Towards a new model for early warning signals for systemic financial fragility and near crises: an application to OECD countries

Using a signal extraction framework and looking at OECD countries over a 30 year period this paper attempts to identify a number of variables significant in predicting near-crises as a pre-cursor to full-fledged crises. These include growth in pension assets as an indicator for the development of liquidity bubbles, equity market dividend yields as a proxy for corporate balance sheet health, banking sector assets growth and relative size to GDP. We also study the development of asset price bubbles through an equity markets indicator and a house price indicator. Finally we also look at a banking sector funding stability indicator and liquidity indicator on a micro-level. Simultaneously, a dynamic research design improves on previous static set-ups and enhances the model predictive power and applicability to different time periods. This paper shows that as early as 2004, clear signals were being given for a number of countries that vulnerabilities were building up with out-of-sample performance better than in-sample in terms of overall noise to signal ratios, showing a significant improvement compared to earlier work. EWS design has significant implications for financial stability and financial regulation.