King Canute has had a bad press. The monarch sat on the beach on his throne with the deliberate intention of demonstrating to his courtiers that he could not stop the waves from coming in.
But in popular thinking, he is the deranged king who believed he could control the sea.
In this spirit, step forward two modern Queen Canutes, Nicola Sturgeon and Jacinda Ardern of New Zealand. Both appear to think they can eliminate Covid-19.
Our own Matt Hancock is showing dangerous signs of succumbing to this syndrome.
On a more sober note, it is certainly true that new cases are rising not just in the UK but across Europe – except in Sweden.
But there are major differences between this wave of infections and that experienced in March and April.
The key question is not really how many people might get Covid-19. It is how many might die as a result. In the jargon, this is the case fatality rate (CFR), the probability of dying from the disease if you catch it.
As ever, the Oxford Centre for Evidence Based Medicine is a font of wisdom.
A month ago, the Oxford researchers showed that in the UK the CFR had fallen from six per cent in the summer to just 1.5 per cent.
This could of course be due in whole or in part to the fact that the majority of infections are now in the young, who are at essentially no risk at all themselves.
But the Oxford group showed last week that something even more important is going on. They analysed data from Germany, which is more detailed and specific in terms of ages than in the UK. The results are striking.
In the 60 to 79 age group, in the March and April period the CFR was nine per cent. By July and August this had fallen to just two per cent.
In the very vulnerable group of the over 80s, in March and April the CFR was a frightening 29 per cent. By July and August this was down to 11 per cent.
So deaths remain very low not just because it is mainly the young now catching Covid-19 or because the elderly are shielding. Both of these are true.
More fundamentally, fatality rates amongst those who actually have the virus have fallen sharply. Treatment has improved. Social distancing means less strong doses are being caught. Whatever the reason, CFR is down.
Government advisors and health care professionals appear not to have taken this on board. They speak and act as if a second wave will be as lethal as the first.
Some might think they are as mad as the King Canute of popular legend. It is more likely that they are simply suffering from confirmation bias.
This is the tendency to search for, interpret, and recall information in a way that confirms or supports one’s prior beliefs or values.
They have long thought that a second wave would be devastating. The emerging evidence should not be allowed to get in the way of this. Their irrational behaviour is costing us all dearly.
As published in City AM Wednesday 9th September 2020
Image: Covid Testing by via GiiPe via Wikimedia
Decisions, whether by individuals, companies or governments, are often made with imperfect and incomplete information.
This is so obvious as to hardly seem worth stating. But for well over a century economic theory assumed that decisions were made with complete information. Economists knew full well that this was not always the case. The problem was that building a formal model without this assumption was a formidable task.
It was solved in the late 1960s by the American economist George Akerlof. His paper, the enigmatically titled “Market for Lemons”, won him the Nobel Prize. The article is probably the most cited in the entire history of economics.
The coronavirus crisis provides an almost textbook example of how different amounts and quality of information affect the decisions which are made.
In March, initial data from China suggested the death rate was between three and four per cent, a real pandemic. The chaos in the health system in northern Italy was a prominent feature in the media.
Little wonder that, against Boris Johnson’s libertarian instincts, a compulsory lockdown was introduced. Little wonder that people reacted with fear to the virus.
Six months later, whilst there is still much to discover about Covid, far more is now known.
It is well established that almost everyone who has died had some serious pre-existing health condition. Perhaps even more importantly, virtually no-one under the age of 40, or even 50, has died.
Many people have reacted to this information in an entirely rational way.
Those who perceive themselves as being either elderly or having a health problem continue to shield. The young see they are at essentially no risk at all and start to behave as they did before the virus appeared.
In Scotland, for example, it seems that around half the new infections are in mere whippersnappers under the age of 25. More generally across the UK, although the reported number of new cases is rising, hospital admissions remain very low because new infections are concentrated amongst the young.
These behavioural responses have arisen entirely spontaneously as the quality and quantity of information has improved. Many have weighed the costs and benefits in the light of this information and have made rational decisions.
The idea that people would change their behaviour during a pandemic was of course almost entirely absent from the mathematical models used by epidemiologists. It is this which led to absurd predictions both of half a million deaths in the UK in the first place and of fears of a second wave with even more deaths.
