Posts Tagged ‘LSE Comment’

No Recourse to Public Funds: More than a quick fix needed for immigration rules

The Prime Minister recently appeared surprised to hear that most non-EEA migrants with temporary leave to remain in the UK are not eligible to claim benefits, and hinted at a review in the light of COVID-19. Alan Manning argues the system needs long-term reform, not just a quick fix during the pandemic.

At the recent Commons’ Liaison Committee meeting, Boris Johnson seemed unaware that most non-EEA migrants with only temporary leave to remain are subject to ‘no recourse to public funds’ – NRPF in the jargon. Ordinarily, these migrants are not eligible to claim welfare benefits so are totally dependent on work for income. But these are not ordinary times and, through no fault of their own, work may have disappeared and destitution looms. The case for some temporary changes to NRPF are overwhelming. But there also questions about NRPF conditions in more normal times.

As a Home Office factsheet points out, most comparable countries – including countries like Canada that have a relatively open migration regime – have some form of NRPF condition for migrants on time-limited visas. The reason for this is that most people feel a sense of responsibility to provide a minimum level of income to those who are permanent members of their society but less responsibility to members of other societies. We might argue that we should spend more on international aid but relatively few argue we should provide UK-level benefits to everyone in the world. And ‘temporary’ migrants occupy an intermediate position: here for the moment but perhaps not for the long term.

For the vast majority of those on Tier 2 work visas subject to minimum salary levels, NRPF conditions probably matter little because salaries are sufficiently high. Those on the Tier 5 Youth Mobility Visa are only here for two years maximum and have no right to bring dependents, and it is probably reasonable that they support themselves financially. Similarly for those on study visas who are probably paying at least £10k in fees and have limited rights to bring dependents. It is for those on family visas where there are a number of concerns.

First, the idea that many of these ‘temporary’ migrants are really temporary may well be something of a fiction. Some are better described as ‘temporarily temporary’ as there is a high probability that they will become permanent members of UK society. The latest Migrant Journey Statistics report finds that 86% of migrants granted family visas in 2014 had valid leave or settlement at the end of 2019, compared to 24% of those who came with work visas and 17% who came with study visas.

This leads on to the second problem with the current system: some migrants may be spending very long periods of time subject to NRPF. Changes to the immigration rules in recent years have lengthened the time to settlement for many migrants on the family route. There are two routes to settlement via a family claim (as a partner or a parent) – a five-year and a ten-year route. The former is for those who meet all the language and financial requirements which include a minimum income level of £18,600 (higher if there are children). The government argues this restriction on the right of British citizens to marry who they want and to live in the UK is justified by a public interest in the level of taxes. It is probably the case that, on average, those coming on the family route receive more in benefits and public services in the long-run than they pay in taxes but this is a human rights migration path not an economic one.

The ten-year route is for those who do not meet the requirements for the five-year route but have a legal claim to remain the UK. On both routes the visa has to be renewed every two and a half years with associated fees payable. It seems likely that the numbers on the ten-year route have been rising but it is hard to be sure because detailed figures of the stock of people on different types of visas is often not published. Of those given settlement in 2019 (when any NRPF condition would be removed) who initially came on family visas, one-third of those getting settlement have been in the UK more than five years. The fraction taking longer than five years is likely to rise in the future as the ten-year route becomes more common. For example, the number of extensions of the right to stay in the UK for those on family visas rose from 22,000 in 2010 to over 115,000 in 2019, suggesting there are now more family migrants in the UK requiring extensions.

We don’t have statistics on the living standards of these long-term temporary migrants but it is likely that they are low. A family migrant may be on the ten-year route because they don’t meet the necessary language or financial requirements in which case their earning capacity in the UK is likely to be limited; some of them may well be amongst the poorest members of our society. There are provisions to remove NRPF if it would cause destitution but the take-up of those entitled may be low. If people are entitled to help, the system should help them to claim it; it is doubtful the current procedures pass that test even if they have been made more user-friendly recently.

The third area of concern is that there may well be many children, including many British-born and future British citizens in these low-income households. Poverty is bad but child poverty is worse. When it affects those who are very likely to become British citizens the long-term, damage done by child poverty affects us all. NRPF may be a bad investment even on long-term financial terms which is the usual justification for the restrictions. Again, we don’t know the scale of the problem – research by the Children’s Society suggests there may be tens of thousands of children affected.

