LPRC - Episode 23 – Environmental Criminology: Places and Crime ft. Grant Drawve

Episode Date: April 25, 2019

The post Episode 23 – Environmental Criminology: Places and Crime ft. Grant Drawve appeared first on Loss Prevention Research Council.

Hi everyone, welcome to Crime Science. In this podcast, we aim to explore the science of crime and the practical application of the science for loss prevention and asset protection practitioners, as well as other professionals. Co-host Dr. Reid Hayes of the Loss Prevention Research Council and Tom Nehan of ControlTech discuss a wide range of topics with industry experts, thought leaders, solution providers, and many more. On today's episode, our featured guest, Grant Drav of the University of Arkansas, discusses place and crime, environmental criminology, risk terrain modeling, and much more. On today's episode, our featured guest, Grant Drav of the University of Arkansas, discusses place and crime, environmental criminology, risk terrain modeling, and much more. We would like to thank Bosch for making this episode possible. Be a leader in loss prevention by implementing integrated solutions that enhance safety, reduce shrink, and help to improve merchandising, operations, and customer service. Bosch integrated security and communication solutions span zones one through four in the LPRC zones of influence while enriching the customer experience and delivering valuable data to help

increase retail profitability. Learn more by visiting Bosch online at boschsecurity.com. All right, well, welcome everybody again to another episode of LPRC's Crime Science. Today, I'm joined as always by my co-host Tom Meehan, Chief Strategy Officer for Control Tech, longtime LPAP practitioner, and we have a special guest today. We've got Dr. Grant Draw of the University of Arkansas. Grant is on the tenure track, assistant professor there, but most importantly, at least to me and to this podcast and our team here at the LPRC and at UF, is grants special expertise with environmental criminology and place and crime more specifically and trying to understand, you know, what goes on in a specific

place and around that place at different scales really helps tell the story and help us understand why some places just have more problems than others. But most importantly, what are the indications from what we're seeing that can help us make the people in that place safer and more secure? So with that, Grant, if I could, I'm going to start off with a couple questions. I think the first one was, you know you personally get involved in crime prevention and criminology and in research on that end? Not surprisingly, I never planned on this at all. I never planned on going to grad school, getting a PhD, or even working in research as a whole. When I started my college education, I was in construction management, architecture.

During the time of the recession, I didn't really think I had much of an outlook, so I switched over. And at the time, it was called administration of justice, so a very kind of archaic feel to it. I was always interested in criminal justice, crime occurrence, victimization, and more so specifically juvenile delinquency so nothing even with the spatial realm I had no idea that environments criminology was the thing I had talked into graduate school and my professor that Cincinnati Nick sorrow I don't think he believes me when I tell him I had no idea about the research side or even grad school until he pulled me aside. I did an internship during my

master's with the Illinois State Terrorism Intelligence Center that really got me hooked on the crime analysis side of criminal justice. With that, I took a crime mapping class by I would say my mentor and that really got me into this realm in general, Jim LeBeau. And after that, I was hooked. I mean, once I was seeing things visually through GIS looking at pattern analysis I was hooked from a data standpoint things started make sense more statistics from a spatial perspective just made more sense to me going through it at that point I viewed it as I'm detecting patterns so if you find patterns within beta at that point from a crime perspective that's

preventable if you're picking up beta, at that point, from a crime perspective, that's preventable. If you're picking up patterns over time, you can change that or hopefully change that. So that's what really got me into the crime prevention side, just in terms of if you can detect patterns, why not try to prevent that from occurring in the long term, short term? And with that, I've worked a lot with criminal justice system agencies law enforcement mostly they have mounds of data very rich data but how do you weed through that and find excuse me how do you find quality information from that how do you build and strategize initiative to lower crime that's where I've enjoyed it is helping them understand just crime analysis detecting patterns within their own jurisdiction

and how to strategize from that with that and academia there's a side of it's more of a cynical view so I'm told is I need to publish my research and I get in the bad habit of working with agencies, helping them answer a question that they have or I have with data, but I never do anything with it. So I've developed this bad habit of, you know, pose questions, answer them, but not really doing anything with it to get it out to the public. So I've been slowly working with agencies on publishing what I do with them. That's always a big part of, I guess, expanding science, what we know. But with that, it's really just kind of grown from there. I was never planning on this, kind of got hooked in this spatial realm,

