. Pan et al. [11] looked specifically at China’s B&R investments (data from MOFCOM) impact on QI in host countries and found significant positive impact for most but not all measures of QI from WGI using the two-step GMM.

3. Empirical Model, Data and Methodology

3.1. Empirical Model

As in [11], the hypothesis of this paper is that China’s B&R investments have a positive impact on the QI of host countries. The model specification, Model (1), is set out below:

(1)${\mathrm{QI}}_{\mathrm{it}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{CISI}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{EV}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{ϵ}}_{\mathrm{it}}$

where i is the host country and t is the year. The β variables are coefficients, CISI is China’s FDI stock intensity (percentage of local GDP) in a host country, EV is a vector of explanatory variables capturing a number of key features of a host country (details in Section 3.2 below), μ is the country specific error term and $\mathsf{ϵ}$ is the idiosyncratic shock term.

Studies including [36,44] have shown the bidirectional causality between QI and FDI. There is also the persistence of institutions, and [45] provided evidence on multiyear QI spillover effects. Any of these would lead to potential endogeneity biases and, therefore, as in [11,46], a QI lagged term is added to Model (1) above to arrive at Model (2) below.

(2)${\mathrm{QI}}_{\mathrm{it}}={\mathsf{\beta }}_3{\mathrm{QI}}_{\mathrm{it}-1}+{\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{CISI}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{EV}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{ϵ}}_{\mathrm{it}}$

3.2. Data

For the QI variable, the widely used six measures from WGI (over 200 countries and territories from 1996 to 2020) were utilised. Here, QI is defined along the lines of governance, which, for WGI, ‘consists of the traditions and institutions by which authority in a country is exercised’ [47]. WGI measures the perceptions of the citizens of a country/territory for the following six categories:

• Voice and Accountability (VA): ability to select own government, freedom of expression and association, together with having a free media.

• Political Stability and Absence of Violence/Terrorism (PV): occurrence of political instability and/or politically motivated violence, including terrorism.

• Government Effectiveness (GE): focus on the quality of public services, civil service and policy formulation/implementation, degree of civil service political interference and government’s commitment to actioning policies.

• Regulatory Quality (RQ): assessment of government’s ability to formulate/implement sound policies and regulations that encourage private sector development.

• Rule of Law (RL): trust in the rules of society, especially contract enforcement, property rights, the police and the courts, along with control of crime and violence.

• Control of Corruption (CC): from abuse of public power for private gain (both small and large) to the disproportionate influence on government by certain groups (e.g., elites, private interests).

The methodology of WGI and analytical issues are set out in [48]. Key notable features include the over 30 data sources from surveys of households and firms and assessments from organisations (governmental, nongovernmental and commercial providers). The WGI dataset was derived solely from perceptions-based measures of governance. Each aggregate measure was constructed using the unobserved components model statistical tool generating results in units of the standard normal distribution ranging from −2.5 to +2.5, together with standard errors; higher values represent better governance. The choice of unit also meant that the world average for each year was zero. The dataset allowed for meaningful relative annual and across-time comparison of countries, but obviously not for global average as this was set at zero for each year. WGI aggregate measures are also reported in percentile rank form (0 = lowest to 100 = highest). This latter reporting was used for presentation in the WGI’s website [49] and was chosen for the empirical work in this paper as well.

For CISI, instead of MOFCOM, data on China’s investment in B&R countries were taken from the CGIT database, which the publisher states is the only public dataset in the world covering both China’s investment and construction activities outside the country [13]. CGIT covers announced large Chinese transactions at current USD since 2005. The use of CGIT is supported by [50] who found that ‘predicted’ FDI from a gravity model exerts significant influence on QI. The cumulative sums of annual announced Chinese investments provided the stock of China’s investment in B&R host countries. CISI, as in [11], was measured as China’s investment stock in a host country as a percentage of the latter’s GDP. Other studies used FDI annual flow instead of stock (e.g., [51,52]), but, as FDI flows tend to be intermittent and volatile, stock was a preferred measure here as in [30,31].

The list of B&R countries (just over 140) was taken from The Green Belt and Road Initiative Centre (Green BRI Centre) [53], as a listing is not available yet on the official B&R website. The Green BRI Centre, promoting an ecologically friendly and green B&R, is part of the International Institute of Green Finance at the Central University of Finance and Economics in Beijing. Not all B&R countries have investments from China as tracked by CGIT, which reduced the number of countries in the dataset to 92. To maximise the number of datapoints, data from 2005, the year CGIT started, were incorporated. This is reasonable, as China had been investing overseas long before the B&R initiative. The number of countries was reduced further to 45 (see Table 1 below) with the removal of war-torn countries (Afghanistan, Iraq and Yemen), western countries and Singapore (with very high governance standards) and countries with less than four investment observations across the sample period 2005–2020. The latter provided a minimum of two uncorrelated datapoints per country for our regression tests.

The explanatory variables vector EV follows [11] and includes the following:

• GDP per capita (GPC): a country’s gross domestic product at current USD divided by its population at mid-year.

• FDI intensity (FISI): the total FDI stock (including China’s) in a host country divided by its GDP. To be consistent with the derivation of CISI, FDI stock is the cumulative sum of FDI inflows into the host country starting from 2005.

• Trade openness (TO): the sum of export and import of goods and services as a percentage of GDP.

• Internet penetration (IT): the percentage of the population using the internet with a user defined as an individual who had accessed the internet via any devices (from mobile phones to digital televisions) in the previous three months.

• Natural resource contribution (NRC): the total natural resources rents from oil, natural gas, coal (hard and soft), mineral and forest, as a percentage of GDP.

Data for all the above variables, except FISI (indirect), were taken directly from WDI [54]. This dataset, which was compiled from officially recognised international sources, is the World Bank’s primary collection of development indicators. WDI contains a large number of indicators and many alternatives could easily be added to or replace the above, but to minimise differences from [11], similar explanatory variables were chosen.

