 Hello everyone and welcome to this webinar about data that you can use for investigating political behaviour in the UK. I'm Dr Zarek Inhill and I work for the UK Data Service based at the University of Manchester. I'm going to be just doing the introductions and helping out in the background with this webinar. Our two speakers today are going to be John Mellon, the Halsworth Research Fellow at the University of Manchester. I'm Dr Nicole Martin, who is an lecturer in politics also at the University of Manchester. This is the third of a series of webinars about different data that can be used on different topics. Okay then, so I think we should move on to John, who is going to be our first speaker. Hi, thanks everyone for joining us and thanks Sarah for introducing me. So I'm going to talk about the British Election Study, so I'm one of the co-directors of the British Election Study. So first of all, let's start with the most basic question, what is the BES, the British Election Study? So the British Election Study has covered every election since 1964. It only covers Great Britain, so not Northern Ireland, and our team has run the survey since 2014 and we're currently in the 2019-2023 grant funded by the ESRC. So in this talk, I'm going to go through three types of data that's available that you might want to use for various types of questions. So a key distinction here is between probability samples and non-probability samples. So the face-to-face surveys are what we call a probability sample. This is where you have some known population such as addresses in the UK that you randomly sample from, maybe with stratification. And then you follow up with those sampled addresses as closely as possible to maximise the response rate. Statistically, there's a lot of good properties of probability samples. Assuming the response rate is high and assuming that you have a large enough sample, the true results from a survey should converge on the true results in general as the sample gets larger. Now, this contrasts with non-probability samples, which are generally recruited through online platforms such as YouGov. And these people self-select into them. They go and sign up to take surveys. And then to make inferences from that data, you have to have a model and understanding of the selection process. And it means that you don't have the same certainty about the inferences you're making from it. However, as we'll go into, there are a lot of advantages of the non-probability data as well. So I'm going to talk a bit about the BES face-to-face surveys, this very high-quality gold standard data, and then also the BES internet panel, which is this much more flexible, very large, 19-wave internet panel. But with the caveat that it is this non-probability survey, so I'm going to talk about the trade-offs between those. And then I'm going to talk about the long-term data from the BES going all the way back to the first survey in 1963 in advance of the 1964 election. So if you want to get hold of the data, BritishElectionStudy.com is the place to go for the most recent data. So eventually all the data will be deposited at the UK data service. But in the meantime, the data that's collected most recently, say for 2019, is available for our own website. It's completely free to download. You do need to sign up, but it's very straightforward. And if you don't know how to use existing statistical software, you can still go and analyse some of the data through a data playground tool that's for people who don't have the statistical software skills. So there's a lot of different things we ask about in the BES surveys. So obviously vote choice, parties, leaders, attitudes, Brexit, identities, values, all of the rest of these, and many other things besides obviously demographics as well. So that's just to give you a little bit of a flavour of the kind of things we ask. I'd be really happy to answer in the Q&A if anyone has questions about things that they can't find in the survey or that they wonder if we ask. Very happy to put you in the right direction on any of those. So to download the BES data, you start by going to our website, click on the data tab. That will bring you to this page here. Now it starts in the data playground, and that's the place to go and do analysis of data up until 2015, if you don't know how to use statistical software. But today I'm going to be talking more about the case where you want to go and analyse the data in more detail yourself. And to do that, you click on the data download button here. Now, when you get that, you'll be brought to the data download page, which gives you various options. So we'll start with the face-to-face data. From that, you click on the cross-sectional data. So the BES face-to-face probability surveys, they vary a bit in size, but generally, there are around 3,000 respondents randomly sampled by address across 300 constituencies. We're currently in the process of conducting the 2019 survey, and given that there is a massive lockdown because of coronavirus, we're currently working to move the face-to-face survey on to online for the remaining surveys. So we'll still have that random probability structure to the sampling, but people will be asked to take the surveys online rather than face-to-face field work, which, again, in the current climate, face-to-face is not a great thing you want to be doing. The other great advantage of the face-to-face surveys is that we have validated votes. This means that we take the respondents who answered these surveys and we match them to the electoral register and look at whether or not they actually voted in the election. So this means that if they accidentally forgot they didn't vote, or they felt embarrassed about it and didn't mention that they didn't vote, we can actually check against the official administrative record to check whether or not they did. And so this gets around the social desirability bias to say you voted. The probability surveys include many items that have been asked continuously since 1964, and they also include the CSES, that's the Comparative Study of Electoral Systems module, that can be used to compare the British data to many, many countries around the world in a comparative dataset. If you're interested in doing