During this evidence to the Commons’ Liaison Committee, the Prime Minister said “I will find out how many there are in that position and we will see what we can do to help them”.  I hope he is true to his word, looking not just at the immediate problems but the longer-term issues.


About the Author

Alan Manning is Professor of Economics at the LSE and an Associate at LSE’s Centre for Economic Performance.




All articles posted on this blog give the views of the author(s), and not the position of LSE British Politics and Policy, nor of the London School of Economics and Political Science. Featured image credit: by Al Soot on Unsplash.

Abolishing the NHS surcharge for health and care workers is not as generous as it sounds – a pay rise is needed

The government was recently been forced to re-think its plans to continue charging workers in health and social care for using the NHS. But how many pay the surcharge and how much would it cost to abolish it? In answering these questions, Alan Manning explains why the gesture is not as as generous as many might think; more needs to be done to address the poor pay and conditions in this sector.

Currently, most migrants from outside the EEA who are on a visa of six months or more have to pay the Immigration Health Surcharge, which starts at £400p.a. though an increase to £624 is planned for October. This includes workers in the NHS and social care on whose work we have all depended in the pandemic.

Understandably, levying this charge on these workers at this time seemed to many a lack of generosity of spirit. The government, after some delay, has eventually decided to exempt workers in NHS and social care, seemingly permanently. But how generous is this gesture depends on how many migrants in these sectors are actually paying this charge. It is hard even for the government to answer this question because many of the relevant migrant workers are not on work visas, so the Home Office would not know what work they are doing. But here are some rough estimates:

Table 1 shows the fraction of UK-born, EEA-born, and non-EEA-born migrants in the sectors of ‘Human Health Activities’ (including but quite a lot wider than the NHS) and ‘Residential Care Activities’.

These figures come from the Labour Force Survey are a bit different from some other sources.  The migrant shares are much higher in some parts of the UK, notably London. EEA migrants do not pay the surcharge (but new arrivals will do so in the future under the government’s plans). Not all non-EEA migrants do either – it is only those on visas, essentially those who do not have permanent leave to remain in the UK. Those who have become UK citizens do not have to pay.  In health, 63% of non-EEA migrants report being a UK citizen; in care it is 46%.

Even among non-EEA migrants who are not British citizens, those with settlement do not pay. It is harder to estimate this proportion but one indication is length of time in the UK. For non-EEA migrants who are not UK citizens, the distribution of time in the UK is reported as:

Half of this group have been in the UK more than ten years when other data sources suggest most migrants will have settlement (though there are always exceptions). In fact, many of those coming under the family or work route (common in these sectors) will have settlement after five years. It seems likely that no more than half of the non-EEA migrants who are not UK citizens have to pay the charge.

These estimates imply that in health, only 2.1% of workers are paying the surcharge, 12% of migrant workers in the sector. In total, this is 50,000 workers though this includes many who are not in the NHS. Assuming each worker has an average of one dependent who also have to pay the charge (a high estimate I suspect) the total amount it will cost to waive the charge in the health sector would be about £41m annually. For the care sector, the estimates imply that 3.8% of workers are paying the surcharge, 18% of migrant workers in the sector. This amounts to about 39,000 workers paying £31m in the charge.

So, the total cost is estimated to be at most £70m annually and I again suspect this is an over-estimate. A nice gesture, and important for those who currently pay the charge, but perhaps not as generous as many might think.

When Boris Johnson says the charge has raised £900m, he may be summing over many years and including everyone who pays the charge, most of whom are probably students. To exempt health and care workers would cost very much less.

There is a flip side to this coin, however. It costs little because fewer workers in these sectors than many think are paying the charge, even among migrant workers. For those that do pay it, waiving it or abolishing it is an important gesture but perhaps we need to do more to thank workers in these sectors, both migrants and the UK-born.

My suggestion would be to make a start by immediately raising the pay of care assistants. Some supermarkets have already paid bonuses to their staff and the Welsh Government has promised a one-off £500 bonus for care workers. The UK Government has spent plenty of money on other things. But it should be a permanent rise in salary, not just a one-off bonus. That may necessitate resolving the problems of financing the sector, but everyone knows this needs to be done. The poor pay and conditions in this sector were a national scandal before the pandemic and the primary cause of the high level of vacancies. Raising the pay of care assistants would not just be an appropriate gesture of thanks for the present but a good investment for the future.


