Why an atlas of public safety
Every year the rankings of Italy’s “most dangerous” cities come out, and every year they become a headline. Then the figures on crime move into politics, where both sides of the spectrum exploit them or ignore them: those who want to prove there is an emergency and those who want to prove there isn’t. So they reach the public in pieces, selective, anecdotal, allusive: a year picked because it suits, a news story taken for a trend, a number out of context, an innuendo in place of a figure.
The atlas tries to put things in perspective, in space and in time. It takes the same official data and looks at them up close, city by city and crime by crime, so that everyone can see for themselves how things have gone where they live. As befits an atlas, its analyses are latitudinal and longitudinal: they compare the cities with one another and follow each one year by year, since 2006.
Then the view goes beyond reported crime. The deep dives are nine pages on the evergreen issues of justice and public safety in Italy, and on the unresolved questions that public debate has dragged along for years. How full are the prisons, and how many people die in them? How many go in before a verdict? How long does a trial last, and how many end under the statute of limitations? How many crimes remain without a known offender? How many of the returns ordered are actually carried out? How unsafe do we feel, and how unsafe are we? To each question the official data answer as far as they can, and the atlas also says where they stop.
Public data only
Every number comes from public sources: ISTAT for reported crimes, population, perceived risk and prisons, Eurostat for returns, the Ministry of Justice and the national Guarantor for the rights of people deprived of liberty for pre-trial custody and prison by prison, the Ministry of Justice also for trials, the statute of limitations and sentences served outside prison (its data may be used for non-commercial purposes, citing the source), and the Comune di Milano (Milano city council) for neighbourhoods. For deaths in prison, the official counts sit next to those of Ristretti Orizzonti and Antigone, which count from outside and are not official data. No estimates and no models: where a figure does not exist, as with crimes neighbourhood by neighbourhood, the atlas says so instead of making it up.
The numbers and their limits
A report is not the same as a crime committed, nor is it a conviction, and a rate calculated on residents weighs more heavily where commuters and tourists arrive every day. The charts point this out next to the numbers, and a change that could be down to chance, or that rests on fewer than 20 cases in one of the two years, stays grey: neither red nor blue.
What people report and what they fear
Alongside the reports is how unsafe people feel in the area where they live. The two don’t always go together, and looking at them side by side helps in reading the debate on public safety.
Every town, not just the provincial capitals
Just search for your town to open its profile: for the 108 provincial capitals with the municipality’s own data, for all the others with the official figures for their province, and the profile says so straight away.
Open and reproducible
The code for the page and the scripts that download and rebuild the data are public on GitHub: anyone can check the sums, redo them and report a mistake.
Jacopo Ottaviani
Data journalist and computer scientist.
His projects have been published by Der Spiegel, El País, The Guardian, Daily Maverick and Internazionale; with The Migrants’ Files he won the Data Journalism Award in 2014, and with Map Makoko the Sigma Award in 2021. He is an ICFJ Knight Fellow and trains journalists and communicators around the world.
The questions everyone asks
The answers in a few lines, with the same official data as the atlas’s charts. Tap a question to read the answer: each one leads to the section that explains it.
How to read these numbers
These are reports, not every crime committed
The data count the crimes that the police reported to the judicial authorities. Many crimes are never reported, and the propensity to report varies from crime to crime and from place to place: a car theft is almost always reported, for the insurance, domestic violence much less often. A rise in reports can mean more crime, but also more trust, or that reporting has become easier.
A report is not a conviction
A report says that an event reached the police and, through them, the judicial authorities, not that anyone is guilty of it. Reports and convictions are not counted in the same way: reports are events, counted in the year they are reported and in the comune where they happened; convictions are people, counted in the year the conviction becomes final, often years later, and not comune by comune. Along the way the offence can change name, for example from attempted murder to assault, and many cases close before any trial, for example because the offender remains unknown. A first-instance sentence is not yet a final conviction: it becomes one when it can no longer be appealed, usually after the appeal and the Court of Cassation, and until then the accused is not considered guilty. This is why no public source says how many reports end in a conviction, and the atlas does not calculate it.
