CCTV Researcher Eric Piza Interviewed

Few academic researchers study video surveillance / CCTV. One of the most notable is John Jay College of Criminal Justice, City University of New York associate professor Dr. Eric Piza who we interviewed.

Inside this interview, Dr. Piza explains:

• Has the Effectiveness of Video Surveillance Improved over Time
• Have Improved Analytics Made Video Surveillance More Practical
• Advice to "Smart"/"Safe" City Projects
• CCTV Best Practices and Policies
• CCTV Has No Inherent Deterrence Value
• Why CCTV Is Better at Preventing Auto Theft Than at Preventing Murder
• CCTV Does Not Displace Crime
• Use of CCTV to Preemptively Catch Suspects
• CCTV as Crime Resolution Instead of Crime Prevention
• Does Face Recognition Facilitate Crime Prevention
• Why There Is Little Research on the Effects of Face Recognition
• Beware of the Hidden Costs of CCTV
• Blanketing Cities in CCTV Cameras Does Not Help

This interview is part of a running series of IPVM interviews of professionals of note. For previous IPVM interview coverage, please see:

• LTS CEO Interviewed
• The Kiosk Market Pivots To Temperature Screening (Interviewed)
• Low Voltage Nation Wants to "Help You Carve Out A Fulfilling Career" Interviewed
• Motorola Solutions Speaks on Avigilon / Video Strategy
• Indonesia Security Association Chairman Interview

Has the Effectiveness of Video Surveillance Improved over Time

IPVM: Video surveillance technology has changed and improved over time. From your research, has the effectiveness of video surveillance also changed over time?

Eric Piza: Based on a recent effort that my colleagues and I did, where we conducted what is referred to as a meta-analysis of all of the existing CCTV evaluation research, we came to a couple of conclusions. One was that overall, the studies do suggest that video surveillance has a modest but significant impact on crime and that also over time, in certain contexts, in certain countries, the effect has actually increased. For example, the last time that type of study was done in 2009, the United States, all of the video surveillance studies conducted in the United States didn't find an overall benefit in the US. So essentially, the main conclusion last time this effort was done was video surveillance cameras don't reduce crime in the US. Our latest effort which was published in 2019, a decade after the prior one, found that actually surveillance cameras do have a notable effect on crime in the context of the United States. So again, that's one example to see that over time the effectiveness of cameras has changed quite a bit.

Have Improved Analytics Made Video Surveillance More Practical

IPVM: How will the development of analytics change the cost-effectiveness and value of video surveillance? Specifically, I have in mind two of your studies where a San Francisco police officer was quoted and criticized the value of the cameras because there is such a large volume of footage and there's not enough time and it's so slow, but analytics can help cut down on that. So how have the analytics changed the cost-effectiveness from your perspective?

EP: Well I think that's one area where analytics and things like machine learning and computer vision algorithms can have a big effect. If you look at the overall theme in a lot of my research, it's that what gets overlooked far too often in technology interventions is the importance of the human operator. We expect technology to fix all of our problems by itself, but it doesn't quite work like that. The technology has to be used by a human being in an effective manner to achieve our goals and in the context of this conversation to generate significant crime reductions. So essentially, a problem is that we task human operators with watching far too many cameras in this example. And it's not realistic for a human being to sufficiently, proactively monitor so much footage. So machine learning and again related software-level technologies can go a long way to help focus operator attention on the exact places and more importantly when at those exact places they should be focused.

Advice to "Smart"/"Safe" City Projects

IPVM: In some of your 2019 research, it was found that CCTV schemes operated by security personnel or by a mix of security and police generated a significant increase in crime reduction over schemes operated solely by police. What advice would you give to those involved in what the industry calls "smart city" or "safe city" projects?

