Bulletin of the World Health Organization

Does recession reduce global health aid? Evidence from 15 high-income countries, 1975–2007

David Stuckler a, Sanjay Basu b, Stephanie W Wang c & Martin McKee d

a. Harvard University, Department of Global Health and Population, 677 Huntington Avenue, Boston, MA, 02115, United States of America (USA).
b. University of California in San Francisco, Department of Medicine & San Francisco General Hospital, San Francisco, USA.
c. California Institute of Technology, Division of Humanities and Social Sciences, Pasadena, USA.
d. London School of Hygiene & Tropical Medicine, European Centre on Health of Societies in Transition, London, England.

Correspondence to David Stuckler (e-mail: dstuckler@hsph.harvard.edu).

(Submitted: 20 June 2010 – Revised version received: 06 January 2011 – Accepted: 11 January 2011 – Published online: 25 February 2011.)

Bulletin of the World Health Organization 2011;89:252-257. doi: 10.2471/BLT.10.080663

Introduction

There is fear that the ongoing global economic recession will lead high-income countries to reduce commitments or disbursements for development assistance for health (DAH). The World Bank’s chief economist for Africa, noting that the Group of Eight failed to meet its 2005 commitment to double aid to Africa, has expressed this fear as follows: “If during the boom period they were unable to meet these commitments, I'm wondering what is going to happen now that we're in a deep recession.”1

Yet so far the issue remains controversial. Although development assistance has been relatively protected from cuts since the start of the 2008 economic recession, many industrialized countries are now embarking on severe austerity measures. Some, such as the United Kingdom of Great Britain and Northern Ireland, have stated that they will protect overseas development assistance; others have been less explicit. Italy and Ireland have reduced their allocations for development assistance by 56% and 10%, respectively,2 while other countries, such as Germany and the United States of America (USA), increased their spending on development assistance when the current financial crisis arose.3 Australia’s AusAID has shown sustained commitment by creating a Global Economic Crisis Taskforce to support recipient countries facing additional economic problems resulting from the financial crisis.4 Despite a General Assembly resolution stating that the United Nation’s central emergency response fund should receive 500 million United States dollars (US$) annually to address disasters such as the earthquake in Haiti and the floods in Pakistan, only US$ 358 million were raised for the current year. China, despite being the world’s second-largest economy, pledged US$ 500 000; Italy promised US$ 1 500 000.5

Non-state actors such as the Bill & Melinda Gates Foundation, the William J Clinton Foundation and the Carter Center also play an important role in determining financial flows to global health programmes. Endowments invested in equities may also be adversely impacted by recession. Grants (for all purposes) made by foundations based in the United States fell by 8.9% in 2009,6 creating concerns about delayed implementation or disbursement of funds as well as potential future reductions in funding. Uncertainty in future funding can greatly affect the capacity of global agencies such as the World Health Organization (WHO) to plan global health programmes effectively.

An overall reduction in commitments or disbursements could be particularly destabilizing for countries that use such aid to support ongoing health-care infrastructure development or to sustain existing health-care programmes7 Among African nations, more than one-third of annual health spending may come from donor financing.

In this paper we examine what has happened during previous economic downturns and hypothesize that recessions lead to declines in official DAH. To test this hypothesis, we examined the records of 15 high-income European Union countries belonging to the Organisation for Economic Co-operation and Development (OECD) for the period between 1975 and 2007.

Methods

We investigated the relationship between economic downturns and the official DAH committed and provided by 15 OECD countries between 1975 and 2007 using data from the Creditor Reporting System.8 The countries included were Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden and the United Kingdom of Great Britain and Northern Ireland. We identified economic downturns in three ways. As a first step we dated business cycles drawing on the methods specified by Camacho et al.9 As a second step we detected fluctuations in per capita gross domestic product (GDP) expressed in purchasing-power-parity dollars (i.e. the economic growth rate in percentage terms, which is effectively the same as taking the difference in the natural log of successive measures of GDP).10 As a third step we attempted to measure not just whether or not a recession existed (coded dichotomously, yes or no), but also its depth and scale. To do so we used continuous measures, such as the unemployment rate, to correlate the severity of the recession with the degree of reduction in DAH. The importance of this step is exemplified by the current recession in the USA: if GDP alone is used as a measure of a downturn (standard practice by the National Bureau of Economic Research), the USA appears to be out of recession, since as we write its GDP is again rising. However, this would fail to capture the economic hardship confronting ordinary people who are losing jobs, a factor that is likely to influence political debates on global aid commitments. Specifically, job loss may put pressure on policy-makers to increase domestic social welfare expenditure over international aid. Thus, our third measure of the severity of the crisis was captured by fluctuations in unemployment rates.

We modelled the percentage change (relative difference) in commitments and disbursements for DAH as a function of each of the three economic variables described in the previous paragraph (a method of first-differences, although our results were consistent when we modelled using the current year’s levels of aid and the absolute yearly differences rather than yearly percentage changes). The rationale for using a percentage difference equation was twofold: first, we hypothesized that a recession (a negative change in GDP) would result in a reduction in DAH (a negative change in DAH); second, our statistical tests suggest that the data series are weakly integrated (using Dickey-Fuller tests), which indicates that there is trending in the pre-2008 data. Had we evaluated the level of aid instead, we could have found spurious associations with the state of the economy, as both the GDP and DAH were rising relatively steadily for reasons that may be causally unrelated. As a further robustness check we added a trend variable to account for the longer-term pattern of rising DAH in response to the scaling up of targets for global aid, which enabled us to better isolate the association of a recession with changes in aid. Finally, to mitigate the impact of small numbers and zero values (for the 69 country–years in which no health aid was committed), we constrained the sample to more than a doubling in funding for DAH, although we also found that this constraint had no effect on the results.

