David Paton Profile picture
Professor of Industrial Economics Nottingham University Business School (views expressed are my own)
7 Apr
This morning, @BBCr4today had a piece on how badly Brazil had managed the pandemic, featuring a doctor bemoaning that they had never had a national lockdown "like you did in the UK". No challenge from the presenter of course …
… Brazil has seen a big increase in deaths in recent weeks, but no mention of the fact that their cumulative rate is still below hard-lockdown Peru and way lower than lockdown UK ...
… and no mention that the 7-day average of new reported cases in Brazil have been decreasing since 27 March, whilst the estimated R rate has been decreasing since mid-March ...
Read 6 tweets
25 Mar
Lots to enjoy in this thread about the end of the Texas mask mandate on 10 March. See also @ianmSC 's other thread on Mississippi which removed the mask mandate earlier on 3 March.
What stands out is how it is a matter of blind faith for so many commentators, including leading health experts, that these sort of restrictions must be effective. Evidence and data just have no impact on them.
In the UK, getting rid of the mask mandate should be first on the list for the road map, not relegated to a "review" in June. The risk of any impact on infection rates is very low, whilst the benefits would be very high ...
Read 7 tweets
24 Mar
Yes, disgraceful for the PM to signal support for such unjust discrimination. However, if it is genuinely up to the business, I suspect the free market will soon consign the suggestion to the dustbin ...
... Given the costs of checking & risk of legal challenges on grounds of discrimination, the only reason a pub would deliberately turn unvaccinated away is if there is really big demand from the public for vaccinated-only pubs ...
... people may well say they like the idea in response to an opinion poll but I would be surprised if there was much demand in practice. ...
Read 6 tweets
24 Mar
Update to various Covid-19 indicators for England.

Another big drop in deaths, hospitalisations down too.

Positive tests down & good to see no further rise in school age: data to 19 Mar, last day of in-school testing.

Zoe has been up & down: down a bit today (data to 20 Mar).
Positive tests for school ages by specimen date will be even harder to interpret over next few days due to start of at home mass LFD testing on Sun 21 March. This is meant to be followed up by a confirmatory PCR test so bear in mind ...
... an LFD positive followed by a PCR positive will only count as 1 new positive but:

1. The case will be allocated to the LFD specimen date but when the PCR test is in, that will be deleted & the case added to the PCR specimen date ...
Read 4 tweets
18 Mar
Update to various Covid-19 indicators for England:

• Another big drop in deaths.
• No further rise in Triage whilst Zoe down again.
• Positive tests down (just) overall, no change in school ages

A few other points to note …
1. Daily admissions drops back to 72 for Midlands (16 Mar), making me more suspicious that the leap to 116 on the 15th may have been a data problem.

2. 7-day ave for positive tests goes up to last Sat. Not a “school mass testing” day so not surprising no further increase ...
3. I've switched back to reporting 4-19 for school ages (rather than 10-19) as testing has also ramped up in primaries (though not as much as secondaries).

In any case, 4-19 more relevant going forward in seeing any impact of school opening itself (rather than mass testing) ...
Read 4 tweets
27 Feb
This slide from yesterday’s press conference is a bit naughty in my opinion.

It is clearly trying to imply that infections only fell after the start of 5 Jan national lockdown.

The problem is ...
... the graph does not show new infections (incidence) but the total number who would test positive at any one time (prevalence).

Incidence of new positive tests will have peaked earlier than prevalence & the peak of new infections will have been earlier still.
ONS modelled estimates actually place the prevalence peak at 1st Jan implying an infection peak probably even before Christmas.

That's a bit earlier than implied by other indicators (Zoe, actual positive tests etc.) but all show infections falling well before national lockdown.
Read 5 tweets
21 Feb
Although cases have plummeted worldwide since Christmas, most countries in Europe have started to see increases (small so far) in past couple of weeks.

This is true in places with few restrictions (Sweden/Finland) as well as those with strict lockdowns (Denmark/Belgium/Austria).
Czechia is a good example: more or less in lockdown since mid-October though with quite a few specific tightening & relaxations at different times.

Restrictions ramped up even more on 30 Jan, yet cases are rising faster than in most countries. Here is the comparison with Sweden.
Not all countries have seen an increase, e.g.:

Portugal (lockdown) & Spain (no lockdown, varying regional restrictions): still seeing decreases.

