Valentina Corradi (University of Surrey) on "Improved Tests for Forecast Comparison" (with Sainan Jin and Norman Swanson)

Valentina Corradi (University of Surrey) on
Speaker:  Valentina Corradi - University of Surrey
  Wednesday, April 5, 2017 at 12:30 PM Polo Santa Marta, Via Cantarane 24, Room 1.59
Jin, Corradi and Swanson (2016) develop a forecast evaluation methodology which is robust to the choice of the loss function and they establish a map between loss function robust forecast evaluation and stochastic dominance.
However, their tests are not uniformly valid and have correct asymptotic size only under the least favorable case under the null. Since tests for stochastic dominance can be seen as tests for infinitely many moment inequalities, we use tools from Andrews and Shi (2013, 2017) to develop test for robust forecast comparison which are uniformly asymptotically valid and asymptotically non-conservative. The (Many) Moment Inequalities literature mainly focus on the independent case. In our set-up, forecast error are typically non martingale difference sequences, because of dynamic misspecification. We establish uniform convergence (over error support) of HAC estimators and of their bootstrap counterpart. Furthermore, we extend Generalized Moment Selection tests for the presence of non-vanishing recursive and rolling estimation error. The suggested testing procedure is used to evaluate SPF, and in particular to assess whether there are subset of forecasters which produce more accurate prediction, in a loss free manner.

Programme Director
Roberto Renò

External reference
Luigi Grossi

Publication date
May 19, 2016