دانلود رایگان مقاله تفسیر برآورد بایاس پیش بینی

abstract

This paper resolves differences in results and interpretation between Ericsson’s (2017) and Gamber and Liebner’s (2017) assessments of forecasts of U.S. gross federal debt. As Gamber and Liebner (2017) discuss, heteroscedasticity could explain the empirical results in Ericsson (2017). However, the combined evidence in Ericsson (2017) and Gamber and Liebner (2017) supports the interpretation that these forecasts have significant timevarying biases. Both Ericsson (2017) and Gamber and Liebner (2017) advocate using impulse indicator saturation in empirical modeling. Published by Elsevier B.V. on behalf of International Institute of Forecasters.

5. Conclusions

Gamber and Liebner (2017) raise important issues concerning the interpretation of empirical results, particularly when employing impulse indicator saturation. In the discussion above, the analysis of alternative model specifications and the calculation of empirical power functions highlight consequences for IIS when the null hypothesis is incorrect. Specifically, IIS has power to detect many empirical features, including heteroscedasticity, structural breaks, outliers, and omitted variables. As a practical implication, the evidence in Ericsson (2017) and Gamber and Liebner (2017) supports the interpretation that U.S. government agencies’ forecasts of U.S. gross federal debt have time-varying biases.