 Ie, byddwn i amsri ei hollwch i'r pwysigol, that is from Dionys Ignatin's, Seren Farat and James Stockton, all from the University of Huzisfield. So Dionys is a Senior Lecturer at the University and a frequent UK crime data user conference goer and Seren and James are both post graduates at the University of Huzisfield. Felly, rydw i'n merthyn ni'n ddweud y ddweud. Rydw i'n credu y ffordd y ffordd bwyd, mae'r ddweud hamdag am 5 munud, wrth gwrs yw ymddi'r hyn yn dweud o'r dda. Felly y pwysoch, bobl gwnaeth a chynnydd yr hyn. On yn ystod yn fath yn ystod ein bod yn ddechrau ffyrwydol. Rydw i roedd yn gwneud eich ffyrwydol, ni'n ddechrau'r ddweud. Rydw i'n credu i'r hyn i ddweud o'r ffyrwydol? I like to move around quite a lot. OK, I'll move around less than that's fine. So the idea which we have here is kind of linked actually to the first sentence by Catherine when you said that CSEW and the reports on it are based on incidents and the little bit of prevalence. That's exactly what this actually is about. The idea that we need to move away quite far from that notion if we are to succeed while just seeing the picture how it actually is and be a bit more practical about it. Now of course concentration is not a perfect measure. We understand that just saying how many victimizations a victim has experienced doesn't mean how severe they are, doesn't mean how they perceive them, doesn't really mean anything in a perfect way. But concentration is the measure that this paper really is about. I will be doing most of the talking and Sarah and James will present one graph and look scared on a side while doing so, which is fine. We've all been there. So the metrics which we usually come across as you will know in just about any report will be your incidents, victimizations per population. So how many crimes a certain group of people has experienced and majority of reports will be about that. CSCW reports will be about that. Police reports will be anything will be largely incidents. Then occasionally you get prevalence, which is victims per population. Again, we don't know which victims. We don't know whether it's a Kelly or whether it's a Jenny. It just speaks about victims in a population. Fair enough, but to be fair, that's also a relatively rare measure as well. Now concentration is the one which seems to be missing from majority of discussion. And we know from decades of research by can peas, who's very dear. I'm sure it's in any people in this room by Maki by, you know, a lot of people that repeat victimization is very much at the core of policing initiatives and of where we go from that. So concentration is the one that we are suggesting is one missing and two really needs to stop being missing quite so much. So, why is that the case? Because high incidents or high prevalence can be interpreted in any way that you like them to be. And we know that during the 1990s, during the crime peak and subsequent drop in my PhD, ages ago was specifically on that, that there is no real way to interpret a specific crime figure. A high incidence can be good or bad. It depends how you read it. There's no real crime figure that can be only good or only bad. So let's say we do have high incidence, a lot of crimes in a population. You can read it in a good way. Police are great. Police are finding all of this crime so police are doing a peachy job or you can read it badly. They look at all the crime which now there is. It's not a good or bad. It's either. It depends how you read it. You take prevalence. A lot of people are experiencing victimization in a population. Well, that might mean one, people are now going to police, which is great. People are ready to speak out and ready to interact. Satisfaction with policemen must be great or everyone is suffering. You know, they mean very little. You can't really demonstrate success or failure with high or low incidence or higher or low prevalence because they're really, it's entirely your viewpoint how you read them. It can be good or bad. And that's where it also links to hotspots as well. I mean, there's a lot of policing practices based on hotspots and not just in UK. America is based around hotspots. Everything is based around hotspots. But hotspots ignore the notion that it's not just prevalence. It's also concentration. Hotspots are prevalence based. You know, you have a smaller area and that area has more incidence or prevalence. Usually prevalence, hence it is being policed more for whatever reasons. And yes, this doesn't account currently for the crime type or anything of the sort. But it's actually hot dots. We published a paper a while ago specifically, it was called from hotspots to hot dots. The idea that you can have an area which looks like a hotspot, but it's actually only a hotspot due to a few, very few, very, very repeat slash multiple victims. CSUW allows to analyze kind of both, but whichever one you speak about. So prevalence and incidence, which is usually used to identify hotspot because that's how it would be, can actually be exchanged for a concentration measure, which will make some hotspots actually not really hotspots. They will make them into hot dots. So let's say you do have a hotspot which has a lot of victims. Great. That's quite hard to police because there's a lot of victims. It's very resource intensive and it will be a scattered approach end of the day. If you do have hot dots instead of hotspots, very few people experiencing the bulk of crime, which we know that's the case, it has been for the last two decades, very easy to spot using statistical predictive measures and resource efficient. It's much easier to target them and they're usually quite happy being supported by the criminal justice system because they've been suffering quite a lot. So yes, as I said, it links back to the sentence by years. So a majority of reports, including the ones from CSCW, pretty much most figures are on incidence. We