The Extraordinary Madness of Crowds

“Not allowed to travel. Police breaking up public gatherings. Empty shelves in shops. Have to queue for everything. I wonder why my parents keep acting like they’ve seen it all before?”

"Men, it has been well said, think in herds; it will be seen that they go mad in herds, while they only recover their senses slowly, and one by one." Memoirs of Extraordinary Popular Delusions and the Madness of Crowds, Charles Mackay, 1852


Interpretation and Opinion

Graph Update to week 25

2020-07-14 16:19

Selected graphs below updated to week 27.

Where? comparing full-year 2018 with year-to-date 2020

NHS England, Trend

Age Profile

Working Age and Younger.

2020-04-23 14:31

To answer a question "…there is an uptick in the 15-44 group; I wonder what that line would look like scaled up?"


Graph Update to week 15

2020-04-21 11:44

Graphs below updated to week 15.

Week 15 all deaths 18,516; respiratory deaths 1,810; COVID-19 6,213. The five-year average for week 15 is 10,520 so on the face of it an excess of 8,000 deaths. Ignoring the extent of any double-counting of the respiratory and COVID-19 categories that 8,000 could be viewed as being the sum of those two categories. The respiratory death numbers are not unusual for a late peak in a mild winter, so taken alone, they might seem just the effect of WuFlu. Year to date there have been 184,960 deaths, 10,000 more than the five-year average. Which, as it happens, is near-enough the number of COVID-19 deaths to-date (10,335). So, up to week 15, this is a very average year, apart from 10,000 (5.6%) extra deaths from COVID-19.

Date of Report v. Actual Date

2020-04-21 11:42

Graph update to week 14

2020-04-14 11:21

Graphs below updated to week 14.

COVID-19 deaths in week 14 3,475, but Respiratory Deaths are 2,106, so at least 1,369 are 'with' not 'of'.

Week 14 2020 is very high for this late in the flu season, yet the weekly total of 16,387 is only 150 more than the peak of 16,237 seen in week 2 of 2015

Year to date total 166,444 this is higher than the 10 year average at week 14 which is 161,330.

Deaths by age show exceptional rise in middle-age deaths 45–64. but young (under 45) not affected.

This is bigger than your normal flu, and it has hit more than just the very old. How many with no underlying problem is not clear.

Does this justify total lockdown?

What should we be looking for?

As expected an increase in week 13, but nothing more than a late peaking normal flu season. The PHE figure for 6th April has daily deaths at 439, down from a peak two days earlier of 708. Looking at the Respiratory Disease figures that big jump in Covid-19 to 509 hasn't shown as a corresponding big jump in the respiratory deaths. so perhaps if it wasn't WuFlu it would have been some other flu.

But… says ONS:

Covid-19 deaths aren't a subset of respiratory deaths, but neither are they all additional to it. Covid-19 deaths that are classed as respiratory deaths will be included in the respiratory death numbers. This would, you'd presume, be a majority. But there will be others that won't count as respiratory deaths, because ONS is counting as a C-19 death all deaths of anyone who had, or was supected of having, C-19, even if they died by, say, falling off a ladder.

So those COVID-19 deaths are with not of Look first and foremost to that overall total, is it any worse than has gone unremarked in the previous decade?

What should we be looking for?

I expect there will be an uptick in week thirteen, but even if the all-deaths figure kicks up by a couple of thousand it will still be no more than a bad flu year. Week 13 is something of an anomaly in that if you look at the highest weekly figure in the period 2010—19 that week is quite low. A statistical quirk, probably caused by the date of Easter and hence the drop in registrations in bank-holiday weeks.

The ONS weekly provisional figures do not show causes other than Respiratory diseases. Analysing death certificates and getting the classification right cannot be done accurately and in a hurry. Showing Respiratory Disease is an exception, justified because *every winter* it is that cause that pushes up the death rate.

What should we be looking for?

In answer to a question from PJF:

What should we be looking for that would falsify “nothing much”?

My attitude to statistics is that if I can’t find a way to present it in a way that is obvious using a simple honest graph then there probably isn’t anything significant. I try to distance myself from colleagues who spatter their papers with talk of confidence intervals and such like. I am a numbers man, always have been, but I take them as a guide only. Step back, half-close your eyes, does it have the look and feel of truthiness?

