Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Volatility forecasting using news sentiment

Samuelsson, Axel LU (2026) STAH11 20252
Department of Statistics
Abstract
This study investigates whether incorporating news-based sentiment into the
GARCH-MIDAS framework improves volatility forecasts for the OMXS30 in-
dex. The sentiment measure is constructed from Swedish business news articles
published in Dagens Industri (DI) between 1995 and 2024, using Latent Dirichlet
Allocation (LDA) to identify thematic topics and topic-representative articles.
The resulting sentiment measures are aggregated to a monthly frequency and in-
cluded as exogenous variables in the MIDAS component. Out-of-sample results
indicate that several sentiment-based specifications yield lower forecast errors
for total variance relative to the benchmark model, according to the Diebold-
Mariano test. However, only one... (More)
This study investigates whether incorporating news-based sentiment into the
GARCH-MIDAS framework improves volatility forecasts for the OMXS30 in-
dex. The sentiment measure is constructed from Swedish business news articles
published in Dagens Industri (DI) between 1995 and 2024, using Latent Dirichlet
Allocation (LDA) to identify thematic topics and topic-representative articles.
The resulting sentiment measures are aggregated to a monthly frequency and in-
cluded as exogenous variables in the MIDAS component. Out-of-sample results
indicate that several sentiment-based specifications yield lower forecast errors
for total variance relative to the benchmark model, according to the Diebold-
Mariano test. However, only one specification yields lower forecast errors with
respect to the long-term variance component, and none of the improvements are
statistically significant. (Less)
Please use this url to cite or link to this publication:
author
Samuelsson, Axel LU
supervisor
organization
course
STAH11 20252
year
type
M2 - Bachelor Degree
subject
keywords
GARCH-MIDAS, Latent Dirichlet Allocation, News sentiment, Volatility forecasting, OMXS30
language
English
id
9224402
date added to LUP
2026-04-17 14:45:35
date last changed
2026-04-17 14:45:35
@misc{9224402,
  abstract     = {{This study investigates whether incorporating news-based sentiment into the
GARCH-MIDAS framework improves volatility forecasts for the OMXS30 in-
dex. The sentiment measure is constructed from Swedish business news articles
published in Dagens Industri (DI) between 1995 and 2024, using Latent Dirichlet
Allocation (LDA) to identify thematic topics and topic-representative articles.
The resulting sentiment measures are aggregated to a monthly frequency and in-
cluded as exogenous variables in the MIDAS component. Out-of-sample results
indicate that several sentiment-based specifications yield lower forecast errors
for total variance relative to the benchmark model, according to the Diebold-
Mariano test. However, only one specification yields lower forecast errors with
respect to the long-term variance component, and none of the improvements are
statistically significant.}},
  author       = {{Samuelsson, Axel}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Volatility forecasting using news sentiment}},
  year         = {{2026}},
}