Chapter

3 Aid Volatility and Macroeconomic Policies

Author(s):
Charalambos Tsangarides, Carlo Cottarelli, Gian Milesi-Ferretti, and Atish Ghosh
Published Date:
September 2008
Share
  • ShareShare
Show Summary Details

At the G-8 meeting in Scotland in July 2005, world leaders announced a US$50 billion increase in official development assistance to poor countries. This surge in aid—aimed at achieving the Millennium Development Goals1—has focused policymakers’ and researchers’ attention on the macroeconomic challenges that need to be addressed to maximize the beneficial effects of large aid inflows (see, for instance, Heller, 2005, and Berg and others, 2007).

There are two key macroeconomic concerns associated with the absorption of large aid flows. First, volatile aid flows may increase macroeconomic volatility, thereby reducing welfare. Second, aid flows may lead to overvalued real exchange rates and hurt export industries that are crucial for economic growth.

Our chapter contributes to this debate by studying empirically whether macroeconomic policies are effective in smoothing the effects of volatile aid flows. Specifically, we show that macroeconomic policies can help reduce the trade balance volatility induced by aid flows. By using an instrumental-variables strategy, we are able to isolate the causal effect of aid flows on the trade balance—a variable that reflects the variability of both imports and exports—and the offsetting effect of macroeconomic policies. We also discuss our results in light of recent theories and draw policy implications.

Aid flows to poor countries are often volatile and unpredictable. These features reduce both the effectiveness and the potential welfare benefits of aid. Bulíř and Hamann (2003,2007) and Celasun and Walliser (2007) document aid volatility, and Arellano and others (2005) show how volatile aid flows may result in substantial welfare losses (see also Prati and Tressel, 2006). This finding is in line with Pallage and Robe’s (2003) estimate of a median welfare cost of business cycles in developing countries between 10 and 30 times that of the United States. Other studies—including Ramey and Ramey (1995) and Aghion and others (2004)—have documented the negative effect of volatility on long-term growth. This literature leaves open the question of whether macroeconomic policies could reduce the transmission of aid volatility to the trade balance, consumption, and growth. Our chapter fills this gap by focusing on how monetary and fiscal policy can influence the link between aid volatility and the trade balance.

Smoothing aggregate demand over time is a first rationale for policy intervention aimed at reducing the response of the trade balance to aid shocks. In countries where the capacity to borrow against future income and to save unexpected aid flows is limited, volatile aid flows tend to exacerbate aggregate demand volatility and, in turn, import fluctuations. Such volatility is unlikely to be optimal. Saving part of temporary increases in aid flows in anticipation of future aid shortfalls can be welfare improving for several reasons, including the need to build reserve buffers against future negative shocks and to adjust fiscal spending paths to a country’s absorptive capacity.

But smoothing aggregate demand fluctuations is not the only reason for trying to weaken the link between the volatility of aid flows and that of the trade balance. Real exchange rate fluctuations and ups and downs in exports are likely to accompany trade balance changes caused by aid volatility. Such volatility of exports ultimately hurts export capacity and growth potential. The underlying mechanism is the following. When large aid flows are spent on goods and services produced in the domestic economy, the price of nontraded goods rises relative to the price of traded goods (the real exchange rate), and export-oriented, high-value-added sectors lose competitiveness. To the extent that these effects are not fully reversed when foreign aid falls back after a surge, aid volatility reduces growth and welfare in aid-receiving economies. Saving part of the initial aid surge may help forestall this phenomenon, which is usually referred to as Dutch disease.

Rajan and Subramanian (2005b) confirm that Dutch disease is a real concern by showing that, in countries that received more aid in the 1980s and 1990s, export-oriented, labor-intensive manufacturing industries grew more slowly than other industries. Similarly, Prati and Tressel (2006) find that foreign aid inflows depress overall exports of poor countries, as Dutch disease would imply, and that macroeconomic policies tend to mitigate these adverse effects. They do not, however, find any negative effect of aid disbursed when countries experience large exogenous shocks (droughts, large negative commodity price shocks, hurricanes, or earthquakes) or during postwar reconstruction. Kang, Prati, and Rebucci (2007) find that the effects of “global” aid shocks on output vary across countries, confirming evidence of Dutch disease in only about half of aid recipient countries and showing how Dutch disease effects tend to be associated with overvalued real exchange rates.