that, you should go to the CSES website because that has the combined file that lets you do the comparison very easily, and the BES is part of that. So once you go and click on the file you want, you'll be asked to register. You just need to give your name, an email, and you reason for using it. You also need to read very carefully the BES terms and conditions. I really encourage you to read them. They're not long, but they're really important. And the most important condition there is that you should never try to identify any respondent. You should never try to de-anonymise the data. And you also can't go and share the data onwards. You have to go and always download it through the website. But once you've done that, it's very simple to download. So I'm just going to give a really quick example of one small question you might want to ask using highly representative data like these. So is social media representative of real life? So an increasing number of social science papers have kind of assumed that social media data is representative. They'll use Twitter and Facebook data to study election forecasting, people's reactions to events, public opinion. But all of those studies are implicitly assuming that what's happening on Facebook and Twitter is representative of the wider population. And we can examine this assumption using the high quality face-to-face data. So first of all, we can look at the age distribution. So in the green here, that's Facebook or Twitter users. And the blue is the non-users with the pink being the overall population. And we can see that for both Twitter and Facebook, their populations skew far younger than the general population. If you're studying social media data, you're studying young people, not the general population. And in terms of education, it's not much better. You're studying people with more undergraduate degrees, more postgraduate degrees, more A-level qualifications. If you're studying Facebook users or Twitter users than the general population, they're very overeducated compared to the normal population. Which means that what's going on on those platforms represents a small slice of Britain, not the whole country. So our conclusions from this paper were that Twitter and Facebook data are not representative of the British public. But we actually have a surprising result, which is once you correct for age, gender and education, the platform's users are actually not that unrepresentative. So it's really these demographic imbalances that are the main difference between the general population and social media users. So the next set of data we're going to look at is the BES internet panel. So the panel currently has 19 waves, so that's the same respondents interviewed up to 19 times over the last five or six years. And on average, we have about 30,000 respondents interviewed in every wave. It's sampled through YouGov, and again, this is a self-selecting sample. The YouGov tries to make it demographically representative with weighting and selection, but you do have to bear in mind that people actively side up for this panel. So you have to go and correct for that and understand it as a limitation. There's funding for three further waves of the panel. And one of the exciting things that we do in the internet panel is that we allow users to propose modules to field to a quarter of the sample at a time. And you can go into submit proposals at thebritishelectionsstudy.com website. We're also planning to link up various data to the internet panel in future. Facebook data with permission through the Social Science One collaboration. Twitter data through the Twitter API, again, with permission of the respondents. And we also have asked during the 2019 campaign what TV programmes the respondents watched every day during the campaign and what newspapers they read. And we'll be able to link that to the content of that media using the BBC's balance tracking data and the Factiva database. So that's potentially some really exciting data that will be linked in that you'll be able to look at to understand exposure to various political information. So just to kind of go over the advantages and disadvantages of this data, it's a long running panel. So respondents have been interviewed at every politically relevant event in Britain of the last five years. Some of these respondents even go back to the 2010 election or even the 2005 elections. You have them over a very long period of time and you can look at change within individuals. That's a really important part of this data. The samples are very large for each wave. That also means that the samples over time remain reasonably large. So you can do a reasonable analysis of how people voted at free subsequent elections. The data includes pre-election, campaign and post-election waves around each election. So you get really detailed information about how people were shifting at election times. The panel data means that there's limited recall bias as well. If you're just using a cross-section, you might be wanting to know how people changed since the previous election, but you're relying on their recall five years later potentially. And that means that they may have forgotten. They may have decided that they don't want to tell you anymore about how they voted back then because they don't agree with it anymore. Whereas with the internet panel, you're relying on what they said at the time, which is more accurate on the whole than what they say five years later. And there's a very wide range of questions asked too. Now, there's only really one major disadvantage and it's again this point that it's not a probability sample. Waiting can go and make it more representative, but you need to go and always have that in the back of your mind that there could be unrepresentativeness that could affect the results. And you should be always alert to that as a possibility. So to download the data, again you go to this download data page and you click on panel study data. Now, this is going to take you to a page with several different options. So let me take you through what these all are. The top one here is a single wave file. So this is just the data that was taken in Wave 19. That's the post-election wave of the 2019 election. However, it doesn't