The views expressed here are my own, not those of my employers.
Summary
The annual conflict deaths as a fraction of the global population decreased 0.121 OOM/​century from 1400 to 2000 (coefficient of determination of 8.45 %), and the annual epidemic/​pandemic deaths as a fraction of the global population decreased 0.459 OOM/​century from 1500 to 2023 (38.5 %).
The variances over the last 10 years of the decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population have very weak trends over the aforementioned periods (1.90 % and 1.61 %).
Introduction
Some argue the risk of human extinction has been increasing, and will increase a lot in the next few years or decades. Toby Ord’s The Precipice is a major example. From an outside view perspective, such increases would be more likely given increasing annual deaths as a fraction of the global population. In contrast, I concluded the logarithm of the annual:
Conflict deaths as a fraction of the global population have trended downwards from 1400 to 2000 (coefficient of determination of 8.45 %), although it is unclear to me whether the direction of the trend is resilient against changes in the modelling of the underreporting of historical conflict deaths.
Epidemic/​Pandemic deaths as a fraction of the global population have trended downwards from 1500 to 2023 (38.5 %), and I guess the direction of the trend is resilient against changes in the modelling of the underreporting of historical epidemic/​pandemic deaths.
However, one could fairly object that what matters for tail risk is not the logarithm of the annual deaths as a fraction of the global population, but its variance. I studied this in this post.
Methods
I run linear regressions of:
The decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population. I had already run similar regressions in the previous posts, where I used the natural instead of decimal logarithm. Which one is used does not affect the fit, but the slope of the regressions involving the decimal logarithm have a more straightforward interpretation.
The variance over the last 10 years of the decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population. By variance over the last 10 years, I mean that among the (varying) reference year and 9 (= 10 − 1) before it.
Linear regression of the decimal logarithm of the annual conflict deaths as a fraction of the global population on the year
Slope (OOM/​century)
Intercept
Coefficient of determination
-0.121
-1.48
8.45 %
Linear regression of the variance over the last 10 years of the decimal logarithm of the annual conflict deaths as a fraction of the global population on the year
Slope (OOM^2/​century)
Intercept
Coefficient of determination
0.0209
-0.192
1.90 %
Epidemics/​Pandemics
Linear regression of the decimal logarithm of the annual epidemic/​pandemic deaths as a fraction of the global population on the year
Slope (OOM/​century)
Intercept
Coefficient of determination
-0.459
4.44
38.5 %
Linear regression of the variance over the last 10 years of the decimal logarithm of the annual epidemic/​pandemic deaths as a fraction of the global population on the year
Slope (OOM^2/​century)
Intercept
Coefficient of determination
-0.0336
0.882
1.61 %
Graphs
Conflicts
Epidemics/​Pandemics
Discussion
The annual conflict deaths as a fraction of the global population decreased 0.121 OOM/​century from 1400 to 2000 (coefficient of determination of 8.45 %), and the annual epidemic/​pandemic deaths as a fraction of the global population decreased 0.459 OOM/​century (38.5 %).
The variances over the last 10 years of the decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population have very weak trends over the aforementioned periods (1.90 % and 1.61 %). I think downwards or weakly upwards trends were to be expected. Conflict and epidemic/​pandemic deaths as a fraction of the global population follow heavy-tailed distributions, whose mean increases with variance, so such a fraction going down tends to imply a decreasing variance.
Variance of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population
The views expressed here are my own, not those of my employers.
Summary
The annual conflict deaths as a fraction of the global population decreased 0.121 OOM/​century from 1400 to 2000 (coefficient of determination of 8.45 %), and the annual epidemic/​pandemic deaths as a fraction of the global population decreased 0.459 OOM/​century from 1500 to 2023 (38.5 %).
The variances over the last 10 years of the decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population have very weak trends over the aforementioned periods (1.90 % and 1.61 %).
Introduction
Some argue the risk of human extinction has been increasing, and will increase a lot in the next few years or decades. Toby Ord’s The Precipice is a major example. From an outside view perspective, such increases would be more likely given increasing annual deaths as a fraction of the global population. In contrast, I concluded the logarithm of the annual:
Conflict deaths as a fraction of the global population have trended downwards from 1400 to 2000 (coefficient of determination of 8.45 %), although it is unclear to me whether the direction of the trend is resilient against changes in the modelling of the underreporting of historical conflict deaths.
Epidemic/​Pandemic deaths as a fraction of the global population have trended downwards from 1500 to 2023 (38.5 %), and I guess the direction of the trend is resilient against changes in the modelling of the underreporting of historical epidemic/​pandemic deaths.
However, one could fairly object that what matters for tail risk is not the logarithm of the annual deaths as a fraction of the global population, but its variance. I studied this in this post.
Methods
I run linear regressions of:
The decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population. I had already run similar regressions in the previous posts, where I used the natural instead of decimal logarithm. Which one is used does not affect the fit, but the slope of the regressions involving the decimal logarithm have a more straightforward interpretation.
The variance over the last 10 years of the decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population. By variance over the last 10 years, I mean that among the (varying) reference year and 9 (= 10 − 1) before it.
Results
The calculations and results are in this sheet.
Linear regressions
Conflicts
Linear regression of the decimal logarithm of the annual conflict deaths as a fraction of the global population on the year
Linear regression of the variance over the last 10 years of the decimal logarithm of the annual conflict deaths as a fraction of the global population on the year
Epidemics/​Pandemics
Linear regression of the decimal logarithm of the annual epidemic/​pandemic deaths as a fraction of the global population on the year
Linear regression of the variance over the last 10 years of the decimal logarithm of the annual epidemic/​pandemic deaths as a fraction of the global population on the year
Graphs
Conflicts
Epidemics/​Pandemics
Discussion
The annual conflict deaths as a fraction of the global population decreased 0.121 OOM/​century from 1400 to 2000 (coefficient of determination of 8.45 %), and the annual epidemic/​pandemic deaths as a fraction of the global population decreased 0.459 OOM/​century (38.5 %).
The variances over the last 10 years of the decimal logarithm of the annual conflict and epidemic/​pandemic deaths as a fraction of the global population have very weak trends over the aforementioned periods (1.90 % and 1.61 %). I think downwards or weakly upwards trends were to be expected. Conflict and epidemic/​pandemic deaths as a fraction of the global population follow heavy-tailed distributions, whose mean increases with variance, so such a fraction going down tends to imply a decreasing variance.