Thanks for all the great thoughts about my question from the mailbag episode!
The point about optimum instability is interesting, I was just reading a piece that touched on that in the recent Brad Roberts counterforce collection (maybe it was yours?). One thing that occurs to me here: we now have a very strong conventional deterrent against Russia, compared with the Cold War situation. It seems like that would mean that the point of optimum instability is "lower" (more stable) than the point of optimum stability was back then?
Thanks for the follow up Austin. I noticed that you didn’t answer the question about counterforce and crisis instability in the podcast, so it’s good to see your response here. I remain unpersuaded, but at least it’s a serious response.
I also wanted to follow up on the P5 glossary with a link (https://2009-2017.state.gov/documents/organization/243293.pdf) and a comment. I was involved in the glossary effort and found it both time-consuming and mostly pointless. China led the effort and seemed to want to maximize length rather than adding value. That’s why there are many entries copied from IAEA glossaries. It would have been useful to spend more time discussing differences in thinking about terms like deterrence and stability rather than set aside terms on which there were significant divergences. For now, I think the greatest value of the P5 process is to discuss doctrines and identify actions to reduce escalation risks.
Mark- thanks for this. Since you are unpersuaded, let me ask you a question. If you were in the leader of a country involved in a conflict with the United States and you believed the United States had the ability to limit or eliminate your ability to retaliate but you did not have a similar ability- the position I posit the Soviets were in by the late 80s if not sooner- what would you order your military target in a large scale first strike in response to the use or lose pressure? Striking many U.S. cities would invite a massive retaliation. Targeting U.S. nuclear forces would be ineffective and actually help remaining U.S. forces finish disarming your forces, with plenty left to coerce war termination by threatening your assured destruction. Bear in mind this is in a conflict that is currently conventional only or conventional with only limited nuclear use.
So what's your move in that situation? Suicide or assisted disarmament? Or do you have some other targets in mind? I will note that for most leaders launch under attack/launch on warning (LUA/LOW) options appear to be the preferable response- but it has its own problems. I plan to post more on LUA/LOW later this year.
On P5- my own experience of the P5 (2018-2025) as a grumpy Joint Staff officer is summed up well by your "both time-consuming and mostly pointless" remark. The only useful doctrine exchanges with the Russians I participated in were the bilateral Strategic Stability/Security Dialogues.
Austin, my view of deterrence is that it is much more visceral and that the fine tuning is mostly beside the point. And nothing I have read or heard about the fine tuning has changed that view. One side’s damage limitation is the other side’s first strike.
"One side's damage limitation is the other side's first strike"
No argument- but it does not obviate my question. If you intend to use because you fear to lose- you still have to target something(s). That is not fine-tuning that is basic military operations, of which nuclear operations are an example. If your options for things to target result in either dynamic disarmament or suicide it seems highly irrational to use, which is why leaders typically prefer having LUA/LOW (or early warning counterstrike) options despite their drawbacks.
I don’t spend much time analyzing deterrence scenarios, and I’m deeply skeptical of the value of doing so. I think they misrepresent the fundamental nature of nuclear deterrence.
Again, I am skeptical that the scenarios we focus on capture the reality of a potential crisis. Presenting the problem in a particular way narrows the field of view and may blind us to dangers don't understand or haven't thought of. I say that as a person who generally thinks it's important to do the analysis.
Sorry, on this... is it really obvious that it's irrational to target US silos, though? You're not *just* disarming yourself, you're causing a decent amount of "collateral" civilian casualties, which the US cares more about than you the dictator do. And you might hope you could do that without provoking the US to go countervalue in response. So counterforce attacks could be used as a sort of de facto punishment.
This certainly doesn't mean the Russians would want to go after US weapons with an all-out attack, but maybe a medium-large one, also holding a good fraction of their arsenal in reserve for launch on warning.
If we are talking about the situation I posited in my post- which is that much of U.S. counterforce capability is resident on subs- launching an attack on U.S. ICBMs (leave aside bombers) would require a minimum of 400 warheads to target every silo (and not any NC2/NC3 facilities). If each Russian warhead was 95% reliable and had a 90% chance of landing sufficiently close to a U.S. silo to destroy it then each of those 400 warheads would have a .95x.9= .86 probability of destroying the silo it was aimed at. This is probably an unrealistically favorable set of assumptions for Russian systems- and roughly 50 U.S. silos would still survive (.86. x400).
This attack would consume at least a third of Russia's deployed ICBM warheads- probably its aging SS-18s mod 6 (~45 deployed) which can carry up to 10 warheads. So to get ~350 US silos (currently 1 warhead each) would consume at least 40 heavy ICBMs at 10 warheads each. A decent exchange on launchers (40 to get 350) but a crummy exchange on warheads (400 to get 350). If the U.S. MIRVs the ICBMs this might be more attractive.
Following this the United States would retain hundreds of warheads on its submarine force to target 40 fewer Russian ICBMs and 400 fewer warheads. And of course this assumes the United States does not launch its ICBMs under attack. This course of action seems... unwise.
