Lochanachit, Sirasit (2020) Estimating socioeconomic indicators using online data. PhD thesis, University of Warwick.

Preview

PDF WRAP_Theses_Lochanachit_2020.pdf - Submitted Version - Requires a PDF viewer. Download (16Mb) | Preview

Request Changes to record.

Abstract

Policymakers and businesses need a good understanding of the current state of society to make fully informed decisions. In contrast to traditional approaches to measuring human behaviour, which can be expensive, time-consuming and subject to delay, data on collective online behaviour, such as what people are searching for on Google, is available publicly, rapidly and at low cost. Studies into online behaviour may therefore be able to provide useful insights into collective human behaviour in the real world.

Here, we investigate whether online data from social media platforms, such as Instagram and Twitter, and search engine data, specifically data from Google, can help estimate key characteristics of society. In particular, we seek to infer the number of people speaking various languages across different urban areas based on publicly exchanged messages on the photo-sharing platform Instagram. We find that such data can help estimate the spatial distribution of language usage in Greater London. In a parallel analysis, we investigate whether Twitter data is similarly useful. However, our results suggest that data from Instagram is more valuable, as a higher number of posts to the service contain location data.

We also investigate whether online data can be used to help estimate economic activity. Specifically, we focus on unemployment rates in the United Kingdom and draw on data retrieved from Google Trends. Our findings reveal that Google search data can help generate quicker estimates of the current level of unemployment before official data is released. We also find that, according to some performance metrics, a variable selection technique based on an elastic net can improve model performance.

This thesis highlights the potential for inferences generated from online data to complement official statistics, for example by providing quicker estimates before official figures are released. We suggest that rapid, low-cost measurements of collective human behaviour from publicly available data may provide valuable new insights for policymakers and businesses alike.

Item Type:	Thesis or Dissertation (PhD)
Subjects:	H Social Sciences > HM Sociology H Social Sciences > HT Communities. Classes. Races Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software
Library of Congress Subject Headings (LCSH):	User-generated content, Internet users -- Psychology, Online social networks -- Psychological aspects, Language and languages -- Variation, Unemployment
Official Date:	June 2020
Dates:	Date Event June 2020 UNSPECIFIED
Institution:	University of Warwick
Theses Department:	Warwick Business School
Thesis Type:	PhD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	Moat, Suzy ; Preis, Toby
Sponsors:	Thailand. Government
Format of File:	pdf
Extent:	xv, 152 leaves : illustrations, charts
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item

Downloads