EP: Well, I think just to be mindful that video surveillance strategies have different components to them. So someone has to monitor the footage, either those people or someone else has to identify incidents of concern within that footage and it has to be reported to a dispatcher who has to make decisions about whether or not police should be sent to the scene. You have officers responding to the scene who have to address whatever incident was reported and I think the tendency has been to consider all of those jobs as kind of the same thing. So, for example, if police are going to be the ones responding to the events then maybe they should be the ones watching the cameras too. And some of our research has suggested that maybe that's not the case. Maybe we leave the response aspect of video surveillance to police officers and train other personnel who may not be as expensive, frankly, especially in the American context, police officer salaries are higher than security salaries. So maybe we could have other people besides the police do the observing and leave the response and law enforcement actions to the police officers.

CCTV Best Practices and Policies

IPVM: You mentioned that the policies and also the practice guiding the use of surveillance cameras are more important than the mere presence of the cameras themselves. What would you consider to be the best practices and policies for the effective use of CCTV to effect crime reduction?

EP: I think anything that reduces opportunities for successful criminal behavior. So, I'll try not to go on too much of a tangent with this. So essentially, when we're putting up a camera, there are certain assumptions that we're making if we're saying that a camera is going to reduce crime. And the primary assumption is, if people are being "watched," they're not going to feel comfortable committing crime because they feel that they'll be apprehended. Now kind of embedded with the assumption that people don't talk about enough is that the camera is actually a real risk increase. So we're assuming that putting up a camera is actually going to deliver on those promises. It's not just that people feel like they're more at-risk, we're kind of suggesting that they are more at-risk. And if offenders call the bluff of police by continuing to commit crime in front of cameras, then the police better show that the risk is real or not much deterrence effects will take place.

In a lot of my research that I conducted in partnership with the Newark Police Department we saw a lot of examples of that. The city put up video surveillance cameras in crime hotspots and places like open-air drug markets and what we found was that in the first couple of years of the program, crime didn't really change at all outside of a very slight reduction of auto theft. All of the priority crimes the city was interested in preventing didn't go down after the cameras went in. So then we started watching the footage systematically and we spent a lot of time understanding the behavior of the people caught on camera and what we found was, despite the fact that they were in very close proximity of the cameras, a lot of offenders in the city felt pretty comfortable committing crime right in front of the cameras. And when we dug deeper into that we actually found out that that wasn't, despite what contemporary belief might be, that wasn't stupid behavior. That behavior was actually based in fact becasue given the nature by which the police department observed the cameras and given the nature by which officers were dispatched to deal with crimes observed on cameras, the real risk of punishment didn't increase all that much just because the city spent all of this money putting cameras up. That's a primary problem, when you put cameras up by themselves but don't do other things to actually increase the risk to offenders from crime in the view of cameras then you're going to have a problem.

CCTV Has No Inherent Deterrence Value

IPVM: We had looked at a short interview that you had given in Stockholm back in 2019 and you mentioned specifically the sight of a camera is often not enough to scare criminals in and of itself and to deter crime.

That said, there are academics, as your research show, and many politicians around the world, and then security professionals, of course, who are eager to sell their products and solutions, and they all seem to agree that there is an inherent deterrence value to CCTV systems, that the camera being there is a deterrent. So, what is flawed about this notion of the deterrence value?

EP: What's flawed is that people are deterred by the mere visible prescence of an intervention. And what's also flawed about that thinking is thinking that surveillance cameras by themselves can do the work of policing necessary to actually communicate the true risk increase.

To put it simply, a video surveillance camera can't stop a crime in progress. It's not like the video camera turns into the Transformer robot and goes out there and starts taking care of things itself. The police are tasked to respond and do those things. So when we put up cameras and the police don't hold up their end of the bargain by effectively monitoring the cameras or effectively responding to crimes captured on cameras, it's not going to take long for people in the community, especially potential offenders, to learn that information and connect the dots to say 'yeah there's a camera there, but I'm in these neighborhoods and I'm committing just as much crime as I used to commit and the cops aren't around any more frequently, so maybe the camera is, you know, just a camera? Maybe they are actually not seeing me commit crime on this camera or if they are seeing me, they obviously can't respond to every single infraction that is observed on the camera.' So I think this whole notion of cameras are effective, period, end of sentence, suffers from that type of fallacy in thinking. We kind of forget what we know from decades of research on criminal offending and what's important in deterrence and what makes effective situational crime prevention. We forget all of those lessons when we say just put a camera up and no-one is going to be "dumb" enough to commit crime in front it. [emphasis added]

Why CCTV Is Better at Preventing Auto Theft Than at Preventing Murder

IPVM: With respect to serious and violent crime, what are your findings of why are there limitations [with CCTV], and perhaps why would someone feel less likely to steal a car when a camera is there but not feel less likely to murder someone when a camera is there?