3.3. Methodology

First, the panel dataset was tested for serial correlation to ascertain the appropriateness of pooled regression that would then be compared with panel regressions. For Section 4.1, a panel regression with fixed effects (FE) was checked against random effects (RE), reflecting the country heterogeneity effect. An FE model assumes the country specific effect is correlated with the regressors, whilst for RE, the country specific effect is idiosyncratic. FE estimators are only valid for in-sample inferences whilst RE estimators can provide outside sample inferences.

As noted in Section 3.1, Model (1) used in Section 4.1 has potential endogeneity biases. Hence, the inclusion of the lagged term in Model (2) is used in Section 4.2 to reflect endogeneity with estimations using GMM. As the panel dataset (see Section 3.2) has the number of groups (45 countries) far exceeding the sample time periods (16 years), the two-step difference GMM estimator of [55] was used. GMM is susceptible to instruments proliferation which could overfit instrumented regressors leading to biased estimators [56]. Kiviet et al. [17] noted the difficulty for researchers to make a ‘reasoned choice’ from the numerous implementations available due to the flexibility of GMM. To overcome this, a sequential model selection process adapted from [16,19] was followed and the detailed steps are set out in Appendix A.

As can be seen in Table 2, variables have significantly different scales and hence were standardised before empirical testing. For Model (1) in Section 4.1, the dependent variables are, respectively, the six measures of QI (VA, PV, GE, RQ, RL and CC) and the explanatory variables are CISI, GPC, FISI, TO, IT and NRC. For Model (2) in Section 4.2, an additional lagged term of each of the QI was added to the explanatory variables. Empirical analyses were conducted using the statistical software Stata [57].

4. Empirical Results

4.1. Model (1)

The panel dataset was checked for first-order autocorrelation and the H₀ of no autocorrelation was rejected for each dependent variable (see Table A1 in Appendix B); hence, pooled regression was not employed. For each measure of QI, a fixed effects (FE) model specification was tested against the random effects (RE) counterpart. The H₀ RE specification was rejected for all QI measures (see Table A2 in Appendix B). Then, each respective regression (FE) was rerun with robust errors. Table 3 below reports the final regression results for Model (1), where all variables were standardised. For all explanatory variables with more than one significant coefficient (here defined as having p-value < 0.05), the directional impact on different measures of QI is consistent. For zCISI (China’s investment stock intensity), it has significant positive impact on zVA (Voice and Accountability), zRL (Rule of Law) and zCC (Control of Corruption), three out of six of the QI measures. The effect on zRL is consistent with the results of [40] and other studies quoted in Section 2. For zCC, this could be indicative of a successful spillover effect of China’s anticorruption drive since 2012 under President Xi. Given the positive result of multiple international surveys (Harvard University’s Ash Center [58], Edelmann Trust Barometer [59] and the World Value Survey [60]) on Chinese citizens’ view of their government, one would expect a positive impact on zGE (Government Effectiveness) due to zCISI. However, there is no significant effect, perhaps reflecting China’s long-standing and strongly held foreign policy of noninterference in the internal affairs of other countries. Another surprise is the positive impact on zVA, as China is a one-party state, but this could be reflective of the country’s enormous success in mass poverty alleviation [61], leading to citizens feeling their voices are being heard [62]. This approach may have filtered through the country’s B&R dealings. Of all the explanatory variables, zGPC (GDP per capita) returns the largest impact on QI with a standard deviation movement in zGPC, resulting in a 20% and 17% standard deviation impact on zRQ (Regulatory Quality) and zRL, respectively. Both institutional measures require human capital investment and higher incomes would allow countries to nurture their talents, but also to bring in outside expertise. Overall FDI stock intensity (zFISI), contrary to most of the recent literature, shows no significant influence on any QI measure. zTO (Trade Openness) has a significant positive effect on zRQ. This reflects, perhaps, that with a greater share of overseas trade in a country’s GDP, it needs to follow the rules and regulations of international commerce more closely. zIT (Internet Penetration) has no significant relationship with any QI measure. The zNRC (Natural Resource Contribution) impact across two (zVA and zGE) of the six QI measures are significantly negative, consistent with the ‘natural resource curse’ (extensive literature). However, given the endogeneity issues (see Section 3.1) and the low R-squared for all six QI measures, especially zPV (which also failed the overall F-test, although it has no significant coefficient), one should not read too much into the regression results.

4.2. Model (2)

Table 4 below sets out the two-step difference GMM estimation results, using Stata module xtdpdgmm [19]. Additional model specification tests are reported in Table A3 below (Appendix B). Test results indicate the model specification is adequate with the [63] test (AB) for autocorrelation of the first-differenced residuals and the [64,65] J-statistic test (SH) for the validity of the overidentifying restrictions. AB H₀ for no autocorrelation of order one was rejected and H₀ for no autocorrelation of order two was not rejected for all measures of QI. This is required as the differenced model specifications with a lagged dependent variable serving as part of the explanatory variables set will, by design, have first order correlation, but not second order. SH H₀ was not rejected; this tests the validity of the overidentifying restrictions of the GMM estimation approach (using lagged variables as instruments) where the model by design has more equations than variables (overidentification). All p-values are within the suggested range (0.1–0.2) of [16], apart from zRQ, where the p-value is above the value range 0.1–0.2, but is still below the 0.25 level suggested by [66]. With GMM estimation, it is easy to increase the instruments used by including more lags of the explanatory variables leading to instrument proliferation, resulting in an increasingly higher p-value that could reach the improbable value of one. Excessive instruments lead to the overfitting of endogenous variables without removing the underlying endogeneity. To reduce this Type I false positive risk, Kiviet [16] and Roodman [66] recommended the p-value range 0.1–0.2 and 0.1–0.25, respectively.

The coefficients of all lagged standardised QI measures are all significant, confirming endogeneity as noted in Section 3.1. China’s investment stock intensity (zCISI) has only one significant impact on the QI measure zCC (positive) versus the results in Section 4.1. The influence here is much more significant, with one standard deviation movement in China’s direct investment stock intensity resulting in a 0.144 standard deviation increase in the Control of Corruption (zCC) measure in host countries. Similarly, GDP per capita continues to have significant positive impact, but now only on the QI measure zRL and not zRQ. Overall FDI stock intensity reports a negative significant influence on QI measure zGE, but this is small at a 4.6% standard deviation. zTO (Trade Openness) returns no significant coefficients. Internet usage penetration (zIT) continues to have no significant influence on any QI measure. The negative significant coefficient for zNRC on zGE is now smaller, whilst that for zVA is now not significant versus the results in Section 4.1. Additionally, zNRC has a new significant negative relationship with zRL for the current model.