About Author

Alan Manning is Professor of Economics at the LSE and an Associate at LSE’s Centre for Economic Performance.




All articles posted on this blog give the views of the author(s), and not the position of LSE British Politics and Policy, nor of the London School of Economics and Political Science. Featured image credit: by Stephanie Martin on Unsplash.

A modern ‘Rasputin’, or the UK’s ‘Vice-Premier’? Whichever view you take, Cummings’s role is unprecedented

In the Dominic Cummings row, developments are not only unprecedent because they concern an unelected adviser but also because they are impacting on the country’s ability to deal with the health and financial crises. This scandal is thus affecting lives and livelihoods, argues Patrick Dunleavy.

Dominic Cummings set out to be, and has certainly become, a sharply polarizing force in British politics. After his COVID-19 travel debacles, much of the country takes a fiercely hostile view of his immanent sense of privilege, including but not restricted to his many anti-Brexit opponents. For them, Cummings is a Rasputin-like figure, a sinister and shadowy influencer with a mysterious hold over a dazed and confused Boris Johnson (the Tsar in this version) and his scheming adjunct (Cummings’s first and most committed sponsor, Michael Gove).

Like Rasputin, Cummings picks gratuitous fights with all around him, insults every established elite, briefs relentlessly against his enemies, is unworried at being pervasively loathed, and is completely convinced of his untouchability because of his hold over the PM. In a developed form, some versions of this view take Cummings as an embodiment of many malignant dark forces on the right, pushing for an unrestrained populist assault on any constitutional element force constraining No.10’s privatizing freedom of action.

Almost as extreme in a different way, however, is the view amongst some Conservatives, all Johnson-regime loyalists, and the far-right neo-liberal think tanks and pressure groups, that Cummings must somehow be defended at all costs because of his critical importance for the current government. In this perspective, his role is akin to that of Vice-Premier, providing a Dick Cheney-like Brexiteer and free marketerr spine for an otherwise out-of-his-depth PM. The 25 May press conference in the 10 Downing Street Rose Garden, with Cummings scripting and delivering his own defence, gave enormous substance to this VP role, affording him a prominence, prestige, and official backing never before given to any cabinet minister in trouble of this kind. There is no UK constitutional role for ‘the PM’s chief aide’, nor are there any established political mechanisms for holding such a ‘Vice-Premier’ to account.

The fact that Cummings evinced absolutely no regret and gave no apologies for his overtly lockdown-breaking behaviour has already occasioned a lot of comment. But in addition, he deliberately garlanded his account of his actions with many details that signalled ‘unreliable narrator’ as clearly as any postmodernist novel – such as testing his potentially problematic eyesight by driving 60 miles to a beauty spot; and portraying his initial midnight dash to Durham as the actions of a man alone, without resources, friendless in London, and on his last possible throw to save his family.

How ministerial resignations are different

Compare all this with the accepted playbook of ministerial resignations charted by Keith Dowding and his co-authors in Accounting for Ministers. Using detailed statistical analysis and a comprehensive survey of cases, they showed that a minister who got into trouble was normally left to fight their corner alone at first, with the PM and 10 Downing Street referring critics to the minister’s explanation of their behaviour, while conspicuously omitting to pledge anything like full support or confidence in them. After a few days, media attention would either subside because the story had no legs, or be kicked into the short grass of a Cabinet Secretary inquiry, or the minister would respond to support in their party crumbling away – penning protestations of loyalty to the PM in their letter of resignation, reaffirming their innocence but leaving so that their conduct should not displace the story of the government’s successes further from public view.

Dowding et al demonstrated that such ‘grasping the revolver’ moments tended to coincide spookily with polling evidence that the minister’s alleged misdeeds were imperiling public support for the PM or the government, following which No 10 statements would communicate a certain impersonal coolness and detachment, until the minister took the hint to go. Similarly, where an investigation had been triggered and misconduct allegations even partially confirmed, the minister got no second chance to state their case – merely the intimation that it is better to resign first than be sacked. In a jostling cabinet of 23, with many more junior ministers knocking on the door, plenty of rivals are usually prepared to help No 10 brief against a black sheep minister who resists pressure to go. And replacing the culprit once gone rarely causes any lasting inconvenience to a Prime Minister.

