1st Edition
Subjective Well-Being and Social Media
Subjective Well-Being and Social Media shows how, by exploiting the unprecedented amount of information provided by the social networking sites, it is possible to build new composite indicators of subjective well-being. These new social media indicators are complementary to official statistics and surveys, whose data are collected at very low temporary and geographical resolution.
The book also explains in full details how to solve the problem of selection bias coming from social media data. Mixing textual analysis, machine learning and time series analysis, the book also shows how to extract both the structural and the temporary components of subjective well-being.
Cross-country analysis confirms that well-being is a complex phenomenon that is governed by macroeconomic and health factors, ageing, temporary shocks and cultural and psychological aspects. As an example, the last part of the book focuses on the impact of the prolonged stress due to the COVID-19 pandemic on subjective well-being in both Japan and Italy. Through a data science approach, the results show that a consistent and persistent drop occurred throughout 2020 in the overall level of well-being in both countries.
The methodology presented in this book:
- enables social scientists and policy makers to know what people think about the quality of their own life, minimizing the bias induced by the interaction between the researcher and the observed individuals;
- being language-free, it allows for comparing the well-being perceived in different linguistic and socio-cultural contexts, disentangling differences due to objective events and life conditions from dissimilarities related to social norms or language specificities;
- provides a solution to the problem of selection bias in social media data through a systematic approach based on time-space small area estimation models.
The book comes also with replication R scripts and data.
Stefano M. Iacus is full professor of Statistics at the University of Milan, on leave at the Joint Research Centre of the European Commission. Former R-core member (1999-2017) and R Foundation Member.
Giuseppe Porro is full professor of Economic Policy at the University of Insubria.
An earlier version of this project was awarded the Italian Institute of Statistics-Google prize for "official statistics and big data".
- Subjective and Social Well-being
- Text and Sentiment Analysis
- Extracting Subjective Well-Being from Textual Data
- How to Control for Bias in Social Media
- Subjective Well-Being and the COVID- Pandemic
Introduction
Subjective Well-Being
Objective Measures
Multidimensional Indicators
Surveys
Social Networking Sites and Data at Scale
What You’ll Find (and What You’ll Not) in This Book
Wellbeing, Well Being or Well-Being?
Gross Domestic Product
Well-being as a Multidimensional Notion
The Capability Approach
Empirical Limitations of the Capability Approach
Multidimensional Well-Being Indicators
HDI: Human Development Index
BLI: Better Life Index
HPI: Happy Planet Index
BES: Benessere Equo Sostenibile (Fair Sustainable Well-Being)
CIW: Canadian Index of Well-Being
Other Initiatives for Measuring Well-Being
GNH: Gross National Happiness
Pros and Cons of Multidimensional Indicators
Self-Reported Well-Being
Gallup Surveys
Gallup World Poll
Gallup-Sharecare and Global Well-Being Index
Well-Being Research Based On Gallup Data
European Social Survey
World Values Survey
European Quality of Life Survey
How to Collect (and Interpret) Self-Reported Evaluations
Social Networking Sites and Well-Being
Sentiment Analysis
Evaluating Subjective Well-Being on the Web
Pros and cons of large-scale data from SNS
International and Intercultural Comparisons
Subjective or Social Well-Being?
Glossary
Text Analysis
Main Principles of Text Analysis
Different Types of Estimation and Targets
From Texts to Numbers: How Computers Cruch Documents
Modelling the Data Coming for Social Networks
Review of Unsupervised Methods
Scoring Methods: Wordfish, Wordscores and LLS
Continuous Space Word Representation: WordVec
Cluster Analysis
Topic Models
Review of Machine Learning Methods
Decision Trees and Random Forests
Support Vector Machines
Artificial Neural Networks
Estimation of Aggregated Distribution
The Need of Aggregated Estimation: Reversing the Point of View
The ReadMe Solution to the Inverse Problem
The iSA Algorithm
Main Advantages of iSA over the ReadMe Approach
The iSAX Algorithm for Sequential Sampling
Empirical Comparison of Machine Learning Methods
Confidence Intervals
Conclusions
Glossary
From SNS Data to Subjective Well-Being Indexes
Pros & Cons of Twitter Data
The Hedonometer
The Gross National Happiness Index
The World Well-Being Project
The Twitter Subjective Well-Being Index
Qualitative Analysis of Texts
Data Filtering for Training-set Construction
General Coding Rules
Specific Coding Rules
How to Construct the Index
The Data Collection
Some Cultural Elements of SNS Communication in Japan
Preliminary Analysis of the SWB-I & SWB-J Indexes
Cross-Country Analysis - with Structural Equation Modelling
Interpretation of the Structural Equation Model
Glossary
Representativeness and Selection Bias of Social Media
Small Area Estimation Method
Weighting Strategy
The Space-Time SAE Model with Weights
An Application to the Study of Well-Being at Work
Data and Variables
The Construction of the Weights
Official Statistics to Anchor the Model
Results of the SAE Model
A Weighted Measure of Well-Being at Work
The Estimated Measure of Well-Being at Work From the SAE Model
Comparison with Official Statistics
Conclusions
Glossary
The Year and Well-Being
The Effect of Lockdown on Gross National Happiness Index
Hedonometer and the COVID- Pandemic
The World Well-Being Project and Tracking of Symptoms During the Pandemic
The Decline of SWB-I & SWB-J During COVID-Related Studies
Data Collection of Potential Determinants of the SBW Indexes
COVID- Spread Data
Financial Data
Air Quality Data
Google Search Data
Google Mobility Data
Facebook Survey Data
Restriction Measures Data
What Impacted The Subjective Well-Being Indexes?
Preliminary Correlation Analysis
Monthly Regression Analysis
Dynamic Elastic Net Analysis
Analysis of the Italian Data
Analysis of the Japanese Data
Comparative Analysis of the Dynamic Elastic Net Results
Structural Equation Modeling
Evidence from the Structural Equation Modeling
Summary of the Results
Conclusions
Glossary
Biography
Stefano M. Iacus is full professor of Statistics at the University of Milan, on leave at the Joint Research Centre of the European Commission. Former R-core member (1999-2017) and R Foundation Member.
Giuseppe Porro is full professor of Economic Policy at the University of Insubria.
An earlier version of this project was awarded the Italian Institute of Statistics-Google prize for "official statistics and big data".
"Besides considering the problem of well-being estimation per se, the book presents a great compendium of methods helpful for students and specialists working on various projects which need getting big data from the net sources for statistical research in social studies."
-Stan Lipovetsky in Technometrics, October 2021"[...] the authors present a detailed introduction to the concept of subjective well-being, citing the work both of psychologists and economists. An account of the methods used to measure subjective well-being, and in particular those relevant to social network data is given, making this work of interest to a wide range of researchers and advanced students, including economists, psychologists, statisticians and data scientists. An exposition of the technical issues involved in text and sentiment analysis, as well as a thorough account of existing techniques and methodologies, provides the necessary background for those new to this area. ... Closely referenced and clearly written, researcher’s and advanced students in all related fields, will find this a useful, informative and eminently readable book."
- Dawn Holmes in Journal of the Royal Statistical Society, Series A, June 2022