The robustness of the results in Section 4.2 is provided by the sequential model selection process, which provided a systematic step-by-step approach to building the final model specifications, as noted earlier. In addition, this sequential model selection was rerun using FDI flows instead of FDI stock. The results are shown in Table A4 (Appendix C). The FDI flow results are consistent with those of FDI stock for China’s investment flow intensity in B&R host countries (positive impact on zCC, although the influence is smaller) and GDP per capita (significant positive coefficient for zRL and now also for zRQ). This shows that estimations using either China’s foreign investment stock or flow support the hypothesis that the country’s investment positively influence a host’s country institutional quality.

5. Conclusions

This paper conceptually replicates the study of [11] on whether China’s investment in B&R countries improves host country institutional quality. The novel approach is the use of datasets compiled by non-China sources and the application of a sequential model approach adopted from [16,19]. The model specification chosen for all six QI measures passes the AB first/second order autocorrelation test and are all within the SH overidentification test p-value range suggested by [16,66]. The empirical results broadly support the conclusion of [11] that China’s B&R investment has a positive impact on some of the WGI’s measures of QI. For this paper, the consistent impact of China’s investment stock and flow intensities (as measured by this study) on host countries’ QI, across Model (1) in Section 4.1 and Model (2) in Section 4.2, together with the latter’s robustness check, is on the single measure of zCC (Control of Corruption). This may be surprising at first, but can be understood given that anticorruption is a recurring top policy of President Xi since 2012. As state owned/controlled institutions are the main drivers of the B&R initiative, it would be astonishing if these institutions did not ‘toe the line’ in their overseas activities. Additionally, China has been consistently promoting itself as a responsible international player focussing on economic development through the B&R initiative. Hence, there would be across-the-board pressure to ensure that overseas activities adopt international best practices. The lack of impact on other QI measures can be interpreted as due to China’s long-held policy of noninterference in other countries’ internal affairs.

Another contribution of this paper is the support it lends to the hypothesis that a country’s GDP per capita is a key determinant of QI. The results from the two models and the robustness check, for some measures of QI (Regulatory Quality and Rule of Law), agree with [67] (using different measures of QI from ICRG) for GDP per capita of a country to be an important determinant (positive) of QI. Economic growth leads to better institutions, as the availability of resources allows for the building and strengthening of institutions.

This study adds to the body of literature on the impact of China’s B&R initiative across the globe. As noted by a number of authors, data from the initiative are still limited. As shown in the CGIT database, some B&R countries have had only limited investment, whilst others have more regular investments and new countries continue to join the initiative. This will remain a rich area for academic studies, and with the launch of the US-led G7’s Build Back Better World [68] and the EU’s Global Gateway [69], both in 2021, would provide opportunities for comparative studies between the former and B&R in the future.

Author Contributions

Conceptualisation, H.Y. and J.H.; methodology, H.Y.; software, H.Y.; validation, H.Y. and J.H.; formal analysis, H.Y.; investigation, H.Y.; resources, H.Y.; data curation, H.Y.; writing—original draft preparation, H.Y.; writing—review and editing, J.H.; visualisation, H.Y.; supervision, J.H.; project administration, H.Y. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Publicly available datasets were analysed in this study. These data can be found here: https://www.aei.org/china-global-investment-tracker/ (accessed on 19 November 2021); http://info.worldbank.org/governance/wgi/ (accessed on 24 November 2021); http://green-bri.com/countries-of-the-belt-and-road-initiative-bri/ (accessed on 19 November 2021); https://databank.worldbank.org/source/world-development-indicators (accessed on 23 November 2021). Stata do-files are available from the corresponding author on request.

Acknowledgments

The authors are grateful to editors and reviewers of this journal for providing valuable feedback to this paper, and the providers of the public datasets used in this study.

Conflicts of Interest

The authors declare no conflict of interest.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Figure and Tables

Figure 1. Country ranking. Source: www.transparency.org/cpi (accessed on 10 November 2021).

Table 1

Selected B&R countries.

East Asia and Pacific
Indonesia	South Korea	Myanmar	Malaysia	Vietnam
Cambodia	Laos	Mongolia	Thailand
South Asia
Bangladesh	Sri Lanka	Nepal	Pakistan
Middle East and North Africa
United Arab Emirates	Egypt	Iran
Europe and Central Asia
Greece	Kazakhstan	Portugal	Serbia	Uzbekistan
Hungary	Poland	Russian Federation	Turkey
Sub-Saharan Africa
Angola	Congo–Brazzaville	Guinea	Nigeria	South Africa
Cameroon	Ethiopia	Kenya	Sierra Leone	Zambia
Congo	Ghana	Namibia	Chad	Zimbabwe
Latin America and Caribbean
Chile	Ecuador	Guyana	Peru	Venezuela

Source: Authors’ selection based on data from CGIT and Green BRI Center.

Table 2

Summary statistics.

Variables	Description	Obs	Mean	Std Dev	Min	Max
VA	Voice and Accountability	720	35.35	23.51	0.48	91.83
PV	Political Stability	720	33.89	23.14	0.47	93.75
GE	Government Effectiveness	720	40.05	24.71	0.95	90.87
RQ	Regulatory Quality	720	38.39	25.24	0.00	92.72
RL	Rule of Law	720	35.87	24.51	0.00	89.47
CC	Control of Corruption	720	34.46	24.10	0.47	91.47
CISI	China FDI Stock Intensity	713	0.06	0.14	0.00	1.15
GPC	GDP per Capita	713	6263	8089	162	44,499
FISI	FDI Stock Intensity	668	0.22	0.32	0.00	2.35
TO	Trade Openness	683	75.97	38.10	0.17	210.40
IT	Internet Penetration	658	30.80	27.12	0.07	100.00
NRC	National Resource Contribution	669	10.58	10.71	0.02	58.65

Source: Authors’ calculation based on data from CGIT, WDI and WGI.

