Disclaimer: This page is not intended to defend the Swedish approach to dealing with the Corona crisis or criticize other countries. I just thought it important to make the relevant data accessible.
This page was published on June 27, 2020. I have not made any significant changes to the text since then, but figures have been updated. Apparently, the initial analysis was more or less correct.
Update November 28, 2020: As it turns out, Sweden did not manage to avoid a second wave. However, it seems to be a lot less severe than in other countries. I have just updated the plots correspondingly.
Nobody knows the total number of COVID-19 infections in Sweden, since far too little testing is being done. It therefore makes sense to consider two other numbers, which are well known and easy to measure:
- The number of new patients admitted to intensive care with COVID-19
- The number of deceased with COVID-19
Assuming the dangerousness of the virus to be roughly constant, with the same mortality rate, these curves should track the total number of infections with some delay.
The data plotted below are publicly available from the Swedish Health Authorities. I will update the plots regularly.

The solid lines are 7-day rolling averages.

The conclusion seems unavoidable: the number of infections in Sweden is decreasing. Since this is not due to a lockdown, people have either changed their behavior or developed immunity. In other words, COVID-19 behaves like every other virus outbreak in human history.
BTW, the excess mortality tracks the above curves. There is no excess mortality in Sweden at this point.
The plot below shows the number of confirmed cases, intensive care admittance (IC), and deaths. Cases started spiking on February 27 (black curve). Lockdown measures were implemented from March 10 to March 30 (red shaded area). Intensive care admittance and deaths started spiking with a two week delay, around March 10 (red and blue curves). The green shaded area shows the lockdowns measures shifted by two weeks. One would expect the measures to have an affect on the number of deaths and IC during this time span.

Let’s apply Occam’s Razor: Around April 13, there were 100 people per day dying in Sweden. Two weeks earlier, there were less than 500 confirmed infections daily. Since it is difficult to believe that more than 20% of the infected actually died, testing probably completely failed to keep up with the spread of the disease. In other words, the real number of infections was at least one order of magnitude higher (literally off the chart) and the shape of the black curve is an artifact of the testing.
Finally, below is a somewhat speculative diagram. Given that the number of deaths decreased exponentially from mid-April to mid-Jun, I used this slope to estimate when the number of deaths will be down to one per day. This would happen early December. Fortunately, the actual number of deaths seems to drop much faster, which could be due to better weather and summer vacations. It could also be due to a delay in reporting the data. Or herd immunity?

It will be interesting to see if Sweden can avoid a second wave with this approach.