2.1 Economic freedom and the environment (Linkage T2)

It is generally accepted that economic freedom provides incentives that can result in an effective use of resources. Several researchers have noted a positive relationship between economic freedom and environmental quality or policy (Barrett and Grady, Reference Barrett and Grady2000; Fredriksson and Gaston, Reference Fredriksson and Gaston2000; Neumayer, Reference Neumayer2002). Bate and Montgomery (Reference Bate, Montgomery and Booth2005) suggest that higher levels of economic freedom (the freedom to trade, enter into contracts and start a business) foster the introduction of new and clean energy technologies in developing countries. However, it is also possible that strong intellectual property rights that are associated with greater economic freedom inhibit the diffusion of green technologies (Hall and Helmers, Reference Hall and Helmers2010). Similarly, if there are policies in developing countries that distort factor prices, such as subsidies for energy use or protection of domestic industries, the adoption and diffusion of new technology will be hampered.

Carlsson and Lundström (Reference Carlsson and Lundström2001) empirically test three hypotheses to explain the impact of economic freedom on carbon emissions – the efficiency effect, the trade regulation effect and the stability effect. Under the efficiency effect, higher levels of economic freedom lead to efficient and competitive markets, which in turn lead to more efficient political regulation, and reduce carbon emissions. Trade deregulation/liberalization has an indeterminate impact on carbon emissions. On the one hand, trade liberalization can result in more effective resource allocation due to competitive pressures in international markets and thus reduce emissions. On the other hand, in developing countries with less stringent environmental regulations, such liberalization can also result in a pollution haven effect where dirty industries relocate to countries with relatively lax environmental regulations. Finally, the stability effect can lead to more efficient investment and consumption decisions and encourage long-term investment which might have a less than salubrious impact on the environment. In a cross-country study of 75 countries over 1975–1995, Carlsson and Lundström (Reference Carlsson and Lundström2001) find that efficiency and trade regulation effects were significantly associated with decreasing carbon emissions and the stability effect had no discernible impact.

Wood and Herzog (Reference Wood and Herzog2014) estimate the direct causal effect of economic freedom on two indicators of air pollution (fine particulate matter concentrations and CO₂ emissions) for 105 countries. Controlling for income, political institutions and endogeneity in the income-pollution relationship, their results indicate strong evidence of a negative relationship between economic freedom and particulate matter. There is a negative but less robust impact of economic freedom on carbon emissions. Using a system GMM estimator with a panel dataset for 24 African countries over the 1995–2013 period, Adesina and Mwamba (Reference Adesina and Mwamba2019) demonstrate that increases in economic freedom are associated with improvements in environmental quality, measured using CO₂ emissions.

2.2 Corruption and the environment (Linkage T3)

Broadly defined as ‘the misuse of public office for private gain’ (Rose-Ackerman, Reference Rose-Ackerman1999; Treisman, Reference Treisman2000; Kunicova and Rose-Ackerman, Reference Kunicova and Rose-Ackerman2005; Transparency International, 2017), corruption plays a significant role in the degradation of the environment. There are two main linkages between corruption and the environment. First, government officials may control access to scarce valuable natural resources and can sell this access. Second, environmental issues such as environmental management, conservation, and enforcement institutions often receive insufficient funding, which creates opportunities for illegal activities. Scholars also argue that corruption can influence environmental degradation via relaxation of stringent environmental regulations (Lopez and Mitra, Reference Lopez and Mitra2000; Damania et al., Reference Damania, Fredriksson and List2003; Fredriksson et al., Reference Fredriksson, Vollebergh and Dijkgraaf2004). Winbourne (Reference Winbourne2002) notes that corruption contributes to the development of environmentally damaging policies and practices, and an inequitable allocation of environmental resources.

