Forecasting in cointegrated systems
研究了在I(1)变量线性系统中施加单位根和协整约束对预测精度的影响，给出了多步预测误差方差的新近公式，并通过蒙特卡洛模拟和英国M1需求预测实例进行了验证。
Abstract We consider the implications for forecast accuracy of imposing unit roots and cointegrating restrictions in linear systems of I (1) variables in levels, differences, and cointegrated combinations. Asymptotic formulae are obtained for multi‐step forecast error variances for each representation. Alternative measures of forecast accuracy are discussed. Finite sample behaviour in a bivariate model is studied by Monte Carlo using control variables. We also analyse the interaction between unit roots and cointegrating restrictions and intercepts in the DGP. Some of the issues are illustrated with an empirical example of forecasting the demand for M1 in the UK.