6

The Macroeconomic Impact of Aid Volatility

John Hudson* and Paul Mosley**

ABSTRACT
We analyse the impact of aid volatility on GDP shares of expenditure. Given the level of aid, positive and negative volatility reduce investment and government expenditure shares. But the former reduces import share and the latter increases consumers’ expenditure share.

*Department of Economics and International Development, University of Bath, Bath, BA2 7AY, UK, email:

** Department of Economics, University of Sheffield, 9 Mappin Street, Sheffield, S1 4DT, UK, email:

Keywords: aid volatility, natural disasters, expenditure shares,

JEL Classification: F35, O16, E20.


The Macroeconomic Impact of Aid Volatility

1. Introduction

Despite, a growing perception that aid is now effective in promoting growth (Collier and Dollar 2004, Mosley et al. 2004), aid volatility is a matter of concern. The volatility of revenue inflows, a high proportion of which are aid in the case of the poorest countries, may result in volatility of expenditure and instability of policy (Rodrik 1990). Lensink and Morrissey (2000), argue that volatility damages the macro-economic effectiveness of aid. Mosley and Suleiman (forthcoming) show that the ability of the recipient public sector to implement coherent investment programmes and fiscal policies is reduced by aid volatility. Hudson and Mosley (2006) also conclude that aid’s impact on growth is reduced by both positive and negative volatility. However, the most comprehensive critique is that by the IMF economists Ales Bulir and Javier Hamann (B-H; 2003, 2005).

Our purpose in this paper is to go beyond this work and to ask how aid volatility impacts on the various components of GDP. We distinguish between positive and negative aid volatility. The former represents sudden surges of aid and may reduce effectiveness due to constraints on absorptive capacity. However, if aid is not fully used as intended, does it simply not get spent at all, or spent inefficiently, or does it seep into some other form of expenditure? In the former case import share may also fall. Negative aid volatility represents sudden declines in aid and may result in projects being abruptly postponed or even terminated and disrupt government budgetary planning, again reducing aid effectiveness.

Our expectations are that capacity constraints will be most related to investment and least to consumers’ expenditure, whilst the disruption caused by negative volatility will affect the least flexible type of expenditure – again consumers’ may have more degrees of freedom in rearranging expenditure flows than is exhibited by investment projects. We assume expenditure of the j’th component to be a function of aid, aid volatility (Vj) and other variables (Xi):

Yj = fj(Aj, Vj,Xi) (1)

Yj/Y is then:

Yj/Y = fj(Aj, Vj, Xi)/f(A, V, Xi) (2)

Throughout the absence of a subscript indicates the variable relates to the whole economy. Differentiating this ratio and re-arranging we get:

∂(Yj/Y)/∂V/(Yj/Y) = [(∂fj /∂Vj)/Yj ∂Vj/∂V –∂f/∂V/Y] (3)

If ∂Vj/∂V=1, the impact of volatility on the j’th expenditure share will depend upon the proportionate impact of volatility on the sector, i.e. (∂fj/∂Vj)/Yj relative to the proportionate impact of volatility on the whole economy (∂f/∂V/Y). Where the former exceeds the latter, aid volatility will increase Yj/Y. In the case of positive volatility this will reflect the different impact of absorptive capacity constraints between the different types of expenditure. In the case of negative volatility, ∂fj/∂Vj will partially depend upon the ability of the j’th sector to substitute other forms of revenue for the shortfall in aid.

A further crucial factor is ∂Vj/∂V. If ∂Vj/∂V<1 then the j’th sector will be protected from the full impact of volatility. For negative volatility, this may be so for two reasons. First donors may seek to ‘protect’ this type of expenditure. Secondly the government, e.g., may have the ability to shift the aid budget so as to cushion the j’th expenditure component from adverse aid shocks. If this is the case then it follows that there must be some sector, the m’th, for which ∂Vm∂V > 1. In this case it is likely that the adverse impact of aid volatility on the j’th sector will be less than for the economy as a whole, in which case aid volatility will increase Yj/Y and vice versa for Ym/Y.

