CREATES

In search of a job: Forecasting employment growth using Google Trends

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

Standard

In search of a job: Forecasting employment growth using Google Trends. / Borup, Daniel; Montes Schütte, Erik Christian.

In: Journal of Business and Economic Statistics, Vol. 40, No. 1, 2022, p. 186-200.

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

Harvard

APA

CBE

MLA

Vancouver

Borup D, Montes Schütte EC. In search of a job: Forecasting employment growth using Google Trends. Journal of Business and Economic Statistics. 2022;40(1):186-200. Epub 2020 Aug. doi: 10.1080/07350015.2020.1791133

Author

Borup, Daniel ; Montes Schütte, Erik Christian. / In search of a job: Forecasting employment growth using Google Trends. In: Journal of Business and Economic Statistics. 2022 ; Vol. 40, No. 1. pp. 186-200.

Bibtex

@article{3ce129ed587949b78af23fd50b018b67,
title = "In search of a job:: Forecasting employment growth using Google Trends",
abstract = "We show that Google search activity on relevant terms is a strong out-of-sample predictor for future employment growth in the US over the period 2004-2019 at both short and long horizons. Starting from an initial search term “jobs”, we construct a large panel of 172 variables using Google{\textquoteright}s own algorithms to find semantically related search queries. The best Google Trends model achieves an out-of-sample R2 between 29% and 62% at horizons spanning from one month to one year ahead, strongly outperforming benchmarks based on a single search query or a large set of macroeconomic, financial, and sentiment predictors. This strong predictability is due to heterogeneity in search terms and extends to industry-level and state-level employment growth using state-level specific search activity. Encompassing tests indicate that when the Google Trends panel is exploited using a non-linear model, it fully encompasses the macroeconomic forecasts and provides significant information in excess of those.",
keywords = "Elastic Net, Google Trends, Machine learning, Random forests, Targeting predictors, U.S. employment growth, ECONOMIC TIME-SERIES, TESTS, PERFORMANCE, BUSINESS, S, U, PREDICTORS, SELECTION, employment growth",
author = "Daniel Borup and {Montes Sch{\"u}tte}, {Erik Christian}",
year = "2022",
doi = "10.1080/07350015.2020.1791133",
language = "English",
volume = "40",
pages = "186--200",
journal = "Journal of Business and Economic Statistics",
issn = "0735-0015",
publisher = "Taylor & Francis Inc.",
number = "1",

}

RIS

TY - JOUR

T1 - In search of a job:

T2 - Forecasting employment growth using Google Trends

AU - Borup, Daniel

AU - Montes Schütte, Erik Christian

PY - 2022

Y1 - 2022

N2 - We show that Google search activity on relevant terms is a strong out-of-sample predictor for future employment growth in the US over the period 2004-2019 at both short and long horizons. Starting from an initial search term “jobs”, we construct a large panel of 172 variables using Google’s own algorithms to find semantically related search queries. The best Google Trends model achieves an out-of-sample R2 between 29% and 62% at horizons spanning from one month to one year ahead, strongly outperforming benchmarks based on a single search query or a large set of macroeconomic, financial, and sentiment predictors. This strong predictability is due to heterogeneity in search terms and extends to industry-level and state-level employment growth using state-level specific search activity. Encompassing tests indicate that when the Google Trends panel is exploited using a non-linear model, it fully encompasses the macroeconomic forecasts and provides significant information in excess of those.

AB - We show that Google search activity on relevant terms is a strong out-of-sample predictor for future employment growth in the US over the period 2004-2019 at both short and long horizons. Starting from an initial search term “jobs”, we construct a large panel of 172 variables using Google’s own algorithms to find semantically related search queries. The best Google Trends model achieves an out-of-sample R2 between 29% and 62% at horizons spanning from one month to one year ahead, strongly outperforming benchmarks based on a single search query or a large set of macroeconomic, financial, and sentiment predictors. This strong predictability is due to heterogeneity in search terms and extends to industry-level and state-level employment growth using state-level specific search activity. Encompassing tests indicate that when the Google Trends panel is exploited using a non-linear model, it fully encompasses the macroeconomic forecasts and provides significant information in excess of those.

KW - Elastic Net

KW - Google Trends

KW - Machine learning

KW - Random forests

KW - Targeting predictors

KW - U.S. employment growth

KW - ECONOMIC TIME-SERIES

KW - TESTS

KW - PERFORMANCE

KW - BUSINESS

KW - S

KW - U

KW - PREDICTORS

KW - SELECTION

KW - employment growth

UR - http://www.scopus.com/inward/record.url?scp=85088963535&partnerID=8YFLogxK

U2 - 10.1080/07350015.2020.1791133

DO - 10.1080/07350015.2020.1791133

M3 - Journal article

VL - 40

SP - 186

EP - 200

JO - Journal of Business and Economic Statistics

JF - Journal of Business and Economic Statistics

SN - 0735-0015

IS - 1

ER -