Yet now, apparently, Cummings is sui generis. He has no rivals or possible replacements as the Svengali of Downing Street. Instead many compliant hapless senior ministers obediently rushed to tweet whatever they are told to do at the behest of the PM’s chief aide, even when ignorant of much about the situation. This is an Armando Iannucci playbook straight out of his ‘The Thick of It’, which increasingly reads as the go-to manual for anyone wanting to understand the Johnson-Cummings regime. In the process, most of the cabinet seem to have succeeded only in publicly identifying themselves as ‘meat in the room’ – a phrase made famous in another Iannucci classic, describing people whose role was just to pad out the table at big meetings and repeat the regime line in public.

Sometimes ministers do survive potential resignation scandals. When Culture Secretary in 2012, Jeremy Hunt was in charge of deciding whether to allow the News Corporation takeover of BSkyB, which Murdoch was desperate to do at the time. During this period, Adam Smith, an advisor in Hunt’s office, sent hundreds of emails to NewsCorp’s media liaison section, apparently a flagrant breach of Hunt’s quasi-judicial role in this decision. An inquiry later whitewashed Hunt, who apparently knew nothing of his aide’s activities. Against the odds, Hunt survived to become a long-lasting Health Secretary and eventual challenger to Johnson for the Tory leadership in 2019 – keeping his copybook conspicuously clean of further blots along the way.

The looming context

What makes the Cummings case so different too, is the unprecedented salience of his controversial behaviours. The UK has suffered a huge coronavirus disruption for weeks, and lost more than 60,000 people to the virus. The explanation of what has gone so catastrophically wrong so quickly, and which will affect the economy and public finances for at least a decade ahead, is still emerging. But it seems safe to say an ongoing omni-crisis has been created at many levels:

  • An initial and wholly familiar UK core executive policy disaster where the Johnson-Cummings regime from January to late March minimized the coming threat and gratuitously delayed taking action, partly while believing in Cummings’s alleged dalliance with ‘herd immunity’ solutions.
  • A public service delivery disaster that was unplanned and unintended. An austerity-vulnerable and hopelessly under-equipped NHS was ‘protected’ only by repelling many cases from hospitals, and leaving the social care sector to completely fend for itself in caring for the very elderly.
  • Structural failures built into the public services by a decade of Cameron-Clegg-May austerity, the failure to undertake any other remedial policy changes during the Brexit era (2016-19), and the withering away of local government and public health capabilities.
  • A second core executive policy fiasco in the ludicrous over-centralization of coronavirus crisis management efforts in No 10, with Cummings’s cronies and associates brought in from the private sector, which has created huge problems. The established bureaucracies in the civil service have produced substantial and reasonably well-working initiatives, like the Treasury/HMRC’s furlough scheme and aid to the self-employed, and DWP’s massive expansion of job-seeker benefits. All three massive policy initiatives were successfully implemented at speed. In acute contrast, Cummings and his Downing Street ‘irregulars’, feuding non-stop with the shattered post-Lansley/post-Hunt Department of Health, have produced only a series of feeble, late, non-working initiatives – the Dyson ventilators, the defective PPE flown in from Turkey, the ‘NHS volunteers’ recruited and left untasked in their tens of thousands, the contact-tracing app that does not work. Many No 10 efforts seem to have gone solely into dreaming up bogus, one-day ‘good news’ stories for limping ministerial briefings – a kind of ‘crisis management by think tank’ strategy.
  • A third level of core executive policy fiasco has now been added by the decision to bulldoze past the Cummings scandal, whatever the cost to public trust, and the wider mismanagement of easing lockdown messages by a health-impaired Johnson.

Normally, the costs of blatant misconduct within the core executive are exclusively political and occasionally administrative: a blow to government popularity, a short downward blip in a Whitehall department’s performance, or another dent in public trust in government. The possible costs of the Cummings scandal, and of Johnson’s decision to brazen his way through it, are again historically unique – for they could yet be counted in terms of more body bags and an even more shattered economy in the weeks ahead, and in the autumn and winter to come.


