Summary
In a new work, we examine how media language about economic news can explain and forecast suicides in England and Wales. This topic is particularly relevant due to the significant delay in official suicide statistics. We analyse over 200,000 news articles from six major UK newspapers spanning 2001 to 2015, using sentiment analysis to assess the language used in economic commentary. We focus on daily indicators of negative emotions commonly linked to poor mental health to explain and predict daily suicide figures. Our findings reveal that highly negative economic commentaries in newspapers, especially those expressing strong fear and despair, predict increased suicide rates. This suggests that intense negative media sentiment could serve as an early indicator of worsening mental health in a population.
In an era where the media wields profound influence, a new study** I co-authored with Elisa Tosetti (University of Padua) and Giorgio Vittadini (University of Bicocca), has emphasised a connection between the language used in economic news and the incidence of suicides in England and Wales. At the heart of this research lies a quest to understand how the media’s portrayal of economic matters intersects with mental health trends and how it might even provide insights into anticipating and addressing the alarming rise in suicides. The new British government’s commitment to mental health is evident not only from its manifesto pledge but also from The King's Speech delivered a few days ago. The speech highlighted that the government is committed to enhance mental health services for young people and to ensure that mental health receives the same attention and focus as physical health. Furthermore, ministers will introduce legislation to modernize the Mental Health Act, making it suitable for the twenty-first century.
We use daily data on the total number of suicides in England and Wales from January 1, 2001, to December 31, 2015. This data, provided by the Office for National Statistics (ONS) following a data request, is not broken down by geographic area to protect the identities of individuals in regions with a small number of suicides. The ONS releases suicide data with a 2-3 year lag due to the time required for coroners' reports and the examination of circumstances leading to death. Even after this lag, the data may be updated as new information on past deaths becomes available.
To reduce intra-year seasonal patterns, we have aggregated the variable number of suicides at weekly level. Further, we have divided the total number of suicides by total population, and expressed suicide rates in 100,000 population. Figure 1 depicts the temporal evolution of weekly suicide numbers per 100,000 population. The shaded grey area represents the recession periods, as identified by the OECD recession indicator for the Euro Area. The graph clearly shows an increase in suicide rates during these recession periods, with a significant peak in 2013. Although the UK was not in recession during this time, several European countries experienced severe economic instability due to the sovereign debt crisis that followed the 2008 Great Recession. This economic turmoil likely affected both the economic situation and suicide rates in the UK through contagion in business failures. Additionally, in July 2013, the UK experienced a major heatwave with temperatures exceeding 30°C for seven consecutive days, the most significant heatwave since July 2006. This rise in temperature could be linked to the increase in suicide numbers observed during that period.
Figure 1: Temporal evolution of weekly suicide rate per 100,000 population in the England and Wales
(The shaded gray areas indicate the OECD based Recession Indicator)
A unique aspect of our study is the extraction of news articles from Dow Jones, sourced from six major UK newspapers: The Guardian, The Daily Mirror, The Evening Standard, The Observer, The Sunday Times, and The Times. We selected these newspapers because they are among the most widely read generalist national newspapers in the country. Our focus is on articles categorized by Dow Jones as discussing economic topics. This selection resulted in a total of 201,807 news articles published between January 1, 2001, and December 31, 2015.
Figure 2 shows the temporal evolution of the total number of words (expressed in thousands) in the selected articles, calculated on a weekly basis. Notably, there is a sharp increase in the number of words over time leading up to 2010, suggesting that the economic crisis resulting from the 2008 global financial collapse garnered significant and sustained media attention. After 2010, there is a reduction in the number of words extracted from the articles, followed by another sharp rise in 2015, possibly driven by discussions surrounding the 2016 Brexit Referendum.
Figure 2: Total number of words from UK economic news over time (expressed in 1,000s of words)
(The shaded gray areas indicate the OECD based Recession Indicator)
Employing a dictionary-based method, we conduct sentiment analysis with the aim of discerning the emotional content of the language used when discussing economic news. Through this method, we extract daily indicators spotlighting negative emotions commonly linked with poor mental health and depression.
Using these emotional indicators, we set out to understand the link between media language and national daily suicide figures. In our analysis, we control for various influential variables recognised by existing literature as pivotal determinants of suicide rates. Among these variables are economic growth, unemployment rates, maximum daily temperature, and episodes of extreme temperature fluctuations.
We find a significant association between highly negative comments concerning economic conditions within newspaper articles and a surge in suicide numbers. This association is particularly pronounced when the language used evoked emotions of fear and despair. The study also finds that the emotional indicator expressing despair extracted from news language is a potent forecaster of suicide figures. This finding suggests that media language carrying intense, negative sentiments might serve as an early warning system for the deterioration of a population’s mental health.
Our findings shed light on the subtle yet profound ways the media can impact society’s mental well-being. By understanding these dynamics, we open doors to proactive support and intervention.
In a world where official suicide statistics are often slow to be published, the study posits an innovative solution: monitoring shifts in media sentiment as a means of early detection. This, in turn, could equip decision-makers with a powerful tool for foreseeing trends in suicides and responding more swiftly with targeted resources for mental health care.
A limitation of this study is the absence of information on individual circumstances that may lead to suicide, such as post-traumatic stress, bereavement, separation, divorce, or bullying. Additionally, we lack details regarding socioeconomic characteristics like income and education. Such detailed information is typically not released, especially at the daily level, to protect individuals' identities. Moreover, as is often the case in analysing mortality data, some suicides might be misclassified as accidents.
By Francesco Moscone
Francesco Moscone, with a bachelor's and master's degree in economics from the University of Essex and a PhD in health economics from King's College London, is a Full Professor of Business Economics at Brunel University London and an Associate Professor of Public Economics at Ca' Foscari University of Venice. He has taught at the University of Cambridge and the University of Leicester and has been a researcher at the London School of Economics and a visiting scholar at the University of Berkeley. He has published in numerous international scientific journals in economics and medicine and has won several European, Italian, and British research grants.
*E. Tosetti, F. Moscone, G. Vittadini (2023), Going Beyond Official Statistics: the Role of Economic News in Predicting Suicides, forthcoming in Economics and Human Biology