Based on a cross section of aid-receiving countries, we show that macroeconomic policies are effective in mitigating trade balance volatility. Specifically, we show that, with openness to trade controlled for, trade balance volatility is significantly smaller in countries where the central bank’s net domestic assets fall in response to a surge in aid flows—an outcome that may reflect both monetary and fiscal policy responses to aid flows. This result holds when we control for (1) country-specific commodity export price shocks and (2) the endogeneity bias that donors’ response to volatile exports and imports might generate. Our findings are consistent with the country case studies of Berg and others (2007), who show that aid recipient countries often try to prevent aid flows from appreciating the real exchange rate by not spending the aid in the year it is received or by sterilizing the monetary expansion associated with the aid-financed increase in public spending.2

The chapter is organized as follows. The second section describes the empirical evidence. The third section discusses the results in light of recent theories and highlights policy implications.

Empirical Evidence

Data Sources

The evidence presented in this chapter is based on an unbalanced panel data set, over the period 1960–2003, of 41 aid-receiving countries whose median aid-to-GDP ratio is greater than 4 percent and population is above one million.3 Aid data are from the Organization for Economic Cooperation and Development Development Assistance Committee database, and other data are from the IMF’s International Financial Statistics. We also construct a commodity price index using data from the IMF’s World Economic Outlook database. The list of countries is reported in Table 3.1.

Econometric Analysis

We present cross-sectional evidence of the role of macroeconomic policies in reducing aid-induced macroeconomic volatility in aid-receiving countries. We show that in countries in which macroeconomic policies are more “countercyclical” to aid flows, aid-induced trade volatility is significantly smaller. This result is robust to controlling for (1) the volatility of commodity export prices, (2) the degree of trade openness, and (3) reverse causality, possibly due to changes in aid flows in response to trade balance shocks.

The empirical model is the following:

The dependent variable (TBvoli) is the standard deviation of the trade balance–to–GDP ratio. The explanatory variables are (1) the standard deviation of the aid-to-GDP ratio (Std_AID); (2) a scaling variable for trade openness (Mean_MX), measured as the average ratio of the sum of exports and imports to GDP; and (3) the standard deviation of a country-specific (trade-weighted) real commodity export price index (Std_COMM). Std_AID_Cni is the standard deviation of the aid-to-GDP ratio multiplied by a dummy variable equal to one whenever a country follows macroeconomic policies that are countercyclical to aid flows (that is, when the correlation between aid flows and changes in net domestic assets is negative). We subtract a country-specific linear trend from all explanatory variables before computing the statistics.

Table 3.1.The Sample(Countries with more than one million inhabitants and median aid flows above 4 percent of GDP)
AlbaniaKenya
AngolaKyrgyz Republic
ArmeniaLao People’s Democratic Republic
AzerbaijanLesotho
BangladeshMadagascar
BeninMalawi
BoliviaMali
BotswanaMauritania
Burkina FasoMongolia
BurundiNamibia
CambodiaNepal
CameroonNicaragua
Central African RepublicNiger
ChadPapua New Guinea
Congo, Republic ofRwanda
EthiopiaSenegal
GuineaSierra Leone
HaitiSri Lanka
HondurasTanzania
JordanTogo
Vietnam

Tables 3.2 and 3.3 report respectively summary statistics and bivariate correlations. Table 3.2 shows that the countries in the sample are very open to trade (mean openness of 53 percent of GDP), have received substantial amounts of aid flows (country median of 10 percent of GDP on average), and face substantial variation in their trade balances (standard deviation of 6 percent of GDP on average) and in aid flows (standard deviation of 4 percent of GDP on average). Table 3.3 shows that countries that face greater shocks to their trade balance receive more aid, face larger fluctuations in aid flows, and are more open to trade on average.