About the Author

Patrick Dunleavy is Professor of Political Science and Public Policy at the LSE, and co-author of The UK’s Changing Democracy (LSE Press, 2018) – free to download here. His next book is Maximizing the Impacts of Academic Research (co-authored), forthcoming from Macmillan in October 2020. He tweets at @PJDunleavy and @Write4Research.

All articles posted on this blog give the views of the author(s), and not the position of LSE British Politics and Policy, nor of the London School of Economics and Political Science. Featured image credit: Minute’s Silence Held For Key Workers Who Have Died During Coronavirus Pandemic by Number 10 on Flickr (BY-NC-ND).

The Cummings row undermines the sense of collective solidarity on which the lockdown relies

The so-far widespread compliance with lockdown measures is driven by social identity and collective responsibility, new data confirms. This substantiates further the argument that by defending Dominic Cummings, the government risks undermining the fight against the virus, write Jonathan Jackson, Reka Solymosi, Chris Posch, Ben Bradford, Zoe Hobson, Arabella Kyprianides, and Julia Yesberg.

Soon after the easing of lockdown measures on 13 May, Boris Johnson is said to have quipped to colleagues: ‘I’ve learnt that it’s much easier to take people’s freedoms away than give them back.’ With the government now announcing the next stages of lockdown release (e.g. more shops will reopen on June 15), officials may be wondering whether widespread fear of the virus will leave people reluctant to take advantage of their new-found freedoms and get the economy moving again.

Over the next few weeks and months, it is likely that lockdown will be replaced with contact-tracing, testing, and quarantine. Implementing this will require high levels of public support, and there is early evidence that people comply with public health measures not because of individualised fear, but because of a sense of shared identity and common fate with others. Thinking about the issue in this way turns the question of compliance on its head: it shifts the focus away from individual risk and responsibility, towards ensuring that people collectively adhere to health measures on behalf of the common good.

It is no wonder, then, that the Dominic Cummings scandal is stimulating so much heated debate. To comply with the rules is to signal to others a sense of solidarity. To go against them implies that ‘there is one rule for them and another rule for us‘, and some are wondering whether the actions of the PM’s Chief Strategist – backed by the PM himself – risks damaging public trust and compromising widespread solidarity in the fight against the deadly virus.

What does ‘the science’ say about fear and group bonds/coordination?

We have previously presented evidence that self-reported adherence to lockdown requirements was rooted not in fear of the virus, police or law, but in a widespread sense of duty and solidarity. We analysed data from the first wave of a multi-wave panel study to track the experiences, attitudes, and behaviours of 1,200 people recruited on the platform Prolific Academic – 300 living in London and 100 living in each of Edinburgh, Newcastle, Cardiff, Leeds, Liverpool, Manchester, Birmingham, Sheffield and Glasgow.

We now have second wave data, collected on 11-14 May, just on the cusp of the easing of lockdown. For enthusiasts of longitudinal research, the attrition rate was an astonishing 8%, with 92% of people taking part in wave 1 continuing in the study to wave 2.

Wave 2 fielded questions designed to reflect some of the nuances of emotional and behavioural responses to risk. We find that just under two-thirds (64%) said that they had felt worried about getting COVID-19 in the past three weeks. Of the 699 people who said they had worried, around a half (51%) said that their quality of life had been reduced either ‘not at all’ or ‘a little’ by their worry about getting COVID-19.

Criminological work has found that, while some instances of worry can be destructive and paralysing, some people and some communities have the potential and the willingness to convert worry about crime into constructive action. In our study, the majority of people said ‘yes’ (91%) when we asked ‘do you take any precautions against getting COVID-19?’. These precautions seemed to make people feel safer as a result (81% said ‘moderately’, ‘quite a bit’ and ‘very much’), but for some it also reduced their quality of life (55% said ‘moderately’, ‘quite a bit’ and ‘very much’).

Based on their worries about catching the virus, as well as the self-reported effect of their worries and precautions on quality of life, we can divide research participants into one of four groups:

  • The ‘unworried’ group (36%): those who had not worried once about catching COVID-19 over the previous three weeks;
  • The ‘seemingly confused’ group (3%): those who had worried about catching COVID-19 but did not take any precautions against the virus;
  • The ‘worried but wellbeing-unaffected’ group (31%): those who had worried, took precautions, and quality of life was not affected;
  • The ‘worried and wellbeing-affected’ group (30%): those who had worried, took precautions, and quality of life was affected.