Table 3

Fixed effects panel regression.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	zVA	zPV	zGE	zRQ	zRL	zCC
zCISI	0.0644 **	0.0755	0.0199	0.0200	0.0368 **	0.0596 ***
	(0.0305)	(0.0398)	(0.0199)	(0.0256)	(0.0160)	(0.0138)
zGPC	−0.0194	0.0813	0.0997	0.204 ***	0.170 **	0.0465
	(0.0491)	(0.0832)	(0.0760)	(0.0672)	(0.0824)	(0.0617)
zFISI	−0.0698	−0.0142	−0.0199	−0.0134	−0.0257	−0.0190
	(0.0373)	(0.0501)	(0.0331)	(0.0356)	(0.0423)	(0.0381)
zTO	−0.0220	0.0161	0.0865	0.123 **	0.0467	0.0327
	(0.0465)	(0.0764)	(0.0432)	(0.0546)	(0.0618)	(0.0614)
zIT	−0.0401	−0.0133	0.0791	−0.0206	0.0482	−0.00621
	(0.0398)	(0.0768)	(0.0432)	(0.0414)	(0.0452)	(0.0435)
zNRC	−0.0738 **	−0.0680	−0.0980 ***	−0.0530	−0.0671	−0.0190
	(0.0285)	(0.0540)	(0.0317)	(0.0295)	(0.0344)	(0.0354)
constant	0.00192	0.00930	0.0161 ***	0.0305 ***	0.00585	0.00459
	(0.00315)	(0.00506)	(0.00434)	(0.00417)	(0.00459)	(0.00380)
Observations	615	615	615	615	615	615
R-squared	0.098	0.036	0.165	0.094	0.131	0.038
Number of ccode	44	44	44	44	44	44
F-value	2.853	1.617	5.322	2.693	9.226	17.910
Prob > F	0.020	0.166	0.000	0.026	0.000	0.000

Robust standard errors in parentheses, ccode = country code; *** p < 0.01, ** p < 0.05.

Table 4

Two-step difference GMM estimation.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	zVA	zPV	zGE	zRQ	zRL	zCC
L.zVA	0.490 ***
	(0.120)
L.zPV		0.870 ***
		(0.109)
L.zGE			0.588 ***
			(0.154)
L.zRQ				0.761 ***
				(0.215)
L.zRL					0.895 ***
					(0.118)
L.zCC						0.781 ***
						(0.102)
zCISI	0.0711	0.0476	0.0240	−0.0478	0.00885	0.144 **
	(0.0722)	(0.0428)	(0.0514)	(0.0474)	(0.0358)	(0.0641)
zGPC	0.0477	−0.0260	0.0437	0.0733	0.118 ***	0.0379
	(0.0271)	(0.0719)	(0.0354)	(0.0381)	(0.0254)	(0.0384)
zFISI	−0.0455	−0.0146	−0.0460 ***	−0.0194	−0.0408	−0.0452
	(0.0243)	(0.0403)	(0.0177)	(0.0263)	(0.0212)	(0.0245)
zTO	−0.0212	−0.0158	0.00357	0.00664	0.0265	−0.00376
	(0.0219)	(0.0285)	(0.0331)	(0.0234)	(0.0252)	(0.0382)
zIT	−0.0108	−0.0124	0.0522	0.0278	0.0138	−0.0224
	(0.0208)	(0.0464)	(0.0270)	(0.0226)	(0.0227)	(0.0252)
zNRC	−0.0288	−0.0342	−0.0247 **	−0.0174	−0.0356 ***	−0.00346
	(0.0209)	(0.0344)	(0.0122)	(0.0168)	(0.0123)	(0.0210)
Observations	574	574	574	574	574	574
Number of ccode	44	44	44	44	44	44
AR1	−3.275	−4.639	−3.200	−3.221	−4.705	−4.037
AR1 p-value	0.001	0.000	0.001	0.001	0.000	0.000
AR2	0.534	0.799	0.973	1.104	−0.00929	−0.680
AR2 p-value	0.594	0.424	0.330	0.270	0.993	0.497
SH2s3w chi2	39.22	40.08	42.19	37.34	39.79	39.53
SH2s3wProb > chi2	0.178	0.155	0.107	0.237	0.162	0.169
SHDF	32	32	32	32	32	32

Robust standard errors in parentheses; L.variable = lagged one period (t-1); ccode = country code; AR1/2 = AB autocorrelation of first differenced residuals with order 1 and 2, respectively; SH2s3w = SH overidentifying restrictions 2-steps with 3 updates to weighting matrix test; SHDF = SH overidentification test degrees of freedom; *** p < 0.01, ** p < 0.05.

Appendix A

The following sequential model selection approach for two-step GMM difference model estimation is adapted from [16,19].

• Select an initial model

Model (2) from Section 3.1 above serves as the initial model with all regressors treated as endogenous. Time effects were checked and if the coefficients were insignificant, then they were dropped to reduce instrument count. Exclusion of time effects are also supported by [45]. The authors show the multiyear spillover effects of FDI on host countries’ QI, reflecting that changes to QI tend to take time/years to take effect. Standard level instruments and robust standard errors [70] were also chosen.

• 2.

  Shortlist valid model permutations

Arellano and Bond [63] test for autocorrelation of the first-differenced residuals to the third lag was used to shortlist model permutations. Permutations failing the test were dropped.

• 3.

  Regressors lags check

Andrews and Lu [71] model and moment selection criteria (MMSC) based on likelihood selection criteria (Akaike [AIC], Bayesian [BIC] and Hannan–Quinn [HQIC] information criteria) were used to select the preferred model specification.

• 4.

  Curtailing instruments

The [64,65] J-statistic test was used to determine the validity of the overidentifying (overid) restrictions. Preferred instruments selection would be for p-values to be within the suggested range 0.1 and 0.2 from [16]. The lag selection which had the maximum number of p-values within the range across the six measures of QI was chosen.

• 5.

