Forecasting the outflow from non-maturity deposits using astressed seasonal autoregressive Monte Carlo simulation

Abstract: Non-maturity deposits (NMD) are saving accounts without a predefined maturity, whichmeans that depositors can withdraw or deposit any amount freely. On the other hand,banks have an option to freely alter deposit rates. This embedded option makes NMDshard to predict and NMDs are important for banks to model since they often rely onthese instruments for medium-long term funding. This thesis seeks to provide a Swedish bank with a flexible framework to forecast theoutflow of deposit volume under the stressed condition that no future deposits will bemade. The deposit volume will be modelled using a seasonal autoregressive model andthe forecast generated by a stressed Monte Carlo simulation that only allows the forecastof withdrawals. Under the Swedish Financial Services Authority’s regulations models for NMDs are onlyapplicable on a certain segment of the total volume known as Core Deposit. This is theamount of the total deposit volume unlikely to be withdrawn even under adverse changesin interest rates. Several approaches on how to estimate this volume are evaluated in thethesis. The best approach is found to be an Account Based approach where each depositis flagged as Core or Non-Core dependent on the account’s sensitivity to interest ratechanges.