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.

This article evaluates and compares the forecasting performance of three international organizations: the United Nations, the International Monetary Fund and the World Bank. The annual forecasts made by the United Nations in the period of 1981-2011 are found to be fairly robust, in terms of bias and efficiency. In comparison, the forecasting performance of the United Nations is found to be marginally better than the other two organizations during the period of 2000-2012. However, the forecasts of all these organizations missed the Great Recession of 2009 by a large margin.

There are two major failings in policy interventions in the crisis in the US and Europe: the reluctance to remove the debt overhang through timely, orderly and comprehensive restructuring and the shift to fiscal austerity after an initial reflation. These have resulted in excessive reliance on monetary means with central banks entering uncharted policy waters, including zero-bound interest rates and the acquisition of long-term public and private bonds. This ultra-easy monetary policy has not been very effective in reducing the debt overhang and stimulating spending. It has, however, generated financial fragility, at home and abroad, particularly in the case of the US as the issuer of…