and doors just continue to kind of open as time progressed. That's good stuff. And I think that that's what's always important for a lot of our listeners, especially the students that tune in, is nothing in real life is linear. It's not a straight line, right, Grant? And I know Tom would agree that there are very, very few people that we even run across that end up doing exactly what they thought and probably not even really close to what they started out thinking they might do. So it sounds like your story's pretty similar with that. And

real quick, I want to dive into, you did a lot of your training at Rutgers. And to us that are out there in the criminology area, we think a lot of Rutgers, but particularly the environmental aspect of the criminology. And again, a lot of Rutgers, but particularly the environmental aspect of the criminology. And again, I know our listeners know the environment means, yes, we'd like to protect whales and parrots and things like that and use these techniques for that. But what we're talking about is we can't shape the genomics or child raising or early childhood experience in peer groups of people around us, but we can help shape our environment online or in person to maybe influence behavior there. But can you tell me a little bit about

Rutgers, what you learned there and the focus there and things that you took away that were helpful so far and you think are going to be helpful for the community at large? Oh, yeah. I mean, I was given the opportunity to have a postdoc position at Rutgers, and I was managing, at the time, a Project Safe Neighborhood initiative with those that's aimed at reducing gang and gun violence, and that one was in Jersey City. So with that, I worked with Joel Kaplan and Les Kennedy and Paul Boxer, and with that was a place-based policing initiative. So with that, it was, for me, project experience that I had not been exposed to before, really handling community group meetings, data analysis,

working with police departments as well, all in that area, and we had a lot of fun interactions and meetings. So what we did and what I took away from kind of my time at Rutgers was understanding from much more of the research side and practitioners, so how you transform that knowledge that we as researchers have and get that to the

practitioners in the field. So to me that was a very foreign approach. I didn't know how to do that and that's why we're now we have translational criminology trying to bring together academics and researchers. I think that's a hashtag on Twitter, pracademics, trying to merge the two fields. But a big part was trying to better understand how place matters in research, and especially with diverse outcomes, crime being one of it. That's always a big part of, yes, people are committing crimes or being victimized, but with a lot of research and it continues to show, place does matter. So with that, working with Joel and Les, with Chris Trey, my line RTM especially,

that allowed me to kind of throw, I'd say, darts at the board and see what stuck, throw ideas back and forth. Yeah, you hit it pretty well that we're looking at place and crime and how critical that is. And that Rutgers, because of some of faculty and the other grad students and others that are there and the tools that were developed there, are focused and are thinking that way. Yeah. So with Rutgers, I mean, they're very well known for their environmental criminology, the history of who's been there, what they're doing now. It's been phenomenal. I mean, it's really pushed the envelope of understanding crime in place, larger spatial temporal patterns of crime. With Kill and Less, a lot of people

associate them with RTM. They are the developers of terrain modeling. For those who are not familiar with RTM, it's a spatial analytical tool that assists in understanding what features from the physical landscape, such as bars, parks, bus stops, relate to a spatial outcome. With that, oftentimes we look at crime. You can change it as long as it's a spatial outcome. You can look at how the built environment affects some type of outcome. So with that, I was drawn immediately to that from a spatial standpoint. At that time, when their first big paper came out, a lot of research has looked at hotspots.

So crime clusters, you can use crime to predict crime that's great but a lot of those research articles themselves they go into great descriptions of what's going on in those hotspots is it a hotel is the bus stop it is a bar so you're understanding what's contributing to the formation of a hotspot but that's post hoc after the fact when when RTM is taking essentially what we know from that literature and literature on just single effects of bars on crime, parks, bus stops, and looking at it at the forefront, using those as factors to develop relationships with crime occurrence.

With that, you can use that in a predictive model or a manner of we know what aspects of the built environment are now impacting occurrence of crime so with that our team continues to expand I am blessed to still work with Joel unless on quite a few projects related to RTM we have one now with some faculty members here at UA and one at IUPUI on using our team for terrorism events in the US. Can we detect spatial patterns for that? Can they be predicted based on the physical landscape and even the social demographics side of larger neighborhood effects community context across the US? So, Rutgers itself has been instrumental. The students they

have there has been great on the development from a crime and place perspective. You can see them at many top universities, programs, as academics, even place an agency. So our Rutgers, for me and my career, changed it dramatically in terms of job opportunities, training, skill set. It opened the door for many, many opportunities, much of which I am still benefiting from now. Fantastic. Tom, let me go over to you. What are some thoughts, some questions for Grant? So I think it's all very fascinating.