include any information from other waves of the survey. Now, compare this to the second one, which is the combined Wave 1 to 19 internet panel. So this includes all the responses everyone gave to any wave of the BES all the way back to 2014. We generally recommend using this because it's the most flexible, it's the most complete data, and it will allow you to look at change over time. So if possible, use that data. Now, the one disadvantage is that it's several gigabytes, so some people's computers will just refuse to actually even load it. But if it's possible for you to use the combined file, we strongly recommend that you do so. The final one is just the open-ended response data. There's various open-text questions we ask. And in those cases, that data is provided separately because it's less used than other variables, and it's, again, several gigabytes. But if you need that data, it's also available. Just to give a little more information about what the survey looks like around elections. For every electoral event and also the EU referendum, we ask a pre-election wave, a rolling campaign wave, and a post-election wave around each of these. So the pre-election wave measures characteristics that don't change in the short term from day to day, things like someone's A, well, not their age, their education, their income, things like that. They wouldn't be expected to change in a matter of weeks. We also measure the baseline for how things might change over the course of the campaign. Then in the campaign wave, we interview 1,000 respondents from the previous wave every single day of the campaign. And this means that we are able to track the attitudes of all the respondents in the pre-election wave as they change over the course of the campaign. And that allows us to look at change over time, really reliably on this period, and look at how the campaign is affecting people's attitudes, their perceptions of the parties. And finally, we have the post-election wave. And this is focused on measuring the dependent variables or the outcome variables of the election. So vote choice, participation, satisfaction with democracy. So I've created a handy table here which shows for each election which wave corresponds to pre-wave, the campaign wave, and the post-wave. And so if you want to look at any of these electoral events, you can go and use those particular waves to capture that. And obviously these are the same respondents across waves, so you could also compare, say, how the campaign went in 2015 and 2017 or 2019. If you use the data, as I've said, you need to use weights. And it's really important because the weights can make a big difference. Where possible, you should be using the weights that are weight underscore new underscore something. That's because these weights include corrections for political attention and education, which are two big biases in the role of data. And these were two of the problems that led to the 2015 polling miss that have now been corrected in the newer weights. So use these wherever possible. Now there are still some waves and wave combinations that use the weight underscore full. We're trying to phase these out. We haven't quite got all of them transferred over yet, but eventually these will disappear entirely. But you may have to use them in a few cases. But just bear in mind that they don't have these attention and education corrections. So it's relatively straightforward to understand what weight to use for a particular wave. Just use the one with the wave number at the end. But for panel weights, it's a little more complicated. So in this case, if we had a variable, a weight called weights underscore new underscore wave six, wave 13. This refers to the weights covering respondents who took wave six and wave 13, regardless of what waves they took in between time. Even if they only took those two waves, they're included in that weight. Now contrast this with weight underscore new underscore wave six underscore wave 13. In that case, that's the weight that covers all respondents who took all of the waves six, seven, eight, nine, 10, 11, 12, 13. And in that case, that's a lot fewer respondents because if they missed any of those waves in between, they're not included in that weight. So if you're interested in respondents who took wave six and wave 13, the post-election waves of 2015 and 2017, you just need the first weight. If instead you're interested in how people changed over the whole course of the period of time in between those two waves, you're going to need instead the one below. So in general, try and choose the weights that covers just the waves you need because that will give you a bigger sample size of respondents and mean that you have more power for your statistical analysis. We also provide profile data, and this is data that's from Ugov. So they go and collect information on their respondents constantly, essentially. And on average, each piece of information is updated about every six months. Now what we do is we include a lot of this information in the actual BES file, even though it wasn't collected by us. And this is great because it gives us a lot more relatively up-to-date demographic information, background information about all of these respondents that we wouldn't be able to ask in every wave otherwise. The downside to it is that there is no guarantee that the respondent was asked that question exactly at the time of that wave. They were asked it in the last six months, but they're not asked it at that very moment. So this means that you have to be a little bit careful about using this profile data to look at change over time. If you have, say, waves seven and eight, which are the pre-EU referendum wave and the campaign wave, they're only asked two weeks apart. There may be an education variable in both of those waves, but you can't assume that those were two separate questions. Probably it's the same data in both of those. Currently, the variables for profile are tagged as profile underscore something, but we're actually revamping the way we do all the profile information, and it will change to P underscore something in future releases. It's also going to be better documented. At the moment, it's not always 100% clear what information was asked as profile