Time-varying model averaging

Structural changes often occur in economics and finance due to preferences, technologies, institutional arrangements, policies, crises, etc. Improving forecast accuracy of economic time series with structural is a long-standing problem. Model averaging aims at providing an insurance against selecting poor model. All existing model approaches the literature are designed constant (non-time-varying) combination weights. Little attention has been paid time-varying averaging, which more realistic under changes. This paper proposes novel estimator selects optimal weights by minimizing local jackknife criterion. It shown that proposed (TVJMA) asymptotically sense achieving lowest possible squared error loss class estimators. Under set regularity assumptions, TVJMA Th-consistent. A simulation study empirical application highlight merits relative variety popular estimators selection.