So I, in my previous life, actually was responsible for data for quite a few years, and we worked with a data scientist and actually the Austin Research Council to create a shortage or shrink indicator model as well as a violence indication model and so RTM was something that we we looked at but I'll you know we had a data scientist and some really smart people they're taking all of what you're talking about today I understand it because I I went through it but if you were talking to a regular retailer, someone who has an analytics team, but is trying to really start using environment into a risk model, what are some basic do's and

don'ts if you could share or principles of, you know, and I know that's a loaded question, but the starting point of, you know, what you see as the confusion for folks. Confusion in terms of what you're putting in a model or trying to understand just the how the built environment is correlated with crime? Probably both. So what, you know, what you're putting in the model and then how the environment impacts it. Oh, yeah. So for a lot of our models that we build it's oftentimes based on crime data crime that's known to the police so with that there's a

copy on not all crime is reported to police so the dark figure crime is unknown so that is always a limitation of what we do and depending on the crime types some are going to be reported more than others with that you have to understand what you're putting in a model from, you have to understand what you're putting in a model from the get-go of what you're putting in and how to translate on the back end. If you're trying to look at how the built environment and crime are related, the first place to start obviously is one work with a police department. There's a huge outpouring of online data portals

that have access to crime data. So you can see it typically these have a link to some type of map so you can already look by address you get an understanding of just place places where crime is hot when you're trying to make connections to where crime is occurring I start with students in Excel so actually has mapping capabilities I don't know if people know about, but it's a super handy tool of you can drill into map crime in Excel and really look at what's going on. So you can type in an address, zoom in to see if you have a stored location that you're trying to look at. You can see what types of crime are occurring around that. You can create a simple density map,

a hotspot map of the crime that you put in. And when you're trying to build those connections, you have to understand what else is in the environment. So a standalone store is going to be different than a store that's in a strip mall that also has fast food places around it to where there's more than one retail chain or multiple types of built environment factors. It could be multiple bus stops, fast food, large retailers, all in very close proximity. So people are coming to that area for multiple reasons. With that, crime opportunities increase just based on sheer interactions of potential victims and offenders. With opportunities arising like that, you have to understand why people are using that

space. And that's where understanding the crime data is important to understand what's being reported, but also how people are using space in general. If people are waiting at a bus stop, that's a different thing getting off at a bus stop, I want to say walking to a grocery store or to a liquor mart, liquor store, convenience store, bodega, or fast food, that type of thing. No, no, that's good. And so I thought, Grant, another question would be, you know, maybe explain if you could, you mentioned RTM, risk terrain modeling. Clearly you've worked with it. You've helped enhance and improve it. You've done it for research purposes.

You've done it to help local law enforcement get a little more focused and effective and things like that. So maybe could you, what are some of the, and this kind of touches, I think, a little bit on what Tom's talking about, but what are some of the inputs into the program and how do those help us make sense of the variance in problems that we might expect at a given place? Yeah, so with RTM, I personally stick to oftentimes measures of the built environment. So accessing data on bus stops, grocery stores, vacant properties, liquor stores, grocery stores, a multitude of think about the urban environment, so the landscape. And when you put those in, you're trying to find how those spatially correlate with say crime occurring with that you can think of your old overhead projector teaching type devices where each sheet is a different layer so how do bus stops

relate to crime how do grocery stores relate to crime how to liquor stores but now if you lay those over each other when you have bus stops that are in close proximity to grocery stores risk increase with that you have multiple factors you want to test that to see the co-location of multiple risk factors contribute to crime occurrence so that's what RTM is trying to identify is one of these factors first correlated with crime and if so when you're building your best model so what combination of factors gives you the best let's say the most variance explained how much other

is related to crime so you can build your own is it a bus stop and close parking to a liquor store that's also close to large retail chains, with that are big box retail chains. If you have those three in similar locations, the risk of crime is likely going to be increased. So with that, you know what's contributing to crime occurrence based on individual level relationships, but it's enhanced or increased when they're in close proximity to one another versus being a standalone facility itself. So, you know, let me, one thing we look at, Grant, is we look at

