Activity Number:
|
573
- Time Series:Testing and Forecasting
|
Type:
|
Topic Contributed
|
Date/Time:
|
Thursday, August 6, 2020 : 3:00 PM to 4:50 PM
|
Sponsor:
|
Business and Economic Statistics Section
|
Abstract #313065
|
|
Title:
|
Testing for Differences in Path Forecast Accuracy
|
Author(s):
|
Andrew Martinez*
|
Companies:
|
U.S. Department of the Treasury
|
Keywords:
|
joint density;
log determinant;
loss function;
mean square error
|
Abstract:
|
Path forecasts help guide many important policy decisions from monetary policy to weather responses. However, analyses of forecast accuracy often focus on one horizon at a time and ignore the dynamics along the path. I propose two tests of path forecast accuracy which specify the joint density as a general loss function to capture differences in forecast dynamics. Simulations highlight the benefits and trade-offs of the proposed tests over existing alternatives. I apply the path tests in two cases: (1) evaluating macroeconomic path forecasts; and (2) the accuracy of Hurricane Irma’s predicted path in 2017. The results illustrate that differences in the forecast-error dynamics drive overall differences in path forecast accuracy.
|
Authors who are presenting talks have a * after their name.