Italy (restrictions but no lockdown & some recent relaxing) & France (significant restrictions, no lockdown): fairly stable for now.
Read 10 tweets
15 Feb
Two things stand out from the data & evidence:

1. When there is a surge in infections, cases eventually come down whether or not there is a lockdown/significant restrictions.
2. Lockdowns/many restrictions have little (not necessarily no) impact on hospitalisations & deaths.
Some find this hard to accept, as it seems so obvious:

Transmission comes from contact, restrictions reduce contact, so surely restrictions must reduce transmission & hence admissions & deaths?

But there are lots of reasons why restrictions may have less benefit than expected:
1. People change their behaviour voluntarily in response to changes in infections.

You choose to take less risk when infections surge, more risk when infections low, but just as importantly ...
Read 11 tweets
14 Feb
Restrictions were imposed to stop health services being overwhelmed.

Govt policy on reopening now seems to be based on an assumption that opening 1 thing inevitably increases transmission & so something else has to be kept shut.

There is little support for this in the data ...
E.g. A number of academic papers suggest opening schools has little or no impact on transmission rates:




So why should pubs stay shut longer because schools are open? Or retail or gyms?
The evidence of significant transmission from outdoor sports is v. weak, even contact sports like rugby: bjsm.bmj.com/content/early/…

So why should allowing sport be conditional on anything else?
Read 5 tweets
12 Feb
I think @ChrisGiles_ is basing his point on schools on the fact that prevalence is down more amongst school age children than 70+.

But if you look at the timing of the trends, I don't think the ONS data really scream out that closing schools reduces infections.
Changes in prevalence will reflect changes in infections about 2 weeks earlier. Take age 2-year 6 group (mainly primary). Estimated prevalence increased until a peak on 24 Dec (implying infections peaked about 10 Dec). So infections started falling at a time schools were open.
Prevalence falls until 6 Jan (i.e. infections about 23 Dec) & then increases again. i.e. infections in primary kids started increasing again when schools were shut for holidays.
Read 8 tweets
7 Feb
One reason I am sceptical about impact of Tier 3&4 is decline in infections from end-Dec is part of the normal longer run trend we see v. often: when cases rise, rate of growth increases, then declines & eventually goes negative. A bit like R declining but still > 1 for a while.
Here is rate of growth for Zoe estimates for England.

Growth rate turns -ve (i.e. estimated infections start to fall) on data from 3 Jan (implying infections falling perhaps 5-6 days earlier). But growth rate started falling from 16 Dec, i.e. before school hols or Tier 4.
V. little sign of tier changes/lockdown accelerating the trend. In fact, the inflection point around 23-28th Dec happens just when effect of school holidays/T3 then T4 in London/SE/E might have kicked in. Similarly, no sign of any accelerating of trend after lockdown.
Read 8 tweets
5 Feb
This is an interesting response by @cjsnowdon to @toadmeister but many of the examples he cites do not support his argument that infections start falling after lockdowns.

Here’s a closer look at 3 of them: France (31st Oct), Ireland (24th-31st Dec) & England (5th Nov)
1. France: 7-day cases were falling from 3rd Nov (not the 8th as @cjsnowdon states), positivity from 2nd, hospitalisations from the 5th. Given lag, these all show infections falling well before 31st October ...
France's estimated R-rate started falling from 24 Oct & there is no indication lockdown accelerated the decrease.

The 31st October French lockdown is not a good example of lockdown effectiveness.
Read 13 tweets
4 Feb
A self-proclaimed fact checking website run by a Conservative MP & Govt policy advisor still states “Cases were not falling before the January 2021 lockdown”. They refer to 3 pieces of evidence ...
... 1. Weekly positive test data:

Weekly data obscures the daily peak but even so, PHE surveillance show peak was in wk 53 (28 Dec-3 Jan), i.e. pre-national lockdown, consistent with the daily data. assets.publishing.service.gov.uk/government/upl…
2. "ONS infection survey ... rose to 1 in 50 by Jan 2 2021":

they don’t clarify this is prevalence (total currently infected) which will peak much later than incidence (new infections). As it happens, ONS daily estimates suggest even prevalence peaked pre-lockdown.
Read 6 tweets
31 Jan
Two things annoy me a little in the school debate. First, well-intentioned comments like "schools should be last to close & first to open".

We never debate when supermarkets should close. It is just accepted that they are essential & must stay open. So it should be with schools.
Second is the suggestion that there should be a trade off: if schools stay open something else has to close.

That reasoning assumes schools being open plays a significant role in infection spread. In fact, the evidence for that is, at best, weak. e.g.:
Read 8 tweets
21 Jan
In his recent article, @cjsnowdon claims that England's November lockdown was a good example of lockdown effectiveness. Let’s look at that claim using the ONS death-by-date registration data which have now been updated for the relevant period.
Although there is some uncertainty over the average lag between infections and death, we should expect any effect of lockdown to be visible in the deaths series after about three weeks.
If the 5th November lockdown had been effective, we might expect a beneficial effect on the deaths trend from about 26 Nov and an adverse effect from about 23 Dec, reflecting the relaxation on 2 Dec.