found one on prevalence by what we looked at. Maybe we didn't read it properly. We do apologize. We did not find any on concentration. Did we read it correctly? I see some knots there. So we did read it correctly. And that's kind of what we're getting at here. You know, this isn't one of those statistical presentations with fancy significance laws and all of that. This is merely about the way we approach these things. Also, this is not a criticism of CSCW. CSCW allows all of those analyses. You know, one of the things we had published again a few years ago was specifically about how regrettable it is that data sets like that don't get analyzed enough and certainly not in any manner that depicts the necessity because data sets like that don't exist everywhere. You look at majority of Central Europe and certainly Eastern Europe, we don't have those data sets there. You know, this is a lot of money and this is a very high quality data set and yes, it has into drawbacks. But what doesn't, the regrettable moment is that it doesn't get analyzed enough. And as I say, CSCW allows to measure these things. It has uncapped screener questions, which we use for the things we will present here today. Capped victim forums, whether it's series or not, it will have longitudinal data, which even more amazingly allows to look at concentration measures. You know, just sitting over there drinking some coffee and eating a cake, which I didn't like. We already had wonderful ideas. You know, take immigrant populations. So one of my PhD students is doing things on immigrant populations with longitudinal data like that. We can see how quickly immigrant populations become criminologically identical to native populations over the course of their stay in the UK. How interesting is that? I think that's fascinating. I'll write that paper, or maybe you will, because now you know about it. Point being, there are so many interesting things we can look at here. And again, we can link all of this further and expand the theory around it to reports and non-reports. Concentration is only as good as interactions with police that follow. If there is no interactions, then the concentration will keep growing. Reasons for non-reports, which are linked with seriousness, which are linked with emotions, angry victims, which, if somebody is being repeatedly victimized and nobody helps them, there will be either very angry or very fearful, most likely very angry over the course of it. It will link to that. And even time lags between victimizations. All of that's in the data set. All of that can be analyzed and you know, we've published a couple of papers here and there, but it feels like way more needs to be done. So, a couple of the things which we will present as little stats graphs, because we're fancy statisticians and all, a little comparison of 1982 versus 2018. We started writing it a while ago, so that's the year we used. What has happened between those two years and why they should be different if all of these things have had an impact? Well, one, crime drop. We know that 1994, 1996, massive crime peak, subsequent crime drop. These two years should be quite different. Much, much lower incidence in 2018. Hopefully better policing. It's been 40 years of data. One would hope that policing has improved whether it's based on resources or just knowledge, education of police. We are one of the universities that has professional policing courses. We are educating police a lot more academically than we used to many years ago. All of these things. We should be better. Are we better? Let's take a look at one of the relatively simple charts and allow Sarin to explain it to us and see if we are any different. Hello, everyone. I'll try not to stutter. This is my first conference. So this graph is comparing 1982 and 2018. It's looking at the probability of repeat victimisation after each victimisation. So, for example, just by looking at the graph overall, you can see that the probability is increasing no matter what. So, for example, I realise that the numbers are a little bit small. So, for 1982, after one victimisation, the probability of a second one is 42%. And then after the second one, it's 58%. And one would assume with the crime drop and all these other things that Danes has just discussed that in 2018 we would see a massive drop. But looking at the two lines between those two years, they look relatively similar. So, what are we doing to change that? For example, the numbers are slightly lower between probabilities. For example, in 2018, instead of it being 42% after the first victimisation, it's dropped 17%. But ideally, as researchers data on this, we want to see it actually decrease over the number of victimisations and the probability reducing instead of actually increasing. Yes, thank you very much. And especially at this end, at the very, very repeat victims and the very, very high concentration, the probability of getting the next one is over 2 thirds. So, essentially, it won't stop until somebody stops the repeat. So, a little point from that chart is basically, the lower the incidence, the more important that should be to us. If we take everything into account that I said on the slide before the chat, if everything is better and incidence is lower and prevalence is lower, when we know that's the case, this graph also shows that's the case, how come still the very, very repeat incidence is just as likely? How come the very, very small proportion of people who are these repeat victims, and some of the papers we published about 10 years ago, it showed that 1% of victims accounts for about a third of all victimisations, a third in 1%. Well, that sounds very, very useful to know, very easy to predict because it's 1% and they're quite different sociodemographically from the rest of the population. It feels like such an easy task to do. But anyhow, with lower incidence, this is more important because instead of looking at a lot of victims and picking the ones who are repeat from a lot of victims, now we have fewer. So, surely those fewer ones should be easier to identify. Anyhow, we got one more graph which James will present here as well. Again, it poses a very similar idea, so I hope that's it on to you. Thank you. So, as Dennis explained, this is quite similar to what we were discussing before, the observed versus expected ratio in victimisation. Now this shows the results of a Poisson test and as you can see, we're ranging the exact same data set from 1982 through to 2018. Now, a Poisson ranges from 1 to infinity for this, you are looking at either 1, 2 or 3 instances of victimisation and I should say we can't really differentiate between multiple or repeat victimisation. So, in these instances somebody could have been victimised three times in a day or three times in a year. But if you look at the graph, year on year, you can see that for the line for three victimisations, over time it holds relatively steady until about 2000, 2002 and then it starts to exponentially increase relative to either 1 or 2 victimisations. Essentially what the graph is telling you is that from 2001 onwards, people that were victimised three times started to very much exceed the expected ratio. So instances of three victimisations repeat or otherwise became much more prevalent than we initially expected. Linking back to what everybody else has been saying, crime rates are going down but repeat instances seem to be on the increase. Just to add to that, basically anything that would be at a 1, what happens is the same as what would be expected statistically. Anything that's below 1 is what's happening is less than what's expected and anything that's above 1 is the problem because something's happening more than it would be statistically expected. Now if we added the 4, 5 and 6, it would go right over there. So the very, very repeat, and yes it's a very simple kind of statistical observed versus expected kind of test, but it shows very quickly and quite severely that over time, as the incidence drops because this is based on incidence, that's how the expected values are calculated, the very high repeats, and in this case they're not even very high, it's just the three repeats, they go way above what's expected. The 4 would go here, the 5 would go here and the 6 would be nearly vertical. It's just that the chart would also look kind of messy. So the ratio severely, severely increases with the number of victimizations experienced per repeat victim and as time passes by. Currently we are at the time when there are fewer of these victims, but the concentration is way past what would be expected statistically. So a couple of self-critiques because we couldn't just be happy with ourselves, couldn't we? We might say that CSCW has changed over time and these comparisons are futile, but regardless of the changes, the stats used here, there are screener questions, there are uncapped screener questions that are being used here and majority of that has stayed at least more or less consistently the same over time. Tendency of victims to telescope reports. There's no particular reason for us to discuss that in any greater detail than what's on the slide and that can be addressed with the wonderful changes that are being proposed by CSCW anyway. And let's say we can't trust the high reports of high-rate victims. Somebody says they've been victimized 20 times, 30 times and so on and they will cluster around those numbers, but of course they will. Anything that we have much of, we will cluster around tens or fives. You know, as I say, the readers might have to say how many times you've ordered pizza unless it's one, two or three. It's probably a five, ten or 20 or 365 if you're lucky. So we don't believe that self-critique of that sort applies all too much to our thing. I only have one slide left, how am I doing with time? Doing good. Okay, I can do that in three. So the key points which we want to make, first one, we believe that the presentation of current data needs a little bit of revision. Not the data itself, we love CSCW, my whole career is based around CSCW. That didn't exist, I probably wouldn't either. Well, it exists, just a lot more miserable. So it's the presentation of data and the absence of concentration measure in particular that seems to be severely lacking. We know that repeat victimization research is very 90s and zeros and a little bit 2010s, but the problem hasn't gone away or changed if anything is simplified. Two, practitioners who have conducted research already could redo it using concentration measures. It doesn't mean that the research is wrong. It means that many of the successes, successes claimed and failures admitted to, might not be successes or failures after all. They might be nothing, they might be the opposite. Only only more research would show. Then we need to continue approaching crime reduction with concentration focus. Police doesn't have that much money and I don't think it's getting much more of it. And it's not just money, it's resources in every way. It's time, it's education, it's personnel. So we cannot focus on incidents and prevalence. A lot of prevalence means massive, massive need in investment which is not possible. Focus on concentration and a sensible one, as the fourth point says. We're not saying here that we can get rid of all repeat victimization but a sensible focus on repeat slash multiple incidents is a lot more resource efficient, a lot more likely to work. And the prediction measures are there. There are plenty of publications by the very colleagues we all know and love that allow to predict who the repeat victims are, we're super victims if you want to call them how they're called across the border. So why are we not doing that? Well, that's a question for us to wonder and for us to solve. And that was me. Thank you very much.