Survation famously got their 2017 General Election prediction pretty near right when all the other pollsters floundered. So in 2019, when nearly every pollster was calling it for the Conservatives as a clear win, a frequent refrain from those rooting for Jeremy Corbyn was “I’ll wait for Survation”. The hope and belief was that once again YouGov, Gallup and the rest were getting it wrong, that Survation would ride to the rescue with a positive poll and then on Election Day the Tories would not only be run close but even routed. That hope was crushed when Boris Johnson won an 80 seat majority.
Of course, it isn’t just Labour supporters who have fallen into that trap over the years. Anyone who closely follows politics will have their favourite pollster. Sometimes that view will be based on who seemed to get closest last time round but often we choose the pollster who tells us what we want to hear. We are suckers for confirmation bias. So Conservative voters will lean towards pollsters that give them bigger leads, Labour as we saw prefer those that show them very much still in the game and Lib Dem supporters … well … umm …
Of course, the reality is that just because they got their predictions close last time is no guarantee that they will be the most accurate this time round especially since all of the other companies review their methodology when things are badly out. Furthermore, what is the point of YouGov showing your party in the lead only for the crushing disappointment of discovering they were wrong on polling day? You see, as politicians are fond of saying “there’s only one poll that matters.” It’s not that the pollsters aren’t helpful in giving an indication of trends, its just that they are fallible.
We are seeing something similar at the moment and indeed have throughout COVID. Each day you can go to the PHE stats page and it will give you lots of data about is happening. The headline figure that everyone is watching is the one showing the number of new cases reported each day. Now that data is not infallible. It’s based on the number of people coming forward to get a lateral flow or pcr test but we know two things.
- You can get false positives
- Not everyone comes forward to get a test
There are other people and other sources of data providing information about what might be happening. The ONS publishes regular survey data and estimates of case prevalence. Meanwhile the ZOE App enables people to sign up and regularly log whether or not they have symptoms. It then uses this information in conjunction with test results in order to give an estimate of the number of new cases. Currently the app is showing 60k new cases daily whilst the Government website indicates that 20-30k of people are testing positive.
Unsurprisingly we see people choosing their preferred data source and questioning the others. When cases were increasing rapidly, those who were sceptical about COVID and lockdowns would frequently insist that we could not trust the Government data because of course if you tested more people then you would get more cases. Now the boot is on the other foot and as cases have come down so those who believe that the Government shouldn’t have lifted restrictions and therefore believe we cannot be seeing improvements yet insist that its down to less testing happening now.
Just as we were told to wait for Survation we are now told to wait for ZOE. Yet why should we assume that the App’s estimates are more accurate than the test results? Indeed ZOE have announced that cases are levelling off several times over the past few months only to discover that this was a false dawn. The app’s model has struggled at times because it is dependent on having a significant number of unvaccinated users and so has to adjust its algorithms accordingly when they are missing. Indeed, at one point ZOE even managed to show less cases than people who had tested positive. So over the past few weeks we’ve seen its methodology changed.
It is possible, likely indeed that cases are lower than what the Government testing results are showing but it is equally possible that they are much lower than the ZOE app is saying. Perhaps the truth lies somewhere between the two. This doesn’t mean that either source is completely useless or irrelevant. They may not give us the exact number of cases but they do help us to observe trends and the number of cases reported has helped us to forecast the likely number of hospitalisations we will see in the next few weeks because admissions have reliably turned out to be 2-3% of the number of positive tests 10 days earlier.
Indeed, hospital admissions are probably the best indicator of how things are going. This is for two reasons. First of all, it gives us genuine hard data concerning serious illness. Secondly, this realy is the litmus test of where we are in the pandemic because it doesn’t matter too much if hundreds of thousands of people experience mild illness. It does matter if they end up in hospital both because that puts pressure on the NHS and it increases the risk of death. So admissions are the equivalent of election day.
For what it’s worth here’s my latest chart of admissions against predictions
You will see that actual admissions are lower than my prediction line but basically following the same trend. This is because my prediction was based on a 2.5% rate but it has at times been slightly higher than that and other times including now slightly lower. But there is a general alignment with what we might expect if the trends suggested by Government data are corrected. Indeed, I understand latest data from English hospitals may be beginning to reflect a slow down in cases. If we see a significant drop in admissions in a few weeks then we will be able to confirm that the apparent drop in cases now was real.
None of that means that cases will continue falling or that they won’t begin to increase again. As this chart shows, it is possible that in a best case scenario we saw the third wave peak 2 weeks back at about 54k reported new cases each day.
We may then see cases fall to much less frightening levels by early August. Equally, it is possible that we may see cases begin to increase again as a result of the 19th July. As this chart shared by James Ward shows, it does look like the re-opening of nightclubs and relaxation of restrictions on the hospitality industry is starting to have an effect on those in the 15-19 and 20-30 age groups.
The big unknown is whether the risk of higher cases from nightclub attendance will be offset by downward pressure from school holidays, sunny weather and higher vaccination rates. The jury is still out.
The lessons here are
- Just because evidence provided is fallible doesn’t mean that it is useless. It just needs careful handling.
- Beware confirmation bias. We tend to look for evidence that supports our preferences and opinions. We need to be ready to hear all the evidence including when it challenges our assumptions.
Of course as always this has application beyond COVID. My particular concern is to see people discipled in faith. Over the years I’ve frequently found that people will seek out those willing to give advice that supports their decisions. Some people will go round a group of people until they get someone to tell them what they want to hear. Others will pick and choose who to go to for advice knowing that one person will be warm and supportive and another sceptical and challenging.
We even do this with God’s Word. We cherry pick the Bible verses that tell us what we want to hear and avoid those that make us uncomfortable, challenge us and disagree with us. Of course the difference here is that God’s Word is infallible.
This is why I keep coming back to the question “Are you prepared to let God’s Word disagree with you?” We need to hear those things that are not what we would like to hear. We need to escape from confirmation bias. It is better to hear unwelcome truth from God’s Word now than to be cushioned from its rebuke or warning now only to face the harsh consequences of our choices later.