Some individuals who are at little risk remain paralysed by fear. They are still relying on outdated information and so behave in non-rational ways.
But it is governments which right now are the least rational of all. Nicola Sturgeon wants to eradicate Covid when even New Zealand, a country which can seal its borders, could not.
And governments persist in incurring the costs of lockdown, when people have shown that they are capable of behaving in sensible and rational ways without government interference.
As published in City AM Wednesday 9th September 2020
Image: Closed Due to Covid-19 Sign by via Pixabay
Office workers continue to display reluctance to return to their workplaces, despite encouragement from the government for them to head back.
The immediate consequences for the service jobs in cities which depend on people commuting into the office are apparent, hence the government drive. But is office work a good thing for the workers themselves?
An important concept in economics is that of revealed preference. Economists believe that preferences are revealed not in surveys, but by the decisions which people actually make.
So could it be that the Covid crisis has given office workers the chance to reveal their true preferences in terms of work-life balance?
This is almost certainly true for the senior staff who make the decisions. They will typically live in large houses in agreeable surroundings, with space to dedicate a room as an office.
It is much less obviously the case for many younger staff. Working with your laptop on your knees in a bedsit, unable to socialise with colleagues, may lead you to prefer the commute instead. But there are constraints on being able to reveal this preference, such as potentially annoying your boss who enjoys working from home.
Furloughed workers, meanwhile, have merely revealed a preference to be paid a large part of their regular pay and not work at all. This is particularly true of public sector workers, who receive their full salary regardless. They have no incentive to change their work patterns.
An obvious incentive for companies is that, if working from home persists, they can save on office costs. They may even be able to adjust salaries downwards, especially for staff who commuted long distance from cheaper locations.
In the very short term, there will be little, if any, loss of productivity to offset this. Most office jobs consist of performing routine, well understood tasks. Within the discipline of an established framework, some people may even be able to do their jobs more efficiently at home, encountering fewer distractions.
This is true even for jobs which require analytical skills. The business model of a number of large consultancies, for example, can be thought of as follows.
The company hires bright young graduates, who come equipped with a stock of the latest ideas — what economists call human capital. The short-term pressures in the big consultancies to make money are so intense that they have little chance to refresh this during their careers. Essentially, they run down their human capital. By the time they make partner or director in their forties, they are operating on half-remembered ideas from graduate courses.
This is why the company requires a steady flow of new recruits, to refresh the business. The challenge with remote working over time will of course be to not only integrate new young staff into the organisation, but to ensure that their ideas percolate.
Productivity will grow more slowly over time if extensive homeworking persists. In part, this will be due to factors internal to the firm. The tacit knowledge and creativity sparked by informal exchanges will be lost.
The real loss, though, is through factors external to any individual company. A huge amount of evidence shows that the higher the density of employment in an area, the higher its productivity.
The drive to get people back into their offices, therefore, is about far more than saving city cafes and restaurants. The government should incentivise firms to get staff back to work, in order to avoid lost productivity — for the entire economy.
As published in City AM Wednesday 2nd September 2020
Image: London Underground by via Pikist
Did Sweden get it right in its response to Covid? There is increasing interest in this question.
Contrary to widespread belief, the Swedes did introduce a few legally enforceable restrictions on behaviour. For example, public gatherings of more than 50 people were forbidden in March. Private ones were exempt from the ban.
But, overall, compared to other European countries, Sweden had effectively no lockdown.
At the start of the outbreak in March, the epidemiologists in the UK were quick off the mark with their dire predictions.
Professor Ferguson and the Imperial College team, in a highly influential paper of 16 March, argued that without any policy changes “we predict approximately 510,000 deaths in Great Britain”.
They also suggested that the peak of mortality would take place after approximately three months.
The peak of mortality was reached after just three weeks, not three months — on 8 April in fact.
A model based on the Imperial one was calibrated on Swedish data in April. The outcome was a prediction of some 80,000 deaths.
To date, the number of deaths in Sweden due to Covid is less than 6,000.
The estimates which emerge from standard models of epidemiology clearly cannot be relied upon.
