News Media and Suicide: Using Big Data Techniques to Assess the Long-Term Impact.

Dear editor, Suicide is a major public health issue around the world, being the second leading cause of death in young people (especially under age 24) (WHO, n.d.). Media coverage of suicide has been studied to be influential not only in a negative way—known as the Werther effect—but also (and more recently studied) as having a protective effect—called the Papageno effect (Niederkrotenthaler et al., 2010). The harmful influence of the reporting of suicide in media worsens with certain ways that content and suicide deaths are depicted (Sisask & Värnik, 2012). We have been using a network of servers to extract massive quantities of data from public newspaper web pages on the Internet using a technique called web crawling. We obtained 37,263 news articles thematically related to suicide from the main Spanish newspapers. We applied sentiment analysis (which rates texts in virtue of their negative or positive meaning) to calculate the average negative sentiment of each suicide piece of news. In addition, monthly suicide rates were obtained from the Spanish National mortality rates, available in the National Institute of Statistics (INE) (INE. Instituto Nacional de Estadística, n.d.). We crossed data, and the results can be observed in Figure 1. Two interesting observations arose: a similarity of trend between monthly rates of suicides and monthly negativity of suicide news and the lagged correlation between those variables. Figures 1(a) and 1(b) show that circa 2010 there was a valley (global minima) and a subsequent large increase in both the number of suicides per month and the negative average sentiment of suicide news. This increase can be explained by the economic recession due to the 2008 financial crisis. Figure 1(c) is a scatter plot of the average negative sentiment of suicide news in a given month (x-axis) and the number of suicides three months later (y-axis). This figure underlines the positive correlation between the two time series in Figures 1(a) and 1(c) with a 0.27 Pearson correlation. A Granger causality test confirmed that the time series are Granger causal, which indicates that higher negative sentiment of suicide news can be related to higher suicide deaths numbers after three months. This leads to demonstrate a long-term dependence that may be related to a cumulative impact of news media on this topic. To sum up, the reporting of suicide news in Spanish newspapers has an impact on suicide rates. This finding may be influenced by