EP: Thas's a good question. So, it probably has something to do with the setting. As you alluded to, motor vehicle theft seems to be the crime type that's most directly impacted by video surveillance cameras. So, it may be that most motor vehichle crimes occur in some kind of well-confined environment, so like a parking lot, for example, it has definitive boundaries, it's kind of easy to see all of it at a single point in time. Even street-level motor vehicle thefts, you go home and you park a car in one place, you can't park a car in the middle of the intersection, in the sidewalk, so it's a well-defined area that may be easy to view via a CCTV camera as compared to a robbery or an assault which can happen, within reason, it can happen anywhere. You could have a fight or rob someone in an intersection or in a dark alley or around the corner. A lot of those crime types are probably a little easier hidden from the view of the camera, whereas a motor vehicle theft, given the very street-level nature of it, may be more out in the open.

CCTV Does Not Displace Crime

IPVM: I read another one of your 2018 articles that talks specifically about Newark and it was mentioned in the text that a high proportion of violent crime occurred indoors, hidden from the view of the CCTV camera. Have you found a phenomenon of the presence of cameras displacing or moving crime out of the view of the CCTV camera? So, the camera is outside and instead of the violent crime or serious crime happening within the sight of the camera, maybe it happens more commonly indoors or in other neighborhoods, as you had mentioned in Newark, the cameras were put into some very high-crime, hotspot neighborhoods or locations. Can you comment on that?

EP: In my original research in Newark, we didn't find any evidence of crime displacement. We looked at if different crime types went up after reductions, we looked at the areas immediately surrounding video surveillance camera locations and we didn't find any evidence of displacement in Newark. And also, the larger literature too, so in our systematic review out of 50 of the studies that we looked at tested for the presence of displacement and only six of those found any evidence of displacement. So, displacement, based on the published research, doesn't seem to be a tremendous concern.

Use of CCTV to Preemptively Catch Suspects

IPVM: In a December 2020 conference the Mayor of Santiago, the capital of Chile, had mentioned that in his city, his cameras include software that allow local Chilean officials to track what they deem to be 'atypical' movements [in Spanish]:

In the conference, he gave an example of a person running in a pedestrian zone and the mayor later said that, to roughly quote him, that people running in pedestrian zones are more likely to be criminals, his words, so what we're wondering is, your research has discussed crime caught by, recorded, or captured by CCTV cameras but are there any findings, is there any research that back up or lend any credence to these claims made by the mayor of Santiago that you can use CCTV not just to record and then later investigate crime that has happened but to preemptively track what some may deem as atypical or unusual behavior and then use that tracking to prevent crime. Is there any evidence behind that?

EP: Well, it depends. So, in one study that I conducted, we looked at footage of serious violent crime incidents that occurred in public places. And we watched the preceeding several minutes, up to a half hour or so. We looked at that footage too, in addition to footage of the actual crime. And what we found was, the vast majority of these events were preceeeded by incidents that police would have had legal grounds to respond to. So, for example, we had shootings that started off as small shoving matches and stuff like that. We had stabbings that started off as disputes over drug transactions. So, in that sense, yeah, the video footage, if you get enough of it, and if you're able to comb through and systematically analyze what's in there, you can learn a little bit about the situational dynamics of crime and maybe use that for future crime prevention purposes.

But something as innocuous as anyone running in public is, probably, I think that's probably stretching it a little bit. But the overall concept of using camera footage to learn a little bit about how crime is committed, I think that's a valid point.