the main effects of things, you know, things about that place or people, and then the interaction of those things. And to try and understand, you know, I'll give you a very quick example with signage. We found that, you know, the text size doesn't matter on its own. If you have a symbol, a logo on a sign, the size of that doesn't really matter. But when you look at them both, that there's a ratio between size of the logo, let's say it's an eye looking at you, and then this says, the text says, we're watching you, that there's actually a ratio between the two. So when they interact with each other, you have a pretty profound effect. So the sign can actually affect behavior. And so because of that, we can figure out how to manipulate

the size, the color, those things about the sign, where we place it, how often we replace it, and so on to affect behavior. And I know you just kind of talked about RTM. It primarily looks at the built environment. And I was going to ask you, can you describe, you know, other things about the environment that aren't necessarily structural or built and how those interact with each other? Oh, yeah. I mean, think about any city as a whole. We oftentimes, especially if you're going somewhere new, almost a tourism effect, the idea is, or you ask a concierge or someone local, where do you want to stay away from? So you get a description there oftentimes of the seediness or what makes an area bad. But it's also just general geography.

If you think about how rivers or streams or lakes separate areas, so that dictates how people move in an environment itself, how hilly it is, the walkability factor of it. You can also think about the socio-demographics of it as well. There's tons of research that looks at the neighborhood effects on-prime, the social structure that comes from the American Census or the American Community Survey where you can get that data. So there's a combination of social factors. You can look at an actual geographic land, actual landscape and features in the built environment. So it's a combination of both edge effects from the landscape of how even say interstates are positioned.

Those can have edge effects versus the social boundaries as well as the built environment effect as well. So you have a multitude of landscape features. RTM focuses usually on the built environment. I've done some research with a colleague here at UA, Sean Thomas, that integrates RTM and more of your social disorganization research. So counting forward not only the built environment, but also the social impact of neighborhoods or the social structure of neighborhoods. Did you find sort of, okay, there may be some main effects of the built and then the structures and then the social environment and then how those two interact and clash, if you will.

Have you guys found that kind of evidence? Yeah. So we did this in Little Rock, and the first idea was due to the main effect. So the built environment mattered. It predicted crime significantly. Social structure disadvantage worked as well. And then we did a follow-up paper of some interactions. So the idea was as risk increases,

the risk of opportunity for crime increases through RTM, we thought that would be higher in neighborhoods that are more disadvantaged. What we found was through the interaction in the most disadvantaged neighborhoods, risk really didn't matter at all. It was the social structure of a neighborhood that mattered more than the actual built environment risk itself. That was a very nuanced finding for us. We're working on a couple other papers now for different cities to see if this holds. But it shows that depending on the larger social structure of a neighborhood, the built environment could have an effect just depending on the level of disadvantage.

Excellent. So those are the kind of results that not only are providing some scientific empirical evidence, but that others in us that didn't conduct the research can start to build on to shape and see what other evidence we can add and how we might use that in the different types of places and spaces that we're working on. Tom, let me go over to you. Further thoughts and questions. I think I have a lot of questions that are probably more opinion. And I know that we're always talking about science.

But how can I pose this without it coming off negative? We're in a world where the words like prescriptive analytics and data modeling and artificial intelligence and machine learning have transitioned from what before were really seen as academic or mathematic terms into marketing terms and this almost overuse of data is happening in your world of being a scientist and a researcher, what are some things that you can, you know, some advice you could give? And this is probably more opinion based so that you know you don't get lost in

that storm of data. Well that's great. I mean I, I use Twitter for academic purposes and especially in the past couple months the idea of just artificial intelligence and the bias within it is a huge topic right now but if you think about the area we're in data is everywhere I mean just access to data data collection your phone knows more about you than you know yourself for the most part there's a quite a bit of I would say conversation that needs to occur around what does the data actually capture so what do you have access to and what can you do with it at the end of the day an AI program a

statistical model the most advanced application you want to use with it still has to be interpreted and used by humans. So there's still, at the end of the day, a human element to it. So a big part of predictive analytics with crime is, so we're predicting where crime is going to occur, potentially. At the end of the day, if it's policing that's going to be used, police are going to go to an area based on an analytical program. With that, what type of policing is being used? For the longest time, I've heard multiple officers give the idea of, we know where crime is occurring. It's been occurring there 10, 15, 20 years.

