MIME-Version: 1.0 Content-Location: file:///C:/2B195D13/burnsideanddollarnotes.htm Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset="us-ascii" Aid, Policies, and Growth

Aid, Policies, and Gr= owth

Burnside and Dollar  AER (2000)

 

Hypothesis about aid:=   Aid affects growth, conditional up= on policies.

 

Direct effect –= sound policies lead to faster growth.

Indirect effect ̵= 1; sound policies lead to aid being more effective leading to faster growth.

 

In countries with bad policies aid is dissipated in ‘unproductive government expenditure’.

 

 

If international capi= tal markets are imperfect, aid can have an important impact as an income transf= er.

 

56 countries, 6 4 year time periods, 1970-73 to 1990-93.

 

Two key questions:

 

= 1)&n= bsp; Is the effect of aid on growth conditional upon economic policies?

= 2)&n= bsp; Do donor governments and agencies allocate more = aid to countries with good policies?

 

i= indexes countries

t= indexes time

git is per capita real GDP growth

 yit is log of initial real per capita GDP

ait is aid relative to GDP

pit is a vector of policies that impact growth=

zit is a vector of other exogenous variables that m= ight impact growth

Fixed time effects a and g are in there to control for worldwide business cycles.

 

Are the error terms correlated? 

Donors respond to neg= ative growth shocks by increasing aid so a negative correlation?

Strategic interests &= #8211; if country has a boom, more aid to ‘get in on the party’ or get some of the credit so a positive correlation?

 

2= SLS as a response.<= /p>

 

GROWTH EQUATION:=

The policy variables:=

 

The Sachs and Warner = openness dummy.

 

Inflation<= /span>

 

B= udget surplus and government consumption, both as a share of GDP. Since these are highly negatively correlated, t= hey go with just budget surplus.

 

 

Other variables:=

 

M= easure of institutional quality that captures security of property rights and efficie= ncy of the government bureaucracy.  Use 1980 value for Knack and Keefe= r.

 

 

Also use ethnolinguistic diversity.  =

 

A= ssassinations variable.<= /p>

 

Money supply (M2) ove= r GDP as a proxy for development of the financial system – lagged one period in response to endogeneity.

 

At first had educatio= n, but no explanatory power and not available for all countries so dropped.

 

R= egional dummies.

 

THE AID EQUATION=

 

L= og of initial income.

 

Log of population

 

A group of variables = to capture donors’ strategic interest:&= nbsp; SSA, Franc zone, Egypt, Central American dummy.  Arms imports relative to total imports (lagged).  Also include policy variables in h= ere to see if policy influences aid.

 

They want to be able = to estimate (1) and (2).

 

However, (1) turns ou= t to be hard to estimate with aid interactions.&nb= sp; So, they take a two step process:&n= bsp; Estimate with no aid terms to fix the coefficients of the policy ind= ex.  Then transform the multiple variab= le policy measures to a single index using the results.  Use this for the interaction.  It is a restricted version of (1) = and (2) that we end up with, as the vector of policy measures (and coefficients= ) is turned into a single scalar.

 

So we end up with (4)= and (5) to estimate.

 

Growth (lhs) and aid = (lhs in 5 and rhs in 4 on the interaction term) are endogenous.

 

Policy is ambiguous.<= o:p>

 

See table A1 for coun= tries.

 

Mean growth rate was = 1.2% for the 275 observations, 1.1% for the low-income sample.

 

Mean value of Aid/GDP= was 1.6%

 

See table 3 on page 8= 54.

 

This suggests, as had= earlier studies, that aid has no significant impact on growth.

 

T= able three, column one, use coefficients to construct policy index and add 1.28 = to it to center on the mean. 

 

Contrast OLS to 2SLS = to investigate endogeneity issue.

 

Table four on page 856, now use policy index rather than policy measures = in different dimensions.

 

Bottom rows are where= the interesting story emerges. 

 

O= utliers perhaps an issue, so they drop a few and rerun.   Table four, column five presents the results.

 

Then drop the middle = income data, leaving 40 countries and 189 observations.

 

What explains the all= ocation of aid?

 

See table 8. 

 

S= maller and poorer.

 

Egypt<= span style=3D'font-size:18.0pt'>.

 

P= olicy with mixed results.

 

F= inally, in IV, aid and government consumption as a share of GDP.  

 

Aid associated with d= onor interests, primarily bilateral aid, increases government consumption.  =

 

Easterly, Levine, and= Roodman in AER, comment 20= 04.

 

Expand the data set c= overage of countries, add in one more time period.<= /o:p>

 

D= ifferent method for removing outliers, but same kind of idea.

 

Earlier paper:  275 observations in 56 countries.<= span style=3D'mso-spacerun:yes'>  New data set, 356 observations in = 62 countries.

 

Reproduce methodology= . 

 

Add in Burkina Faso, Congo, Iran, Jordan, Ma= li, Myanmar, Papua New Guinea, and Uganda.

 

Both sets find a posi= tive coefficient for the aid squared interaction term.