CCTV as Crime Resolution Instead of Crime Prevention

IPVM: This question came from 2018 research that spoke specifically about body-worn video cameras, that crimes captured by CCTV are significantly more likely to be closed by an enforcement action than those crimes reported via other means, including by telephone, 9-1-1, so what we would like to ask is about the messaging, so what it seems as if, and my perception from reviewing your research is that CCTV solutions do have a good deal of efficacy as a crime resolution tool. So, what would you say or what is your opinion on shifting the public messaging that politicians and manufacturers give and discussing CCTV primarily as a crime resolution tool instead of as a crime prevention tool?

EP: That's a really good question and the research evidence unfortunately doesn't tell us as much as we need about that. The vast majority of CCTV research has looked at the crime prevention aspects of the technology. It's only been in the last five years or so where researchers have made an attempt to really look at the investigative benefits of the technology. So we don't nearly know as much as we should from a scientific perspective about exactly how the surveillance cameras can improve investigations. Of the studies I'm thinking of I think it's pretty mixed, for instance there was one study that looked at metal theft along a railroad in the UK and I think they found some benefits. There was another study conducted in Dallas that found some benefits, I think there was an Australian study that didn't find much of a benefit either way. But again, I'm not able to speak to that as much because the research community hasn't looked at it as much as it has looked at crime prevention.

Does Face Recognition Facilitate Crime Prevention

IPVM: Regarding face recognition, setting aside the ethical concerns momentarily, just from a pure crime prevention angle, what is the value of face recognition? What do you think of cities banning facial recognition?

EP: To me, I'd like to see much more research on facial recognition, because right now, what we're going off of in this public debate is pretty much just opinion. In our systematic review, for example, we looked closely at each study and none of the studies that we looked at used facial recognition alongside video surveillance cameras, so we have no scientifc evidence on whether or not that's a good idea. There's also precious little research on the actual facial recognition software itself, so we're in what I would classify as a very research-poor environment when it comes to facial recognition technology. So my hope would be that the research community picks up the slack and starts to generate some scientific evidence around this technology and that hopefully that evidence can be used to inform some of these types of public conversations. Regardless what side you come down on, whether you're pro- or anti-facial recognition, I think we just don't have enough scientific evidence to say which one of those sides is actually more correct than the other one. So, my hope would be that the research knowledge improves around this and that maybe cities who have moved to ban it can re-open a dialogue just to confirm that they've come down on the right side of this. [emphasis added]

Why There Is Little Research on the Effects of Face Recognition

IPVM: Why do you think that the research community hasn't looked into as much? Just because it's a new technology?

EP: It's a new technology, the outcome that you're interested in measuring isn't very easy to measure. For example, there's a reason why there's been so much research on the effect of CCTV on crime occurrences, because most medium to large police departments have robust data systems to measure the occurrence of crime. So measuring whether or not crime went up in the view of a camera, it's, I don't want to say it's easy to do, because these are very large data sets, they take a lot of cleaning, but it's possible. This whole notion of, is facial recognition accurate in identifying suspects, that requires an agency to be out there systematically recording the facial image of suspects in a readily-available format for researchers to engage with. And that's just not the case. That's not something that's readily available. And I'm not saying it should be, that's very sensitive information, but that also means that it's a little bit out of the realm of the research community, so it's a little difficult to measure.

That's probably also why we don't know as much about case closure. Because, again, police departments do a much better job of tracking the occurrence of crime in terms of whether or not it occurred as compared to tracking what happened to a case as it goes through the system.

IPVM: That makes sense because, especially in the US, a lot of times, closing it moves beyond the police and then goes towards the court system, the legal system, which is beyond their control.

EP: Exactly. So that would require a police data set to continuously look backwards to update their cases and, I mean a lot of police databases are just not very good to be perfectly frank, so that's actually more difficult than it should be.

Beware of the Hidden Costs of CCTV

IPVM: How might technology change the value proposition of video surveillance?

EP: What do you mean by value proposition, can you explain that?

IPVM: Does it makes sense for a city to deploy video surveillance. In your research, I think it was mentioned specifically the vast majority of cities over 250,000 people in the US deploy some form of CCTV system. In Brazil, I think there are about 5,500 municipalities and the majority of them, at this point in time, don't have a video surveillance system, but as the future goes on, as technology improves becomes cheaper that uptake will increase and it will become, at least from a monetary standpoint, more feasible for more cities and for smaller cities to deploy CCTV. On the technological point, these can become cheaper but from the crime prevention angle, in terms of preventing crime or closing cases, how can the technology change that value proposition?