So using the same methodologies to try to lower crime, and oftentimes the mentality is just through arrest. So through arrest, we're going to lower crime. But if it's still occurring in the same places, that hasn't been working well. You're not changing the environment within that. The behaviors, the norms, the attractiveness of those places remain beyond just the arrest. So with that, analytics can give you great information. But how do humans use that? I think that's where the conversation has really started to pick up of no matter what type of data you have, no data is pure. There's going to be bias in how it's collected, especially if you look at criminal justice and crime-related data. There's bias within it. There's a number of podcasts, documentaries, research articles on

bias within data. We have to be cognizant of that. Some fascinating things I've seen with a movement towards using predictive analytics from a social science perspective, especially with crime or criminal justice law enforcement agencies, is developing almost a citizen-resident oversight committee or to review how police departments are using this to where it's not infringing on American residential rights, constitutional rights themselves. So we have this big idea that this algorithm is going to save the world. It's going to be the next best thing. But we don't know how to use that from a human perspective. Yes, it identifies where a relationship, but what are we going to do about it? That's where I think it's almost a data overload.

At this point, we have too much data, not enough value in it. And that's where it comes down to, theoretically, does it make sense to have this in a model that you're trying to test? What does this give you? What does this provide you? If you're throwing data in just because you have it, that's going to be hard to interpret. You don't know what you're doing with it at the back end so say is relationship relationship comes back as significant you don't know why you put the data in in the first place how are you supposed to interpret it on the back end so what

data you are using and accessing needs to have some type of thought process obviously a theoretical understanding of how the two should be related helps. But if you're throwing everything in but the kitchen sink, that's going to be problematic from any type of predictive analytics, even just simple crime analysis data analytics. You need to understand the data before you even put it in a model. So that's helpful, Grant. I think, like you say, we've got to weigh a lot of things. And I know as scientists, we're basically on a mission here. And I know, particularly with our teams at UF and at the LPRC, that we're here to help protect vulnerable people. vulnerable people. And, you know, and then we go from there. And so we are always concerned about public opinion, about any kind of legal or litigation type concerns. But at the end of the day, we're trying to help very vulnerable people be safer from victimizers and, you know, from

becoming victims. So, you know, these are incredible topics, but I really am interested in, like you said, it's not like we have a lack of data now and not even, you know, pretty good data, but everything has some bias to it. Everything has a lot of unintentional error in other ways as well. And so how do we help narrow down? But I know with a father and grandfather that are physicians, it's the same issue and it always will be. We've got all this incredible imaging data and all kinds of measurements data and everything going on, but it's fraught with error and there is inherent bias in what goes in and then how we interpret what we're seeing and on and on.

But the real world is here and the real world relatively can be dangerous. And so we've got to do something and we've got to figure it out together. So we are very, very appreciative of you and your colleagues and others, particularly in the social sciences and in particular,

you know, environmental criminology saying, look, we're going to help develop but continue to dial in the tools that the practitioner needs because, you know, there's real theft, fraud, and violence going on out there. And in our ecosystem of 120 plus retail chains that are working together, you know, with us, they're seeing it all day, every day in their parking lots and in their store places as their people travel. They're the people that are bringing them their goods as they deliver goods to people's homes and things like that. So a lot going on. I guess one other question I had was, what are some other complementary

place in crime analytical tools that you've used or you're familiar with? You've talked about RTM, risk-trained modeling. What are some other that are out there that you're familiar with? So I've used quite a few over the years. Even when I teach, even when I work with departments, I usually view it as what is easily accessible, little to no cost. I've mentioned it before, but Microsoft Excel does so much now in terms of data analysis. They have an extension for a data analysis tool pack where you can do a lot of crime analysis, data analysis within Microsoft Excel. That is just fascinating. If you want to get into more advanced statistical programs and techniques, you have R, which is free, but it's a steep learning curve. You have SAS data. I was trained during my master's and PhD

on SPSS. I'm now delving deep into R, which is kind of scary to think about but it's a good one out there another free one that's through the National Institute of Justice is crime stat I've used that for years it has numerous spatial temporal analytical tool a heck of a manual to go through tons of examples I use geota quite a bit and then as well as the near-repeat calculator. Those are kind of my go-to software programs, I would say, I use frequently. And before I let you go on that, that's good insight. You mentioned near-repeats. Could you kind of quickly reinforce with our audience what are repeats, near-repeats, phantoms, and so forth,

