In the June 21st update, I showed a chart with the US states I was most concerned about. Florida, Texas and Arizona were showing worrying early signs of increases, while other states such as New York were keeping control of the virus. Since then the situation has developed as I feared.
You could interpret this chart as the UK doing a good job and I would certainly agree it shows the positive effect of the lockdown, but I would also argue that this is a warning against complacency. Florida for example opened their bars in early June and 6 weeks later we can see the consequences.
Yet anti-lockdown commentators cite the lack of large increases in the number of deaths in these US states as a cause for celebration. During the outbreak in New York, deaths rose at the same time as positive case results as in the graph below. Some argue this shows the virus is different this time, perhaps less virulent, perhaps we know how to treat it now, maybe it’s less deadly during the summer.
Unfortunately, this is mainly a misunderstanding of the statistics. The biggest reason deaths have not risen is that they have not risen yet. For an individual case, there is obviously a time period between getting infected with Covid, when you can test positive, and dying.
During the first outbreaks in London and New York, most tests were being done on people already very sick, perhaps on admission to hospital. Essentially, we only began to notice the virus once people were already dying. Without widespread testing across the community, we saw no lag between reported cases and deaths but given more comprehensive testing systems, I would expect this time difference to be larger now.
As cases in these problem states started to rise after June 21st i.e. 18 days ago. I would expect to see the death rate to start rising around now. Since this is the first time we have had significant test data in advance of the deaths it is not easy to make a confident projection. We do not know the lag we should use, we know positive tests but not the number of people infected who have not been tested and we do not know the mortality rate. Below I use a simple projection with a 2-week lag between positive tests and deaths, and a case mortality rate of 2% which is roughly what it was before the recent spike. In this projection we might expect deaths per day total in Arizona, Florida, and Texas to reach 400 per day in a couple of weeks’ time as in the chart below.
There may indeed be a lower mortality rate with improved treatments as we learn more about the disease. There is also a positive demographic effect in the short term as the recent cases have been concentrated in younger people who have a much lower mortality rate. This does not actually affect the overall mortality rate of course but it will make it look less deadly at first glance. If the death rate does not rise over the next few weeks, then I will revisit this debate to see if there is something new actually happening with Covid.
Applied Macro 