Our basic model was thus:

Hi,t = α + βRECi,t + εi,t

where i is the country, t is the year, ∆H represents the percentage change in commitments or disbursements for DAH (i.e. rate of growth), REC includes a series of measures of economic downturn (including the economic growth rate and percentage point changes in unemployment rate) and ε is the error term. To reflect the fact that countries were not independently sampled, we clustered standard errors (SEs) for robustness to autocorrelation (noting that autocorrelation affects the SE but does not bias the coefficient estimate11). Data were analysed using STATA v 10.1 (StataCorp. LP, College Station, USA). All data and statistical codes are available from the authors upon request.

Results

Table 1 shows the results of our basic models. We found no statistically significant association between commitments in DAH on the one hand and a state of recession, fluctuations in GDP per capita or changes in unemployment rates in donor countries on the other. These findings were unchanged after holding fixed differences between countries constant (correcting for time-invariant differences in surveillance) and after accounting for longitudinal trends in DAH commitments (correcting for patterns in DAH commitment changes that were already in motion before the onset of recession; Table 1).

While countries may have remained committed to DAH, they may have disbursed less than they had committed themselves to give. Disbursements recorded between 1995 and 2007, when they were first reliably incorporated into the OECD Creditor Reporting System database, were approximately 4% lower than commitments to DAH (on average, US$ 122 per capita in commitments versus US$ 117 per capita in disbursements).8 Overall, health commitments and health disbursements were strongly correlated (r = 0.82; P < 0.0001; number of country–years = 314). As shown in Table 2, we found no association between recessionary years, GDP downturns or unemployment increases and disbursements for health assistance; changes in disbursements for DAH correlated with changes in commitments to deliver DAH, and disbursements for DAH were not reduced during recessions.

It is, of course, possible that recessions exert a delayed effect on health spending. This may occur, for example, when budgets including commitments to deliver DAH have been set in advance for the year in which the recession occurs. Table 3 shows the results of statistical tests examining the delayed effects of the selected economic indicators on DAH. Again, we found no statistically significant association between any of the three measures of economic downturn and commitments in international assistance for health when incorporating time lags. Small negative coefficients marked the relationship between real changes in GDP and aid commitments using one- and two-year lags; however, they were not statistically or jointly significant (F(2,14) = 0.15; P = 0.86).

On the other hand, fluctuations in the previous year’s employment rate were significantly associated with falls in DAH commitments. For every percentage point increase in unemployment over the previous year there was a drop of 13.8% in DAH. However, for every percentage point increase in unemployment there was a 5.7% increase in aid after a two-year lag and an 8.5% increase in aid observable in the concurrent year; cumulatively (14.1%) this cancelled out the drop in aid observed after a lag. Thus, there may be some cyclical volatility in DAH associated with economic fluctuations, although the overall change in the amount of DAH is not statistically significant.

Discussion

Overall, we found no robust evidence to support our hypothesis that recessions lead high-income countries to reduce their total DAH commitments or disbursements, either immediately after recession begins or within a two-year period. Before interpreting these findings, we must note several possible statistical reasons for the lack of a significant association between indicators of recession and aid flows. Current data samples could lack the statistical power needed to detect any effect of recession if it is small. However, our sample represented more than 200 country–years and is therefore highly likely to detect any effect other than a very small one. A high degree of measurement error is another possibility. All models yielded very low R2 values, which suggests a lot of unexplained variation both within and between countries. Measurement error could also lead to attenuation bias,13 although the Creditor Reporting System has put considerable effort into standardizing reporting to improve comparability among countries. We have also evaluated within-country variations, which would not be impacted by between-country differences in surveillance methods. Such differences are where the greatest artefactual variation in measurement is likely to occur.

We have reported the average estimated association between recessions and DAH, and the average may conceal variation in how countries respond to financial downturns. For example, Sweden’s health disbursements were US$ 74 million per capita in 1990 and dropped slightly, to US$ 73 million, in 1991. However, they rose to US$ 99 million in 1992. On the other hand, Finland experienced a more substantial reduction of about one-third in GDP from 1991 to 1992 and reduced its aid commitments by a similar magnitude. Reduced commitments by some donors may be compensated for by other donor agencies, resulting in a null effect. Several non-state actors have increased aid commitments during the current recession. For example, Warren Buffett, a major donor to the Gates Foundation, saw his fortune increase by more than US$ 10 billion during 201014 and the overall resources available to the Gates Foundation (which depend on the stock market performance of its endowment investments) increased in 2010 after a short-term decline the previous year.15,16 Nonetheless, WHO staff report that some foundations have taken measures to offset the possibility of a future endowment decline, such as a delayed delivery of funds and other strategies that could unintentionally disrupt the effective planning of global health programmes.5

The above notwithstanding, our findings suggest that high-income countries’ economic performance is not likely to be a significant determinant of DAH commitments or disbursements, as shown by past experience. Instead, they point to the importance of other global health debates and to political factors in determining aid allocations. One contemporary risk is that donor agencies will behave in response to a political climate calling for generalized austerity that fosters the erroneous belief that a reduction in aid is the inevitable consequence of recession. Our evidence is consistent with a potential mimetic effect, whereby donors’ aid decisions are influenced by those of other donors. Thus, we may be facing a self-fulfilling prophecy. Future research should investigate the factors that may be driving health aid allocations and their potential influence on the reliability of development aid.


Competing interests:

None declared.

References

Share