E M M A N U E L F R O T & M A R I A P E R R O T T A
C H A L L E N G E S F O R A I D E F F E C T I V E N E S S C O N F E R E N C E
S T O C K H O L M M A Y 3 0 , 2 0 1 1
Aid Effectiveness: New
Instruments, New Results?
Aid effectiveness
Aid effectiveness is a central matter for development agencies
Officially, aid has a positive role. Monterrey consensus:
“ODA is critical to the achievement of the development goals and targets of the Millennium Declaration”
However the academic community seems much less convinced
Aid effectiveness
Large academic literature on the relationship between aid and GDP
No robust conclusions
Aid skeptics take that as further evidence that aid does not help poor countries
Easterly (2007): “Was Development Assistance A Mistake?”
Why is it so difficult to reach robust conclusions?
When trying to prove aid has an effect on GDP
growth, economists measure the correlation between aid and GDP growth:
-2-1012GDP growth
2 4 6 8 10
Log ODA
Why is it so difficult to reach robust conclusions?
But this correlation does not tell us anything about causation
Countries with high growth are expected to receive less aid → apparent negative relationship between aid and growth
We need to use econometric techniques to isolate the causal link between aid and GDP growth
Aid ? GDP growth
Instrumental variables
In order to establish causality, instrumental variables are used
An instrument is a variable that:
Affects aid
Affects GDP growth only through aid (it is exogenous)
Such an instrument is quite difficult to find!
Instrumental variables in the aid literature
The whole aid literature uses similar instruments
But these are not convincing. Basically, all the papers strongly rely on population being a valid instrument
That assumes:
Population affects aid quantities
Population affects GDP growth only through aid
What next?
Are there no valid instruments?
Is there no answer to the aid to growth question?
Some argue econometric techniques should be abandoned on this topic
But:
That leaves the debate to loose arguments picked to support one’s position (aid skeptics vs. aid enthusiasts)
It also implies nothing can be precisely, rigorously said about the efficiency of the past 40 years of aid spending
This paper
We discard instruments used in the past
We build a new instrument
As far as possible, we address all the econometric issues raised in the past against this literature
A new instrument
A donor aid portfolio is made of several countries (usually more than 100)
Each donor-recipient partnership started at a different date:
For instance, Sweden started to give aid to Liberia in 1962, but only in 1976 to Madagascar
For each donor-recipient pair, we find the first year aid was disbursed: the entry date
A new instrument
Then we use a key observation: donors give relatively more aid to countries with earlier entry dates
This is true even after many years
Entry dates and aid shares
Predicted aid
Predicted aid corresponds to the aid quantity a country should on average receive, given its entry dates
Predicted aid is our instrument
Is it a good instrument?
First, there is no reverse causality: a negative shock to GDP causes higher aid, but not higher predicted aid
Second, it is correlated with actual aid
Third, we claim, and provide evidence, that it only affects GDP growth through aid: it is exogenous to GDP growth
Aid-growth relationship
Aid-growth relationship
Comparison with existing literature
A meta-analysis of the literature finds that “the best estimate we can make of the elasticity of the real product to aid is about 0.13”
If aid increases by 1 %, GDP increases by 0.13 %
We find a positive, statistically significant effect, but ten times smaller
Conclusion
Mixed results: positive effect, but really small
It implies past aid efforts did help, on average, to increase GDP, but not by a large quantity
It clarifies the political debate:
It quantifies the trade-off between the costs and benefits of aid spending
It also shows what we can expect from aid if disbursed as it used to be
Results consistent with arguments heard more and more often:
Aid sometimes work and sometimes does not, so on average the effect is rather small
A fruitful research area is to understand when it works
Aid is not sufficient to bring development