Does ‘fear’ shape lockdown compliance?

This categorisation allows us to assess the relationship between emotional and behavioural responses to risk and lockdown compliance. To measure lockdown compliance (just before the recent easing), we asked participants ‘How often during the past week have you engaged in each of the following behaviours during the COVID-19 outbreak?’:

  • ‘socialised in person with friends or relatives whom you don’t live with?’ (74% said never, 17% rarely, 6% sometimes and 3% often or very often),
  • ‘went out for a walk, run, or cycle and spent more than a few minutes sitting somewhere to relax?’ (57% said never, 15% rarely, 15% sometimes, and 13% often or very often), and
  • ‘travelled for leisure (e.g. driven somewhere to go for a walk)?’ (82% said never, 10% rarely, 5% sometimes and 3% often or very often).

It’s not about fear of the virus or fear of the police

We find no difference in levels of lockdown compliance, comparing the ‘unworried’, the ‘seemingly confused’ and the ‘worried but wellbeing-affected’. Some people admit bending the rules, but for these groups the presence or absence of worry about catching COVID-19 does not seem part of the explanation. We find that the ‘worried but wellbeing-unaffected’ group actually had higher levels of lockdown compliance, adjusting for the many other factors in the statistical model.

As in wave 1 (data collected in late April), we find no evidence that deterrence plays a role in compliance. Unsurprisingly, given the police’s ‘enforcement as last resort’ policy centred on procedural fairness, perceptions of the likelihood that the police would step in (if people were flouting the rules) decreased between late April and early to mid-May. But while levels of lockdown compliance were also lower in wave 2 than they were in wave 1, this decrease in perceptual deterrence does not seem to explain the decrease in compliance.

Compliance is about social identity and collective responsibility rooted in legal requirement

Just like in late April, we find that social norms backed up by symbolic legal force are crucial.

Norms guide behaviour and attitudes in a number of different ways: people look to the behaviour of others to determine what is normal, beneficial, and accepted; people benefit from acting in certain ways through social approval and refrain from acting in certain ways via social disapproval; and by defining ‘who we are’, norms define social groups, and are especially powerful when people identify with the particular social group. By helping to make people accountable to each other, norms solve collective action problems.

There is also symbolic import to the fact that something is made legally required/prohibited. By turning social distancing into legal requirement, the legal system acts as an expressive agent: it sends the message to the nation that the threat is to the group rather than the individual, and that we collectively need to take the virus seriously. It also clarifies how citizens need to act to fight the pandemic.

This seems to have worked among our sample, with 94% agreeing with the statement ‘by making it a legal requirement, the government sent the message that social distancing is important to fight the pandemic’ and 92% of people agreeing with the statement ‘introducing the social distancing rules helped communicate to the public the need to do what we can to stop the pandemic from spreading’. Stronger agreements with these statements was associated with greater compliance.

Our research participants also seem to act as expressive agents in response to legal requirement. We find that 87% of people agreed with the statement ‘observing the social distancing laws shows other people in my community that I care for their safety’ and 82% of people agreed with the statement ‘following the social distancing rules helps me feel that I am part of the collective fight against the pandemic’. The law seemed to have helped frame the threat and the solution at the group rather than the individual level.

According to our multivariate analysis, when people comply with lockdown law, they signal to each other a sense of collective solidarity and shared identity in a way that works in addition to the role that social norms play. Acting in unison binds people together, especially when there is a legal requirement to coordinate at the group level against a common threat.

Routes ahead

So what does all this mean as lockdown eases, as track and tracing begins, and as the message from government is increasingly one of individual choice rather than collective solidarity? At a time when the PM’s defence of Dominic Cummings risks undermining the sense that ‘we truly are all in it together’, what are some plausible roads of travel?

Our findings suggest, at least looking back to the height of lockdown when the restrictions were rigid, that it was not fear of COVID-19 that drove adherence to the public health measures that are needed to control the virus. It was, instead, a sense of shared identity and collective responsibility, backed up by the extraordinarily popular laws underpinning lockdown.

