COVID-19: PROPRIETARY MODELS SET TO MONITOR PANDEMIC EVOLUTION
At the end of 2020, a new coronavirus started spreading in China. A few months later, it was already all around the world, changing the life and habits of billions of people.
- In this report we present stylized facts and regularities of the spread of the virus deriving from 4 proprietary statistical models which focus on: short-term forecast of contagions and deaths, longer term Gaussian model for deaths estimate and, finally, current estimate for total real cases.
- We advocate basing cases estimates on deaths as a stable measure, less affected by testing issues than cases.
- The evolution in the pattern of contagion rates in Western countries is typical and quite predictable: an initial outbreak with very high contagion rates followed by linear moderation as habits change and containment measures step in.
- Some Eastern countries which succeeded in containing the virus have patterns that either skip the initial acute phase (Japan) or contain it very rapidly (Korea), leading to dramatically lower total cases and lesser strain on the Health system.
- Gaussian models of the epidemic’s evolution predict the peaks well (and hence the maximum strain on intensive care) but the dying out phase is slower than the bell-shape curves would predict.
- The actual number of infected people is a multiple of the official figure, and can be roughly estimated at around 100 times the official deaths.
- The typical trajectory of an eventual new outbreak (second wave) would probably skip the initial acute phase due to the improved degree of cultural and material preparation, looking more like the Japanese case than what was experienced by Western countries.