  Check for additional predetermined and exogenous regressors

Permutations for different regressors as endogenous, predetermined and exogenous were run together with the tests from List 3 above and List 4 above. An additional underidentification (underid) test from [72] was included. However, there is currently scant literature providing guidance on suitable selection criteria. There is tension between underid and overid. To pass the underid H₀ at p < 0.05 would likely result in failing the overid H₀ test. The test results were then averaged for each model specification to allow for the selection of the best match for all six measures of QI. A pragmatic approach was adopted selecting the model that minimises the average MMSC test results and at same time are within reasonable ranges for both overid and underid tests.

• 6.

  Check for additional instruments from [73]

The incremental overid test was used for selection, with the suggested p-value range of 0.5 to 0.7 for regressor groups from [16].

• 7.

  Check for nonlinear moment conditions from [74]

The same test from List 6 above was used, but as there is only one additional instrument, the suggested p-value range 0.4 to 0.6 from [16] was used.

Appendix B

Model specification tests:

Table A1

Wooldridge panel data autocorrelation test.

	(1)	(2)	(3)	(4)	(5)	(6)
D.DepVar	zVA	zPV	zGE	zRQ	zRL	zCC
Observations	522	522	522	522	522	522
R-squared	0.000	0.000	0.018	0.002	0.001	0.000
Wooldridge F	77.61	127.60	40.23	68.96	112.20	118.80
WProb > chi2	0.000	0.000	0.000	0.000	0.000	0.000

W = Wooldridge.

Table A2

Hausman specification test.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	zVA	zPV	zGE	zRQ	zRL	zCC
zCISI	0.063	0.069	0.014	0.017	0.033	0.055
	(0.013)	(0.023)	(0.014)	(0.013)	(0.013)	(0.015)
zGPC	0.022	0.219	0.205	0.268	0.231	0.149
	(0.034)	(0.058)	(0.036)	(0.035)	(0.036)	(0.040)
zFISI	−0.072	−0.010	−0.033	−0.020	−0.031	−0.029
	(0.017)	(0.032)	(0.019)	(0.018)	(0.019)	(0.021)
zTO	−0.009	0.096	0.130	0.150	0.075	0.070
	(0.022)	(0.040)	(0.024)	(0.023)	(0.024)	(0.027)
zIT	−0.042	−0.019	0.083	−0.021	0.047	−0.006
	(0.017)	(0.033)	(0.020)	(0.019)	(0.019)	(0.021)
zNRC	−0.083	−0.085	−0.121	−0.072	−0.085	−0.039
	(0.019)	(0.034)	(0.021)	(0.020)	(0.020)	(0.023)
constant	−0.018	0.003	−0.006	0.002	−0.004	−0.007
	(0.113)	(0.101)	(0.076)	(0.086)	(0.091)	(0.096)
Observations	615	615	615	615	615	615
Number of ccode	44	44	44	44	44	44
Hausman chi2	44.11	36.67	68.46	56.26	47.31	58.80
HProb > chi2	0.000	0.000	0.000	0.000	0.000	0.000

Standard errors in parentheses, H = Hausman.

Table A3

Additional 2-step GMM specification tests.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	zVA	zPV	zGE	zRQ	zRL	zCC
AR3	−0.876	−1.451	0.360	0.462	−0.633	0.451
AR3 p-value	0.381	0.147	0.719	0.644	0.527	0.652
SH2s2w chi2	31.52	31.95	36.87	32.36	32.89	33.20
2s2wProb > chi2	0.491	0.469	0.254	0.449	0.423	0.409
uid chi2	39.38	40.26	40.69	35.16	35.76	35.10
uidProb > chi2	0.206	0.180	0.168	0.366	0.340	0.369
uidDF	33	33	33	33	33	33
L2depvarDF	4	4	4	4	4	4
L2depvar	0.687	0.0271	0.523	0.356	0.413	0.438
L0zCISIDF	5	5	5	5	5	5
L0zCISI	0.536	0.224	0.494	0.639	0.740	0.428
L2zFISIDF	6	6	6	6	6	6
L2zFISI	0.448	0.069	0.585	0.708	0.976	0.041
L0z4DF	24	24	24	24	24	24
L0z4	0.646	0.560	0.160	0.232	0.196	0.404

AR3 = AB autocorrelation of first differenced residuals with order 3; SH2s2w = SH overidentifying restrictions 2-steps with 2 updates to weighting matrix test; uid = underidentification; DF = degrees of freedom; L2 = lag 2 periods, L0 = current.

Appendix C

Robustness check:

Table A4

Two-step difference GMM estimation.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	zVA	zPV	zGE	zRQ	zRL	zCC
L.zVA	0.596 ***
	(0.102)
L.zPV		0.805 ***
		(0.116)
L.zGE			0.632 ***
			(0.0854)
L.zRQ				0.776 ***
				(0.116)
L.zRL					0.998 ***
					(0.118)
L.zCC						0.816 ***
						(0.0891)
zCIFI	0.0138	0.0445	−0.0132	−0.0118	0.00683	0.0265 **
	(0.0186)	(0.0241)	(0.0122)	(0.00950)	(0.00880)	(0.0121)
zGPC	0.00235	−0.172	0.0391	0.0999 **	0.145 **	−0.0463
	(0.0582)	(0.128)	(0.0670)	(0.0412)	(0.0643)	(0.134)
zFIFI	−0.00617	−0.00815	−0.00619	0.00348	0.00279	−0.00170
	(0.00729)	(0.00957)	(0.00923)	(0.00671)	(0.00510)	(0.00554)
zTO	−0.0946 **	−0.201 **	0.0756	0.0347	0.0353	−0.0987
	(0.0471)	(0.0928)	(0.0618)	(0.0434)	(0.0532)	(0.0850)
zIT	−0.000456	0.0755	0.0129	0.0131	0.0420	−0.0267
	(0.0232)	(0.0611)	(0.0323)	(0.0323)	(0.0385)	(0.0519)
zNRC	−0.0200	0.0405	−0.0451	−0.0111	−0.00809	0.00112
	(0.0193)	(0.0368)	(0.0259)	(0.0211)	(0.0182)	(0.0441)
Observations	574	574	574	574	574	574
Number of ccode	44	44	44	44	44	44
AR1	−3.950	−4.171	−4.183	−3.868	−4.688	−3.932
AR1 p-value	0.000	0.000	0.000	0.000	0.000	0.000
AR2	0.102	0.635	0.871	1.180	−0.0424	−0.988
AR2 p-value	0.919	0.525	0.383	0.238	0.966	0.323
SH2s3w chi2	42.36	43.68	43.67	43.90	43.99	44
2s3wProb > chi2	0.216	0.177	0.178	0.172	0.169	0.169
SHDF	36	36	36	36	36	36

