As of 2 October 2020, the following method is being implemented by CBS:
According to CBS’ definition, excess mortality occurs wheen the observed number of deaths over a particular period is higher than the number that is expected for that same period.
The expected mortality - in the absence of an epidemic - has been estimated based on the observed number of deaths over the past five years. First, mortality on a weekly basis is established for each of the preceding years. Subsequently, CBS determines an average weekly mortality figure for each week and the six surrounding weeks. This average weekly mortality is an approximation of the expected weekly mortality. Based on its population forecast, the expected mortality is then rescaled to take account of population growth, ageing and increasing life expectancy. The margins around expected mortality are estimates, based on the observed spread of weekly mortality over the past five years.
CBS is implementing a different method as of 2 October 2020
Expected mortality had no coronavirus epidemic occurred has been estimated based on the number of deaths in the preceding weeks, adjusted for seasonal factors. For week 11, it is assumed that the expected weekly mortality was the same as in weeks 3 to 10. The seasonal factors are adjusted for the average weekly temperature. The assumption is that every one degree Celsius above the average weekly temperature in weeks 3 through 10 results in 1 percent decline in mortality. Once the maximum temperature rises above 16.5 degrees Celsius, mortality rises again by 2 percent per degree Celsius. For weeks 11 to 14, the correction was very small due to the minor temperature differences relative to weeks 3 to 10. From week 15 onwards, it was warmer and the corrections amounted to several percentage points. As of week 25, the maximum temperature exceeded 16.5 degrees Celsius (with the exception of week 28). The average temperature fell below 16.5 degrees again as of week 35.
Reason for this change
Expected mortality as calculated according to the method that was in use prior to 2 October 2020 relied heavily on the observation of mortality figures over weeks 3 to 10 inclusive of 2020. It was a sound method for mortality during the first wave of the coronavirus epidemic, but the method was not robust enough to determine excess mortality over a longer period of time. That is why, as of 2 October 2020, CBS uses a long-term average to serve as the basis for expected mortality. This will include a range indication of the commonly determined margins around weekly mortality. These have been estimated based on the range distribution in previous years.