The Fallacy of Differencing to Reduce Multicollinearity
从理论上证明，对时间序列数据做差分处理以减少回归自变量间的多重共线性，在考虑扰动项影响后不可能成功；即使忽略扰动项，其直觉依据也存在缺陷。
Abstract It is shown analytically that differencing time‐series data for the purpose of reducing multicollinearity in the data set for the independent variables of a regression equation cannot possibly succeed when its effect on the disturbance term is taken into account. In addition, the intuitive basis used to justify first differencing of multicollinear data is demonstrated to contain a flaw, even when the effects of differencing the disturbance term are ignored.