Illicit Financial Flows & Fragile States

Illicit financial flows1 (IFF) have been identified as a major socio-economic obstacle in the path to sustainable development. Regarding IFFs, all the academic studies done so far have focussed on the superset of developing countries, without giving special attention to the fragile and conflict-affected ones. Moreover, the literature on the economic consequences of IFFs is still in the embryonic stages and the majority of scientific studies (Collier et al., 2001, 2004; Le and Zak, 2006) have mainly focussed on the causes and effects of capital flight2 (Herkenrath, 2014). While several theoretical reflections
and case studies focus on the interrelationship between IFFs and state fragility, there have been no attempts to empirically quantify the IFF volume and exact dynamics underlying this interrelationship, primarily, due to a lack of reliable data on IFFs in fragile and conflict-affected states.

Besides understanding the interlinkages between IFFs and state fragility, another important focus of this Ph.D. thesis would be to bolster the toolkit of policymakers committed to curbing IFFs. In view of that, this research will demonstrate the untapped utility of Machine Learning (ML) based models in detecting and preventing IFFs by creating a prototype model using disaggregated trade and administrative transaction data. The model shall combine the price filter analysis method (commonly used for identifying trade mispricing; de Boyrie et al., 2007; Hong et al., 2014) with ML-based anomaly
detection algorithms such as Gradient Boosting Machine (GBM), Support Vector Machine (SVM), LSTM Recurrent Neural Network, etc. (commonly used in the financial and banking sector for identifying credit card frauds or for detecting network intrusion in cybersecurity). Such a hybrid model could recognise known and unknown patterns of trade mispricing and has a huge potential for improved IFF detection, estimation, and eventual reduction.