Regression results are reported in Table 3.4. This first set of estimates shows that a higher volatility of aid flows in a country is significantly associated with a higher volatility of its trade balance (columns (1) and (6)).

In columns (2) and (7), we add as an additional regressor Std_AID multiplied by the dummy for countercyclical macroeconomic policies that is described above. The estimated coefficient of this variable (Std_AID_Cn) is negative and significant, indicating that the link between aid volatility and trade balance volatility is significantly weaker in countries that sterilize the monetary impact of aid inflows. (This sterilization can be achieved by relying either on monetary policy or on fiscal policy, as discussed in Berg and others, 2007.)

Table 3.2.Summary Statistics

(Based on regression (1) of Table 3.4)

VariableMeanStandard DeviationMinimumMaximum
Trade balance–to–GDP ratio (country median)–0.010.05–0.180.11
Country–specific standard deviation of trade balance–to–GDP ratio0.060.040.010.20
Country–specific standard deviation of aid-to-balance–to–GDP ratio0.040.020.010.10
Openness (exports + Imports/GDP), country mean0.530.270.191.19
Country–specific standard deviation of commodity price index33.3819.086.0492.16
Aid–to–GDP ratio, country median0.100.050.040.22
Dummy equal to 1 for macroeconomic policies countercyclical to aid flows0.590.5001
Number of observations: 41

The remaining columns in Table 3.4 confirm the robustness of our results when we control for possible endogeneity biases, caused either by omitted variables or by reverse causality. Indeed, some exogenous factors other than aid inflows (such as weather fluctuations, wars, or shocks to export demand) may increase trade balance volatility and, in turn, induce more volatile aid inflows as donors modify their policies to respond to these shocks. In this case, causality would go from the volatility of the trade balance to that of aid inflows. In cross-sectional regressions, there is no standardized and straightforward way to add controls in Equation (3.1) that could proxy for these omitted sources of aid volatility.

Table 3.3.Bivariate Correlations Among Main Variables
Trade Balance–to–CDP Ratio (Country Median)Country–Specific Standard Deviation of Trade Balance–to–CDP RatioOpenness (Exports + mports/CDP), Country MeanCountry–Specific Standard Deviation of Commodity Price IndexCountry–Specific Standard Deviation of Aid–to–CDP RatioAid–to–CDP Ratio, Country Median
Trade balance–to–CDP ratio (country median)1
Country–specific standard deviation of–0.24761
trade balance–to–CDP ratio0.1186
Openness (exports + Imports/GDP),–0.22570.81621
country mean0.15590.0000
Country–specific standard deviation of0.2334–0.1480–0.41561
commodity price index0.14190.35570.0069
Country–specific standard deviation of–0.25310.32700.08460.10391
aid–to–CDP ratio0.11030.03690.59890.5180
Aid–to–CDP ratio, country median–0.14700.33760.2282–0.30790.52381
0.35900.03090.15140.05020.0004
Note: p-values are reported in italics.
Note: p-values are reported in italics.

To address this problem, we need instrumental variables that are unrelated to trade balance developments but that can explain the volatility of aid flows. In columns (3) and (8), we use Rajan and Subramanian’s (2005a) instrument for aid levels (that is, the predicted value of aid based on exogenous determinants of donors’ aid policies such as colonial relationships and donors’ budgetary cycles), which happens to be correlated not only with aid levels, but also with their standard deviations. In columns (4) and (9), we add as a second instrument a measure of the dispersion in individual donors’ aid policies that we find to be positively correlated with aid volatility. The use of two instruments allows us to statistically test their validity using an overidentifying restriction test (J-statistics). To construct the second instrument, we first compute, for each country and year, the year-on-year change in the aid flow from each donor.4 We then compute the absolute difference between that donor-specific change in aid flows and the average change across donors for that country and year. Finally, we compute the average of this measure across years for each country. This instrument would take a value of zero if all donors increased or reduced their aid to a given country from year to year by the same amount. The greater the dispersion of the donor-specific changes around the average annual change of the aid flow, the higher is the value taken by this instrument.5 In columns (5) and (10), we show that our results remain broadly unchanged when we also instrument trade openness (Mean_MX) using Frankel and Romer’s (1999) instrument constructed from gravity regressions.

