Abstract:
The present study suggests a model for predicting liquidity gap, based on source and cost of funds approach concerning the daily time series data (25 March 2009 to 19 March 2018), in order to control and manage the liquidity risk. Using the family of autoregressive conditional heteroscedasticity models, the behavior of bank liquidity gap is modeled and predicted. The results show that the APGARCH with the Johnson-SU distribution is the most suitable model for explaining the liquidity gap behavior. Based on the rolling window method the more accurate model has been selected to be the APGARCH model with T-Student distribution which provides the least error in forecasting liquidity gap.
Machine summary:
Modeling the Liquidity Gap in a Private Bank Esmaiel Abounoori* Ali Sadeghzadeh Yazdi† Alireza Erfani‡ Received: 9 Oct 2018 Approved: 8 Sep 2019 The present study suggests a model for predicting liquidity gap, based on source and cost of funds approach concerning the daily time series data (25 March 2009 to 19 March 2018), in order to control and manage the liquidity risk.
In order to identify the process of data production and to control and manage the liquidity risk of a private bank, we try modeling and predicting the behavior of the liquidity gap series (the difference of resource and cost of funds) based on cash flow prediction using Generalized Autoregressive Conditional Heteroskedasticity models, In this way, some of the observed characteristics of the liquidity gap series such as Heteroskedasticity, Fat Tails, Volatility Clustering, Leverage Effect, Volatility Feedback and Long Memory in the modeling process are considered.
Divandari, Lucas and Mousavi (2004) design a prediction model for liquidity management of financial institutions within the framework of a usury-free banking system in order to control liquidity risk.
The results of the predicted error statistics for the mean time series and the volatility of the liquidity gap, applying the rolling window, indicate that the APGARCH model with t-student distribution has a better performance in predicting the liquidity gap of the bank in periods of 10, 20 and 30 days.
The results of this comparison indicate that the APGARCH model with the t-student distribution has the least error in the prediction of liquidity gap in the time horizons of research based on all three MAE, RMSE and TIC criteria.