Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Forecasting Election Results: A Bayesian Frequentist Comparison

Erik, Oldehed LU (2019) STAH11 20182
Department of Statistics
Abstract
We present a Bayesian and frequentist comparison when forecasting elections through polls. Our focus is on studying the differences of these approaches in forecasting elections. An evaluation of the fit is performed using the odds ratio. We propose a frequentist methodology for prediction horizons three months ahead while a Bayesian methodology may be slightly more accurate for shorter prediction horizons. The contribution of this paper lies in shedding light on the importance of the prediction horizon when choosing between a Bayesian or frequentist methodology to forecasting election results.
Please use this url to cite or link to this publication:
author
Erik, Oldehed LU
supervisor
organization
course
STAH11 20182
year
type
M2 - Bachelor Degree
subject
keywords
Bayesian forecasting, frequentist forecasting, non-homogeneous hidden Markov models, autoregression, kernel smoothing
language
English
id
8972533
date added to LUP
2019-04-04 09:17:42
date last changed
2019-04-04 09:17:42
@misc{8972533,
  abstract     = {{We present a Bayesian and frequentist comparison when forecasting elections through polls. Our focus is on studying the differences of these approaches in forecasting elections. An evaluation of the fit is performed using the odds ratio. We propose a frequentist methodology for prediction horizons three months ahead while a Bayesian methodology may be slightly more accurate for shorter prediction horizons. The contribution of this paper lies in shedding light on the importance of the prediction horizon when choosing between a Bayesian or frequentist methodology to forecasting election results.}},
  author       = {{Erik, Oldehed}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Forecasting Election Results: A Bayesian Frequentist Comparison}},
  year         = {{2019}},
}