I generally agree with other commenters who have established a ~40–50% base rate: I reached the same conclusion via a Poisson approximation, though I'd like to introduce an alternative data source. Others have linked data or news sources which seem to be based on work by J. M. Powell and C. L. Thyne, which concludes that there have been 104 successful coups in Africa since 1960.
However, Powell & Thyne's bar for coup success is that the coup must survive for 7 days. There's no such stipulation in the resolution criteria, though I find it somewhat ambiguous what "maintains authority" means. This 7-day rule means that the 1971 Sudanese coup is not counted — but for resolution of this question it may or may not be.
An alternative source is the Cline Center's Coup d'État Project which tracks coups (including self-coups). The Cline centre talk of "realized" vs. "unrealized" coup, where a "realized coup" is defined as:
A coup that achieves the effective removal or displacement of the targeted incumbent actor or that ends their ability to direct national governance. The plotters need not accomplish the effective takeover of the state apparatus.
To my reading, this is probably a bit weaker than the resolution source. Perhaps the Powell & Thyne criteria is closer, but I think we should look at both, and then make a judgement.
A Poisson approximation from the Cline data gives me around 70% for "Yes".
Since there may be some amount of autocorrelation in the timing of coups in a given country (e.g. maybe a recent coup makes another coup in the short-term more likely, or maybe less likely) I also computed base rates by considering the historical fraction of 6 months periods on which this question would resolve 'Yes' for both sources. We get:
Date
-Today [P&T]
-Today [Cline]
Decade [P&T]
Decade [Cline]
1960
53%
66%
65%
80%
1970
51%
64%
67%
82%
1980
48%
59%
67%
70%
1990
43%
56%
53%
75%
2000
39%
49%
39%
48%
2010
42%
53%
29%
45%
2020
85%
83%
85%*
83%*
It looks like the 2020s have been quite active in terms of coups so far (e.g. all the coups in the Sahel) which could indicate a real uptick in the number of coups. It's not clear whether this uptick is 'significant', and it's not clear whether it will be sustained. But I'd generally put 48% as a good base rate.
Why go through all this again? I think this is one of the cases where the base rate is likely to be a fairly good estimate, in the absence of specific indicators of developing coups (that are likely to be successful). We may be able to identify possible points for coups (e.g. specific elections) but Africa is big enough that staying close to the base rate is a good idea. Also, by the nature of coups, there are often few explicit warning signs, so treating them as generated from a random process is probably fine. Considering more data and trying to refine the base rate is likely to be useful.
I think though I'd be keeping my eyes out for the following scenarios:
Possible self-coups following elections. Though historically these have been quite rare: the Cline dataset records 7 successful African self-coups since 1960 (Uganda '66, Lesotho '70, Chad '75, Burundi '76, Lesotho '90, Algeria '91, Lesotho '94, Tunisia '21).
Why do you think you're right?
I generally agree with other commenters who have established a ~40–50% base rate: I reached the same conclusion via a Poisson approximation, though I'd like to introduce an alternative data source. Others have linked data or news sources which seem to be based on work by J. M. Powell and C. L. Thyne, which concludes that there have been 104 successful coups in Africa since 1960.
However, Powell & Thyne's bar for coup success is that the coup must survive for 7 days. There's no such stipulation in the resolution criteria, though I find it somewhat ambiguous what "maintains authority" means. This 7-day rule means that the 1971 Sudanese coup is not counted — but for resolution of this question it may or may not be.
An alternative source is the Cline Center's Coup d'État Project which tracks coups (including self-coups). The Cline centre talk of "realized" vs. "unrealized" coup, where a "realized coup" is defined as:
To my reading, this is probably a bit weaker than the resolution source. Perhaps the Powell & Thyne criteria is closer, but I think we should look at both, and then make a judgement.
A Poisson approximation from the Cline data gives me around 70% for "Yes".
Since there may be some amount of autocorrelation in the timing of coups in a given country (e.g. maybe a recent coup makes another coup in the short-term more likely, or maybe less likely) I also computed base rates by considering the historical fraction of 6 months periods on which this question would resolve 'Yes' for both sources. We get:
It looks like the 2020s have been quite active in terms of coups so far (e.g. all the coups in the Sahel) which could indicate a real uptick in the number of coups. It's not clear whether this uptick is 'significant', and it's not clear whether it will be sustained. But I'd generally put 48% as a good base rate.
Why go through all this again? I think this is one of the cases where the base rate is likely to be a fairly good estimate, in the absence of specific indicators of developing coups (that are likely to be successful). We may be able to identify possible points for coups (e.g. specific elections) but Africa is big enough that staying close to the base rate is a good idea. Also, by the nature of coups, there are often few explicit warning signs, so treating them as generated from a random process is probably fine. Considering more data and trying to refine the base rate is likely to be useful.
I think though I'd be keeping my eyes out for the following scenarios:
Why might you be wrong?