Mixed-frequency machine learning: Nowcasting and backcasting weekly initial claims with daily internet search volume data

Daniel Borup; David E. Rapach; Erik Christian Montes Schütte

doi:10.1016/j.ijforecast.2022.05.005

Mixed-frequency machine learning: Nowcasting and backcasting weekly initial claims with daily internet search volume data

Daniel Borup, David E. Rapach^*, Erik Christian Montes Schütte

^*Corresponding author for this work

Department of Economics and Business Economics

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaper › Journal article › Research › peer-review

Abstract

We propose an out-of-sample prediction approach that combines unrestricted mixed-data sampling with machine learning (mixed-frequency machine learning, MFML). We use the MFML approach to generate a sequence of nowcasts and backcasts of weekly unemployment insurance initial claims based on a rich trove of daily Google Trends search volume data for terms related to unemployment. The predictions are based on linear models estimated via the LASSO and elastic net, nonlinear models based on artificial neural networks, and ensembles of linear and nonlinear models. Nowcasts and backcasts of weekly initial claims based on models that incorporate the information in the daily Google Trends search volume data substantially outperform those based on models that ignore the information. Predictive accuracy increases as the nowcasts and backcasts include more recent daily Google Trends data. The relevance of daily Google Trends data for predicting weekly initial claims is strongly linked to the COVID-19 crisis.

Original language	English
Journal	International Journal of Forecasting
Volume	39
Issue	3
Pages (from-to)	1122-1144
Number of pages	23
ISSN	0169-2070
DOIs	https://doi.org/10.1016/j.ijforecast.2022.05.005
Publication status	Published - Jul 2023

Keywords

Elastic net
Internet search
LASSO
Mixed-frequency data
Neural network
Unemployment insurance
Variable importance

Access to Document

10.1016/j.ijforecast.2022.05.005

Cite this

@article{4b35abd3f13148579de64489a0bdbb5c,

title = "Mixed-frequency machine learning: Nowcasting and backcasting weekly initial claims with daily internet search volume data",

abstract = "We propose an out-of-sample prediction approach that combines unrestricted mixed-data sampling with machine learning (mixed-frequency machine learning, MFML). We use the MFML approach to generate a sequence of nowcasts and backcasts of weekly unemployment insurance initial claims based on a rich trove of daily Google Trends search volume data for terms related to unemployment. The predictions are based on linear models estimated via the LASSO and elastic net, nonlinear models based on artificial neural networks, and ensembles of linear and nonlinear models. Nowcasts and backcasts of weekly initial claims based on models that incorporate the information in the daily Google Trends search volume data substantially outperform those based on models that ignore the information. Predictive accuracy increases as the nowcasts and backcasts include more recent daily Google Trends data. The relevance of daily Google Trends data for predicting weekly initial claims is strongly linked to the COVID-19 crisis.",

keywords = "Elastic net, Internet search, LASSO, Mixed-frequency data, Neural network, Unemployment insurance, Variable importance",

author = "Daniel Borup and Rapach, {David E.} and Sch{\"u}tte, {Erik Christian Montes}",

note = "Publisher Copyright: {\textcopyright} 2022 International Institute of Forecasters",

year = "2023",

month = jul,

doi = "10.1016/j.ijforecast.2022.05.005",

language = "English",

volume = "39",

pages = "1122--1144",

journal = "International Journal of Forecasting",

issn = "0169-2070",

publisher = "Elsevier BV",

number = "3",

}

Mixed-frequency machine learning: Nowcasting and backcasting weekly initial claims with daily internet search volume data. / Borup, Daniel; Rapach, David E.; Schütte, Erik Christian Montes.
In: International Journal of Forecasting, Vol. 39, No. 3, 07.2023, p. 1122-1144.

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaper › Journal article › Research › peer-review

TY - JOUR

T1 - Mixed-frequency machine learning

T2 - Nowcasting and backcasting weekly initial claims with daily internet search volume data

AU - Borup, Daniel

AU - Rapach, David E.

AU - Schütte, Erik Christian Montes

PY - 2023/7

Y1 - 2023/7

N2 - We propose an out-of-sample prediction approach that combines unrestricted mixed-data sampling with machine learning (mixed-frequency machine learning, MFML). We use the MFML approach to generate a sequence of nowcasts and backcasts of weekly unemployment insurance initial claims based on a rich trove of daily Google Trends search volume data for terms related to unemployment. The predictions are based on linear models estimated via the LASSO and elastic net, nonlinear models based on artificial neural networks, and ensembles of linear and nonlinear models. Nowcasts and backcasts of weekly initial claims based on models that incorporate the information in the daily Google Trends search volume data substantially outperform those based on models that ignore the information. Predictive accuracy increases as the nowcasts and backcasts include more recent daily Google Trends data. The relevance of daily Google Trends data for predicting weekly initial claims is strongly linked to the COVID-19 crisis.

AB - We propose an out-of-sample prediction approach that combines unrestricted mixed-data sampling with machine learning (mixed-frequency machine learning, MFML). We use the MFML approach to generate a sequence of nowcasts and backcasts of weekly unemployment insurance initial claims based on a rich trove of daily Google Trends search volume data for terms related to unemployment. The predictions are based on linear models estimated via the LASSO and elastic net, nonlinear models based on artificial neural networks, and ensembles of linear and nonlinear models. Nowcasts and backcasts of weekly initial claims based on models that incorporate the information in the daily Google Trends search volume data substantially outperform those based on models that ignore the information. Predictive accuracy increases as the nowcasts and backcasts include more recent daily Google Trends data. The relevance of daily Google Trends data for predicting weekly initial claims is strongly linked to the COVID-19 crisis.

KW - Elastic net

KW - Internet search

KW - LASSO

KW - Mixed-frequency data

KW - Neural network

KW - Unemployment insurance

KW - Variable importance

UR - http://www.scopus.com/inward/record.url?scp=85132813220&partnerID=8YFLogxK

U2 - 10.1016/j.ijforecast.2022.05.005

DO - 10.1016/j.ijforecast.2022.05.005

M3 - Journal article

AN - SCOPUS:85132813220

SN - 0169-2070

VL - 39

SP - 1122

EP - 1144

JO - International Journal of Forecasting

JF - International Journal of Forecasting

IS - 3

ER -

Mixed-frequency machine learning: Nowcasting and backcasting weekly initial claims with daily internet search volume data

Abstract

Keywords

Access to Document

Other files and links

Embargoed Document

Fingerprint

Cite this