The rate is calculated on residents
The rate is the number of crimes per 100,000 residents, calculated on the average population for the year, as ISTAT does. In cities that take in commuters, students and tourists every day, such as Milano, Firenze, Bologna, Rimini or Venezia, crimes also affect people who don’t live there, so the rate comes out higher than the real likelihood of being a victim for those who do.
Municipality, not province
Each city here is the area of the provincial capital’s municipality, not the province. Monza, Fermo, Andria, Barletta and Trani have data from 2011, the first year in which ISTAT publishes them as capitals of their new provinces.
If your town is not a provincial capital
ISTAT publishes reported crimes municipality by municipality only for the provincial capitals. For every other town the profile shows the official rates for the province it is in: an area that includes the town you searched for but also the provincial capital, where more crimes are usually reported. The list of municipalities is ISTAT’s official one. In Sardegna ISTAT counts crimes by the four historical provinces (Sassari, Nuoro, Cagliari, Oristano): each municipality in Sardegna is assigned to the province it belonged to between 1995 and 2005, according to the codes in ISTAT’s official list of municipalities.
Neighbourhood by neighbourhood: a prototype
The Milano profile has a section on the city’s 88 neighbourhoods (the Nuclei di identità locale, or local identity units, of the Piano di governo del territorio, the city’s land-use plan), with boundaries, residents, density, foreign residents, age, commuters and run-down buildings, as published by the city council. Crimes by neighbourhood, on the other hand, do not exist in any official source: ISTAT and the Ministry of the Interior stop at the municipality. The section is ready to take them on the day the Prefettura (the prefecture), the Questura (the police headquarters) or the city council publishes them.
How a spike is identified
For each city, crime and year we compare the rate with that of the previous three years, skipping 2020 and 2021. We mark a spike when the increase is at least 25%, there are at least 20 cases, and the difference is at least three times larger than the fluctuation you would expect by chance with that number of cases (a rate-ratio test, z ≥ 3). Sharp falls follow the same rule in reverse. On the maps and in the changes, a difference that does not pass the statistical threshold (z below 2) is shown in grey.
Two consequences to bear in mind. Steady growth, like that of online fraud, produces a spike almost every year, so a year’s count of spikes also includes long-term trends. And with the 20-case threshold, rare crimes such as homicide or loan sharking almost never show up as spikes, and only where a year has at least twenty cases.
Years to read with care
In 2020 the lockdowns made almost all street crime collapse. Since 2016 insult (ingiuria) and simple criminal damage have no longer been crimes but civil wrongs: this slightly lowers the total number of crimes and, from that year, reports of criminal damage (−12% in 2016). Rare crimes, such as homicides, should be read over several years: that is why the rankings use the 2022–2024 average.
Sources
Data downloaded from the SDMX API of esploradati.istat.it. The dataset is rebuilt with python3 scripts/build_data.py; the municipalities and provinces with python3 scripts/build_comuni.py. Not from ISTAT are returns (from the Eurostat API, with python3 scripts/build_rimpatri.py), pre-trial measures and compensation (the Ministry of Justice’s reports to Parliament, with python3 scripts/build_cautelari.py) and prison by prison (Ministry of Justice, national Guarantor, Ristretti Orizzonti and Antigone, with python3 scripts/build_istituti.py).
The photos on the deep-dive cards come from Unsplash (Unsplash licence): prisons by Anil Baki Durmus, returns by Alexander Schimmeck, perception by Mohamed Marey, unknown offenders by Alex Suprun, minors by Piotr Osmenda, pre-trial detention by Rai Singh Uriarte (the Supreme Court of Chile, in Santiago), prison by prison by Dan Stephens, trials by Wesley Tingey, outside prison by Ian An.