Many studies have also looked at the impact of corruption on external inflows such as foreign direct investment (FDI). Alesina and Dollar (Reference Alesina and Dollar2000) point out that almost two-thirds of all foreign aid collected goes to government consumption. These funds are therefore allocated by international sources and end up in the hands of government bureaucrats to be distributed in some form to the general public. Many believe that aid increases the incentives of rent-seeking behavior because there is more rent to divide (Krueger, Reference Krueger1974; Murphy et al., Reference Murphy, Shleifer and Vishny1993; Svensson, Reference Svensson2000; Torvik, Reference Torvik2002; Hodler, Reference Hodler2007). The enforcement of environmental regulations in Indonesia and Thailand has often been compromised due to rent-seeking behavior by bureaucrats (Cribb, Reference Cribb and Desai1998; Riggs and Stott, Reference Riggs, Stott and Desai1998). Bureaucrats in these developing countries are more inefficient and more corrupt relative to those in developed countries (Lopez and Mitra, Reference Lopez and Mitra2000).

2.3 Rule of law and the environment (Linkage T4)

Scholars have highlighted the existence of property rights and the rule of lawFootnote ⁵ in the management of natural resources (Deacon, Reference Deacon1994; Bohn and Deacon, Reference Bohn and Deacon2000). Mani and Fredriksson (Reference Mani and Fredriksson2002) develop a theoretical framework to examine the linkages between rule of law and environmental policy formation. Their model suggests that an increase in the degree of rule of law has two opposing effects on environmental policy. Policy decisions implemented according to the law increase the stringency of environmental policy. However, a relatively high rule of law incentivizes polluting firms to increase lobbying efforts for favors from corruptible policymakers. Accordingly, they conclude that an increase in corruptibility of policymakers lowers the stringency of environmental policy.

Chen and Chai (Reference Chen and Chai2010) and Castiglione et al. (Reference Castiglione, Infante and Smirnova2011) have tried to integrate the impact of the rule of law into empirical environmental Kuznets curve (EKC) analyses.Footnote ⁶ Utilizing the rule of law variable from The World Bank's World Governance Indicators (at https://databank.worldbank.org/source/worldwide-governance-indicators), the authors find that both income and the rule of law have positive effects on environmental policy stringency in 71 countries. Castiglione et al. (Reference Castiglione, Infante and Smirnova2011) find that the rule of law has differential effects on carbon emissions in 28 European countries, partly based on the country's sector composition, and whether or not it had a Socialist past. Gani (Reference Gani2012) examines the relation between different good governance indices and CO₂ emissions for 99 developing countries and finds that political stability, rule of law and control for corruption are negatively correlated with CO₂ emissions. Scott (Reference Scott2016) finds a strong correlation between the Yale environmental performance index and the World Justice Protect Rule of Law index and suggests that environmental advocates should pay attention to strengthening rule of law. The author argues that a strong rule of law increases the public's ability to protect the environment through the procedural rights of access to justice and participatory decision-making.

In sum, while the above-discussed literature documents the association between institutional quality indicators and environmental wellbeing (linkages T2, T3 and T4), there are no studies that focus on the mediating impact of institutional quality on the green ODA-environmental wellbeing association (linkages T5, T6 and T7). Further, as presented above, literature on the direct impact of green ODA on the environment is also scant (linkage T1). This caveat motivates our empirical investigation and research question on the environmental impacts of green ODA and whether the efficiency of such aid is mediated by the quality of the institutional environment.

3.1 Model specification

The CO₂ equation that we wish to estimate has the following form:

(1)\begin{align} {{\rm CO}_2}_{it} & = \alpha {{\rm CO}_2}_{i,t - 1} + {\beta _k}{X_{it}} + \gamma {{\rm INS}_{it}} + {\delta _k}{{\rm GODA}_{i,t - k}} \notag \\ & \quad + {\mu _k}{\rm INS} \cdot {\rm GODA}_{{i,t - k}} + {\eta _i} + {v_{it}},\end{align}

