| Summary: | This essay reviews theories on developments during financial booms that lead to financial imbalances and possible crises. Those theories are put to the test with a statistical model that is used to forecast the probabilities of banking crises. We find that domestic indicators of financial imbalances, such as credit to the private sector, real house prices and bank financing as well as their international counterparts, are useful predictors of crises in a multivariate framework. The resulting model displays good crisis predicting properties in sample and seems to combine risks emanating from many sectors logically. Risks stemming from the financial sector in Iceland are low according to the fitted model (as of Q1 2017). The strength of the model lies in combining multiple risk factors into a comprehensive measure. The weakness, which applies in Iceland especially, is that the model does not provide much interpretable information during tranquil times. For that reason, we draw the conclusion that the model is best considered a helpful tool in combination with multiple univariate models but not a measure that can be used in isolation. As with other indicators that are based on historical data, the logistic regression models developed in this thesis are only useful to predict crises that are similar in nature to those that have taken place in the past. Since systemic banking crises happen infrequently it’s necessary to use data from many countries to gather varied enough crisis observations so that the model can anticipate as many future crises as possible.
|
|---|