Robust standard errors in parentheses; L.variable = lagged one period (t-1); CIFI (China investment flow intensity = China FDI flow divide by host country GDP); FIFI (Total FDI flow intensity = FDI flow divide by host country GDP); ccode = country code; AR1/2 = AB autocorrelation of first differenced residuals with order 1 and 2, respectively; SH2s3w = SH overidentifying restrictions 2-steps with 3 updates to weighting matrix test; SHDF = SH overidentification test degrees of freedom; *** p < 0.01, ** p < 0.05.

References

1. B & R Initiative-Belt and Road Portal. Available online: https://eng.yidaiyilu.gov.cn/ztindex.htm (accessed on 7 November 2021).

2. Huang, Y. Understanding China’s Belt & Road Initiative: Motivation, Framework and Assessment. China Econ. Rev.; 2016; 40, pp. 314-321. [DOI: https://dx.doi.org/10.1016/J.CHIECO.2016.07.007]

3. Chohan, U.W. What Is One Belt One Road? A Surplus Recycling Mechanism Approach. The Belt and Road Initiative; Brill Nijhoff: Leiden, The Netherlands, 2018; pp. 205-219. ISBN 9789004373792

4. Office of the Leading Group for the Belt and Road Initiative. Building the Belt and Road: Concept, Practice and China’s Contribution; 1st ed. Foreign Languages Press Co. Ltd: Beijing, China, 2017; ISBN 9787119108100

5. Brautigam, D. A Critical Look at Chinese ‘Debt-Trap Diplomacy’: The Rise of a Meme. Area Dev. Policy; 2020; 5, pp. 1-14. [DOI: https://dx.doi.org/10.1080/23792949.2019.1689828]

6. Singh, A. The Myth of ‘Debt-Trap Diplomacy’ and Realities of Chinese Development Finance. Third World Q.; 2020; 42, pp. 239-253. [DOI: https://dx.doi.org/10.1080/01436597.2020.1807318]

7. World Bank. Belt and Road Economics: Opportunities and Risks of Transport Corridors; World Bank: Washington, DC, USA, 2019.

8. About the OECD—OECD. Available online: https://www.oecd.org/about/ (accessed on 10 November 2021).

9. Public Institutions for SDGs. Available online: https://publicadministration.un.org/en/Themes/Public-Institutions-for-SDGs (accessed on 9 April 2022).

10. The Belt and Road Initiative International Green Development Coalition (BRIGC) | Asia and the Pacific | UNEP—UN Environment Programme. Available online: https://www.unep.org/regions/asia-and-pacific/regional-initiatives/belt-and-road-initiative-international-green (accessed on 21 April 2022).

11. Pan, C.; Wei, W.X.; Muralidharan, E.; Liao, J.; Andreosso-O’Callaghan, B. Does China’s Outward Direct Investment Improve the Institutional Quality of the Belt and Road Countries?. Sustainability; 2020; 12, 415. [DOI: https://dx.doi.org/10.3390/su12010415]

12. Schwarzenberg, A.B. Tracking China’s Global Economic Activities: Data Challenges and Issues for Congress; Congressional Research Service: Washington, DC, USA, 2020.

13. China Global Investment Tracker | American Enterprise Institute—AEI. Available online: https://www.aei.org/china-global-investment-tracker/ (accessed on 19 November 2021).

14. McGann, J.G. 2020 Global Go To Think Tank Index Report; TTCSP: Philadelphia, PA, USA, 2021.

15. Top Influential Think Tanks 2021 | Academic Influence. Available online: https://academicinfluence.com/inflection/influence/influential-think-tanks (accessed on 11 November 2021).

16. Kiviet, J.F. Microeconometric Dynamic Panel Data Methods: Model Specification and Selection Issues. Econom. Stat.; 2020; 13, pp. 16-45. [DOI: https://dx.doi.org/10.1016/j.ecosta.2019.08.003]

17. Kiviet, J.F.; Pleus, M.; Poldermans, R.W. Accuracy and Efficiency of Various GMM Inference Techniques in Dynamic Micro Panel Data Models. Econometrics; 2017; 5, 14. [DOI: https://dx.doi.org/10.3390/econometrics5010014]

18. Cheng, N.; Bang, Y. A Comment on GMM Estimation in IS Research. Proceedings of the ICIS, 2; Munich, Germany, 15–18 December 2019.

19. Kripfganz, S. Generalized Method of Moments Estimation of Linear Dynamic Panel Data Models. Proceedings of the Stata Conference; London, UK, 5 September 2019.

20. Batschauer da Cruz, C.B.; Eliete Floriani, D.; Amal, M. The OLI Paradigm as a Comprehensive Model of FDI Determinants: A Sub-National Approach. Int. J. Emerg. Mark.; 2022; 17, pp. 145-176. [DOI: https://dx.doi.org/10.1108/IJOEM-07-2019-0517]

21. Nielsen, B.B.; Asmussen, C.G.; Weatherall, C.D. The Location Choice of Foreign Direct Investments: Empirical Evidence and Methodological Challenges. J. World Bus.; 2017; 52, pp. 62-82. [DOI: https://dx.doi.org/10.1016/j.jwb.2016.10.006]

22. Kim, J.U.; Aguilera, R.v. Foreign Location Choice: Review and Extensions. Int. J. Manag. Rev.; 2016; 18, pp. 133-159. [DOI: https://dx.doi.org/10.1111/ijmr.12064]

23. Francis, J.; Zheng, C.; Mukherji, A. An Institutional Perspective on Foreign Direct Investment. Manag. Int. Rev.; 2009; 49, pp. 565-583. [DOI: https://dx.doi.org/10.1007/s11575-009-0011-x]