So what actually happened?
Read 10 tweets
19 Jan
Update to Sweden.

Now Christmas testing/reporting effects have worked through, we can see a decrease in positive tests of > 40% since peak just before Christmas.

ICUs also coming down steadily. We can expect deaths to follow though backdating means it is hard to be sure yet. ImageImageImage
Sweden has introduced some more measures over recent months, though nothing like those seen in UK & elsewhere, e.g.:

24 Nov: ban on more than 8 people gathering
7 Dec: schools for 16+ shut
18 Dec: mask guidance (not law) on public transport
24 Dec: some restrictions on bars
Taking account of the lag from infections, the positive test & ICU data suggest infections peaked around 16th December.

Definitely too late for gatherings to be the cause. Probably too late for schools. Clearly too early for the mask guidance or bar restrictions.
Read 5 tweets
18 Jan
Some people have been worried about high numbers of pupils attending schools since 4 Jan.

Children are (generally) not at risk of serious illness but concern was new variant wd cause fast spread in children, which cd then spill over to others.

We now some early data to look at.
First positive tests (7-day ave) for 5-9 & 10-14s. Allowing a 7-day lag to test, infections seem to increase during 1st part of holidays (but beware Christmas testing effects), then decrease from end of Dec. Early days, but so far looks like the decrease has continued since 4 Jan
Next NHS Covid-19 triage for U19s. Lag probably shorter than for tests.

7-day ave trend suggests infections leading to triage decrease before schools shut for Christmas, level off during holidays & then decrease again from about 4 Jan.

(Thanks to @dontbetyet for the chart)
Read 4 tweets
16 Jan
This is an interesting read by @cjsnowdon but not sure his examples necessarily support his case.

Take Ireland: Level 5 lockdown restrictions (pubs, cafes hairdressers shut) started on 24 Dec. The vertical rise in cases started 8 days after that on 2 Jan & continued to 8 Jan. Image
More restrictions were added up until 31st Dec & cases started dropping from the 9th. But you could just as easily claim the 24th Dec restrictions caused the vertical rise on the 2nd as to claim the extra restrictions caused the drop on the 9th.
I don't think Ireland reports by specimen date & also there were reporting lags. So probably pretty hard to be certain about any cause & effect, but it's not exactly a poster case for lockdown restrictions.
Read 7 tweets
13 Jan
Good question! Of course bad news sells for the media but it seems to be Govt strategy not to focus on any good news at the moment.

Same in April when they were highlighting high reported deaths each day & ignoring that deaths-by-date were showing we had passed the peak.
I get that the Govt is worried if people hear things are improving, they may change their behaviour & spread the virus more.

Personally, I would argue it is far better to treat the public with respect & to provide balanced, accurate information whether good or bad news.
At the moment the balanced message would be that infections are decreasing in most parts of the country (especially London) whilst at the same time many of our hospitals remain under immense pressure.
Read 7 tweets
12 Jan
Data over the past week has utterly demolished the case made last week for the latest national lockdown, i.e. that infections (and subsequently hospitalisations) would not decrease without another full lockdown.
Current indicators show infections were decreasing nationally before lockdown 3, the most dramatic falls being in London, SE & E, but most regions decreasing.

(& let’s be clear: the more promising trends were already becoming evident even when lockdown was announced last week).
You can credibly make a case for Level 4 restrictions on the grounds they were having some effect. FWIW, I don't think there is anything in the data to suggest that, but let's put that to one side for now.
Read 6 tweets
12 Jan
Zoe symptom estimate for England down again today (data up to 8th Jan) & now 13% below recent peak.

All regions except NW below peak: London 28% down, East 20%, SE 16%, NE&Y 9%, Mids 8%, SW 4%.

Given infection-symptom lag, decreases will relate to pre-lockdown infections.
The Zoe estimates have been a pretty reliable early indicator but the decrease is also evident in other indicators: NHS triage, positivity % & positive tests have also all decreased nationally & especially in London, East & South East.
NHS triage data goes to 11 Jan so should pick up at least the start of any effect of schools re-opening to key workers from 4th Jan (+ to nearly all primary pupils for 1 day the 4th). So it is good to see U19 7-day ave falling.

n.b. big thanks to @dontbetyet for the chart.
Read 4 tweets