societal issues as sage was doing for the scientific issues? there is a strong case for having the technical economic advice remembering economic advisers often market sensitive. me transparently available to people and, where it is for experts to challenge it, there is a case to be made. my suspicion is a case to be made. my suspicion is this is one that the treasury are fairly unlikely to want to but that is for a large number of legitimate reasons. but, you know, what you say is correct, it would, however, the one bit of what you said i would just be a little bit more cautious of is the idea that the science advice would be trumped by the economic advice. my view is political leadership take both bits of advice and then they should balance them. in a sense that is theirjob, not mine. i balance them. in a sense that is theirjob, not mine.— balance them. in a sense that is theirjob, not mine. their “ob, not mine. i have refer to that theirjob, not mine. i have refer to that direct — theirjob, not mine. i have refer to that direct the, _ theirjob, not mine. i have refer to that direct the, that _ theirjob, not mine. i have refer to that direct the, that is _ theirjob, not mine. i have refer to that direct the, that is a _ theirjob, not mine. i have refer to that direct the, that is a political. that direct the, that is a political decision for ministers. modelling. some witnesses have suggested that there was an overreliance on epidemiological modelling within sage, particularly betweenjanuary and march. sage, particularly between january and march. ~ �* , ., ., ., and march. we're “ust going to leave the code inquiry — and march. we're just going to leave the code inquiry there _ and march. we're just going to leave the code inquiry there for _ and march. we're just going to leave the code inquiry there for a - and march. we're just going to leave the code inquiry there for a moment| the code inquiry there for a moment to bring you some breaking news. we brought you a story a little earlier about four missing teenagers have been missing since sunday morning. well, we have just been missing since sunday morning. well, we havejust been been missing since sunday morning. well, we have just been told been missing since sunday morning. well, we havejust been told by been missing since sunday morning. well, we have just been told by the police to have in fact located the vehicle that the boys were travelling in. it is believed they have gone camping in snowdonia. we do have the four boys names. they were last seen in a silver ford fiesta. mountain rescue teams and the coastguard helicopter had been searching the area and north wales police have appealed for any sightings to be reported butjust to reiterate, officers say that they have now located the vehicle. but we understand no sign of the boys at the moment. there is a live page on the moment. there is a live page on the bbc news website with any updates do head there if you want any latest developments. we are going to return to the covid inquiry and he is continuing to give his evidence. a strong preference for actual data over model data where thatis actual data over model data where that is available. the problem we had and models have many uses. the problem we had early in the pandemic, in the first three months, was we were dealing with very sparse data and data that had to be integrated from lots of different areas. where the actual data didn't tell you a terribly clear story. as the numbers sadly ticked up, and there are many more cases in the uk, then the actual data was relatively easy to interpret. it was much harder earlier on. models have a separate use which is testing out scenarios and saying if you do this what is the likely effect? it's always important and again modellers will also agree with this, it is a cliche of modelling that all models are wrong but some are useful and the point about these models was that they helped to explore and test some policy options. they were not predictions of the future and i think this is where some of the problems arose. these are not meant to be predictions. they were not presented as predictions but they were often interpreted as predictions.— were often interpreted as redictions. �* ., ., predictions. i'm going to come back to the public— predictions. i'm going to come back to the public perception _ predictions. i'm going to come back to the public perception of - predictions. i'm going to come back to the public perception of models l to the public perception of models in a moment but remaining focused on this issue of the extent to which modelling was required to understand the basic data which would inform your advice to government as to what the state of emergency was, how quickly the virus was emerging and in relation to the spread of the virus, its transmission. modelling wasn't required, was it? to inform you of the infection hospitalisation rates, how many people infected would be hospitalised? how many people would die of those who were infected, or the impact the nhs was likely to be. thosejudgments rested upon actual data or short—term scenario planning, just basic standard assessments on what was likely to happen. standard assessments on what was likely to happen-— standard assessments on what was likely to happen. when we had, from late march onwards, _ likely to happen. when we had, from late march onwards, unfortunately, l likely to happen. when we had, from j late march onwards, unfortunately, a lot of people with covid in the uk, a lot of people going to hospital and a lot dying, and a bit later than that we had very good day —— make data flows and we could see where things were going i agree with you that was much better to rest on those as the principal reasons to make decisions, presenting data to ministers and the general public. that was not the situation, though, we had injanuary, february and early march. remembering the numbers at that point were extremely small and, in fact, at that point were extremely small and, infact, we at that point were extremely small and, in fact, we were not picking up very large numbers of them and we look at the decision—making, it had to be based on extrapolation of the true numbers so, for example, early on, there was a very useful analysis done by professor ferguson saying that the numbers in china must be substantially greater than the numbers being reported based on his modelling about what happened if it had left china. i’m modelling about what happened if it had left china.