مقاله انگلیسی رایگان در مورد برآورد پیش بینی بایاس

Using impulse indicator saturation (IIS), Ericsson (2017) tests for and detects economically large and statistically highly significant time-varying biases in forecasts of U.S. gross federal debt over 1984–2012, particularly at turning points in the business cycle. Gamber and Liebner (2017) discuss Ericsson (2017), obtaining different empirical results and offering a different interpretation. The current paper resolves those differences through a re-examination of IIS.

Gamber and Liebner (2017) examine Ericsson’s (2017) choice of IIS’s significance level and interpretation of the estimated bias, concluding that the empirical basis for time-varying bias per se is weaker than claimed, and that the outliers detected by IIS could easily arise from heteroscedasticity rather than from time-varying bias. Because IIS does have power to detect heteroscedasticity, heteroscedasticity could explain the IIS results in Ericsson (2017). However, as Sections 2 and 3 below show,time-varying bias is more consistent with the combined evidence in Ericsson (2017) and Gamber and Liebner (2017). Section 4 comments further on modeling with IIS.