Table 3.4.Foreign Aid and Trade Balance Volatility
Median ODA–to–CDP Ratio > 4 PercentMedian ODA–to–CDP Ratio > 5 Percent
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)
OLSOLSIVIVIVOLSOLSIVIVIV
Std_AlD0.3990.5711.0620.6120.5080.4090.5971.3490.6250.508
(2.70)**(3.41)***(3.40)***(4.63)***(2.96)***(2.35)**(3.10)***(2.90)***(3.70)***(2.36)**
Mean_MX0.1270.1240.1180.1240.1420.1270.1240.1160.1240.147
(8.77)***(10.31)***(8.31)***(10.15)***(4.08)***(8.34)***(9.87)***(6.14)***(9.95)***(4.36)***
Std_COMM0.000390.000440.000350.000420.000530.000390.000470.000310.000440.00056
(2.99)***(3.37)***(1.99)**(3.40)***(2.79)***(2.72)**(3.28)***(–1.37)(3.48)***(2.94)***
Std_AID_Cn–0.272–0.3–0.226–0.222–0.301–0.297–0.257–0.243
(2.31)**(1.79)*(2.22)**(1.89)*(2.46)**(–1.52)(2.48)**(1.97)**
Number of observations41414141383434343432
R20.760.790.710.780.770.760.780.610.780.76
F-test (First stage)
Std_AID5.3414.7714.842.3713.6112.13
Std_AID_Cn19.3331.0127.5117.7427.5221.48
Mean_MX2.292.29
j-statistic (p-value)0.430.200.320.15
Note: Each regression includes a constant; robust t-statistics in parentheses. ODA = official development assistance; OLS = ordinary least squares; IV = instrumental variables (two-stage least squares).

significant at 10 percent;

significant at 5 percent;

significant at 1 percent.

Note: Each regression includes a constant; robust t-statistics in parentheses. ODA = official development assistance; OLS = ordinary least squares; IV = instrumental variables (two-stage least squares).

significant at 10 percent;

significant at 5 percent;

significant at 1 percent.

Standard tests suggest that most of the instruments we have employed are valid. The overidentifying restriction tests are always passed (J-statistics for regressions whose results are reported in columns (4), (5), (9), and (10)). We also check whether our first-stage regressions pass Stock and Yogo’s (2005) tests of the null hypothesis of weak instruments. In the first of these tests, the null hypothesis is that the bias of the instrumental-variables regression is more than 10 percent of the bias of the ordinary least squares regression. For two endogenous variables and four instruments, the critical value for this F-test is 7.56. In the second (more stringent) Stock and Yogo test, the null is that the true significance level is 10 percent when the nominal level is 5 percent. For two endogenous variables and four instruments, the critical value for this F-test is 16.87. The specifications of the regressions whose results are reported in columns (4) and (9) pass both tests rejecting the null of weak instruments. Frankel and Romer’s instrument for trade openness is, instead, confirmed to be a weak instrument, as already pointed out in the literature.

Conclusions

Although there is a vast literature on foreign aid, few researchers have tried to identify macroeconomic policies that enhance its benefits and limit its undesirable consequences.6 This chapter focuses on the transmission of aid volatility to the trade balance, and the mitigating role of monetary and fiscal policies. We show that aid flows can increase the volatility of the trade balance and that macroeconomic policies that are countercyclical to aid flows can help reduce this volatility.

Prati and Tressel (2006) develop a two-sector general equilibrium model consistent with the evidence presented in this chapter. In that model, policymakers can use monetary and fiscal policies to reduce the adverse effects of volatile aid flows by containing aggregate demand during aid surges and using the resulting savings to expand aggregate demand during aid drops. More generally, monetary and fiscal policies can help achieve optimal spending and current account paths. When aid flows are excessively front-loaded, monetary and fiscal policies can improve welfare by increasing national savings in the form of higher international reserves. When aid flows are excessively back-loaded, an expansionary monetary or fiscal policy can improve welfare if the stock of international reserves is large enough. In the presence of externalities (such as those underlying Dutch disease), not only can macroeconomic policies smooth the effects of aid flows over time, but they can also prevent aid flows’ possible negative persistent effects on real growth.