24. North, D.C. Institutions. J. Econ. Perspect.; 1991; 5, pp. 97-112. [DOI: https://dx.doi.org/10.1257/jep.5.1.97]

25. Dunning, J.H. Toward an Eclectic Theory of International Production: Some Empirical Tests. J. Int. Bus. Stud.; 1980; 11, pp. 9-31. [DOI: https://dx.doi.org/10.1057/palgrave.jibs.8490593]

26. Dunning, J.H. Location and the Multinational Enterprise: A Neglected Factor?. J. Int. Bus. Stud.; 1998; 29, pp. 45-66. [DOI: https://dx.doi.org/10.1057/palgrave.jibs.8490024]

27. Dunning, J.H. The Eclectic Paradigm as an Envelope for Economic and Business Theories of MNE Activity. Int. Bus. Rev.; 2000; 9, pp. 163-190. [DOI: https://dx.doi.org/10.1016/S0969-5931(99)00035-9]

28. DiMaggio, P.J.; Powell, W.W. The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in Organizational Fields. Am. Sociol. Rev.; 1983; 48, pp. 147-160. [DOI: https://dx.doi.org/10.2307/2095101]

29. Buchanan, B.G.; Le, Q.V.; Rishi, M. Foreign Direct Investment and Institutional Quality: Some Empirical Evidence. Int. Rev. Financ. Anal.; 2012; 21, pp. 81-89. [DOI: https://dx.doi.org/10.1016/j.irfa.2011.10.001]

30. Bénassy-Quéré, A.; Coupet, M.; Mayer, T. Institutional Determinants of Foreign Direct Investment. World Econ.; 2007; 30, pp. 764-782. [DOI: https://dx.doi.org/10.1111/j.1467-9701.2007.01022.x]

31. Daude, C.; Stein, E. The Quality of Institutions and Foreign Direct Investment. Econ. Politics; 2007; 19, pp. 317-344. [DOI: https://dx.doi.org/10.1111/j.1468-0343.2007.00318.x]

32. Ali, F.A.; Fiess, N.; MacDonald, R. Do Institutions Matter for Foreign Direct Investment?. Open Econ. Rev.; 2010; 21, pp. 201-219. [DOI: https://dx.doi.org/10.1007/s11079-010-9170-4]

33. Fan, J.P.H.; Morck, R.; Xu, L.C.; Yeung, B. Institutions and Foreign Direct Investment: China versus the Rest of the World. World Dev.; 2009; 37, pp. 852-865. [DOI: https://dx.doi.org/10.1016/j.worlddev.2008.07.016]

34. Cantwell, J. Location and the Multinational Enterprise. J. Int. Bus. Stud.; 2009; 40, pp. 35-41. [DOI: https://dx.doi.org/10.1057/jibs.2008.82]

35. Dunning, J.H. Location and the Multinational Enterprise: John Dunning’s Thoughts on Receiving the Journal of International Business Studies 2008 Decade Award. J. Int. Bus. Stud.; 2009; 40, pp. 20-34. [DOI: https://dx.doi.org/10.1057/jibs.2008.75]

36. Fukumi, A.; Nishijima, S. Institutional Quality and Foreign Direct Investment in Latin America and the Caribbean. Appl. Econ.; 2010; 42, pp. 1857-1864. [DOI: https://dx.doi.org/10.1080/00036840701748979]

37. The International Country Risk Guide (ICRG) | PRS Group. Available online: https://www.prsgroup.com/explore-our-products/international-country-risk-guide/ (accessed on 13 November 2021).

38. Freedom in the World | Freedom House. Available online: https://freedomhouse.org/report/freedom-world (accessed on 30 January 2022).

39. Database of Political Institutions | Data Catalog. Available online: https://datacatalog.worldbank.org/search/dataset/0039819 (accessed on 30 January 2022).

40. Ali, F.; Fiess, N.; Macdonald, R. Climbing to the Top? Foreign Direct Investment and Property Rights. Econ. Inq.; 2011; 49, pp. 289-302. [DOI: https://dx.doi.org/10.1111/j.1465-7295.2010.00319.x]

41. Dang, D.A. How Foreign Direct Investment Promote Institutional Quality: Evidence from Vietnam. J. Comp. Econ.; 2013; 41, pp. 1054-1072. [DOI: https://dx.doi.org/10.1016/j.jce.2013.05.010]

42. Long, C.; Yang, J.; Zhang, J. Institutional Impact of Foreign Direct Investment in China. World Dev.; 2015; 66, pp. 31-48. [DOI: https://dx.doi.org/10.1016/j.worlddev.2014.08.001]

43. Demir, F. Effects of FDI Flows on Institutional Development: Does It Matter Where the Investors Are From?. World Dev.; 2016; 78, pp. 341-359. [DOI: https://dx.doi.org/10.1016/j.worlddev.2015.10.001]

44. Huynh, C.M.; Nguyen, V.H.T.; Nguyen, H.B.; Nguyen, P.C. One-Way Effect or Multiple-Way Causality: Foreign Direct Investment, Institutional Quality and Shadow Economy?. Int. Econ. Econ. Policy; 2020; 17, pp. 219-239. [DOI: https://dx.doi.org/10.1007/s10368-019-00454-1]

45. Zhang, Y.; Cheng, Z.; He, Q. Time Lag Analysis of FDI Spillover Effect: Evidence from the Belt and Road Developing Countries Introducing China’s Direct Investment. Int. J. Emerg. Mark.; 2020; 15, pp. 629-650. [DOI: https://dx.doi.org/10.1108/IJOEM-03-2019-0225]

46. Bhargava, A.; Sargan, J.D. Estimating Dynamic Random Effects Models from Panel Data Covering Short Time Periods Panel Data Covering Short Time Periods’. Econometrica; 1983; 51, pp. 1635-1659. [DOI: https://dx.doi.org/10.2307/1912110]

47. WGI 2021 Interactive > Documentation. Available online: https://info.worldbank.org/governance/wgi/Home/Documents (accessed on 18 November 2021).

48. Kaufmann, D.; Kraay, A.; The, M.M.; Bank, W. The Worldwide Governance Indicators Methodology and Analytical Issues; World Bank Group: Washington, DC, USA, 2010.