Looking forward to the coming few weeks and months, we will need people to be collectively willing to act appropriately in terms of social distancing and hygiene, to tell officials who their contacts have been, and to isolate if they get the disease. Voluntary public compliance will be as important as it ever was, as we adjust to greater degree of relaxation. Adherence to guidelines relies on solidarity rather than fear, and we will need people to trust the government if we are to fight the virus and come out of this together.

Our data are consistent with the concerns of key members of SPI-B – that by defending Cummings and failing to communicate effectively, the government risks undermining the message of social solidarity and jeopardising the widespread support that has served the nation so well thus far. Will this occur among our sample of people? Wave 3 will be in the field next week, so watch this space!


About the Authors

Jonathan Jackson is Professor in Research Methodology and Head of the LSE Department of Methodology.

Reka Solymosi isLecturer in Quantitative Methods at the Department of Criminology at University of Manchester.

Chris Posch is a Postdoctoral Research Fellow in the Department of Methodology at LSE.

Ben Bradford is Professor of Global City Policing in the Department of Security and Crime Science at UCL.

Zoe Hobson is a Researcher in the Department of Security and Crime Science at UCL.

Arabella Kyprianides is a Research Fellow in the Department of Security and Crime Science at UCL.

Julia Yesberg is Research Fellow in the Department of Security and Crime Science at UCL.

All articles posted on this blog give the views of the author(s), and not the position of LSE British Politics and Policy, nor of the London School of Economics and Political Science. Featured image credit: by United Nations COVID-19 Response Unsplash.

Estimating how many Britons have already had COVID-19 using self-reported data

Using a random sample survey and self-reported data on infection and symptoms, Patrick Sturgis and Jouni Kuha find a symptomatic infection rate in the general population of Great Britain of approximately 10% by the end of April.

Any attempt to estimate the population infection rate must deal with two key challenges: how to measure infection and who to make the measurements on. Available methods for estimating the population infection rate have different balances of strengths and weaknesses with regard to these matters of measurement and sampling, respectively.

The most discussed approach to measuring population prevalence are the antigen tests carried out daily by Public Health England. These show that (as of 7/05/20) 1,139,626 people had been tested and 206,715 had tested positive – an infection rate of 18.1%. However, because these tests are targeted at groups likely to have contracted the virus, they will tend to over-estimate the population infection rate.

Antigen tests are also of limited value for estimating the population infection rate because they only identify people who are currently infected and will, therefore, miss the small but growing section of the population who have recovered from an earlier infection. For example, an ONS survey of 10,000 households found an infection rate of 0.27% between 27 April and 10 May, which is likely considerably lower than the proportion of people who have ever been infected.

Here, we assess whether random sample surveys using respondent self-reports of infection and symptoms can help get a better picture of the true rate of symptomatic infection in the population. To do this, we analyse data from the Wellcome Trust UK Monitor survey which was carried out on the NatCen probability-based panel between 30 March and 26 April.

The key advantage of probability surveys for estimating the population infection rate is that, in theory at least, the random selection of sample members from the target population will give approximately accurate estimates of population characteristics.

The corresponding disadvantage of surveys is that measuring the infection status of respondents is challenging. The best way to do this would be to administer both viral and antibody tests and, indeed, this is what the government is proposing to do when an accurate antibody test is available. In the meantime, we must rely on respondents telling us whether they think they have had the coronavirus and its associated symptoms and this kind of subjective data is clearly subject to a range of potential errors.

The 2,400 respondents in the Wellcome Trust Monitor sample were asked to give a probabilistic assessment of whether they have had the virus and ten of the most common symptoms of COVID-19. As shown in the Figure below, 0.9% said they had ‘definitely’ had the virus, 12% ‘probably’ had the virus, 43% ‘probably not’ had the virus and 44% had ‘definitely not’ had the virus (0.7% said that they did not know).

Combining those who say they have ‘definitely’ or ‘probably’ had coronavirus, gives a crude estimate of the symptomatic infection rate of 13%. However, some of those who think they have had the virus will not actually have been infected (false positives), while others are likely to have been infected but say they have not (false negatives).

One way of assessing the extent of these reporting errors is to cross-check self-reported infection status with reported symptoms. All respondents were asked whether they had experienced any of ten common COVID-19 symptoms since the start of February.

