Measurement error in a first-order autoregression
The Ordinary Least Squares (OLS) estimator for the slope parameter in a first-order autoregressive model is biased when the variable is measured with error. Such an error may occur with revisions of macroeconomic data. This paper illustrates and proposes a simple procedure to alleviate the bias, and is based on Total Least Squares (TLS). TLS is, in general, consistent, and also works well in small samples. Simulation experiments and an empirical example show the usefulness of this method.
|Keywords||Errors-in-variables, First-order autoregression, OLS, Total Least Squares|
|JEL||Econometric Methods: Single Equation Models; Single Variables: General (jel C20), Model Construction and Estimation (jel C51)|
|Journal||Advances in Decision Sciences|
Franses, Ph.H.B.F. (2020). Measurement error in a first-order autoregression. Advances in Decision Sciences, 24(2), 1–15. Retrieved from http://hdl.handle.net/1765/130342