2. Data and Methodology

The data base covers the period 1977-2001 and relates to all developing countries from the World Bank Development Indicators. However, missing data reduces the number of countries to 131 and focuses the analysis on more recent years. We will regress GDP expenditure shares on a vector of variables including aid, aid volatility, a time trend, OECD growth, population density, lagged GDP per capita and a disaster variable. The latter is included because as with aid volatility, disasters impact negatively on GDP (Hudson and Mosley, 2006). In addition positive aid volatility may be linked to disasters and the two impacts need to be separated. OECD growth partially instruments GDP growth and allows for the possibility of GDP shares being different at different stages of the cycle. All variables are defined in an appendix.

We follow B-H (2003) in measuring volatility by calculating the residuals from the application of a series of Hodrick Prescott filters applied to aid for each recipient country individually. Positive/negative aid volatility are based on the positive/negative residuals from the filter and otherwise zero. The use of this filter is open to dispute in that the resulting series could tend to underestimate volatility, but as B-H comment in practice it makes relatively little difference whether this or some other methodology is used. As in their analysis we used a value for λ of seven. In the original B-H paper the variables were not logged and we follow this practice. However, we take the additional step of normalizing all the variables around their mean. This removes the need to standardize aid revenues by the use of some denominator, thus contaminating the volatility of aid with that of the denominator and any covariance between numerator and denominator. We then multiply the two volatility variables by aid share in year t as the impact of volatility on GDP and its components will depend on the size of the aid budget. Both negative and positive volatility are defined as absolute values – hence a positive coefficient in the regressions indicates that, even for negative volatility, an increase in volatility increases expenditure share.

3. Empirical Results and Interpretation

Insert table 1 about here.

The results are shown in Table 1. From the Lagrange multiplier test statistic we reject the hypothesis of no group-wise heteroscedasticity, whilst the Hausman test favors the use of fixed over random effects[1]. The use of fixed effects largely removes the influence of country specific factors which are either constant or change only gradually (e.g. proportion of land in the tropics and ethnic mix). We focus on the impact of aid, aid volatility and disasters. Aid increases all expenditure shares apart from exports. With respect to investment, government expenditure and indeed imports this is the anticipated effect if aid ‘is effective’, e.g. much of aid is focused on budget support for governments. The impact on consumers’ expenditure could reflect either an element of fungibility or consumers’ believing that the aid will increase future prosperity, responding in a rational manner, as the permanent income hypothesis would suggest. Positive volatility significantly reduces the expenditure shares of both investment and government expenditure at the 1% level. This is suggestive of absorptive capacity constraints. There are no significant impacts on consumption nor exports, but import share is also reduced which may be a consequence of the impacts on investment and government expenditure. Negative volatility significantly reduces investment and government expenditure shares at the 5% and 1% levels respectively, and substantially increases the share of consumption. To a considerable extent our results are consistent with the theoretical analysis of Arellano et al. (2005) which predicts aid shocks will impact mainly on investment as a result of consumption smoothing. Disasters impact significantly in reducing investment share, but increase the shares of government and consumers’ expenditure[2]. This suggests that in the aftermath of a disaster the focus of relief effort is such as to divert attention away from investment projects. In addition, of course, investors may be deterred by susceptibility to disasters such as weather or earthquake related ones which damage infrastructure.

Previous work (Hudson and Mosley, 2006) has indicated that aid’s impact on growth is reduced by both positive and negative volatility. Hence the impact of negative volatility in increasing consumption share simply implies that the impact of negative volatility on consumption is less than on other components of GDP, rather than that consumption responds positively to negative volatility. Negative volatility involves a sudden decline or shortfall in aid and the results suggest the primary impact is on investment and government expenditure. This may be because of the type of aid which is subject (a donor effect) to volatility or because consumers’ are better able to absorb shocks by drawing on savings and/or borrowing than other agents. The results also suggest a limited ability of governments to rearrange revenue flows to reduce the impacts upon their own expenditure priorities. The results with respect to positive volatility suggest that absorptive capacity constraints particularly limit aid’s effectiveness with respect to both investment and government spending. Thus in both cases volatility impacts most adversely on both investment and government shares. In the case of positive volatility changes in the import share tend to restore the macroeconomic balance whilst for negative volatility an increasing consumers’ expenditure share does this.

References

Arellano, C. A. Bulir, T. Lane and L. Lipschitz, 2005, The dynamic implications of foreign aid and its variability, IMF Working Paper WP/05/119 (2005).

Bulir, A., and J. Hamann, 2003, Aid volatility: an empirical assessment’, IMF Staff Papers, 50, 64-89.