With Rickard Nyman, a young Swedish computer scientist currently based at UCL, I have compared the trajectories of deaths in Sweden and the UK to provide an estimate of how many lives lockdown really did save.
The first death in Sweden was registered on 11 March. On that day, there were seven deaths in England and Wales (EW). But our population is nearly seven times that of Sweden, so the death series in the two countries start at the same level, adjusting for population size.
The peak death rate was observed in Sweden on exactly the same day as in EW, 8 April.
Further, the paths followed between 11 March and 8 April are almost exactly the same. For example, deaths reached half their eventual peak value on 1 April in EW and on 31 March in Sweden.
So lockdown had no impact on the build up to the peak.
There is no doubt, however, that the number of deaths fell more sharply here than in Sweden after the maximum level was reached.
Deaths had dropped in EW to half the 8 April peak by the end of April and to only a quarter of the peak by the middle of May. In contrast, in Sweden the “half peak” value was only seen on 12 May, and the quarter by the middle of June.
These differences in the trajectories of the death numbers after their peak in early April tell us how many lives were saved in the UK as a result of lockdown.
It is not 500,000. It is just 20,000, a tiny fraction of the Imperial figure.
So, yes, lockdown did save lives. But the massive economic and social costs, let alone the adverse effect of lockdown on people suffering from other illnesses such as cancer, can hardly be said to justify the saving.
As published in City AM Wednesday 26th August 2020
Image: NHS Heroes by Jernej Furman via Flickr CC BY-SA 2.0
There seems to be a fundamental problem with quangos. Hardly a day seems to go by without some new story of incompetence and mismanagement emerging.
Public Health England (PHE) is at least going to be put out of its misery by health secretary Matt Hancock, and replaced with a new agency specifically focused on pandemics.
An anonymous government minister is attributed as saying early in the crisis “we didn’t really know what PHE actually did, except from time to time they would put their head over the parapet and try and ban something like Coco Pops”.
PHE is by no means the first government body to have a slightly dubious record of effectiveness. The Highways Agency, set up by John Major’s government, was of such legendary uselessness that David Cameron renationalised it just before the 2015 election. It rose again as a “government owned company”, Highways England. As such, it has been responsible for the funereally slow works which install so-called smart motorways.
Network Rail, meanwhile, seems to have the Midas touch — for its contractors and consultants, that is, as it fails to complete Crossrail and the costs of HS2 soar out of control.
And the shambles at Ofqual over A-levels is beyond parody.
The simplest way from the outset would have been to accept, as a unique one-off event, the predicted grades of teachers. That is all that the quango need have done. The problem of who to admit would have been devolved to the universities, which are the bodies with the ultimate financial interest in these matters.
Now, the situation is even more complicated — places have already been assigned based on algorithm-assigned grades before the U-turn, and universities have mere days to work out how they will teach thousands of additional students, rather than months to work out a reasonable system.
In the case of each individual quango, there will be reasons as to why those specific problems occurred. But there are so many examples that there must be some more general principles at work.
As usual, economics can help. There is a long history of work around the concept of what has come to be known as “public choice” theory. The American economist James Buchanan did more than anyone to establish it as a standard and wide-ranging tool of economics, and he was awarded the Nobel Prize for this as long ago as 1986.
Too often in public discourse, a contrast is made between self-interested companies and government bureaucrats working in a seemingly selfless way for the public interest. Indeed, some of the bodies pilloried above were set up with the explicit aim of taking politics out of decision making in the relevant areas.
In essence, Buchanan believed that bureaucrats and politicians behave in the same self-interested way as everyone else. This does not mean that other motives are absent, but that rational self-interest is an important driver of behaviour.
We can see this clearly in the way the teaching unions have reacted to the Covid crisis. Many teachers are dedicated and committed to their pupils. Equally, however, there is a significant minority who do not act in any way as selfless professionals. Their revealed preferences have been to draw full pay and do little to no work, regardless of the consequence for children.
Detaching public policy decisions from direct political control has been very fashionable for the past few decades. The performance of the UK’s quangos shows that it is high time for a change of mind.
