Econometric Issues in Macroeconomic Models With Generated Regressors

Abstract

The paper critically reviews the literature on the econometric issues raised by the use of generated regressors (GR) in empirical models. The economic rationale for the use of GR is considered, with examples being drawn from several macroeconomic examples, including New Classical Macroeconomic (NCM) models which postulate monetary neutrality. Various estimation methods are discussed for models which include 'surprise' or 'unexpected' terms and the strengths and weaknesses of each approach are investigated. Drawing upon the work of McAleer and McKenzie (1991b), situations where the typically inefficient two-step estimation (2SE) method will be efficient are highlighted. Problems of model misspecification and measurement errors are also investigated. An empirical section highlights some of the dangers of using uncorrected 2SE estimation results through a careful consideration of many recent attempts to test the NCM monetary neutrality hypothesis.

The Role of Openness in the Effect of ICT on Governance

“…According to Pagan (1984, p. 242), whereas two-step estimators are consistent and efficient, very few valid inferences can be obtained. The caution of the author is in accordance with a recent stream of literature surrounding the subject, namely: Westerlund and Urbain (2013a); Ba and Ng (2006); McKenzie and McAleer (1997) and Oxley and McAleer (1993).…”

Fighting terrorism in Africa: evidence from bundling and unbundling institutions

“…Consistent with the narrative, whereas a two-step process results in consistent and efficient estimates, not all corresponding inferences are valid. The inferential concern broadly aligns with an abundant supply of literature devoted to articulating the same concern: (Oxley and McAleer 1993;McKenzie and McAleer 1997;Ba and Ng 2006;Westerlund and Urbain 2013a).…”

The Incremental Effect of Education on Corruption: Evidence of Synergy from Lifelong Learning

“…According to the narrative, while two-step estimators are reliable for the most part, they provide few valid inferences. This concern has been confirmed in more contemporary studies, namely: Oxley and McAleer (1993), Ba and Ng (2006), McKenzie and McAleer (1997),…”