Predicting Financial Fragility in Nigeria: Does the Filter Method Matter?

Abstract

This study has applied two distinct methods of analysis to evaluate and compare the predictive ability of certain EWIs of financial crisis when gaps are generated using two filter methods – the Hodrick – Prescott and Kalman Filters. First, the receiver operating characteristics curve where the area under the receiver operating characteristic (AU-ROC) curve and distance to corner were the basis of evaluation and 2, the logistic regression encompassing the estimates for individual indicator parameters, the model Expectation-Prediction Evaluation and the Hosmer-Lemeshow (HL) and Andrews Tests for goodness-of-fit. On the basis of the AU-ROC and Distance to Corner, the study concludes that the credit-to-GDP gap is a predictor of financial crisis in Nigeria. On the other hand, the Logit regression leads to the conclusion that none of the EWIs tested (Credit-to-GDP gap, Nonperforming loans, Loan-to-Deposit ratio and asset prices) could predict financial crisis at the 5% level of significance although credit-to-GDP gap could at 10%. Nevertheless, both the AU-ROC and Logit regression, suggest that credit-to-GDP gap outperforms Non-performing loans, Loan to deposit ratio and asset prices as EWIs of financial crisis in Nigeria. Going by the results of the AU-ROC curve and the Logistic regression, we do not find any significant difference whether gaps are from HP filter or Kalman filter. It is hoped that regulatory authorities apply the EWIs of financial crisis with caution, explore the different methodologies available and identify which EWI, filter method as well as the analytical model suitable for their jurisdiction.