Advanced

Can Simple Combination Strategies Improve Forecasts Of Swedish Inflation?

Leandersson, Andreas LU (2017) NEKP01 20171
Department of Economics
Abstract
In this paper I use a procedure for selecting, combining and forecasting VAR models in order to improve the prediction of Swedish inflation. Three simple average combinations are evaluated with a number of tools. I conclude that a simple combination strategy improve the forecast ability and reduce the root mean square errors. The performance is an improvement over both a simple benchmark and a previous top performing individual model. This result are consistent for shorter horizons, while on longer horizons all models tend to normalize with more similar performance. Furthermore I compare the models forecast performance using the Diebold-Mariano test. I found that the com- bination models performance is not significantly different from the... (More)
In this paper I use a procedure for selecting, combining and forecasting VAR models in order to improve the prediction of Swedish inflation. Three simple average combinations are evaluated with a number of tools. I conclude that a simple combination strategy improve the forecast ability and reduce the root mean square errors. The performance is an improvement over both a simple benchmark and a previous top performing individual model. This result are consistent for shorter horizons, while on longer horizons all models tend to normalize with more similar performance. Furthermore I compare the models forecast performance using the Diebold-Mariano test. I found that the com- bination models performance is not significantly different from the benchmark models. The conclusion therefore is that even if the results are promising, more studies are required to fully evaluate the used procedure in order to include longer sample and for other inflation environments. (Less)
Please use this url to cite or link to this publication:
author
Leandersson, Andreas LU
supervisor
organization
course
NEKP01 20171
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Forecasting, VAR Models, Combination forecasts, Average Combination, Swedish Inflation
language
English
id
8912732
date added to LUP
2017-07-10 14:34:47
date last changed
2017-07-10 14:34:47
@misc{8912732,
  abstract     = {In this paper I use a procedure for selecting, combining and forecasting VAR models in order to improve the prediction of Swedish inflation. Three simple average combinations are evaluated with a number of tools. I conclude that a simple combination strategy improve the forecast ability and reduce the root mean square errors. The performance is an improvement over both a simple benchmark and a previous top performing individual model. This result are consistent for shorter horizons, while on longer horizons all models tend to normalize with more similar performance. Furthermore I compare the models forecast performance using the Diebold-Mariano test. I found that the com- bination models performance is not significantly different from the benchmark models. The conclusion therefore is that even if the results are promising, more studies are required to fully evaluate the used procedure in order to include longer sample and for other inflation environments.},
  author       = {Leandersson, Andreas},
  keyword      = {Forecasting,VAR Models,Combination forecasts,Average Combination,Swedish Inflation},
  language     = {eng},
  note         = {Student Paper},
  title        = {Can Simple Combination Strategies Improve Forecasts Of Swedish Inflation?},
  year         = {2017},
}