Other recent papers have studied the role of policies in maximizing the positive effects of foreign aid. Matsen and Torvik (2005) are interested in the optimal management of exogenous transfers and derive an optimal spending path of natural resources wealth in the presence of Dutch disease. Unlike that of Prati and Tressel, their model does not have a monetary sector, and individuals’ consumption decisions are constrained by an exogenously set current account; Matsen and Torvik also do not consider the question of whether and how macroeconomic policies might replicate an optimal spending path.7

There are, however, limits to the extent to which monetary and fiscal policies can correct the effects of an inappropriate distribution of aid over time. When aid flows are excessively back-loaded, insufficient international reserves can prevent macroeconomic policy from bringing resources forward. Conversely, when aid flows are excessively front-loaded, sterilization costs may reduce the benefits of reserve accumulation. These costs can be large in practice and limit the potential role of monetary policy. If the taxes needed to finance the differential between the interest rates on sterilization bonds and international reserves are distortionary or costly to be levied, sterilization will have welfare costs that should be weighed against the benefits of smaller trade balance volatility and Dutch disease effects. These costs will be even larger if high interest rates depress interest-sensitive private investment that might enhance productivity. In such cases, saving part of the aid for later use, for instance, through government deposits at the central bank, might be a better alternative. But this policy may also be costly. If fiscal surpluses are achieved by postponing the very public investment that is supposed to be financed with the aid increase, the trade balance and Dutch disease effects of aid will be undone, but any related productivity benefit will be lost as well. Lack of coordination between fiscal and monetary authorities can also make the implementation of the policies prescribed in this chapter challenging, when the fiscal authority spends the full increment of aid flows while the central bank sterilizes the monetary effects of aid flows to prevent a real exchange rate appreciation. Moreover, in such a case, aggregate demand pressures might in the end be contained through lower private consumption—which may not be a desirable outcome.

Finally, our results do not provide any indication that an increase in the overall net present value of aid can reduce welfare. They pertain, instead, to the welfare implications of the distribution of a given net present value of aid over time. Indeed, Prati and Tressel (2006)) suggest that the impact of aid on exports depends upon the circumstances that a particular country faces. During periods of large negative shocks, or of reconstruction efforts such as those after a war (including a civil war), aid flows can have positive effects on exports. From this perspective, the declared objective of the donor community to raise official development assistance to 0.7 percent of industrial countries’ GDP from a level that is currently only about one-third of that target can only be welcome.

References

    AghionP.G.-M.AngeletosA.Banerjee and K.Malova2004“Volatility and Growth: Financial Development and the Cyclical Composition of Investment”(unpublished). Available via the Internet:http://elsa.berkeley.edu/users/webfac/auerbach/e231_f04/aghion.pdf

    • Search Google Scholar
    • Export Citation

    ArellanoCristinAlešBulířTimothyLane and LeslieLipschitz2005“The Dynamic Implications of Foreign Aid and Its Variability,”IMF Working Paper 05/119 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    BergAndreAiyarShekharMumtazHussainShaunRoacheTokhirMirzoev and AmberMahone2007The Macroeconomics of Scaling Up Aid—Lessons from Recent ExperiencesIMF Occasional Paper No. 253 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    BulířAleš and A. JavierHamann2003‘‘Aid Volatility: An Empirical Assessment,’’IMF Staff PapersVol. 50 (April) pp. 6489.

    • Search Google Scholar
    • Export Citation

    BulířAleš and A. JavierHamann2007“Volatility of Development Aid: An Update,”IMF Staff PapersVol. 54 (November) pp. 72739.

    • Search Google Scholar
    • Export Citation

    BurnsideC. and D.Dollar2000“Aid, Policies, and Growth,”American Economic ReviewVol. 90 (September) pp. 84768.

    CelasunOy and JanWalliser2007“Predictability and Procyclicality of Aid: Do Fickle Donors Undermine Economic Development?”paper presented at the 46th Panel Meeting of Economic Policy,Lisbon,October1920.

