Editors select winner of 2021 Denis Sargan Econometrics Prize
30 Jan 2022
The editors of The Econometrics Journal decided that the 2021 Denis Sargan Econometrics Prize will be awarded to Xuewen Yu (Purdue University) for his article "Generalized forecast averaging in autoregressions with a near unit root" with Mohitosh Kejriwal in the January 2021 issue of The Econometrics Journal (https://doi.org/10.1093/ectj/utaa006).
Xuewen and Mohitosh's prize winning article develops a new approach to the important problem of forecasting a highly persistent time series. It models such a series as the sum of a deterministic trend and an autoregressive component of possibly unknown order with a root local to unity. It then takes Hansen's (2010) proposal to forecast it using a so called Mallows weighted average of ordinary least squares (OLS) estimators of the constrained (to a unit root) and unconstrained models. It notes that, in a local-to-unity setting, feasible generalized least squares (FGLS) may outperform OLS and therefore explores a FGLS version of Hansen's OLS approach. It first derives the resulting asymptotic (in-sample) mean squared error and the asymptotic one-step-ahead mean squared forecast error in the local-to-unity framework for a known autoregressive lag order. It shows that the corresponding Mallows weights are different from Hansen's OLS ones. It then extends the approach to an unknown lag order, by averaging over models that differ not only in whether a unit root restriction is imposed, but also in their autoregressive lag order. It both provides an asymptotic justification for Hansen's OLS version of this approach and develops the FGLS one. Simulations and an empirical application to forecasting U.S. macroeconomic time series demonstrate that the proposed approach offers considerable improvements over existing forecasting methods.
The 2021 Denis Sargan Econometrics Prize will be presented to Xuewen Yu at the start of Serena Ng's Sargan lecture at the 2022 Royal Economic Society Conference, which is held online from 11-13 April 2022.
The Denis Sargan Econometrics Prize was introduced in 2011 by The Econometrics Journal on behalf of the Royal Economic Society. It is awarded for the best (unsolicited) article published in The Econometrics Journal in a given year by anyone who is within five years of receiving their doctorate. An honorarium of £1000 is awarded to the winning author.
Note: The Denis Sargan Econometrics Prize may be awarded for articles coauthored with senior researchers, provided that the junior author(s) contributed at least equally. In those cases, the Prize is only given to the eligible, junior author(s).