As published in City AM Wednesday 19th August 2020
Image: NHS Heroes by Bill Nicholls via Geograph CC BY-SA 2.0
In an otherwise depressing week, two pieces of very good news emerged from India.
In Mumbai, blood tests conducted by the city authorities on 6,936 randomly selected people found that some 40 per cent had coronavirus antibodies. Just 6,000 deaths have been reported so far in a city of 20 million.
A similar exercise in Delhi (population 29 million) found that around a quarter had had the virus. Only 3,300 deaths have been attributed to Covid-19.
Indian experts are clear about what these figures mean. Asymptomatic infections are a high proportion of the total. Further, the virus death rate is “very low”, to quote the Mumbai study.
These studies fit in with a mounting body of evidence that the virus is nowhere near as deadly as first thought. Even the cautious Centers for Disease Control and Prevention (CDC) in America has revised its estimate of the mortality rate down to 0.26 per cent — comparable to a bad influenza year.
Casting our minds back to the middle of March, the picture seemed completely different. Data from Wuhan suggested that the death rate was between three and four per cent. Our television screens were full of images of the health system in northern Italy being overwhelmed.
Given this particular set of information, it was perfectly understandable for the government to impose a lockdown. Behaviour was already changing rapidly in a spontaneous fashion as people voluntarily limited their social contact, but it would still have been very difficult at the time for the government not to take the action it did.
One of the most balanced set of commentaries throughout the whole crisis has been provided by Professor Carl Heneghan and his team at the Oxford University Centre for Evidence Based Medicine.
It was Heneghan, for example, who recently exposed the fact that Public Health England (PHE) did not let anyone recover from Covid-19 — at least not on paper. If you had recovered from the disease and were subsequently killed in a car crash, say, according to PHE data you died from the virus.
The Centre’s work shows that since the week of 14 June, total deaths in England and Wales have been running below their five-year average — despite an apparent upsurge in Covid cases.
Heneghan’s latest research even casts serious doubt on whether the true number of new cases has in fact risen.
On a technical point, the reported number of new cases follows a fairly shallow upward linear trend — not the exponential rise which would characterise a genuine second wave.
His key point, however, is that the number of tests has increased. Once this is taken into account, the rise in cases disappears completely. On the basis of personal knowledge, I can confirm that this seems to have happened in Rochdale in Greater Manchester. The Council installed walk-in test facilities in the town centre. Shortly afterwards, the number of reported new cases rose.
Covid-19 can be highly unpleasant and it retains the capacity to kill, so it is sensible to take precautions to avoid getting it. But the massive costs of lockdown — both economic and social — can no longer be justified.
As published in City AM Wednesday 5th August 2020
Image: Post Lockdown by Bastian Greshake Tzovaras via Flickr CC BY-SA 2.0
A key concept in modern economics is, to use the jargon term, rational expectations.
The idea has dominated orthodox macroeconomics over the past 30 years. Not all economists have been persuaded of its merits by any means, but nevertheless, its influence has extended far beyond academia, into finance ministries and central banks around the world.
The basic idea is simple, even though the maths of the macroeconomic models which use rational expectations can rapidly become hair-raising.
When forming a view about the future, an individual chooses the model which best describes how the economy works. It is then simply a matter of running the model and using the values which it outputs as your expectations.
An obvious criticism seems to be that economic forecasts are very often wrong, but this is easily handled by the rational expectations enthusiasts. Each individual forecast in a sequence of months or years can be wrong. The key thing is that the errors over time cancel out. On average over time, the forecasts are correct.
The idea is not as absurd as it may appear to the layperson. For example, the Federal Reserve Bank of Philadelphia has published the Survey of Professional Forecasters since 1968.
This does what it describes on the label. It takes a wide range of forecasts by academic, commercial, financial and governmental bodies. Information on the average forecast for, say, GDP growth one year ahead is published, as well as details of the spread around the average.
Remarkably — and exactly in line with rational expectations — comparing the predictions with the actual growth over many years, the errors do indeed balance out. Spectacular errors have been made for individual years, but the over and under-predictions cancel out over time.
A more telling attack is that economists themselves do not seem to agree on what constitutes the correct model of the macroeconomy. Different groups have different models.