    • Search Google Scholar
    • Export Citation

    EasterlyW. R. Levine and D.Roodman2004“Aid, Policies, and Growth: Comment,”American Economic ReviewVol. 94 (June) pp. 77480.

    • Search Google Scholar
    • Export Citation

    FrankelJ. and D.Romer1999“Does Trade Cause Growth?”American Economic ReviewVol. 89 (June) pp. 37999.

    HellerP.2005“Pity the Finance Minister: Issues in Managing a Substantial Scaling-Up of Aid Flows,”IMF Working Paper 05/180 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    HellerP.S.Gupta2002“Challenges in Expanding Development Assistance,”IMF Policy Discussion Paper No. 02/5 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    KangJoon ShikAlessandroPrati and AlessandroRebucci2007“Aid, Exports, and Growth: A Time-Series Perspective on the Dutch Disease Hypothesis” (unpublished; Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    MatsenEgi and RagnarTorvik2005“Optimal Dutch Disease,”Journal of Development EconomicsVol. 78 (December) pp. 494515.

    PallageStephan and Michel A.Robe2003“On the Welfare Cost of Economic Fluctuations in Developing Countries,”International Economic ReviewVol. 44 (May) pp. 67798.

    • Search Google Scholar
    • Export Citation

    PratiAlessandr and ThierryTressel2006“Aid Volatility and Dutch Disease: Is There a Role for Macroeconomic Policies?”IMF Working Paper 06/145 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    RajanR.G. and A.Subramanian2005a“Aid and Growth: What Does the CrossCountry Evidence Really Show?”IMF Working Paper 05/127 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    RajanR.G. and A.Subramanian2005b“What Undermines Aid’s Impact on Growth?”IMF Working Paper 05/126 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    RameyG. and V.A.Ramey1995“Cross-Country Evidence on the Link Between Volatility and Growth,”American Economic ReviewVol. 85 (December) pp. 113851.

    • Search Google Scholar
    • Export Citation

    StockJ.H. and M.Yogo2005“Testing for Weak Instruments in IV Regressions,”inIdentification and Inference for Econometric Models: A Festschrift in Honor of Thomas Rothenberged. by D.W.K.Andrews and J.H.Stock (Cambridge: Cambridge University Press).

    • Search Google Scholar
    • Export Citation
1

Tripling official development assistance is viewed as a necessary step to achieve the Millennium Development Goals by 2015 (Heller and Gupta 2002). The Millennium Development Goals, which emerged from the September 2000 Millennium Declaration at the United Nations, are a set of measurable targets for halving world poverty between 1990 and 2015.

2

To sterilize a monetary policy expansion due to aid inflows, central banks can, for example, reduce their holdings of bonds by selling part of their bond portfolio in the open secondary market or by subscribing for an amount of bonds in the primary market that is smaller than the amount coming due. If domestic bond markets are too thin, central banks can issue their own bonds with similar effects.

3

Given that many countries in the sample have Aid–to–GDP ratios much higher than 4 percent, the average of each country’s median Aid–to–GDP ratio is 10 percent. Our estimation sample excludes outliers identified using a standard procedure. If outliers were included, our results would be stronger and would also hold in larger samples with median Aid–to–GDP ratios greater than 2 or 3 percent.

4

All aid flows are measured in percent of the recipient’s GDP.

5

The identifying assumption on which the validity of this instrument depends is that donors are more likely to change their disbursements from one year to the next in a coordinated fashion when they are responding to shocks. As a consequence, a high dispersion of annual changes in aid disbursements across donors would be a measure of lack of coordination and volatility in donors’ policies unrelated to shocks affecting the recipient country.

6

Easterly, Levine, and Roodman (2004) show that the result of Burnside and Dollar (2000) that aid fosters growth in countries adopting good policies is not robust.

7

In the case of natural resources wealth, the country can set up a special fund—such as oil reserve funds in oil-rich countries. In the case of aid, such an option is unlikely to be available.

    Other Resources Citing This Publication