Grant? Yeah. So the near-repe repeat idea is repeat victimization or repeat occurrence. So a near repeat could be within a close time and space window. So how is it the same address, so truly repeat victimization in a couple days, a week, two weeks, depending on what type of incident you're looking at. So some might be more valuable. What's the likelihood the likelihood of revictimization within a week say for burglary if it heightened for just that residence or is it at the neighborhood effect the larger area around it so the near repeat idea is if something has occurred the likelihood of it occurring in a small time window and in a smaller

scale spatially so So that's what it's really enforcing is, is there repeat a contagion effect to it? Once one occurs, are there multiple that should be suspected afterwards or expected? Perfect. And there you go. We want to dig in and understand why did they hit the first place? Why they not hit other places nearby that seem to be exposed to similar or the same conditions, even the same offender base or social and built environmental factors. So that's interesting. So let me go back over to Tom. Tom, any other questions, comments, suggestions here? No, just for the listener, I think one thing that you really hit is that there are some free tools out there and there's somewhere to start.

So for Excel, if you're thinking of Excel and there was a retailer that wanted to really do some basics, is the data on this tool set the best way to do? Should they be doing clustering? Where's a good start for someone that's going to go to Google? Because the reality is our audience and listening is so vast there are a lot of folks that are going to listen to this and literally go what can I do today you know so if you're saying starting fresh because they're going to Google how to do it what in Excel is is the best way to start if you're if you're thinking of using relationship or mapping to let's just say a risk model of any sort for a retailer, whether it's a

violence risk model or a shrink prediction model, understanding that they're going to get the data, they're going to have to put that. But what's the best kind of starting point to play with, if you will? I think the best starting point is Excel. You know your store locations, and if you're in an area that has publicly accessible crime data, you can just put that in Excel, and it will give you the option to filter or select certain crime types. You can drill into your store locations. If you have multiple ones, you can move around.

It's through they integrate being matched since it is a Microsoft product. Starting there is phenomenal. It gives you the option to make a density map, so the clusters of of it so you can look at how dense crime is in certain areas, what's it look like around different store locations, different streets themselves, or even where crimes not occurring. What's different about those environments that creates a cold spot to where crime is not occurring in certain places but is in others. I do that frequently with especially with departments or agencies that have no analytical team within house I start Excel because typically people know or uncomfortable with Excel when you start

talking about pay for programs or something where they have to go out and learn something brand new there are so many tutorial videos on Excel it is so helpful I still watch them quite a bit even though I'm slowly making some crime specific ones for my students Excel is the go-to one for pivot tables even to show relationships just on time of day and day of week of occurrence the temporal heat map it's a great tool just to start there to understand relationships with crime occurrence from an analytical even crime at play standpoint. Great. Thank you very much. I think that's, you probably will have a bunch of people

who are Googling it right now and taking a look at seeing and throwing their stuff in there, because you'll find that most of the folks that are listening to this from the retail world already have that spreadsheet and are already doing some sort of regression or relationship have that spreadsheet and are already doing some sort of regression or relationship analysis and really looking to get to that next step. So I think that's it for me, Reid. I want to thank you, Grant, and from everybody that's working with us on the anti-violence innovation chain. We want to also thank you, Grant, for all that you're doing to help us on the data and mapping component of that to help us more visually, graphically understand dynamics and understand what happens, what are the effects when we put in some anti-crime measures in certain test locations,

what happens there and nearby. So that's a big thank you for all of us and feel part of the team. And we appreciate your participation today. Thank you both for the time. And if there's any questions, always feel free to reach out. Absolutely. Have a great day. Thanks for listening to the Crime Science Podcast presented by the Loss Prevention Research Council and sponsored by Bosch Security. If you enjoyed today's episode, you can find more crime science episodes and valuable information at lpresearch.org. The content provided in the Crime Science Podcast is for informational purposes only and is not a substitute for legal, financial, or other advice. Views expressed by guests of the Crime Science Podcast are those of the authors and do not reflect the opinions or positions of the Loss

Prevention Research Council.