Great article Nihat Ö. Ayhan to start discussing this issue. As we privately discussed earlier, I’d like to note a few reservations.

I’ve been also tracking the excess death figures in İstanbul since March 1. I’m manually extracting the data day by day. Probably because of the paperwork delays, the most recent figures take about a week to stabilize, so I track each day until the data is stable (the reason for NYT figure to be 2,100 instead of 2,136). As if this was not hard enough, the government now added an extra layer of difficulty by adding a not-so-human-friendly captcha.

Besides Istanbul, I came across the death stats page of Keçiören Municipality, second largest municipality of Ankara with a 940K population. Keçiören has a fairly usable query page, as you can at least query death stats between two dates.

Anyway, Istanbul has an excess death of 2,902 people in 2020, in just 52 days between March 1 and April 21, compared to 2019. That makes a daily average excess of 55,8 deaths. There are 26.67% more deaths in 2020 than in 2019 in this period. The 2020 deaths start to spike right after Turkey announced the first cases in March 10. As presented in an official document, Istanbul accounts for 60% of the country’s Covid-19 case load. Assuming the same representation for mortality, there may be a total of 4,883 excess deaths countrywide.

Keçiören — a fairly large municipality with a population of 940K, representing 16.65% of Ankara — has an excess of 171 deaths for the same period, 24.5% more than 2019.

We don’t have an openly accessible data for the rest of Ankara or İzmir. Bursa and a number of other cities have the same system as İstanbul, meaning the data must be extracted manually, day by day for both years. An unnecessary task at this time. Istanbul & Keçiören are reliable sample bases to form an educated opinion.

After presenting self-gathered evidence for the Turkish excess death figures. I must highlight two issues that I think might be erroneous in your analysis.

1. The stats extracted from the Istanbul Metropolitan Municipality’s Directorate of Cemeteries query page are demise/death figures and not burials, as you present as a possible explanation for the variance. Funerals intended to be buried out of Istanbul are never taken into account in this, not in 2020, not in 2019, ever. Today, Turkish Minister of Health took refuge under the same excuse, despite this factual evidence.
2. Regarding the below NYT chart, I think the presentation is somewhat skewed or at least incomplete. The second column depicts the percentage above normal, which is irrelevant in a comparative context whether governments are cheating about Covid-19 death toll or not. This percentage is important but only to show the true local impact of the pandemic. Also, the last column, is presented as a nominal difference, it should rather be a percentage comparison. We should consider the fact that the number of days taken for each geographic locality is different , a variance percentage takes this into account, but also each geographic location is at a different stage of its own local epidemic, ideally this should also be normalized.

So I rearranged the above table adding a few important parameters. First the variance between excess and the reported death toll should be presented as a percentage. Second, countries are here compared with dense metropolitan cities, where infectiousness and virality levels may be at very different stages.

So the population and population density must somehow be taken into account. Comparing the 3 large metropolitan cities in the table we can fairly say that the variance of NYC seems pretty acceptable and can be explained by rationales 2, 3 or 4 in your article. Jakarta’s variance, despite the city’s high density is undoubtedly a big cheat. Whereas Istanbul, a big but relatively less dense metropolitan city also stands out as an outlier with a variance of 108.7%. So a political camouflage is highly probable.

We should note that, all infectious diseases within a geographic area form some form of a network. So — according to Metcalfe’s Law roughly applicable to any network — the impact of the epidemic is somewhat proportional to the square of the number of infected users, the nodes in the system (n²).

The correct figure for Istanbul is 2,136, not 2,100, hence the updated variance for Istanbul between March 9 & April 12 is 112,3% (2,136/1006) and for a broader range between March 1 & April 21 is even higher at %114 (2,902/1,355).

We may need further research and empirical evidence to prove that, but I highly suspect that the population and population density of a geographic area are highly important in assessing the relative impact of an epidemic in that area, comparing to others.

C.G.-