Aid fungibility and government spending in Africa: evidence from instrumental variable panel quantile regression

ABSTRACT

This study investigates the impacts of foreign aid on government spending for 54 African aid-recipient countries with the distinction between on-budget and off-budget aid at the aggregate level. Applying instrumental variable panel quantile regression, the results indicate that, generally, the on-budget aid is partially fungible but the degree of fungibility decreases in the high-level government spending countries (higher quantile). On the other hand, off-budget aid is generally nonfungible though partial fungibility is significant in the low-level government spending countries (10^th to 30^th quantile). In addition, the impact of total aid is downward biased at various quantiles when the distinction between on-budget and off-budget aid is not made. Besides, the trade openness, the short-run shock on government spending and the quality of governance are also important determinants of government spending.

KEYWORDS:

• Off-budget aid
• fungibility
• government spending
• instrumental variable panel quantile regression

JEL CLASSIFICATION:

• O55
• C23

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Marć (2017) notes that donors are more likely to finance increased government spending than stimulate any reduction in taxes or borrowings.

² McGillivray and Morrissey (2000) provide comprehensive reviews on this relationship..

³ In this study, the ‘flypaper effect’ is used to describe the phenomenon by which the government spending of aid-recipient country increases more than the value of aid..

⁴ Van de Sijpe (2013) provides an economic framework and show that the estimation bias caused by excluding off-budget aid in the analysis.

⁵ Following Chatterjee, Giuliano, and Kaya (2012) the term ‘fungibility at the aggregate level’ is used interchangeably with ‘additionality of aid’. In this study, fungibility refers to fungibility at the aggregate level.

⁶ Galvao (2011) combines the work of Chernozhukov and Hansen (2005; Chernozhukov and Hansen 2006) and propose a quantile regression method for dynamic panel data with endogeneity.

⁷ Technical cooperation, which is obtained from OECD(DAC 2a) database, relates to payments and transfers made to nationals and others in the aid-recipient’s country for consultancy, training and education.

⁸ See Rodrik (1998), McGillivray and Morrissey (2000), Gupta et al. (2004), Rajan and Subramanian (2008) and Marć (2017).

⁹ The difference GMM estimation is employed to control for unobservable heterogeneity and prevents potential endogeneity problems in the dynamic panel model. Test for the absence of second-order serial correlation (Arellano and Bond 1991) and over-identifying restrictions (Hansen test) is conducted and the null hypotheses are also supported.

¹⁰ The SEs are bootstrapped on 500 replications..

¹¹ We replace the two independent variables on-budget and off-budget aid by a single independent variable total aid and perform the same set of regressions. The estimated coefficients of total aid in FE, GMM, 10th, 30th, 50th, 70th and 90th quantile are 0.21, 0.239, 0.07, 0.101, 0.169 0.208 and 1.570, respectively. The statistical significance of total aid in the set of regressions is comparable to the variable on-budget aid..