A striking article published in The Lancet in April 2010 concluded that when governments receive international aid for healthcare projects, their own spending gets “crowded out”, or displaced. That is, when $1 of money gets delivered from the US to help build a hospital in Ethiopia, the Ethiopian government often takes away $0.43 of its own spending on the hospital and puts that money into something else like the military. The article also indicated that this “displacement” doesn’t happen when international aid is given to non-government organizations like charities.
The implications are obvious–is aid making governments less responsible for their own healthcare systems, and should aid be redirected away from governments to private charities? This week, a new article published in PLoS Medicine questioned those conclusions–and the very premise of the original Lancet article. In today’s blog post, we take a look at the data underlying these two articles and the debate they have generated, and ask whether we should be concerned about “crowding out” issues in international health aid.
First, a quick review of the original Lancet article: A group of statisticians assembled data from the International Monetary Fund and the World Health Organization describing domestically and externally financed public health expenditures in 193 countries from 1995 to 2006 (that includes government expenditures on health–GHE–which incorporates government domestic health spending, and development assistance for health–DAH–or international aid). They found that while GHE had increased dramatically over the decade being studied (see Figure below), for every US$1 of DAH to a government, government health expenditures from domestic resources were reduced by $0.43 to $1.14. However, DAH to the non-governmental sector had a positive and significant effect on domestic government health spending (more domestic spending when international aid comes in through an NGO).
The authors were notably quite cautious about conclusions made from the article. They warned that a substantial portion of the data were “imputed”–or estimated via inference, given the lack of reliable data. For example, of $19 billion in DAH disbursed in 2006, the authors were only able to directly trace $5.7 billion of it. They also specified that the displacement effect may be found because government ministries are unable to “absorb” the aid and make use of it, so they can only spend so much money on a given budget at a time given limitations of capacity. Other commentators noted three potential other hypotheses for the displacement: governments might be compensating for exceptional international generosity to the health sector by reallocating government funding to other sectors that are not as well-funded but are necessary (e.g., waste management); governments might be anticipating long-term unreliability of international health aid by stalling possible increases of recurrent health expenditure; or governments might smooth irregular aid flows by spreading aid across several years.
But while these academic discussions continued, the paper’s perception in the development community and in the public eye could best be summarized by an old quote from a World Bank economist: “When the World Bank thinks it is financing an electric power station, it is really financing a brothel.” A recent newspaper columnist entitled an article “Health aid made some countries cut budgets”, and concluded that “When an aid official thinks he [sic] is helping a low-income African patient avoid charges at a health clinic, in reality, he is paying for a shopping trip to Paris for a government minister and his wife.”
The new study
Now, a replication of the Lancet article has called into question some of its fundamental conclusions. In the new analysis, an independent set of scientists used the same dataset to analyze how the displacement result was achieved. What they found is that the association between DAH and displacement of government health expenditures is not robust after exclusion of a small subset of highly-implausible data. The trends are driven by “outliers”, or extreme data points that are likely to be the result of erroneous interpolation of misreporting. For example, in the Figures below, we can see that there is a wide scatter in how GHE relates to DAH, rather than a linear trend implied by a simple linear regression result.
When the authors replicated the Lancet results, they found that the linear relationship between DAH and government health expenditure from domestic sources was lost when data for which the World Health Organization (WHO) and International Monetary Fund (IMF) databases differed by 10-fold or more were removed (i.e., removing data that was likely erroneous or based on poor information), when a small set of implausible data points were removed, and when restricting the sample to eliminate those countries that receive very little DAH as a percentage of GDP (extreme data points that generate artifacts).
The new study also found that the old study left out 51 countries who received DAH, including Russia and much of Eastern Europe, Iraq, Afghanistan, the occupied Palestinian territory, Somalia, and several small Island states. The new analysis also statistically accounted for how different countries have different aid management systems–a statistical technique known as “fixed effects” that takes into account each country’s unique characteristics by creating a dummy variable for each country in the analysis, rather than assuming that each country’s data is a random selection from the distribution of all countries who manage aid similarly (Wonkish side note: fixed effects essentially looks at the overall slope of the spending curve in relationship to aid, to see if there is a within-country relationship between aid and spending. One way to test which type of model is appropriate for an analysis is a Hausman test).
So does displacement exist?
The new study does not eliminate the possibility that diversion exists–it merely shows that there are serious data limitations with making any conclusion about diversion. With those data caveats in mind, one issue to note is that both the original and new study look at correlations between overall health spending and overall levels of aid relative to GDP; alternatively, one can analyze whether short-term changes to the existing level of aid produce changes to the existing level of domestic spending on health, as this might better capture whether diversion is occurring.
When we performed that analysis, we found (subject to all the data limitations mentioned earlier about imputation), that an additional $1 in aid produces 63 cents of diversion. However, as we explained in a prior post, the result really depends on whether or not countries are borrowing from the International Monetary Fund, which advises countries to divert their resources to compensate for the volatility of aid. In countries that were subject to IMF policies, each additional $1 of aid resulted in less than $0.01 added to the health system on average (nearly complete displacement). In countries that did not borrow from the IMF, however, each additional $1 of aid resulted in about $0.45 added to the health system on average.
Of note, this doesn’t mean that one group of countries is committing a mistake or corruption and the other isn’t. We don’t really know if displacement is a “bad” thing. For example, an analysis from Vietnam found no changes in outcomes are by displacement of government funding; this might indicate that governments are appropriately shifting resources where they are more needed and not “redundant”. Furthermore, a careful analysis of direct data from several countries suggested that countries increased their spending in health projects when donors increased spending, suggesting that specific projects will be funded or defunded based on individual project-level concerns rather than as a general rule.
Alternatives to the crazy aid market
A primary conclusion to draw from this debate is that the data on aid and its impact, even on other financial streams of funding (let alone health outcomes) is highly irregular and varied. Furthermore, aid streams seem themselves volatile and preferentially distributed to pet projects.
A novel proposal to generate an alternative to this current state of affairs is to create a Global Fund for Health similar to the Global Fund for AIDS, TB and Malaria but for broader health concerns. The more general fund would provide overall health system and social spending support on a perennial basis (i.e., not for individual projects, but provide a regular infusion of money to counter-act general lack of funds and provide reliable long-term investments that all participating countries pay into and draw out of). In theory, this is like developing an international welfare system analogous to our domestic welfare systems.
When natural disasters or market crashes occur, such as fund could serve as a reliable buffer, offering redistributive social support funds internationally just as food stamps and disability insurance provide buffering systems domestically. And just as we all pay taxes to support food stamps and disability insurance domestically (and we all benefit from them in times of bad luck or illness), rather than viewing these taxes as project-based charity, the Global Fund would offer the same mechanism for money collection and distribution across borders based on ability and need. The debate and discussion about creating such a fund is currently taking place in a conference in Berlin, with hopeful prospects to develop a serious proposal and finance it through capital-collection methods that we described in a previous post.