Error autocorrelation revisited: the AR(1) case
探讨线性回归模型中采用AR(1)误差项所隐含的限制，发现这等价于假设所有变量具有相同的时间结构且互不格兰杰因果，导致残差自相关时OLS和GLS估计量均有偏且不一致。
The aim of the paper is to consider the implicit restrictions imposed when adopting an AR(1) error term in the context of the linear regression model. It is shown that these restrictions amount to assuming a largely identical temporal structure for all the variables involved in the specification. Implicit in this is the assumption that these variables are mutually Granger non-causal. The main implication of this result is that in most cases when residual autocorrelation is detected boththe OLS and GLS estimators are biased and inconsistent.