Development of a forecasting model for deposits of credit unions

The purpose of this thesis is to develop a forecasting model to predict demand deposits and term deposits of credit unions. It begins with a survey of the literature on demand functions for liquid assets. Both single equation models and simultaneous equation systems are summarized. The hypothesis for a likely structural model of credit union financial behaviour is also presented. However, a structural model cannot be estimated because there are no published data on credit unions' interest rates and there is a limited number of observations for the dependent variable. The forecasting technique that is being developed in this thesis is an application of time series analysis. The basic idea behind this approach is to express the time series of demand deposits and of term deposits as a weighted sum of the past values of deposits. The weights in the sum are determined so as to achieve the greatest predictive power by minimizing the mean square error of the forecasts. The data are quarterly time series for demand deposits and for term deposits for each of three credit unions in the Vancouver Region in British Columbia from the second quarter of 1962 to the fourth quarter of 1974. The data are printed in the Appendix. The strength of the mixed autoregressive moving average process (ARIMA) as a forecasting tool for financial intermediaries such as a credit union is evaluated by using a large sample of monthly data of personal demand deposits and personal term deposits of Canadian chartered banks. The best models for each credit union's demand deposits and term deposits are matched against the naive model of a random walk process. They are compared with respect to their minimum mean square error of prediction for the four quarters of 1974. For both the three credit unions and the chartered banks, in all cases the best ARIMA model outperformed all other candidates.Business, Sauder School ofGraduat