Does energy aid reduce CO₂ emission intensities in developing countries?

ABSTRACT

Whether energy aid contributes to the reduction of CO₂ emissions in developing countries is an unresolved question. We investigate whether the OECD Development Assistance Committee members’ energy aid can help reduce CO₂ emission intensities in 64 recipient countries over the period 1995–2014. We find that once lagged energy aid – when measured in terms of once lagged GDP, in absolute terms, or in terms of once lagged total sectoral aid – is effective in reducing the recipients’ subsequent CO₂ emission intensities. Importantly, we find highly varied effects of energy aid on the recipients’ CO₂ emission intensities – fossil fuel-rich (poor) countries stand to benefit least (most) from receiving energy aid. Our findings point towards the merit of energy aid as a policy tool in achieving the CO₂ emission reduction goals and the necessity for bilateral aid donors to take into account the recipients’ fossil fuel abundance when making provisions for energy aid.

KEYWORDS:

• Development assistance
• sectoral aid
• endogeneity
• system GMM estimation

JEL CODES:

• C23
• O13
• Q56

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The OECD bilateral ODA donors as a whole seem not to have used common criteria in selecting their recipients of energy aid. There is a paucity of evidence indicating that the donors’ political affinity or post-colonial ties with recipients are the major drivers of the provision of energy aid. Kim (2019) found that bilateral energy aid donors tend to select those recipients with low energy consumption who are in high need of residential electricity access and renewable energy. On the other hand, Henderson (2019) laid particular stress on a possibility that the recipients’ CO₂ emissions increase because the vast majority of bilateral energy sector aid has gone to fossil fuels.

2 The Arellano–Bond autocorrelation test and the robust estimates of the coefficient standard errors in generalised method of moments estimations assume no contemporaneous correlation in the idiosyncratic disturbances. Time dummies make this assumption more likely to hold (Roodman 2009b).

3 We ignore the population effect because the variable $m$ is defined relative to GDP.

4 Lagging the aid variable once is a cleaner and more transparent way of dealing with reverse causation than searching for a strong and valid instrument (Clemens et al. 2012). Another strategy to deal with endogeneity is to use some external instruments of the aid variable, e.g. the United Nations General Assembly voting positions of donor–recipient dyads. However, it is debatable whether dyadic political proximity is a plausible exogenous driver of the provision of bilateral energy aid. Kim (2019) found that bilateral energy aid donors respond primarily to the recipient’s sustainable energy needs.

5 There is an inbuilt bias towards a positive correlation between the regressors and the regressand when they are expressed relative to GDP. Additional estimations where the aid variable enters in absolute terms are appropriate because the recipients of development aid, regardless of their sizes, may benefit from similarly high aid amounts (Kretschmer, Hübler, and Nunnenkamp 2013).

6 Oil (natural gas, coal) rents are the difference between the value of crude oil (natural gas, both hard and soft coal) production at world prices and total costs of production. The share of total resource rents in GDP is a measure of resource abundance (Atkinson and Hamilton 2003). Bekun, Alola, and Asumadu-Sarkodie (2019) found a positive long-run relationship between natural resource rents and CO₂ emissions in 16 European countries over the period 1996–2014. Adedoyin et al. (2020) found a negative relationship between coal rents and CO₂ emissions per capita in Brazil, Russia, India, China and South Africa over the period 1990–2014.

7 Because the fossil fuel rents of 14 countries in our sample are simply zero, this variable has a highly left-skewed distribution. Therefore, the log transformation ln (x + 1) is used to make it less skewed. Cambodia, Costa Rica, Dominican Republic, El Salvador, Honduras, Jamaica, Kenya, Lebanon, Mauritius, Namibia, Panama, Paraguay, Sri Lanka, and Uruguay have no fossil fuel rents.

8 Bhattacharyya, Intartaglia, and McKay (2018) transformed their data into five-year averages to smooth noise and potential business cycles in the data and tackle the sparse-coverage problem of the data. To minimise the time dimension, we transformed our panel data into four-year averages because the Arellano and Bond (1991) test for both AR(1) and AR(2) based on the residuals from the two-step estimator of the first-differenced equation can be performed only on samples where T ≥ 5.

9 Kretschmer, Hübler, and Nunnenkamp (2013) found a negative impact of total aid ratio on the recipients’ energy intensity. Their finding offers a tentative explanation for the present result that the lagged total aid ratio has a negative impact on the recipients’ subsequent CO₂ emission intensities.

Additional information

Funding

This work was supported by the Japanese Government Ministry of Education, Science, Sports, and Culture (MEXT) [Monbu Kagakusho] Grant-in-Aid for Scientific Research (C) No. 17K03581.