The standard defence is that the best model will eventually prove its superiority and will drive the others out of existence. But the problem here is that this has just not happened.
The concept of rational expectations can be applied directly to the predictions of the epidemiological models. These purport to describe how a virus spreads. So to form a view about the future, make a set of assumptions about the key inputs, and use the forecasts generated by the model.
It should be much easier in epidemiology for the best model to eliminate its competitors than it is in economics. Economics has a wide range of variables to predict, such as inflation, GDP, unemployment, public borrowing, interest rates. The focus of epidemic predictions is much narrower and their models are in general mathematically simpler than those of economics.
The Covid pandemic set up a competition between epidemiological modelling groups of fierce Darwinian intensity. The efforts of many years of academic debate have been concentrated into a handful of months.
But no group seems to have admitted yet that its model is not up to scratch — and huge differences in forecasts persist.
As published in City AM Wednesday 29th July 2020
Image: Opposites via Pxfuel
Another day, another lurid, headline-grabbing number of deaths to expect from Covid-19.
This time, it was a study from the Academy of Medical Sciences. A second wave, we were warned, could kill 120,000 this winter in hospitals alone.
To be fair, this study was a projection rather than a forecast. A forecast is what is thought to be the most likely outcome. A projection looks at what could happen under a particular set of conditions.
The Academy essentially assumed that behaviour would revert to the same as it was before the crisis, with people acting as though the virus had never appeared.
The researchers made their assumptions clear. But entirely predictably, the media seized on the 120,000 figure for deaths. The qualifications made around that number faded into the background.
The plain fact is that the assumptions of the report were wholly implausible. Even if lockdown were lifted completely, behaviour is not going to immediately revert to exactly what it was before the Covid crisis. Will people shake hands? Of course not. The concept of a typical way of life has irrevocably changed.
The entire history of the world can be cited in evidence of this proposition. In the face of an epidemic, people alter their behaviour. They do not need to be told to do so by governments.
Of course, how much behaviours will change is ultimately a matter of judgement. But it is one where the social sciences, including economics, can make a valuable contribution. Epidemiology is currently too important a subject to be left in the hands of the epidemiologists.
It is something of a mystery why the numbers churned out by various epidemiologists retain any credibility. Their models in general take no account of behavioural change when a pandemic occurs.
In March, we all remember the Imperial College study claiming that without lockdown there would be 500,000 deaths in the UK. This was ludicrous — and economists such as Gerard Lyons and I quickly attacked it.
In April, a very similar model was run on Swedish data. It claimed that there would be 40,000 deaths by the beginning of July. In fact, even with no lockdown, there were only some 5,000, the majority of which took place in care homes.
Forecasts such as these make economic forecasts seem like Platonic ideals of precision.
This is far from a mere spat between scientific disciplines. Poor models and their resulting projections can lead to poor policy decisions. They generate a wholly unwarranted climate of fear among the population. This reduces economic activity, meaning less money available to fund health services, and greater poverty with its associated illnesses such as depression.
The number of deaths in England and Wales peaked on 8 April. The average time from infection to death tells us that the number of cases peaked in the week 18–25 March. Lockdown was only introduced on the evening of 23 March. This shows that behaviour had already altered dramatically.
No more credence should be given to epidemiological projections which do not assume behavioural change.
As published in City AM Wednesday 22nd July 2020
Image: UK Government Coronavirus by Gustave iii via Wikimedia CC BY-SA 4.0
The Prime Minister is now demanding that offices reopen to revive economic activity in the centres of towns and cities.
But there is not much sign of a return to work.
The preferences of the workforce are an important factor in the very slow pace of return. Fears expressed about the safety of public transport may or may not be genuine, but it is certainly true that many prefer to avoid the time spent commuting and enjoy the extra leisure time this brings.
But why do offices cluster together in urban centres anyway?
It is easy to see that in the old days industries such as steel and coal clustered geographically. One was a key supplier of the other. Being near at hand minimised transport costs.
Today’s offices span a wide range of diverse industries, from consulting to law to oil companies. The reasons why they locate in close proximity are more subtle.
The views of economists on this are still shaped by the writings of Alfred Marshall. He established the faculty of economics at Cambridge in 1903 and was then probably the world’s leading economist.
