 This is about analysis of well-being. During the first stages of the COVID-19 pandemic, using an understanding society and my colleague Janet and Murthy and I were responsible for this video. So the overall purpose of our study was to investigate the trajectory of well-being in the population over the course of the first wave of the pandemic. And we were also interested in sub-segments of that, whether change in well-being was distributed equally across the population. So there are obvious things that might impact from previous work and sort of theory. Medical conditions, social isolation, financial stress deprivation, these were all predictors for changes in well-being. And we're also further than that interested whether there were community-level characteristics which might protect against decline. So social capital of some form or other being the obvious predictor. There's plenty of work done prior to us publishing our own work, psychological effects of the pandemic, lockdown and contracting COVID. All are kind of relevant to thinking about what might actually be causing changes in well-being. So there was Deheri et al's 21 paper, Psychological Impact Observed, across 18 different countries. So there's an effect that's pan-cultural. And Google trends indicated an increase in terms like loneliness, worry, sadness during the initial stages. That was a paper by Brejure et al. So there was already a precursor in work that was published very quickly into the pandemic. It's really just outlining that. So what do we do? We use the first four waves of the understanding society COVID-19 survey and used wave nine of the main understanding societies as the baseline. So that was where people were before the pandemic hit. And then we're observing changes relatively to that baseline. Our outcome variable was the GHQ caseness score, very commonly used outcome variable for well-being. And then we're looking at the impact of the onset of the pandemic by comparing wave nine of the main survey with wave one of the COVID study. And then looking at the trajectory of changing well-being up to the first relaxation of restrictions, which was effectively in wave four. So we run two different sets of simple OLS regression models. One for the initial response and a separate one for the decline. So we used longitudinal weights inevitably for these particular models. So these are the explanatory variables we thought on the basis of previous work and theoretical concerns that might be relevant to predicting whether an individual had a change in well-being. So deprivation and we have only got the decile by LSOA. So it's a bit coarse-grained and I'll talk about that later. And then community cohesion variable which is one of the survey questions. So that's really looking at that social capital question. And then loneliness as a predictor. So we have loneliness before the start of the the pandemic and how that itself changes during the course of the pandemic. So loneliness obviously very tied in with ideas of well-being and then the baseline well-being. And then two particular variables because obviously one of the impacts of the pandemic was the changing people's income statuses and because of furlough and or even being being laid off. So a financial crisis indicated by these food banks in the last four weeks and an actual reduction in recorded income. Covariates that were thought to be relevant were age, ethnic minority and a deprivation indicator and an existing health conditions indicator. So the mean GHQ case in a score by sex is what we're showing on this diagram and you can see here probably reasonably predictable that there's this initial decline in well-being. So the case in a score, high case in a score equals poorer well-being. And that is having a different effect on on men and women. So women were more acutely affected by the initial effects of the pandemic but the recovery was also stronger as the pandemic progressed. There is of course a well-established difference in overall baseline level of well-being between between men and women and obviously this is self-report based so that we don't know whether this is to do with under-reporting by men or actual real differences between men and women in their in their well-being. However, the relative effects here are obviously noticeable. So our first model here predicting the declining well-being. So this is the basic covariates and what we can see here is that the covariates are having not a great deal of effect. So you see these r-squares at the bottom here, really tiny r-squares, slightly more for the female model than the male model but essentially there is no particular nothing really going on here. So the covariates actually are not having that much of an impact which is slightly surprising and noticeable that health conditions and IMD deciles are not significant at all. Now once you add in the baseline well-being then there is a sudden increase in in the r-square. There's a really quite noticeable change. So this is one of the big predictors of a declining well-being. However, what's happening here is slightly surprising because it's having a negative effect. Now just to get this clear that means that it's actually associated with a smaller decline in well-being than if you have a higher baseline score. So the the effect on well-being is perhaps the opposite of what one might have predicted. Those who were already had poor well-being did not decline as much as those who were their well-being was good. If you add in loneliness as a predictor that effectively is adding another significant amount of to the r-square. So loneliness and baseline well-being are heavy predictors of the decline. So here we are looking at something very specific going on in terms of possible mechanisms because we're actually after is a change in loneliness over the course of the of the pandemic. So during those ways and this probably is the mechanism which is causing this change in well-being and that begins to make some sort of sense because if you have you're in a situation where you've got a lot of social capital and that is then challenged by the pandemic the lockdown essentially separated out. It was a leveler in that sense. Then your change in loneliness will be a natural consequence of that and that might then lead to a change in well-being. So we can see a possible causal mechanism here although obviously this is regression data and we can't read too much into that. So this is the recovery and again the covariates are not really making much headway in terms of the r-square value that we've got here and the predictiveness of the of the recovery. But once we add in the baseline well-being and the initial declining well-being as a predictor then the r-squares jump up again. So the recovery feels a bit like a boomerang effect and the well-being is declining and then the amount of the decline and the the original baseline is then predicting the race of recovery and that that is happening in a strong time. We reversed because we're talking about recovery here we've reversed the coefficient so a positive value means a greater recovery and that again and then we've added in a set of other covariates here to do with somebody who's always lonely as a as a predictor of how well they recover and that is that's a good predictor and then this financial crisis caused by the pandemic is also also a significant predictor and this again increases the power of the model. So our conclusions a declining well-being was first observed at the beginning of the first lockdown period at the beginning of march 2020 no great surprises there but this was matched by a corresponding recovery between April and July as the restrictions were gradually lifted. There's no association between the decline and deprivation nor between deprivation and recovery so this again was a slightly surprising result obviously the granularity of the deprivation indicator may have been an issue here but the strongest predictor of decline in well-being was the baseline score with the counter-insurgency finding that those with the pre-existing poorest well-being the impact on the of the pandemic restrictions on mental health was minimal but those who felt previously well the restrictions was greater and so for recovery the baseline the degree of decline new economic hardship and reporting of loneliness were all important factors so just to say again IMD was at LSOA level and that may not be fine grained enough the findings do conflate several different factors so the background effects of the pandemic the effects of lockdown itself and then personal effects of the pandemic so individuals may have people who have died or have become severely ill they may themselves have become severely ill and these will obviously have very particular effects on one's sense of well-being so there is a question about whether it's the background effects of the pandemic the personal effects of the pandemic or all the specific effects of lockdown and we these are obviously conflated in the data and in the sense one sense of that is this doesn't actually matter it would be good if one was trying to get deeper into this to try and tease those out but there is a kind of almost a collated effect of those things the sample significant attrition between the main survey and the COVID survey we have to be straightforward about that obviously we've applied the weights that are available but there are quite a lot of people who elected not to take part in the COVID survey who are in the main survey and obviously that may be that may be a factor in predicting the change here people who suffer the biggest change in well-being may well have been the ones who have elected not to be in the survey so there may be some biases in here that we have that we can't possibly pick up on that the weights won't help us with so that's something just to be careful about interpreting these findings okay and here's the reference to the paper the published paper and thank you very much for listening