What do official COVID19 death statistics tell us?
The Turkish Ministry of Health announces the death toll due to novel coronavirus cases on a daily basis. These statistics are based on confirmed COVID19 cases. Like other countries, Turkey does not include people who died without proper medical diagnoses into the official statistics. As such, the number of patients who died directly or indirectly due to COVID19 tend to be higher than the numbers in official statistics.
In any country, the degree of underreporting depends on the institutional capacity of the state and the extent to which the pandemic is affecting the locality. One could expect underreported figures to be higher in developing regions and hard-hit locations. In addition to measurement error in excess deaths due to the pandemic, localities with a high prevalence of disease may have a severely overwhelmed health care system, which increases other-cause mortality. Finally, changes in lifestyle affect mortality by reducing all-cause mortality in the general population (e.g., via reductions in motor vehicle accidents, better care for the elderly), at least for those who are spared of the infection.1
Existing studies suggest substantial variation in underreporting in COVID19 deaths across countries. According to The Economist, for example, the total death toll in Italy and the Netherlands is more than twice the official reported numbers; in Spain, UK, and USA, excess mortality due to COVID19 is between 10%-60% of the official reported figures.
Other Provinces
The funeral data are publicly available for another 9 provinces in Turkey. None of these provinces have an excess mortality pattern that is as striking as Istanbul. Perhaps with the exception of Bursa and Sakarya, most provinces follow the same mortality pattern that they did over the previous years for which I collected data. Why the burial statistics in Istanbul are quite different from others is an important question yet impossible to answer with the publicly available data. The Ministry of Health should shed light on the mystery. You can find the data used in this study here and here. The programs that generate the estimated numbers and the figures are on my github page.