A massive experiment that deployed regular police patrols on platforms in the London Underground has shown that four 15-minute patrols a day in some of the capital's most crime-ridden stations reduced reported crime and disorder by 21%.
Researchers from Cambridge University's Institute of Criminology worked with the British Transport Police (BTP) to conduct the experiment across six months in 2011-2012. The findings have been published in the journal Criminology.
The team identified the 115 London stations where reported crime was highest. They randomly allocated 57 of these stations with four daily "doses" of platform patrols - two officers on foot for quarter of an hour - four days a week, and compared the effects to the remaining "untreated" stations.
The researchers found that, while the experiment was running, a total of 3,549 calls to police from the platform came from stations without patrols, compared to 2,817 in the stations receiving a policing "dosage" - a relative difference of 21%.
The team also looked at crime data from the six months prior to the experiment, and found that recorded crime fell 14% overall during the experimental period in those stations treated with the new patrols.
Strikingly, they discovered that the vast majority of reduction in both crime and calls for assistance occurred when these police patrols were absent - some 97% of the measured effect. The criminologists have dubbed this the "London Underground paradox".
"The total crime prevention benefit of police patrols may be greater when they are absent than when they are present," said study co-author Prof Lawrence Sherman. "In the London Underground experiment we see a huge residual effect of brief appearances by patrolling officers after they leave"
"This phantom effect suggests that crime declines when potential offenders are apprehensive about a possible police presence based on recent patrolling patterns - even when there are no police in the vicinity," he said.
"In London stations, it may be that more professional kinds of offenders are particularly sensitive to changes in police presence, such as pickpockets and distraction thieves."
"The London Underground paradox could have implications for debates on police priorities in an age of austerity, such as the benefits of investigating past crimes compared with the benefits of preventing future crimes," Sherman said.
London's Underground opened in 1863, the first underground railway in the world, and provides more than 1.3 billion passenger rides per year.
The majority of crime in the transport network occurs on the trains and in concourse areas. Crime on platforms constitute 11% of the total, and historically platforms have had no regular police patrols.
As such, platforms offered an opportunity to conduct an experiment on spaces within a major metropolis that had never seen proactive police presence - ideal for gauging patrol effectiveness without previous "contamination", say researchers.
"Platforms are small, stable and confined places with finite entry and exit points. These characteristics make them optimal for measuring the localised deterrence effects of police patrols," said first author Dr Barak Ariel.
"We wanted to measure what happens when police patrols are introduced into an urban environment for the first time in over 150 years."
The team targeted "hot spots" - areas where crime is more concentrated, and preventative patrols can have greatest effect - by ranking stations based on the previous year's crime rates, and including the top 115 of Greater London's 270 stations in the experiment.
Researchers also narrowed the experiment's focus based on "hot hours" and "hot days". Previous data showed the sample platforms experienced more crime and calls to police from Wednesday to Saturday between 3pm and 10pm.
Twenty uniformed BTP officers were selected and trained to work exclusively on patrolling the platforms of the "treatment" stations during "hot" days and hours. Each two-person unit was allocated between three and five stations, with platforms patrolled for fifteen minutes four times a day.
Officers were asked to conduct these patrols in a random or unpredictable order within the "hot hours", and encouraged to engage with the public while patrolling.
Police were most effective at preventing platform crime during periods and days when patrols were scheduled - but just 3% of that reduction came when officers were actually scheduled to patrol.
The researchers also found "regional" effects: crime in the rest of the station fell almost as much as crime on platforms during the four days when regular patrols were deployed.
"Our findings indicate that consistent patrols can cause large reductions in both crime and emergency calls in areas that have never before been proactively patrolled by police in this way," added Sherman.
"The more that uniformed police have been there, and the more recently, the less likely future crimes may be to occur."