EP: Good question. So my advice, and I have fewer and fewer cities to give this advice to since everyone has purchased a video surveillance system in North America, but maybe this is relevant for a country like Brazil that hasn't dived completely in yet, so my advice is for a police agency to understand the continuing, invisible cost of a system. So it's one thing to say 'we have x amount of dollars in the budget, cameras cost x amount of dollars, so we can get this many.' You have to keep in mind that that's going to require for you to hire personnel, you're either hiring new personnel to watch the cameras or you're diverting personnel from other tasks to watch cameras. You're going to have demands on public requests, so if someone makes a public information request for footage you have to pay someone to look through everything, download it onto disks, e-mail it out, so there's all these ongoing costs that aren't obvious at first. So you consider that alongside the fact that the mere presence of cameras is not going to solve your crime problem by itself, you're going to have to do some kind of proactive activity in tandem, all of that needs to be taken into consideration before considering how many cameras you can truly afford. And if after doing all of that analysis, a city determines that they can afford 40 cameras, my really hard advice would be stop at installing 40 cameras. Resist the urge to put 100 cameras up, resist the urge to cover your, to blanket your entire city in CCTV because you frankly can't afford to do so because it's not just about paying for the cameras, it's about paying for the cameras as well as all of the other things that need to be in place for video surveillance to be successful. I'm not sure if I directly answered your question but that's kind of where I'd come down on it. [emphasis added]

I don't think the fact that technology is getting cheaper necessarily means that more cities need to invest in massive CCTV systems, I still think they need to take an honest look at it.

Blanketing Cities in CCTV Cameras Does Not Help

IPVM: Have you looked at how China builds video surveillance systems and how they tend to blanket cities with cameras? Are you concerned that will emerge as a model for other countries? Do you think that might be a good thing?

EP: I haven't done any direct research on China, but I don't think the blanket-the-cities approach is the wisest way to deploy this technology. But nonetheless I think that a lot of cities, I think it's hard for cities to resist that urge because if you're a believer in video surveillance technology, if in your heart of hearts you really do think that that's a wise use of public resources, it's only human nature to say, alright, let's get as much of it as possible. Let's cover every single city block because this is such a good idea that we need as much of it as we can afford. The problem is, there's kind of a point of diminshing returns. Twenty cameras isn't necessarily twice as beneficial as 10 cameras. One hundred cameras isn't necessarily five times as beneficial as 20 cameras. At a certain point, it becomes too large for you to actually realize the benefits that you think you're getting out of it. [emphasis added]

Again, it's kind of human nature to think the opposite. So one time, I gave a talk, I was talking about the, I think this was the project that, the one where the San Francisco police officer was quoted, I was presenting the findings of that experiment that we did which essentially for a summer, we paired this subset of surveillance cameras with a proactive team of officers working in unmarked patrol units and essentially what we found was, when we did that, the cameras worked exceptionally well at reducing violent crime, disorder, and narcotics crime, in addition to the type of motor vehicle crime that CCTV has been limited to impacting. So essentially, I gave that presentation and my closing is all about the work, we tried to do it citywide, we identified a manageable subset of cameras, the police department thought long and hard about how many officers they had to dedicate to the intervention, so that's why it worked.

The next speaker, literally the next speaker who was there for my entire talk literally in the first row, local police chief, gets up there and says during his talk, as Dr. Piza's research shows, the more cameras, the better, and my dream is to cover every street corner in this city with cameras, which is literally the opposite of what I said. This guy is a proponent of video surveillance, he was excited to hear that my study found some pretty sizeable crime reduction effects that were caused by cameras, and he couldn't help himself but get excited. And now this guy is probably raising money to try and cover his entire city in cameras despite the fact the research says that's not what you should do.

More Questions? Other Research?

IPVM is interested in maintaining a dialogue with Dr. Piza and other video surveillance / CCTV researchers. What would IPVM members like to see more research or study of?