This is the basic All Deaths by week of registration chart. I have calculated the average for the ten complete years 2010–2019 and also the highest and lowest value for the equivalent week in any of those years. So if the 2020 figure for a week is higher than the plot of that thin red line then it is true that deaths in that week will be the highest in ten years. But each year has its peaks, and significant troughs that are very much influenced by Bank Holiday dates. So I have added a trend line based on the max and min values but uplifted to generally span the peaks. Will there be extra deaths this year or will it be a case of WuFlu rather than some other flu? A late surge in what to end-March was a relatively low year for seasonal deaths? A week or two just toping that trend line as has happened before? If we see something way above that trend, consistently week after week, then that would be something of note.

Look at that death-by-age-group graph below. The young deaths are pretty constant, the only down dip is because the registry offices (and all the services that will need a death certificate) are closed for several days at year end. The 85+ deaths are very seasonal and you can see year on year it is variable —’noisy’. So when that current year line bursts through that noisy band of normal for this time of year, especially for the not so old. then it might be ‘something significant’.

But significant of what? A particular strain of ‘flu? A health system overload? A misdirected overload, to be discovered when deaths from all those conditions which were not treated in a timely manner because the GP was not answering calls, or worsened because patients were discharged too soon in order to free up capacity? Or the very hard to quantify costs of lockdown: economic and social ruin? When the dust settles I suspect a lot of people, especially politicians, will claim credit for things not being as bad as hyped while being very quiet about some very real and lasting damage.

It has been a mild season so far, perhaps there is a pool of elderly at risk right now? Is This What's Behind Italy's Outrageous 10% Mortality Rate From COVID-19? is an interesting speculation.

Weekly provisional figures on deaths registered in England and Wales

I created this series of charts as a response to a blog post by Hector Drummond looking at the ONS statistics for Deaths in England and Wales. He wanted them in line-graph form overlayed the last few years together. So I did just that.

Weekly provisional figures on deaths registered in England and Wales

Five years 2015–2019 plus the average for the previous five years, and 2020 to week 12. The effects of Bank Holidays can be seen in the big dip over Christmas and New Year, the movable feast of Easter, two May Holidays, and one at end August. Deaths always peak in Winter but the rise (Excess winter deaths) varies a lot depending on the severity of the 'Flu Season.

Respiratory Diseases

This chart follows the same form as the all-deaths chart but shows only "Deaths where the underlying cause was respiratory disease (ICD-10 J00-J99)". From week 11 of 2020 this includes "a separate breakdown of the numbers of deaths involving COVID-19. That is, where COVID-19 or suspected COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions".

Weekly Deaths by Age

A set of three charts each showing the full years 2010–2019 and 2020 to date. The vertical scale is the same for each so it can be seen how deaths age 85+ are much higher relative to those for the age 75–84, and 65–74. Note also how the deaths of the very old are not only very seasonal but vary significantly year to year according to the severity of the winter season.

All Age Deaths 2010–2020

And putting all together here are the deaths for all age groups. For clarity the lines are the same colour for all years with 2020 emphasised by a bold line. Death affects mostly the very old in every season, but how many in each year is ever more variable the older you get.

The story so far? 2020 is not an exception.

Notes from the ONS

Counting deaths involving the coronavirus (COVID-19)

And the notes that accompany the weekly registered deaths:

Weekly death figures provide provisional counts of the number of deaths registered in England and Wales for which data are available, and the number of deaths for which the underlying cause was coded to respiratory diseases as defined in the International Statistical Classification of Diseases and Related Health Problems Tenth Revision (ICD-10). From 31 March 2020 these figures also show the number of deaths involving coronavirus (COVID-19), based on any mention of COVID-19 on the death certificate.

Provisional data on deaths registered in each week (ending on a Friday) are compiled at the end of the following week.

Bank Holidays could affect the number of registrations made within those weeks.

The counts of deaths from specific conditons are updated with each weekly publication as the coding of the underlying cause is not always complete at the time of production.

To meet user needs, we publish very timely but provisional counts of death registrations in England and Wales in our Deaths registered weekly in England and Wales, provisional dataset. These are presented by sex, age group and regions (within England), as well as for Wales as a whole. To allow time for registration and processing, these figures are published 11 days after the week ends.

Because of the coronavirus (COVID-19) pandemic, our regular weekly deaths release now provides a separate breakdown of the numbers of deaths involving COVID-19. That is, where COVID-19 or suspected COVID-19 was mentioned anywhere on the death certificate, including in combination with other health conditions. Previously, the number of deaths with an underlying cause of respiratory disease was published a week behind the current week. These will now be published for the current week and revised the following week. Alongside this, a new bulletin summarises the latest weekly information and will be updated each week during the pandemic. A link to the bulletin can be found in the notice box on the weekly deaths dataset page.