Marshall described the tendency of businesses to cluster near each other as “agglomeration”. He gave three key reasons why this colocation is observed.
In addition to the savings on the costs of transporting the materials needed in industrial processes, Marshall developed a theory of labour market pooling, in which firms located near one another can share labour.
Further, he believed that “intellectual spillovers” were important. Firms locate near each other in order to learn and speed up the process of innovation. Think of Silicon Valley, formed nearly a century after Marshall wrote.
A large number of detailed studies in recent decades confirm that these are not just mere theories. They have strong empirical support. The Harvard economist Ed Glaeser, for example calculated that in the US in the 2000s each of Marshall’s three reasons were of roughly equal importance.
There have been very distinct benefits to agglomeration. Throughout the developed world, the greater the density of employment in an area, the higher is its productivity. Head offices contain more highly skilled staff and so will be more productive than the average. But in city centres, their productivity is even higher than their skill levels suggest they should be.
Has Covid-19 changed all this? Or more specifically, has the crisis enabled people to see that new technology could overturn two centuries of experience in urban centres in industrialised countries?
Certainly, tech platforms such as LinkedIn offer the potential for efficient hiring of relevant skills and for employees to discover opportunities through their networks. But new recruits need to be integrated. And younger people probably still need a combination of social and remote interaction to develop their own professional networks.
It is less clear that remote working can encourage innovation in the same way. Much of the informal contacts needed for this cannot be captured by video conferences.
Yes, there will be an increase in working from home. But Marshall’s insights into the benefits of agglomeration still hold true.
As published in City AM Wednesday 15th July 2020
Radical leaders such as Jacinda Ardern in New Zealand and Nicola Sturgeon in Scotland have gained plaudits through their relentless focus on eliminating Covid-19.
But this comes at an obvious economic cost. Tourism is some 15 per cent of New Zealand’s GDP, and major destinations such as Queenstown in the Southern Alps have been devastated. Here in the UK north of the border, Scottish businesses are increasingly frantic about the economic damage done by the rigorous lockdown.
Health experts and epidemiologists have not helped. They have remained firmly enclosed in their own silos of expertise, unable or unwilling to see the broader picture.
But good policy is not made in a vacuum. A key concept in economics is that of trade-offs.
An obvious health related example is road accidents. As a society, we trade off the 1,800 deaths and 250,000 injuries a year in road accidents against the benefits of using vehicles.
Trade-offs between alternatives have always been central to our economic policy and political debate. Lockdown is no different to any other policy. It has both benefits and costs.
The benefits of lockdown were never in doubt: the policy was intended to save lives. But there is now increasing awareness of the potential economic costs. Ironically, some of those most shrill in favour of lockdown are now crying out for jobs to be saved.
David Miles, an Imperial College economist and former member of the Monetary Policy Committee, along with two medical specialists published last week a valuable assessment of the overall costs and benefits of lockdown.
Miles argues that we must normalise how we view Covid-19. Its costs and risks are comparable to other health problems, such as cancer, heart problems and diabetes, where governments have made resource decisions for decades.
The lockdown is a public health policy, and Miles and his colleagues value its impact using the standard tools developed by the National Institute for Health and Care Excellence to guide healthcare decisions in the UK public health system.
A key concept in this is QALYs — quality of life adjusted years. Essentially, the benefits of any policy in terms of QALYs are compared with its costs.
Economists have developed a broad consensus on the value of saving a single quality of life adjusted year. Macabre though it may seem, some metric like this is necessary in order to have any meaningful assessment of the costs and benefits of different decisions.
Miles’ conclusions are stark. The estimate of lives saved by lockdown in his analysis is deliberately chosen to be the one at the very top end of the range of such estimates. This way, his team cannot be said to be underestimating the benefits of lockdown.
Even so, in the authors’ own words: “we find that having extended the lockdown for as long as three months consistently generates costs that are greater — and often dramatically greater — than likely benefits”.
Gerard Lyons and I warned in early April that, while it was necessary to introduce lockdown, it needed to be relaxed rather swiftly because of the costs it would entail. The work of Miles and his colleagues confirms, in impressive detail, this view.