where CO_2it is a per capita CO₂ emission level of country i in time t (hereafter, the subscript it is omitted for all variable explanations); INS proxies institutional quality–corruption, economic freedom and rule of law; GODA_i,t−k is a lagged level of green ODA divided by total population; INS•GODA_i,t−k is the interaction term between institutional quality and green ODA; k is a lag order for green ODA and interaction term (k = 0, 1, 2). The variable X is a vector of exogenous variables that are commonly used as determinants of CO₂ emissions such as per capita level of GDP (the EKC hypothesis mentioned above) and energy intensity (Bae et al., Reference Bae, Li and Rishi2017; Shahbaz et al., Reference Shahbaz, Hoang, Mahalik and David2017). η _i is the individual effect and v _it are residual terms.

Concerning the relation between GDP and CO₂ emissions in equation (1), extant literature on the EKC has tested non-linearity between per capita GDP and CO₂ emissions. We utilized a linear term for per capita GDP for two reasons. First, most studies show that the relationship between per capita income and global pollutants such as CO₂ emissions for most developing countries has a linear relationship. Second, including a square term for per capita GDP will not allow us to use system GMM estimation as the variance-covariance matrix does not achieve full rank.

3.2 Arellano-Bover/Blundell-Bond estimator

The econometric method in estimating the relations between green ODA and carbon emission levels under different institutional qualities employs the two-step system GMM Arellano-Bover/Blundell-Bond estimator with Windmeijer finite sample correction (Windmeijer, Reference Windmeijer2005).Footnote ⁷ A system GMM technique is efficient as it allows for more instruments relative to the Arellano-Bond (Arellano and Bond, Reference Arellano and Bond1991) estimator. The system GMM can include time-invariant regressors, which would have disappeared in difference GMM estimation. Moreover, the regressors are not required to be strictly exogenous (Roodman, Reference Roodman2009). Several studies have utilized system GMM to analyze the effectiveness of the foreign aid (Armah and Nelson, Reference Armah and Nelson2008; Feeny and Ouattara, Reference Feeny and Ouattara2009; Adedokun and Folawewo, Reference Adedokun and Folawewo2017).

Soto (Reference Soto2009) notes that the system GMM estimator is the most precise when confronted with small sample bias, which is critical in this study, given our small sample of green ODA recipient countries. However, the moment conditions of the Arellano-Bover/Blundell-Bond estimator should not exhibit significant AR (2) behavior. If a significant AR (2) statistic is encountered, the lags of endogenous variables will not be appropriate instruments for their current values. Accordingly, we test for autocorrelation and provide the results of the Arellano–Bond test for serial correlation in the tables in sections 4.2 and 4.3.

Consider the instruments that the system GMM estimator uses. In the following dynamic equation,

(2)\begin{equation} {y_{it}} = \rho {y_{i,t - 1}} + x_{it}^{\prime} \beta + {\alpha _{i}} + {\varepsilon _{it}},\end{equation}

${y_{it}}$ is a dependent variable; $x_{it}^{\prime}$ is a row vector of explanatory variables; i = 1, …, N is the index for countries; t = 1, …, T is the index for years; $\rho$ is the coefficient of the lagged endogenous variable ${y_{i,t - 1}}$; $\beta$ is an unknown parameter vector of the k explanatory variables; ${\alpha _i}$ is the individual effect; and ${\varepsilon _{it}}$ are residual terms. Arellano and Bond (Reference Arellano and Bond1991) suggest using a difference equation derived from (2) to eliminate the individual effect ${\alpha _i}$:

(3)\begin{equation} {y_{it}} - {y_{it - 1}}\; = \rho ( {{y_{i,t - 1}} - {y_{i,t - 2}}} ) + (\; x_{it}^{\prime} - x_{it - 1}^{\prime} )\beta + {\varepsilon _{it}} - {\varepsilon _{it - 1}}.\end{equation}

For equation (3) the previous level values are used as instruments. For example, for the final period T,

