Change in Normalized Excess Deaths

Death rates (deaths from all causes) per 100k population. Population and mortality data from Eurostat. The big difference to HMD, ourworldindata etc is that we include "unallocated" deaths (this doesn't change the rankings however, as can be seen when comparing to the exclude unallocated tab). Most countries are not able to exactly date all their deaths, and those deaths end up in the unallocated bucket. The methodology we use to portion out the unallocated deaths is to first distribute an integer number of deaths to each week and then to portion out the residual deaths randomly to weeks that year (the probability of being assigned to a week is determined by how big that week's share was of the total allocated deaths for that year). Note the long reporting lag for Italy. "histo_median" is the median annual number of deaths per 100k population between the years 2015-2019 (so we divide the number of deaths each year by the population for that region in January of that year). The last data point for the population data is Jan-2019, so the figures for 2020 will be slightly off (but at least they're off in the same way for all countries). The table is sorted by descending relative delta, ie. the last normalized death figure divided by the historical median (so a large delta means that the death rate this year is high relative to the historical data, presumably mostly due to covid-19 deaths). If the week column says eg. 49, that means that we calculate all the historical data up to that week as well (ie. we won't include 50, 51 or 52). If a year has a week 53, we count that as unallocated (ie. it will be taken into account as per the methodology described above).