Bulir, A., and J. Hamann, 2005, Volatility of development aid: from the frying pan into the fire? Washington DC: IMF, unpublished paper.

Collier, P. and D. Dollar, 2004, Aid effectiveness, what have we learnt? Economic Journal, 104, F244-F271.

Hudson, J. and P. Mosley, 2006, Aid volatility, policy and development, presented at UNU-WIDER Conference on Aid: Principles, Policies and Performance, Helsinki, June(2006)

Lensink, R. and O. Morrissey, 2000, Aid instability as a measure of uncertainty and the positive impact of aid on growth, Journal of Development Studies, 36, 31-49.

Mosley, P. J. Hudson and A. Verschoor 2004, Aid, poverty reduction and the new conditionality, Economic Journal, 114, 217-243.

Mosley, P. and A. Suleiman (forthcoming) Aid, agriculture and poverty, Review of Development Economics.

Rodrik D., 1990, How should structural adjustment programmes be designed? World Development, 18, 933-947.

Table 1: Impact of Aid Volatility on Expenditure Shares

Investment / Consumption / Government Expenditure / Exports / Imports
Constant
Aid Share
Positive
Volatility
Negative
Volatility
Trend
Log GDPPCt-2
Disaster
Log Population
Density
OECD Growth
Observations
F
B-P LM
HausmanΧ2 / 35.274**
(4.55)
0.204**
(6.84)
-0.087**
(3.38)
-0.162*
(2.26)
-0.00240
(0.41)
2.353**
(3.68)
-0.173*
(2.03)
-8.338**
(4.65)
0.226**
(2.93)
2467
17.25**
7141.1**
32.73** / 127.37**
(14.03)
0.240**
(6.89)
-0.039
(1.31)
0.238**
(2.84)
0.265**
(3.85)
-8.543**
(11.44)
3.028**
(3.05)
3.815
(1.82)
-0.527**
(5.85)
2462
41.59**
12114.5***
16.03** / 27.475**
(5.46)
0.0830**
(4.32)
-0.0658**
(3.95)
-0.0015**
(3.22)
-0.0717
(1.88)
1.422**
(3.43)
2.267**
(4.13)
-5.568**
(4.79)
-0.0267
(0.54)
2465
32.47**
8351.1**
18.98** / 27.36*
(2.53)
0.080
(1.95)
-0.026
(0.34)
-0.160
(1.60)
0.317**
(3.85)
3.672**
(4.11)
0.920
(0.78)
-8.837**
(3.52)
0.455**
(4.25)
2522
30.71**
17138.9**
11.86* / 92.152**
(7.83)
0.518**
(11.56)
-0.157**
(4.00)
-0.077
(0.71)
0.570**
(6.36)
-2.704**
(2.78)
2.199
(1.70)
-13.583**
(4.96)
0.138
(1.19)
2522
29.62**
17980.5**
37.55**

Equations estimated by fixed effects. (.) denotes t statistics and a **/* denotes the variable is significant at the 1%/5% levels. B-P LM denotes the Breusch-Pagan Lagrange multiplier test for random effects, HausmanΧ2 denotes the Hausman test for fixed vs random effects.


DATA Definitions

Aid Sharea Share of aid in GDP (note average GDP in t and t-1).

Positive/Negative Residual when positive/negative (otherwise zero) from a Hodrick

Volatility Prescott filter on normalised aid applied to one country at a time, multiplied by aid share.

GDPPCt-2a GDP per capita in constant US$ lagged two years.

Disasterb Defined as when at least one of the following criteria is fulfilled: (i) 10 or more people reported killed, (ii) 100 people reported affected, (iii) a call for international assistance or (iv) a declaration of a state of emergency. Disasters include floods, earthquakes, epidemics, droughts, famines, windstorms, etc. The variable we use is the percentage of the population affected when it is greater than 10%.

Population Densitya People per squared (land) kilometre.

OECD Growtha Growth rate of the OECD countries.

Sources: a World Development Indicators b EM-DAT: The OFDA/CRED International Disaster Database - www.em-dat.net - Université Catholique de Louvain, Brussels, Belgium

[1] There were no significant differences between fixed and random effects estimates with respect to the impact of aid, aid volatility and disasters.

[2] The variable relates to current and lagged disasters, with a two year lag depth, which is more significant than simply disasters in the current period.