Dangerous interactions: Problems in Interpreting tests of conditional aid effectiveness

David Fielding*, Stephen Knowles

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

There is now a substantial empirical literature examining the determinants of aid effectiveness. A large part of this makes inferences based on a regression incorporating aid (as a share of recipient GDP) interacted with some institutional or policy variable. Recently, some authors have questioned the statistical robustness of such regressions, pointing out that results vary according to the way aid is measured and the estimator applied to the data. Moreover, the regression equations used to test hypotheses about the determinants of aid effectiveness are often introduced without any corresponding formal theory. We explore aid-policy interaction terms in the context of a simple theoretical model, showing how different nonlinearities may be conflated. The resulting difficulties in the interpretation of aid-growth regressions are illustrated in the context of a seminal paper in the conditional aid effectiveness literature. One simple change in the way that aid is measured - in per capita terms rather than as a fraction of GDP - completely changes the regression results. This indicates that adding interaction terms to otherwise linear regression equations is an inadequate way of capturing the nonlinearities in the growth process. Our aim is to re-emphasise the importance of grounding.

Original languageEnglish
Pages (from-to)972-983
Number of pages12
JournalWorld Economy
Volume34
Issue number6
DOIs
Publication statusPublished - 21 Jun 2011

Fingerprint

Dive into the research topics of 'Dangerous interactions: Problems in Interpreting tests of conditional aid effectiveness'. Together they form a unique fingerprint.

Cite this