— had left china. i'm “ust going to ut ou had left china. i'm “ust going to put you there. _ had left china. i'm “ust going to put you there. i _ had left china. i'm just going to put you there. i am _ had left china. i'm just going to put you there. i am so - had left china. i'm just going to put you there. i am so sorry. i had left china. i'm just going to i put you there. i am so sorry. that was not however modelling exercise. you look to the number of flights coming out of wuhan, worked out for the number of people in hospital how many people therefore were likely to be infected, and worked out the infection hospitalisation rate from that. it was not a modelling exercise. that. it was not a modelling exercise-— that. it was not a modelling exercise. a, , ~ ., a, exercise. that sounds like a model to me. exercise. that sounds like a model to me- let — exercise. that sounds like a model to me- let me _ exercise. that sounds like a model to me. let me ask— exercise. that sounds like a model to me. let me ask you _ exercise. that sounds like a model to me. let me ask you this. - exercise. that sounds like a model to me. let me ask you this. it - exercise. that sounds like a model to me. let me ask you this. it is i to me. let me ask you this. it is obvious from — to me. let me ask you this. it is obvious from the _ to me. let me ask you this. it is obvious from the 28th _ to me. let me ask you this. it is obvious from the 28th of - to me. let me ask you this. it is| obvious from the 28th of january obvious from the 28th of january sage meeting, for example, that spy and were asked to advise on the actions of the united could take to slow down the spread of the outbreak. why was it thought necessary to ask modellers to be the vanguard of that response? to give advice to sage about how in practice the government should respond? modelling could never be a substitute for basic epidemiological analysis of death rates, hospitalisation rates and impact on health services. i hospitalisation rates and impact on health services.— health services. i think you're robabl health services. i think you're probably using _ health services. i think you're probably using modelling - health services. i think you're probably using modelling in l health services. i think you're probably using modelling in a j health services. i think you're - probably using modelling in a much narrower sense than i would say a lot of things you've talked about, in my view, depend on models, so, for example, how you calculate a clinical fatality rate, for example, how you calculate a clinicalfatality rate, or for example, how you calculate a clinical fatality rate, or a population level fatality rate is a modelled number, particularly early on when numbers are changing very rapidly. so i think modelling had some quite small and discreet uses and these are laid out quite nicely in several of the witness statements you've already received so i'm not going to go into them in detail so i think quite a lot of the relatively simple data are still model derived. i think that you're talking about a scenario models which actually test out the point about this as they don't propose, they test various approaches and say how these, if you did this if you did that, what, in the view of the model with a big caveat i made that models are not predictions, which are the ones that would have big impacts and which would have big impacts and which would not? you can try to do that without a model but a model will give you a lot of information you otherwise would not have. mas give you a lot of information you otherwise would not have. was this the position — otherwise would not have. was this the position that _ otherwise would not have. was this the position that a _ otherwise would not have. was this the position that a great _ otherwise would not have. was this the position that a great deal - otherwise would not have. was this the position that a great deal of - the position that a great deal of time and energy and resource was spent in february on that sort of future modelling that is to say trying to model what the various contingent outcomes would be or steps that be taken by the government, what impact measures would have, but that... of necessity relatively less time was spent focused on the actuality of the scenario faced by sage on the government which was that there was emerging data from china and from the diamond princess episode and from the basic icl and london school of hygiene and tropical medicine reports telling you what the death rates and hospitalisation rates were likely to be. rates and hospitalisation rates were likel to be. ~ u, rates and hospitalisation rates were likely to be— likely to be. welcome i mean, firstl , likely to be. welcome i mean, firstly. a _ likely to be. welcome i mean, firstly. a lot— likely to be. welcome i mean, firstly, a lot of— likely to be. welcome i mean, firstly, a lot of those - likely to be. welcome i mean, firstly, a lot of those data - likely to be. welcome i mean, firstly, a lot of those data are| firstly, a lot of those data are coming from modelling groups, just to clarify on that point. it is also important to realise there are huge strands of scientific work in parallel with the modelling work. now, the modelling work tends to get a lot of promise in one of the reasons it has lodged itself in the public mind are some of the prominent modelling groups were led by people very good at explaining it in the media to be tended to hear a lot of more of that than they did from virologists or others but actually, alongside this was a very large research and analytical effort across multiple domains. and modelling only one of those. it is an important one. tom modelling only one of those. it is an important one.