The figure below compares the reported rate of each symptom between those who say they have had coronavirus and those who say they have not. By taking the ratios of reported symptoms for these two groups, we find that the most distinctive symptoms are high temperature (ratio = 10), new cough (7.2), loss of taste/smell (8.5), and shortness of breath (5.4). This pattern is consistent with the list of typical symptoms reported by the World Health Organisation and the Centers for Disease Control.

Of those who said they have probably/definitely had the virus, 22% reported having had none of the five most distinctive symptoms, so these seem likely to be false positives. The high apparent rate of false-positives should be a concern for policy-makers, as many in this group are likely to (falsely) believe they are immune from infection and so cannot pass the virus on to others.

On the other hand, 3% of those who said they probably/definitely have not had the virus reported three or more of the five most distinctive symptoms. If we remove the false-positives from the group who report having been infected and incorporate the apparent false-negatives, we obtain an overall symptomatic infection rate of 11%.

A different way of estimating the symptomatic infection rate is to use a statistical model called Latent Class Analysis (LCA). This is a method for detecting groups or clusters that underlie distinctive patterns of response. We applied LCA to the symptoms data, focusing on a model which identifies four clusters of symptoms, as shown in the table below.

The numbers in the cells of the table are the probabilities for the people in each class of reporting each symptom and the numbers in bold are the proportion of the sample in each class. So, for example, 20% of people are in Class 3 and people in Class 3 have a probability of 0.34 of reporting a high temperature.

The pattern of probabilities for each class enables us to identify different symptom clusters. We can label Class 1 (42% of the population) as the non-symptomatic group as they have near-zero probabilities for every symptom. Class 2 (36% of the population) display the symptoms of the common cold but not the main symptoms of COVID-19. People in Class 4 (2% of the population) have a high probability of reporting all ten symptoms, so they show strong signs of having been infected.

Class 3 presents a less clear picture. This group has moderate probabilities of the core symptoms and higher probabilities of the others. So, Class 3 seems to contain a mixture of the infected and the non-infected, but we cannot distinguish between them based on these self-reported symptoms alone. The difficulty of identifying a single distinct cluster of COVID-19 symptoms is in line with research that suggests symptom patterns are quite variable across individuals.

One way to separate out the COVID-19 group in Class 3 is to combine class membership with self-reported likelihood of having been infected. Of the people assigned to Class 3, 34% said they have definitely or probably had COVID-19 (compared to 4%, 9% and 72% of those in classes 1, 2, and 4, respectively). If we combine these respondents with everyone in class 4, we get an estimated symptomatic infection rate of around 9% – similar to the 11% we found using a more ad hoc approach.

Finally, because the Wellcome Trust Monitor survey contains lots of other information about respondents, we can compare infection rates across demographic groups. Interestingly, we find no statistically significant difference in the symptomatic infection rate between men and women, or between people with different levels of income, even though substantial gradients have been found between these groups in COVID-19 fatalities by the Office for National Statistics. It would seem, then, that susceptibility to infection by coronavirus is more evenly distributed across demographic groups than are fatalities.

One exception to this pattern is age.  Here we find that older people are the least likely to report symptomatic infection, with only 5% of the over 70s falling into this group, compared to 15% of those aged 18-29. This suggests that, while older people are most likely to die from a coronavirus infection, they are less likely to become infected in the first place. This seems likely to be a result of older people following government advice to shield themselves from infection by staying at home and avoiding social contact.

The estimates of infection we have presented here should be treated with caution. Their key strength is that they are based on data from a high-quality random sample, their corresponding weakness is that respondent self-reports of infection and symptoms are prone to errors of both omission and commission. And, of course, self-reports can only ever be informative about the symptomatic infection rate, so will miss anyone who has been infected but displayed no symptoms.

We will not know the true rate of COVID-19 infection in the population until the results of the government’s large-scale random testing programme are available. Until then, these kinds of estimates can help to give a sense of the current rate of symptomatic infection in the general population.


Note: the above was first published on LSE COVID-19 and represents the views of the authors only. Featured image credit: by United Nations COVID-19 Response on Unsplash.

About the Authors

Patrick Sturgis is Professor of Quantitative Social Science at the LSE Department of Methodology.



Jouni Kuha is Professor of Social Statistics at the LSE Department of Methodology.




Powered by WordPress | Designed by: index backlink | Thanks to insanity workout, car insurance and cyber security