49. WGI 2021 Interactive > Interactive Data Access. Available online: https://info.worldbank.org/governance/wgi/Home/Reports (accessed on 26 November 2021).

50. Gao, X.; Gu, Z.; Koedijk, K.G. On the Role of Projected FDI Inflows in Shaping Institutions. East Asian Econ. Rev.; 2020; 24, pp. 441-468. [DOI: https://dx.doi.org/10.11644/KIEP.EAER.2020.24.4.387]

51. Jiang, C.X.; Yang, Q.; Li, S.; Wang, Y. The Moderating Effect of Foreign Direct Investment Intensity on Local Firms’ Intangible Resources Investment and Performance Implications: A Case from China. J. Int. Manag.; 2011; 17, pp. 291-302. [DOI: https://dx.doi.org/10.1016/j.intman.2011.03.001]

52. Choi, S.; Furceri, D.; Yoon, C. Policy Uncertainty and Foreign Direct Investment. Rev. Int. Econ.; 2021; 29, pp. 195-227. [DOI: https://dx.doi.org/10.1111/roie.12495]

53. Countries of the Belt and Road Initiative (BRI)—Green Belt and Road Initiative Center. Available online: http://green-bri.com/countries-of-the-belt-and-road-initiative-bri/ (accessed on 19 November 2021).

54. World Development Indicators | DataBank. Available online: https://databank.worldbank.org/source/world-development-indicators (accessed on 23 November 2021).

55. Arellano, M.; Bover, O. Another Look at the Instrumental Variable Estimation of Error-Components Models. J. Econom.; 1995; 68, pp. 29-51. [DOI: https://dx.doi.org/10.1016/0304-4076(94)01642-D]

56. Roodman, D. Practitioners’ Corner: A Note on the Theme of Too Many Instruments. Oxf. Bull. Econ. Stat.; 2009; 71, pp. 135-158. [DOI: https://dx.doi.org/10.1111/j.1468-0084.2008.00542.x]

57. StataCorp. Stata Statistical Software: Release 14.2; StataCorp LP: College Station, TX, USA, 2015.

58. Long-Term Survey Reveals Chinese Government Satisfaction—Harvard Gazette. Available online: https://news.harvard.edu/gazette/story/2020/07/long-term-survey-reveals-chinese-government-satisfaction/ (accessed on 9 February 2022).

59. 2021 Edelman Trust Barometer | Edelman. Available online: https://www.edelman.com/trust/2021-trust-barometer (accessed on 9 February 2022).

60. WVS Database. Available online: https://www.worldvaluessurvey.org/WVSOnline.jsp (accessed on 9 February 2022).

61. Liu, M.; Feng, X.; Wang, S.; Qiu, H. China’s Poverty Alleviation over the Last 40 Years: Successes and Challenges. Aust. J. Agric. Resour. Econ.; 2020; 64, pp. 209-228. [DOI: https://dx.doi.org/10.1111/1467-8489.12353]

62. Zuo, C.; Wang, Z.; Zeng, Q. From Poverty to Trust: Political Implications of the Anti-Poverty Campaign in China. Int. Political Sci. Rev.; 2021; 019251212110017. [DOI: https://dx.doi.org/10.1177/01925121211001759]

63. Arellano, M.; Bond, S. Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Rev. Econ. Stud.; 1991; 58, pp. 277-297. [DOI: https://dx.doi.org/10.2307/2297968]

64. Sargan, J.D. The Estimation of Economic Relationships Using Instrumental Variables. Econometrica; 1958; 26, pp. 393-415. [DOI: https://dx.doi.org/10.2307/1907619]

65. Hansen, L.P. Large Sample Properties of Generalized Method of Moments Estimators. Econometrica; 1982; 50, pp. 1029-1054. [DOI: https://dx.doi.org/10.2307/1912775]

66. Roodman, D. How to Do Xtabond2: An Introduction to Difference and System GMM in Stata; SAGE Journals: New York, NY, USA, 2009; Volume 9.

67. Alonso, J.A.; Garcimartin, C.; Kvedaras, V. Determinants of Institutional Quality: An Empirical Exploration. J. Econ. Policy Reform; 2020; 23, pp. 229-247. [DOI: https://dx.doi.org/10.1080/17487870.2020.1719102]

68. FACT SHEET: President Biden and G7 Leaders Launch Build Back Better World (B3W) Partnership | The White House. Available online: https://www.whitehouse.gov/briefing-room/statements-releases/2021/06/12/fact-sheet-president-biden-and-g7-leaders-launch-build-back-better-world-b3w-partnership/ (accessed on 8 January 2022).

69. Global Gateway: Up to €300 Billion for the European Union’s Strategy to Boost Sustainable Links around the World | International Partnerships. Available online: https://ec.europa.eu/international-partnerships/news/global-gateway-eu300-billion-european-unions-strategy-boost-sustainable-links-around-world_en (accessed on 8 January 2022).

70. Windmeijer, F. A Finite Sample Correction for the Variance of Linear Efficient Two-Step GMM Estimators. J. Econom.; 2005; 126, pp. 25-51. [DOI: https://dx.doi.org/10.1016/j.jeconom.2004.02.005]

71. Andrews, D.W.K.; Lu, B. Consistent Model and Moment Selection Procedures for GMM Estimation with Application to Dynamic Panel Data Models; Elsevier: Amsterdam, The Netherlands, 2001; Volume 101.

72. Windmeijer, F. Testing Underidentification in Linear Models, with Applications to Dynamic Panel and Asset Pricing Models. J. Econom.; 2021; [DOI: https://dx.doi.org/10.1016/j.jeconom.2021.03.007]

73. Blundell, R.; Bond, S. Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. J. Econom.; 1998; 87, pp. 115-143. [DOI: https://dx.doi.org/10.1016/S0304-4076(98)00009-8]

74. Ahn, S.C.; Schmidt, P. Efficient Estimation of Models for Dynamic Panel Data. J. Econom.; 1995; 68, pp. 5-27. [DOI: https://dx.doi.org/10.1016/0304-4076(94)01641-C]