or in the survey, and that will be a lot clearer in the near future. So I'm just going to give a very, very quick flavour of what you can do with this kind of data. So in 2017, there was a really dramatic change during the campaign where Labour Party went from historically low to almost drawing equal with Theresa May's Conservatives during the course of just a few weeks of the campaign, and in fact, the 2017 election ended up seeing the largest two-party share of the vote in decades. So what drove these changes? Was it Brexit positions, or was it sections of the leaders, or something else? So these are sankie plots. The basic idea is to show the flow of the vote from a starting point, in this case, the 2015 election, to an end point, in this case, the pre-election wave of the BES. So this is just before the campaign started properly. I've broken down these flows by whether people voted leave or remain in the EU referendum. So there's a lot going on here, but the key things we see is that on the leaf side, the Conservatives sweep up a high proportion of leaf supporters, including those who voted Labour in 2015, but especially they sweep up a very, very large portion of 2015 UKIP voters. So on the leaf side, Labour loses a lot of votes to the Conservatives, and it also loses a lot to people who are now saying they don't know who they will vote for. On the remaining side, Labour retains its voters a bit better, but it's also losing some to the Lib Dems and a lot to don't know. Also the Conservatives' vote is collapsing towards don't know Lib Dems on that side too. So the Conservatives have consolidated the lead vote. Labour now has a more remaining coalition, but really just because it lost levers so badly, not because it's particularly attractive to remainers. Now we can look at the vote flows from that pre-election wave to the post-election wave in 2017. So this is before the campaign to afterwards. And here we see that the Conservatives continue to consolidate the lead vote, but what's really interesting is that Labour gets a lot of votes back on the leaf side and the remaining side. In fact, it's about equally. So Labour during the 2017 campaign actually manages to recruit levers and remainers and boosts its vote share on both sides. So the odd thing here is that Labour ends up with a very pro-remain coalition without ever differentially attracting remainers. It just lost a lot of levers and then attracted back everyone, which sort of makes sense given that Labour's position on the EU was very unclear during this campaign. So why did Labour do so well at recruiting lost voters over the course of the campaign? Well, I mean, there's a lot of explanations for this, but one of the big ones is the trends in leadership ratings of Corbyn in May using the daily campaign data for the BES internet panel. So we can track how much people liked on the 0-10 scale May and Corbyn over every day of the campaign. And what we see is there's a nearly linear trend upwards in approval of Corbyn and a linear trend downwards in approval of May. We also show later in this paper that these changes are closely related to switching to the Labour Party at the individual level, because, again, we have this great panel data that lets us look at that switching. So overall, we see that polarization along Brexit lines took place before the campaign started. So the campaign was less about Brexit, but the election as a whole was still structured around it. The campaign itself was more about Corbyn and May's leadership. So the final set of data I'm going to look at is the long-term BES data. Again, the BES has been running since 1963. You can download the older panel studies or cross sections from the BES website or UKVS. Unfortunately, the data isn't harmonised for you, so you will have to go and recode the variables yourself, which is quite a bit of work, but very, very worthwhile if you can. So in our book, we go and look at some of these questions of how things have changed over time, particularly about how attached voters are to parties between 1964 and 2017, and how willing they are to switch parties. So using the BES face-to-face cross-sectional data going all the way back to 1964, we're able to see the change in how strongly attached people are to a political party. And when you're back in 1964, nearly half of people have a very strong party idea over here, with very few of them saying they have no party ID. But this declines rapidly in the 70s and continues ever on downwards, whereas the proportion of non-identifiers is increasing over this whole time period. So you end up with a very different world in terms of how strongly people identify with parties between the start of the surveys in the 1960s and the most recent ones. And this is also reflected in how willing people are to switch. Now, this is using the inter-election panels where the same respondents are interviewed at one election on the next. And we could look at the proportion of people who voted in both of those elections who changed parties. Back in the 1960s, you're looking at maybe 12, 13% of people switching. By 2015, that's above 40%. So this is a huge change in the context of British politics and explains a lot of the dramatic changes we've seen over this time period and in recent elections. So I'm just going to conclude with where you can find other political science data in the UK. There's the Scottish Welsh and Northern Ireland election studies available from UKDS. There's the British Social Attitudes Survey. Again, this is the high quality probability sample data. The big advantage of it is it's asked every single year and it asks a lot of the same questions every time. It's not all about politics, but there is some good politics questions available. So also something to look into. There's also the Ethnic Minority British Election Study. So this data is incredibly valuable because the ethnic minority population of Britain is relatively small, so it's hard to examine it in a general survey. But the downside of this data, although it's collected using the best methods, is that the most recent data collection was 2010 and the data is simply just very expensive to collect. So as of now, there hasn't been a new one since then, but we're hoping there will be one in the near future. We also have some linked data available through the BES website. Click on that tab there. So we have expert surveys about the positions of parties and their policies. We also have constituency results available all the way from 2010 to 2019. Great. Thank you very much. Okay, perfect. Thanks very much. We're now going to move on to Nicole's presentation. Okay, so thanks very much, John, for that presentation and Sarah, for facilitating this. I'm going to talk about a slightly different study called Understanding Society now. Although I work at the University of Manchester now, I used to work for the study and still am involved with them as their politics topic champion. So if after this presentation you have further questions about using understanding society and politics research, please feel free just to get in contact with me directly via email, and I'll do my best to help. So this is what I hope to cover, time permitting. So I'll give you an overview of this study as a whole. So what is understanding society and how does it work? Because we're interested in politics, I'm going to tell you about the types of questions that exist in understanding society that would let you study questions about politics, political behaviour, that kind of stuff, and also about which kinds of projects are best suited to understanding society data. I'm going to point you towards documentation and training that is available and also to where you can access the data and then to kind of, perhaps, something a bit more interesting and illustrate what the data can be useful. I'm going to talk about some of the existing published research that people have carried out using the study. So we're going to look at Brexit, conditionality and benefits and also whether newspapers win elections or not. So understanding society's key features. So it's a household panel study and it started in 2009. So it's, I'll tell you about size in a second, but it began in 2009. However, it also has incorporated respondents from a previous household panel study that started in 1991 called the British household panel study. There are 8,000 households in the British household panel study in its last wave in 2008 and they were all invited to join this new study called understanding society in 2009, quite a few did. So it's a household panel study. The key feature is that it's household. So everyone in the household contributes. This means that we collect information, as long as they consent from everybody in participating household. So anyone who's an adult and in our definition that's anyone who's aged 16 or over is invited to complete a questionnaire themselves or an interview themselves that lasts about 40 minutes and collects lots of detailed information about their lives. But if they are 10 to 15 years old, they can also complete a youth questionnaire. So we have particular questions we ask of young people. So we get to understand about their lives. And then finally there are zero to nine-year-olds and their parents give information about them at different stages. But that's not really relevant to us because we don't ask them about their eight-year-olds political identity, for example. So it's a household study. We have information on everyone in the household and then it's a panel study. So instead of getting a new sample every single year we go back to the same households and follow people from those households as their households change over time. So you can follow the same person. There are some respondents from the British household panel study who started in 1991 who are still being interviewed in the current way of understanding society. So we'll have almost 30 years of data from those people quite shortly, which is very exciting. The key features about understanding society beyond the fact that it's a household panel study though are five things which are all quite useful for studying politics. So the first is that it has a large ethnic minority and immigrant boost sample. The first boost sample was added in 2009 when the study started and that's about 4,000 households where have at least one person within them who's from an ethnic minority background. But of course there's been continuing migration to the UK since 2009. So in 2015 a new immigrant and ethnic minority boost sample was added of about 2,500 households with at least one member who was born outside the UK or someone from an ethnic minority background. So if you want to study the political preferences of the one in 10 UK voters, one in 10 British voters who belong to an ethnic minority, then understanding society can help you out with that. We also cover all four countries or all four nations within the UK and that applies to the British household panel study at least from 1999 to 2001 onwards. So you can look at just voters in Wales for instance if that's what you're interested in. It has a very large sample size. So to give you an idea of this, in 2009 it started up with about 40,000 households that had information on about 100,000 people. But of course over time some panel studies are likely, some people say at some point enough decide not to belong to the study anymore. So in the most recent wave of data that's available that covers 2017 to 2018, there were 20,000 households who were in that wave and that's information about 52,000 individuals. Of that there are 36,000 adults who completed a full interview or had a proxy questionnaire done for them. So you have, those are the people you have, 40 minutes worth of questionnaire content on lots of information and the oldest respondent in 2017-19 was 103 years old. There are also 1,700 people aged 80 to 90 in the most recent wave. So that means that because of the large sample size you can conduct research on subgroups for the population that might not be represented in a smaller survey. It's multi-topic. So it's not just a political science study. It studies lots of things about people's income, their jobs, their health, their family structure, their environmental behaviours, whether they recycle or not. And also there's lots of epidemiological information there. So we have biomarker data, health measures and genetics data collected in wave two of the study. So it's very much a broad study with lots of information rather than a specialised election study. And that's what it's good for. And finally there are various levels of geographical data linkage. So if you're interested in politics you can easily link to constituency information say if you want to know people who lived in electoral constituencies where there was a marginal race, a marginal election last time had different political attitudes and those who lived in a safe seat that's something you can do with understanding society. So what data is there on politics? So after elections we collect turnout and vote choice. I will say I'm glad that John mentioned the idea of validated vote, validated turnout because we don't do that in understanding society. So we just ask people and this leads us probably to overestimate the level of turnout that happens. But because it's a broad topic, a broad multi-topic study we have to make decisions like that. And we also ask people if they voted, who did they vote for. The vote choice data after elections is a bit complicated because for the 2010 and 2015 general elections it was not asked of everybody in the sample, only a sub-sample. This is because the field work period for the study occurred over two and a bit years. So if there was an election tomorrow we would have three waves in the field. So for the 2017, 2019 general elections we ask everyone within a year of the election who they voted for. But that means if you want to study the 2017 general election for instance you're going to need to combine data from wave eight and wave nine of understanding society to get all the information we have. General elections were also asked about in the British House for Panel study and we asked about two recent referendums, the Scottish independence referendum and the EU referendum. And we are reintegrating devolved assembly elections into our questionnaire. Away from elections there is also some information on political attitudes. So for instance vote intention or party identity, how many people feel strongly that they identify with political party. Some of these questions are worded differently from how the British election study asks them and that means you can see how the trends differ between the different ways of asking the same question. But importantly they are asked consistently within understanding society and the British House for Panel service. So if you want to compare people in 1995 and 2015 the same person the question will have been asked the same way. And finally I think the other key part of politics data is linking understanding society data to election results. I've already described how you might want to do that. So one thing that you might do with this data you might want to know about the people who work in the higher education sector are more left wing than the population on average. And because of the large sample size and the detailed socioeconomic coding in the data you can create a graph like this. So in this graph the darker kind of colour bars are people who work in the higher education sector and the lighter colour bars are the general population and this was vote intention and this was party identity in Wave 6 in 2014-15. And you can see that for instance the darker red bar shows that more people who work in the higher education sector felt close with the Labour Party than the general population in Wave 6. And finally we also asked young people about their political preferences. So 10-15 year olds get asked these two questions at every couple of years in understanding society and we have about 2,000 to 3,000 observations each wave. So this is really wonderful if you want to know what happens to 12 year olds who say they want to vote for the Conservative Party what jobs they go on and do they go on to do or questions like that. Okay, so if you want to know what variables there are about politics in understanding society then you should go to this link up at the bottom here which will take you to place in the documentation where you can see all the questions in understanding society in a digital panel survey that are to do with politics. So you can go and look through those later on. Things that understanding society is good at is long-term political processes. So the BHPS and understanding society data has been harmonized so it's quite easy to compare the same people over time or people in the same kind of family situation over time using that data from 1991 to the current day. It's also quite good for long-term dynamics so how do people's opinions change over time or how does a change in a household's circumstances affect their political opinions? So these are the panel structure of the data means that you have a much better handle on those kinds of dynamics. But I think the real strength of understanding society is in the sample and the detailed social, economic and health information. So if you want to have a very detailed picture of exactly what people think about British politics then you should use a study that's designed to capture that. You should use something like the British Election Study or perhaps the British Social Attitude Study. But if you want to know for instance what has a change to the way a particular benefit has been administered by the government changes someone's attitudes to politics then you need a study that focuses really on getting the social, economic and health information measured really accurately and that's what understanding society does. And then finally the fact that it's a household context and that it has this large, ethnic minority and immigrant boost sample. Okay, so documentation is quite important and understanding society being this long running study that's very complex for the household component. Documentation itself can be quite complex. So the website is quite simple. It's understandsociety.at.uk and if you get there the documentation section is right at the top. If you want to go there and click on the documentation section you get a menu like this. If you are an experienced data user and you want to dive straight in then maybe you should have a look at the documentation for the main survey so that's the left hand most bubble. But if I would strongly recommend if you're an experienced data user maybe check out the Getting Started Guide which you can access through the link highlighted there. So this will tell you important things about how the data is structured, how the variables are named, where to find information about which weights to use and where to find further resources depending on your needs. So if you're an experienced data user I would recommend the Getting Started Guide. However, if you want a bit more help or you think that you've decided you definitely want to use the study then a really good option is to attend one of the training courses. So these are two-day training courses. They're completely free to attend and they run at the University of Essex. Of course due to the coronavirus in-person training courses have been suspended but they will run online instead in a kind of similar webinar format and they did that recently and I think it went quite well. Now the courses actually run simultaneously for users of four different statistical software programs data, SPSS, SAS and R so don't let the fact that you're a SAS user or an R user put you off attending the course set up to deal with all those different or those different ways of working. But if you don't want to attend a training course there are also self-guided courses available in data, SPSS and SAS. So I strongly recommend those if you're committed to using the data or if for instance if you're doing a PhD you want to get started using panel data for the first time. There's also a user support forum so this is what the user support forum looked like yesterday. This is where if you have a question that isn't answered in the documentation or you need some help or you think there's an error in the data you've reported to us and as you can see from the dates it's very active with the whole team answering questions so that can be quite a good way of getting help. And finally how do you know how to find the studies while they're all released to the UK data service? There's actually a data collection for understanding society that pulls together all the studies that understanding society runs so I haven't told you about everything there is to know about the study but this is where to find this is where to get access to them. If you were to go to this page you'd see these three different levels of access and for most of us we'd want to click on the where I put the red circle and this would show you how to get hold of the main data set understanding society data set for all the ways that are currently released and the British household panel study that is harmonised to be the same. But if you were to click around there you'd notice there are quite a few studies so there are many more studies that would come up on the UK data service page for understanding society so something that you might want to check before you download a data set or apply to use it is is it the main study or is it something else for the innovation panel it's probably that you want the main study. What is the user licence and is there anything specific in the data set name so for instance if the if the data says understanding society marriage for habitation histories that's probably not what you want that's a special data set that's been released separately and the most likely data sets that as a politics researcher you'd want to use is the main understanding society one that's SN6614 so if you type that into the UK data service search function it should come up or the Westminster constituency geocodes or the local authority geocodes there are different levels of data access so for the main one you can download once you've registered with the UK data service and accepted the study conditions but for other levels of access you have to go through a longer application process then it's quite straightforward to go through the UK data service website and it takes you to different stages this is essentially so that we don't release disclosive data too readily especially because there are many different levels of geography that we have available and so you just have to explain your project and justify why you need the more disclosive data so this is a piece of research done by by Luca Bernardi and Rob Jones of the universities Liverpool and SN6 in the UK and they pointed out that depression is the kind of second most common medical condition in England it's second only behind heart disease but it's really considerable when we think about politics and they argue that this is a bit of an oversight so people with clinical depression clinical depression might cause people to think about politics a bit differently especially due to this thing known as status quo bias so people with depression less likely to seek out new information and more pessimistic about the prospect of change various aspects of status quo bias which actually if you think about electoral choice do I want this who do I want to win this election or who do I want to win this referendum this is really quite relevant and they look at the case of Brexit and clinical depression to answer this question so with status quo bias they actually say and this is why it was a bit of a trick question that people with clinical depression and status quo bias you'd expect people with status quo bias to have a dim view of the change involved in leaving the EU and so to be more pro-remain more in favour of keeping the status quo but the referendum result changed the status quo and so what you should see is that after the referendum people with clinical depression so not people who just feel people sad but people with diagnosed with clinical depression should think that leaving the EU is the right thing to do and this is exactly what they see so the dark blue bars on the right hand side of each set of two are people with clinical depression before the referendum they're more pro-remain and after the referendum they're slightly more pro-leaf or at the very least they're no longer more pro-remain and so Luca and Robert Robjons argue that this is this is a whole area of political science waiting to be explored the idea that mental health and health conditions are important in our understanding of politics which nowadays seems quite timely okay and if you want to learn more about research I'm sure you can look in my slides later on but I think I will leave it there