Determinants of market-assessed credit risk
View/
Keywords
Credit ratings ; Sovereign debt ; Information and communication technologies ; ICT ; e-readiness ; Developing countries ; Shadow economy ; NRI index ; CART ; Random forest ; Bagging ; Gradient boosting ; Recurrent neural networkAbstract
A vast body of academic literature unveils as primary determinants of sovereign credit ratings and bond yields, a number of domestic macroeconomic and financial fundamentals, as well as global factors like the international risk appetite and the global liquidity. The scope of this study is to evaluate two phenomena that have not been explored in a great extent in previous research as potential factors of sovereign ratings and rates. The first phenomenon is the shadow economy, a pervasive and widespread feature of economies throughout the world. The second one is the prevalence of information and communication technologies (ICT) that transform every aspect of social and economic life.
The study unfolds in two waves. The first wave, which corresponds to thesis’ first chapter, covers the years 2001-2010 and concentrates only on ICT effects following a parametric model. More specifically, we adopt a modified random effects approach which allows us to distinguish between short and long run effects on a dataset of 65 countries for a time span of ten years. We show that ICT have a significant impact on a country’s credit rating and cost of debt, regardless of the presence of other key variables proposed in the literature. The effect is stronger for non-OECD countries, indicating a pathway for developing countries to improve their access to debt markets. Our conclusions are robust to the advent of the recent financial and debt crisis.
The second wave expands in years 2001-2016, corresponding to thesis’ second chapter and attempts to outline the main effects of shadow economy and ICT penetration on sovereign credit ratings and the cost of debt, along with possible second-order effects between the two variables. The chapter presents a range of machine-learning approaches, including bagging, random forests, gradient-boosting machines, and recurrent neural networks. Furthermore, following recent trends in the emerging field of interpretable ML, such as feature importance and accumulated local effects, we attempt to explain which factors drive the predictions of the so-called ML black box models. We show that policies facilitating the penetration and use of ICT and aiming to curb the shadow economy may exert an asymmetric impact on sovereign ratings and the cost of debt depending on their present magnitudes, not only independently but also in interaction.
The last chapter is a brief presentation of the time-evolving impact of the two phenomena on the Greek sovereign cost of debt through years 2001-2016. A number of local model-agnostic interpretations of predictions regarding Greece is presented in order to identify the magnitude of the attributes that shape the prediction. Policy implications drawn upon research findings and government plans and intentions are also briefly discussed.
Department
Σχολή Ναυτιλίας και Βιομηχανίας. Tμήμα Βιομηχανικής Διοίκησης και TεχνολογίαςNumber of pages
132Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα