 Rwy'n gymhysgol i'r eich eistedd treinr hefyd yn ysgrifennu'n hynny, ond y ffordd rhan o'r ddweud o'r bullwch cyfrnwyr ysgrifennu ymweld. Mae'r ddweud o'r ddweud o'r drwng. Jo os yn cael y ddweud o'r O&S o'r 20 o'r sefydliad, yn ddiweddol, mae'n meddwlchio'r ddweud o'r ddweud i gyfrinwyr yn ysgrifennu, a dy命hwyrwch gweithio cyfwale i gyfwyr Fyngloedd Cymru, o bethau Gweithwyrwch. Mae'r adam yn dweud am gwybodd cyfwyr ddechrau i'r gweithwyr yn ymlaes iddyn nhw i gweithwyr cywyr, gaewch, prefiadau y ffordd, yn ysgrifennu cynghwyr, mae'r cyfwyr yn yswrthyffiydd, rydych chi'n cyfrifwyr sydd wedi'i gweithwyr yn cyfwyr cyfwyr cyth-wyr ffordd. Rydych chi fod yn gweithwyr cyfwyr yn ystod yr amser, i wnaeth daeth gwrth mathosyn mewn ymu'r chyfnidol, i ysgol arllunur o gy富 o danes gan ystod yr amser. I ran ymellafu i gŷnod yma, i hefyd, ac mae'r rhai croes i'r tyfnod hefyd, ac yn ymddangos, mae yma'r prôl ac yn ystod yr erbyn yn ymgrwneud. Mae'n gynnwys yn ochr, mae'n ddigonwys eu meddwl. Here I find myself over ten years later delivering a presentation on changes to the approach to the crime survey for England and Wales. I never really asked her why she said that. I think one of the main kinds of considerations, which is what Cath just mentioned, is the timeline of course and the maintenance of a consistent time series with a consistent methodology behind it and that is extremely important and probably I think one of the main things that she was driving at so it is with a lot of hesitation that you kind of consider changes to the approach and the methodology anyway I might say a little bit more about that as we go to, of course there have been changes to the survey and the survey instruments over time as Cap was saying I think a big one, one of the major ones was the introduction of self completion interviewing for the modules on violence, domestic violence and sexual assault and so on you know a major change and of course but that didn't affect the key estimates because the key estimates were in a different section of the question and independence of that and then there's been various changes to the sample size but it's always been a random probability sample always conducted in more or less the same way using the postal address file and so on and then we come on to the pandemic which is an interesting point in time really because we had to change the methodology and you know it was funny when the pandemic struck in the March and all face to face interviewing was suspended and there was an immediate question what's going to happen to crime over this period, you go over the main measure that we have of crime we can no longer conduct so we were all scrambling around to see what we could do, we were asked to do something in a very short period of time so it was you know I suppose the first thing you always do in a survey is struggle to look for a sample and the only thing that we could use as a sample was people who previously agreed to take part on the crime survey and we had the telephone numbers so that's effectively what we did at that point in time, we took all those people who had agreed to take part on the survey over the previous two years where we had the telephone and we thought well we'll interview those people the one issue with that sample was it wasn't a very large sample and we wanted to get some very quick snapshots so what we thought well we'll introduce a panel design so we'll interview people at three monthly intervals the main reason for going back to people wasn't that we wanted to move to a panel design necessarily but just to maintain the sample because we were going to run out of the sample very quickly anyway we did that and that was the approach we took but we changed the methodology and to some extent that's changed our thinking going forward and then of course face-to-face interviewing was introduced again in October 2022 so I just want to say a little bit about the drivers for the change, the drivers to move, the things that sit behind this transformation programme and the way in which we're transforming it the first of course is the post-pandemic field operations, Kat already mentioned a little bit about that also changing policy requirements, what are those policy requirements that sit behind a survey that kind of dictate the way through them and then of course modern technologies are also a driver so drivers for change to the face-to-face due to field operations, well the cost-of-face interview has increased substantially in recent years and particularly post-pandemic so there's a real financial consideration there about the cost of field operations face-to-face field work is considered the gold standard but the costs associated with it are exorbitant and the main part of that cost is in relation to getting the interview in the first place it's the amount of legwork that people do to go round to the houses to actually get the cooperation the actual interview itself is not where the cost lies, the costs are always associated with actually getting the person to agree to participate I think another thing that also has affected things over time is the panel of interviewers that always maintained our surveys during the pandemic of course when face-to-face interviewing was suspended and a lot of these people were older people I think they gave up being a face-to-face interviewer on surveys and I think the panel reduced significantly and I do think that has affected things since we've gone back to face-to-face interview and then of course societal factors leading to less willingness to take part, I think that of course is a driver so the response rates as Kath mentioned pre-pandemic around 69-70% and they were consistently around that, we used to consistently get around between 70-75% over the last 10 years have now dropped to 42% and I just put the labour force survey figures up there as well because you can see but it's not just the crime survey for England and Wales, it's all government surveys that seem to be affected post-pandemic in relation to it so field operations policy initiatives, well there's been a real drive for greater granularity particularly as part of the UK leveling up agenda which is really targeted policy at high crime areas so there's a need for greater detail which means a bigger sample size or a larger number of interviews on which the estimates are based and then there's various targeted policies for subpopulations and relations, things like violence against women and girls I'm not going to cover that, I'm just covering the main survey changes here and then I think one of the other things to say is that for a lot of surveys the use of administrative data has seen the changing way which people regard social surveys but I don't think that's the case for the crime survey for England and Wales I think the crime survey for England and Wales was always set up as an alternative measure as against police recorded crime and its coverage is very different, its coverage is limited and it does only cover the household population, that's very true but what it also does is it covers those crimes not reported to or recorded by the police and I think that continues to be the case so when you think of the major administrative data source you do see that it has its limitations and I think that's kind of like an important factor for the continuation of the crime survey so what is the transformation programme? Well I think there's two pillars really in relation to it one is the an annual wave formation and that's for two reasons really because of all those field operations and so on if you manage to get an interview with a person and you've put all that time and cost and effort into the person seems really strange that after the first interview you go thank you very much goodbye there's obviously a benefit to actually saying to that person actually would you take part in this survey again there's a real cost benefit in order to do that and then the other side of it is a move towards with new technologies towards multimodal surveying I think we were quite lucky in the transformation programme and when the pandemic hit we moved to a telephone survey operation so we quickly designed or translated I think probably is a better word to say the face to face survey instrument to operate over the telephone so that we realised that we could develop the survey instrument to operate both face to face and via the telephone the second part of the transformation programme is to try to see whether we can make it operate fully multimodally so that includes online and asking those survey questions not only in a kind of face to face way with an interviewer but actually seeing whether we can do that and further reduce costs by moving to some kind of multimodal or online survey the real struggle with that side of things of course is getting incident estimation it's easier to ask people whether they've been a victim of different crime types when you start to ask them how many times in the last year somebody has been a victim of each and when you think about repeat victimisation, multiple victimisations and so on there's a real complexity there and actually if you look at the crime survey instrument and the way in which it operates if you look at the face to face question you see that it relies on a real interaction between the respondent and the interviewer to actually negotiate those repeat which are repeat which are multiple incidents which are series of incidents and so on in order to do that in an online situation is very complex and that's where the real struggle is we've been struggling this for a number of years and that's where the real struggle lies in relation to that part of the transformation programme so where have we got to? well the wave one will continue to operate as a face to face survey and that's one of the real central elements to this and harps back to what I was saying about Jill Matheson that part of the transformation programme we know we can move to a panel design and we know that we can maintain the same estimates in the same way because we are not really changing the way in which the first interview takes place so following the first interview subsequent interviews take place at intervals of one year for a further two to five years we don't know how many years we would continue to panel it depends on response and various other things and there's a lot that sits behind that so I think we have to look at how long we would want the panel to run for there are advantages and disadvantages for those later waves just a little bit more on the panel design what we're doing at the moment is we're conducting wave one interviews in the home and then we're following up at wave two via the phone and we're just about to start wave three but there are other opportunities there opportunities which we haven't taken at the moment so we're concentrating on that middle line that middle section, that's what's going on but you could see how with further development follow up interviews for online self-completion modules as an approach that could be taken and also that you could actually, because it's a panel now that you could actually take some kind of ad hoc studies between waves so there's real opportunities to move to a panel design which we need to think about and we need to take forward at some point but we're just concentrating on that middle band at the moment so we went live in on the panel design yet we went live in October 2022 so September 2023 we received the first year's data from the contractor we've processed that data I will say something about the data the first six months the return to face-to-face interview in the year before the return was quite slow as you might imagine so the first six months data we were kind of suffered from response rates and various other issues with it so the way to data that follows on a year later of course that has a similar effect so we waited for a full year's worth of data before we could really start to interrogate it in terms of looking at the data in terms of bias and attrition and all those various different aspects we've just received the data then November time the work that we've done it says the modal effects and bias we've concentrated on the bias of course to begin with I think I can say that we've probably got our first iteration of a kind of a non-response bias weight at weight so we've kind of modelled the kind of all the characteristics that are associated with victimisation at weight one and put them into the waiting for the for the second wave of interviews I think these are early days on these kinds of approaches and these kind of methodologies we're looking at it we're onto a secondary vision already in relation to that the one thing about the biases though I suppose in relation to it it's not like bias at weight one the bias at weight two at least you know who the non-responders are between the two waves so you can make some kind of modelled adjustment for it you're not reliant on census data which is what we do at weight one so yeah so kind of we're starting to do some work on that and then of course there's going to be various assessment of the data before we actually start to combine these data together properly and produce an estimate which will be different from the wave one on its own it will have to be a text a different approach, different methodology working on understanding that on the multimodal survey the progression I mean we started this a long time ago really it was pre-pandemic when we did the first iteration with the contractor Cantor at the time who now called Varian so they did the first I think probably the first two rounds of work in relation to this in-house where we have O&S methodologists now looking at it there's a series of reports that have been done there's three links to three papers there we've kind of we took kind of a slightly different approach more recently re-looking it says a discovery report but we've actually re-looked at what ourselves and other countries around moving online particularly the NCBS in America and the approach that they've taken in relation to transforming the survey instrument for online capability I will touch on one thing within that and that is actually that there are also improvements to those screener questions that will make a difference to the estimates that the crime survey produces even at wave one it's an interesting thing that if you look at the screener questions on the crime survey for England and Wales the final question in the set is around threats and intimidation when you actually look at those people who say yes to threats and intimidation what you actually see is that a lot of it goes into the classification of violence without injury so what they consider actually to be a threat is actually more akin to a common assault and gets coded as such in the survey instrument when we moved to the TCSEW during the pandemic we were very concerned about harassment at the time so we tweaked the threat question just tweaked it to change the wording to include harassment and what we suddenly saw a substantial increase in the estimates of violence without injury from the crime survey because a lot more people were saying yes I've suffered harassment but actually when you actually look at what they describe it actually comes out as common assault so therefore we are actually looking at the questions and the screener questions and who gets screened into this into these things to see whether we can actually improve the estimate now again back to the Jill Matheson comment at the beginning we have to be really cautious about how we do that and how we implement it and at what stage we've not come close to even considering that in terms of parallel runs and how that actually would be integrated and would operate but all I would say to you as users is that we're very mindful of it and we know we need to give it full consideration a little bit on wave 2 response rates you can see the wave 1 response rate there hovering around the 45% and then the wave 2 response rate not brilliant somewhere around a 30% mark two different measures of it there that's both at wave 2 wave 2 response rate around 30% for those who completed a wave 1 interview and around 50% of those that are actually issued at a sample those who agreed to take part 10 minutes I must have rushed through it a little bit I'm usually quite slow but today I've gone a little bit quicker