Noise Created Through the Interactions of Arbitrageurs and Noise Traders. Estimation of an ARMA Model of Well – Known Closed – End Funds of Major Investment Banks

Abstract

In this article, we have applied an autoregressive moving average, ARMA(2,2) model of order AR(1), AR(2), MA(1), MA(2) and SMA(12) to test the natural logarithmic monthly market returns of the of the closed-end funds of major investment banks such as Van Kampen income trust, Aberdeen Asia Pacific income fund, Credit Suisse asset management income, Scudder high income and Templeton global income. The purpose of this article is to eliminate the seasonality and to compare the forecasted with the actual expectations of the market prices of the closed-end funds that arises from the interaction of arbitrageurs and noise traders. The software that we have used is EViews 6. We have found that the natural logarithmic monthly market price returns of all closed-end funds are a stationary series. There is serial correlation in the natural logarithmic monthly market price returns of all closed-end funds before applying the ARMA model. The natural logarithmic monthly market price returns of all closed-end funds are not normally distributed. All equations except equation (8) show a covariance stationary and invertible ARMA process. The equation of Aberdeen Asia Pacific income fund shows a stationary ARMA, but not invertible. The ARMA model is properly specified as the difference between the actual and the estimated residuals of the autocorrelations and partial autocorrelations function of the correlogram are very small. After applying the ARMA(2,2) model in all closed-end funds, we have found that there is no serial correlation in the residuals of the estimated model. The residuals of all closed-end funds are homoskedastic and there is no additional ARCH effect. We have selected the model with the best forecasting ability in terms of the smallest mean absolute percentage error, the lowest value of the Akaike information criterion, Schwarz criterion, and covariance proportion. We have concluded that the best fit model is Templeton global income, (LNMPTGI) closed-end fund. The Akaike and Schwarz criterion are -4.015 and -3.89 respectively. The mean absolute percentage error is 73.78% and the covariance proportion is 0.37. The one – step ahead forecast of the natural logarithmic monthly price returns of all closed-end funds does not have power to forecast future values. It does not show unanticipated changes in the prices of the closed-end funds that could help the arbitrageurs to craft their investment strategy. The whole dataset is from 31/01/1990 to 31/12/2001. The total observations are 144 and the logarithmic monthly market returns observations are 143. The data was obtained from Thomson Financial Investment View database.