Working Papers

This paper proposes the use of synthetic training data generated by large language models
to improve machine learning SDG classifiers. It shows that supplementing existing training data with
synthetic data produced by the ChatGPT tool improves the performance of the SDGClassy classifier.
This addition of synthetic data is especially useful in building SDG classifiers given the limited availability
of properly labeled data and the complex, interconnected nature of the SDGs. Synthetic data thus enables
more effective machine-learning applications in this context.

From around 2000 onward, donors and recipient governments embarked upon a new aid paradigm. The most important elements include increased selectivity in the aid allocation, more ownership of recipient countries based on nationally elaborated PRSPs, and more donor alignment and harmonization via program-based approaches such as budget support. The paper assesses the theoretical merits of this new paradigm, identifying some contradictions and limitations, and then examines its implementation over the past decade and its results. The empirical results largely confirm the earlier identified weaknesses and limitations. The paper concludes with some suggestions for improving aid practices.

This paper recalls the history of proposed “innovative” mechanisms by which governments could strengthen financial cooperation for development. Such proposals sought more predictable and assured financial flows to facilitate recipient country programming, while also substantially adding to the volume of highly concessional international support for development. International discussions of these proposals mostly began in the 1960s and in many cases continue today, although implementation thus far has been modest. These discussions are contrasted with generally more recent proposals that proponents call “innovative” but that do not share the characteristics of the more radical thinking…