(4)\begin{equation} {y_{iT}} - {y_{iT - 1}}\; = \rho ( {{y_{i,T - 1}} - {y_{i,T - 2}}} ) + (\; x_{iT}^{\prime} - x_{iT - 1}^{\prime} )\beta + {\varepsilon _{iT}} - {\varepsilon _{iT - 1}}.\end{equation}

The available instruments are: ${y_{i,1}},\; {y_{i2,\; \ldots,}} {y_{i,T - 2}},\; x_{i1}^{\prime}, \; x_{i2}^{\prime}, \; \ldots, \; x_{iT - 1}^{\prime}$. In addition to the difference equation, Blundell and Bond (Reference Blundell and Bond1998) suggest using the level equation, where differences are used as valid instruments.

Again, consider the last observation T:

(5)\begin{equation} {y_{iT}} = \rho {y_{i,T - 1}} + x_{iT}^{\prime} \beta + {\alpha _i} + {\varepsilon _{iT}}.\end{equation}

The available instruments are ${\rm d}{y_{i,1}},\; {\rm d}{y_{i2,\; \ldots,}} {\rm d}{y_{i,T - 1}},\; {\rm d}x_{i1}^{\prime}, \; {\rm d}x_{i2}^{\prime}, \; \ldots, {\rm d}x_{iT}^{\prime}$, where d is a difference operator.

We tested square terms for per capita GDP for our specifications by using a fixed effects model (see online appendix tables B1–B4). The results indicate that the square terms were insignificant, or the estimated turning points were above the maximum value of per capita GDP in our data sample. This is consistent with some EKC studies that indicate that developing countries do not display an EKC relationship between per capita GDP and per capita CO₂ emissions.Footnote ⁸

3.3 Data and variables

Table A1 in the online appendix provides definitions, data sources and summary statistics for all variables. The total number of green ODA recipient countries in 2011 was 142. However, as green ODA data is combined with the other data, the number of total observations falls to 86 countries and we obtain an unbalanced panel of 86 countries between 2003 and 2014. The list of countries employed in our analysis is presented in table D1 of the online appendix.

Data on green ODA for recipient countries between 2003 and 2014 were collected from the OECD Rio Marker Creditor Reporting System (CRS). As explained above, data on aid flows by sector are monitored by the OECD's creditor reporting system. The original data include several overlapping dimensions such as climate change mitigation, adaptation, desertification and biodiversity.Footnote ⁹ For the purposes of this paper, we used ‘climate change mitigation aid’ data as our proxy for green ODA as it relates more specifically to our research focus on carbon emissions. We consider mitigation aid flows directed to the three most carbon intensive sectors, namely, energy, transport and industry. The choice of ‘climate mitigation aid’ as a proxy for green ODA is based on the fact that such aid is aimed at carbon emission reductions or stabilization of carbon emissions via the application of new and renewable forms of energy. Such aid flows also capture measures to improve the energy efficiency of existing generators, machines and equipment.Footnote ¹⁰ As reported in tables C1–C3 in the online appendix, we also considered total ODA flows per capita in our estimations. We utilized the real value (2015 US$) for both green ODA and total ODA.

Data on per capita CO₂ emissions was collected from the World Bank's World Development Indicators Footnote ¹¹ for the period between 2003 and 2014 and measured in metric kilos per capita. The data for per capita GDP in constant 2010 US$ and energy intensity were also collected from World Development Indicators. Data on economic freedom was obtained from the Heritage Foundation (at https://www.heritage.org/index/explore). The rule of law index and control of corruption index were collected from Worldwide Governance Indicators (at https://databank.worldbank.org/source/worldwide-governance-indicators). All institutional indices were rescaled from 0 to 1, where a higher value implies better institutional quality.

The interaction term (GODA*INS) in equation (1) is of special interest to our analysis. Specifically, green ODA multiplied by different institutional quality indices will enable us to investigate whether the effectiveness of green aid is mediated by the quality of the institutional environment. Since all indices are scaled from 0 to 1, a lower level of institutional quality index connotes a reduction in the value of the interaction term.