Abstract
Editorial bias and censorship can be quantified studying how the occurrence of the word ‘killed’ (‘morti’ in Italian) changes over time and reported location. To this purpose, we have analyzed the complete online archives of the major US newspaper (The New York Times - NYT) and the three major Italian ones (Il Corriere della Sera - CDS, La Repubblica - REP, La Stampa - STA).
After 1960 we find a common trend of decreasing coverage given to violent events in all three Italian newspapers (NYT is more stable), opposite to the growing perceived threat of violence in Italy.
In all Italian newspapers we also find that the female/male ratio is about 30% and roughly constant over the years, with only La Repubblica showing an increase of reporting of female deaths of about 3% /year . Even accounting for the lower female casualty rate, especially in work-related accidents, this hints to the presence of some gender bias in the reporting of violent deaths.
Historically, we show evidence of censorship in Italian newspapers during WW1 and Italian Fascist regime and estimate that in the period 1923-1943 ’ 57,000 articles (75%) featuring domestic deaths were censored in Italy.
We also find that the number of casualties is often (up to 26%) artificially increased to the next multiple of 5 or 10 to emphasize the importance of the article. The only exception to this editorial practice is found in domestic articles by Italian newspapers during the Fascist regime, another effect of censorship trying to downplay domestic casualties.
Furthermore, we find that in all newspapers, the distribution Nk of the number of articles involving k persons killed is described by a power law Nk =A*k^(-γ) for 2≤k≤1E6. The value of γ decreases
in wartime and increases in peacetime and reflecting how the state of belligerence of a country is being reported.
In foreign events, editorial bias results in a break of the power law for 2≤k≤10 resulting in up to 100% articles missing in comparison to what would be expected by a pure power law distribution, which
describes the distribution of all domestic articles.
The suppression of low casualties articles grows with geographical distance from the publishing nation with a rate higher by a factor 5 in the Italian newspapers than in NYT and by a factor 2 - 4 when
considering only countries in Europe (for the Italian newspapers) or America (for NYT), sign that the geographical distance plays a strong role when reporting among countries that share common social traits.
These techniques can be be applied in a wider context, e.g. toward specific ethnic groups and contribute to quantitatively assess the freedom of press in a given country.