— modelling only one of those. it is an important one. two final points on modelling- _ an important one. two final points on modelling. firstly, _ an important one. two final points on modelling. firstly, can - an important one. two final points on modelling. firstly, can you - on modelling. firstly, can you return to the point you made earlier about the public appreciation of modelling? there is plain evidence before the inquiry that quite inappropriate degree of alarmism was apportioned to some of the scenario forecasting modelling done by icl and professor ferguson and also by the london school of hygiene and tropical medicine. in general terms, was that alarmism and criticism justified in any way at all? i thought sir patrick did an excellent job of laying out his discomfort and my discomfort at trying to explain models in very short form, press briefings because they have to come with large numbers of caveats. and what a never to the happened with models, unfortunately, what you cannot actually argue with the number of people going into hospital, the number of people sadly dying. you can argue with the model so they tended to become a way in which both sides of a polarised debate tended to have that debate with some people saying this is all made up, it is exaggerated, this is just modelling and the modelling is exaggerated and other groups saying the model show this is terrible and bayern to be doing more so the models tended to become the focus for the debates between people who had strong opposing views, because they were more debatable, actually. also because they were not fully understood and a large number of people debating them in public were doing so because they are the position they wanted to advance and they were going to use the model to advance that position almost whatever the model showed. and i think this demonstrates that trying to use modelling outputs in public discourse has to be done with care. it doesn't mean it shouldn't be done but it should be done with great care whereas using actual data is easier. everyone can it. you can test whether it is true or not but you can then interpret it and that is why, personally, ifar preferred, and if you see my presentations of data, i tried wherever possible over need use either existing data or data with very short term projections because i think that is much more straightforward to explain. in much more straightforward to exlain. , a, , explain. in truth, is that why when there was a _ explain. in truth, is that why when there was a change _ explain. in truth, is that why when there was a change of— explain. in truth, is that why when there was a change of strategy, i explain. in truth, is that why when there was a change of strategy, as we have heard and will debate in due course, what has been called the change of strategy occurred around about the 13th of march, what drove that change of strategy was actual data about where we were on the epidemiological... that data about where we were on the epidemiological. . ._ epidemiological... at this point were going _ epidemiological... at this point were going to _ epidemiological... at this point were going to say _ epidemiological... at this point were going to say goodbye i epidemiological... at this point were going to say goodbye to l epidemiological... at this point l were going to say goodbye to our viewers on bbc two. you can carry on watching the covid inquiry if you check on the qr code. the problem we had was both where we thought we were on time and where we thought we were on time and where we thought we were in terms of the force of transmission and therefore the number of measures you need to actually get on top of things change quite significantly once actual data started to flow that was more reliable and that is kind of inevitable. data trumps models every time. that is everybody. and any model is only as strong as the date on which it is based.— model is only as strong as the date on which it is based. coming back to the criticisms _ on which it is based. coming back to the criticisms that _ on which it is based. coming back to the criticisms that were _ on which it is based. coming back to the criticisms that were made i on which it is based. coming back to the criticisms that were made in i on which it is based. coming back to the criticisms that were made in the | the criticisms that were made in the public sphere in relation to models, models numbers of deaths in essence, that may occur in the event that, for example, step is not taken or they may model a variety of mitigation that may or may not be put into place because of because of there's mitigations are put into place and the government does take steps then the number of deaths estimated will not come to pass, all right. finally, some evidence has been given to the inquiry that the modelling that was relied upon by sage failed to give sufficient weight to spontaneous changes in behaviour on the part of the population as opposed to weighing up the likely consequences of government ordained legally backed change in behaviour. what do you say to that? do you think that the issue of voluntary or spontaneous changes in behaviour was correctly understood and put in its place in the correct a sense, in the model, and i'm going to cause deep into my modelling colleagues in the way i'm going to describe this but i'm going to do it in a sense for a general audience. the model can say, for example, what would happen if you reduced interactions between households by 75% or more. that reduced interactions between households by 7596 or more. that is a straightforward _ households by 7596 or more. that is a straightforward thing, _ households by 7596 or more. that is a straightforward thing, in _ households by 7596 or more. that is a straightforward thing, in fact, - households by 7596 or more. that is a straightforward thing, in fact, that i straightforward thing, in fact, that some of the models ask that question. you can then make an assumption, which you can vary, to how far you would get by simply saying please, everybody, stay at home, and how far you can get by adding on to that and the government will insist. there is a perfectly possible to model. that is not actually a particularly difficult to do. we are doing is saying what proportions do i assign to these and it could be you got 100% adherence without any government action or it could be there is quite a big difference between the government insisting on people doing it voluntarily. think one of the problems of course we had in march, in particular, but also at other points and other pandemics is there was no way of being confident, really about what the relative contributions of those would be and by the time you would be confident, you would be several doubling times further along the path. so there wouldn't be time, and a sense, to the back and say, well, that is fine, then, we probably don't need to take more radical steps. would it be helpfulfor mejust to take more radical steps. would it be helpful for me just a to take more radical steps. would it be helpfulfor mejust a bit to take more radical steps. would it be helpful for me just a bit of a better background to this? i don't think so but— better background to this? i don't think so but thank _ better background to this? i don't think so but thank professor. it i think so but thank professor. it would appear, do hereby summarise your position fairly, the question of the weights to be given to volun