An abridged version of this paper is forthcoming in Computational Statistics and Data Analysis Published online 5 Aug. 2015 DOI: 10.1016/j.csda.2015.07.003 Moment Ratio Estimation of Autoregressive/Unit Root Parameters and Autocorrelation-Consistent Standard Errors (Unabridged Version) J. Huston McCulloch Ohio State University (Emeritus) April 6, 2015 Corrections Aug. 1, 2015 ABSTRACT A Moment Ratio estimator is proposed for an AR(p) model of the errors in an OLS regression, that provides standard errors with far less median bias and confidence intervals with far better coverage than conventional alternatives. A unit root, and therefore the absence of cointegration, does not necessarily mean that a correlation between the variables is spurious. The estimator is applied to a quadratic trend model of real GDP. The rate of change of GDP growth is negative with finite standard error but is insignificant. The “output gap” has an infinite standard error and is therefore a statistical illusion. Keywords: Method of Moments, Autoregressive processes, Unit Root Processes, Regression Errors, Consistent Covariance matrix, Real GDP growth JEL codes: C13 (Estimation), C22 (Time Series Models), E17 (Macroeconomic Forecasting and Simulation) URL to PDF of latest version of paper, with color graphs: http://www.econ.ohio-state.edu/jhm/papers/MomentRatioEstimator.pdf Corresponding author: J. Huston McCulloch, 145 W 67th St, New York, NY 10023. Cell: (614) 460-1195. E-mail:
[email protected] 2 1. Introduction Serial correlation is a pervasive problem in time series models in econometrics, as well as in statistics in general. When, as is often the case, positive serial correlation is present in both the errors and the regressors, it has long been well known that the Ordinary Least Squares (OLS) estimates of the standard errors are generally too small, and hence the derived t-statistics too large.1 If the form and parameters of the error serial correlation were known, it would be elementary to compute correct standard errors for OLS regression coefficients.