For sure. But I'm imagining the real objective of the attack wouldn't be to get the ICBMs for its own sake. It would be to kill a million or so US civilians, cause a public panic, hopefully put the country into a mood of "We don't want to die for Estonia/Taiwan/wherever." And the hope would be to accomplish that without provoking the US to escalate to city-busting, the way you probably would if you killed that many people with a countervalue attack.
Or maybe a less risky way to try for the same goal would be to target conventional military bases in the continental US?
If you want to kill a million Americans, attacking ICBM fields is not such a great way to do it. They are almost literally in the middle of nowhere- take the 90th Missile Wing- its silos are closest to towns like Pine Bluffs WY (pop ~1100) and Harrisburg NE (pop ~80). If your goal is to kill Americans you are better off attacking the bases for ICBMs- Cheyenne WY where FE Warren AFB, the 90th MW HQ, is located has a population of... ~66,000 so still a long way from a million but does not require dynamic disarmament.
If you just want to kill a million Americans you are gonna have to hit some larger urban areas aka city busting. So sure target non-nuclear military bases that are near urban areas (Ft. Belvoir here in northern Virginia perhaps) And then the U.S. may go ahead and launch a damage limiting attack anyway to keep you from doing it again. Or may launch a reciprocal limited attack. Or quit. But your nuclear first use against the U.S. homeland has not really solved your use or lose problem unless the U.S. quits after you kill 1 million Americans. You like Russian roulette it seems.
Ah I was thinking of the old First Strike documentary, which predicted casualties in the millions for a Soviet counterforce attack. But yields were higher then and there were more silos in more different parts of the country.
This is a good point when it comes to the Soviets, but in the present day it seems entirely possible that if you're in a position Vladimir Putin could end up in--you believe that Russian defeat or capitulation would likely end with you dead in a coup--then you might well go for the countervalue first strike rather than risk being disarmed. At least then you'd probably survive in a bunker.
Edit: Or perhaps more realistically, you might launch a large-scale nuclear attack against US/NATO conventional forces to try to prevent Russian defeat without escalating to direct attacks against US territory?
It would be a weird world in which one believes you are more likely to survive an attack by the un-attrited nuclear forces of the United States than a coup, bunker or no. The United States has in the past made the destruction of enemy political leadership a key objective of targeting in large scale attack (see for example
Therefore a counter-value first strike would likely mean your bunker was pretty heavily targeted in retaliation by all U.S. nuclear forces, including bombers. This attack could thus presumably include the B61-11 gravity bomb: https://www.nukestrat.com/us/afn/99-29_ACC120897.pdf
So this large scale counter value attack would likely be committing suicide for fear of death, or at least playing Russian roulette with several rounds in the cylinder- to mix Bismarck with the Deer Hunter, analogically speaking. In contrast, Putin has already rather handily put down one quasi-coup in 2023 (that was a fun time to be in Pentagon...). Hardly seems worth rebuilding the glory of Russia to see it all burned down, unless one really has no other choice. LUA/LOW seems preferable I think, despite its problems.
Your point about large scale use against NATO conventional forces is probably correct but would be unlikely in the first instance to involve Russian strategic forces. The U.S. has assessed for some time that Russia has up to 2,000 "non-strategic" (i.e. not New START accountable) nuclear warheads for such eventualities. See
Thank you, Austin... I had the impression from the Miller/Acton CSIS targeting debate that attempting to kill specific Russian leaders isn't a realistic objective. It seems like there's more disagreement among experts about these questions of fact than I thought.
Re the Russian tactical weapons, does the damage limitation approach target these, or does it disregard them because they're not a threat to the continental US?
I didn't see that debate, pretty sure I was washing my hair. Regardless I am not referring to targeting specific leaders (e.g. Putin), I am talking about targeting classes of targets like NC3 or leadership, which are located in facilities (e.g. bunkers, which was your starting point). Whether attacks on those facilities (and those that inhabit them) are successful depends on intelligence about them and weapons that can damage/destroy them. The B61-11 link I sent addresses the latter, the former is harder to say at the unclassified level but U.S. intelligence has a decent track record of finding underground facilities when it is looking. See for example:
So one could decide "the U.S. does not know where my bunker is and/or does not have a weapon that can kill/entomb me in it" so I will load up for Russian roulette with nuclear weapons and go big countervalue first. But this seems unlikely at least in the conventional/early limited nuclear use phase, particularly when a LUA/LOW option is available.
On the question of "experts"- disagreement among them is normal and fine. I will note not all experts are created equal. More importantly, look at arguments and especially evidence experts employ. "In God we trust, all others must bring data" is a good maxim. This is why I include links to evidence in most of my posts.
Russian NSNW- damage limitation is about denying assured destruction so weapon that are of limited range/yield would not be a factor. Damage limitation would include, to the extent practicable, weapons that threaten allies with assured destruction but the principal logic is about U.S. willingness to run risks on behalf of allies- the core of extended deterrence.
That makes sense--the main point is to reduce the downside to the US sufficiently so that we have incentives to follow through on extended deterrence.
Thanks for putting all this information out there. I teach nuclear ethics at Michigan, and I thought I was pretty plugged in (for a philosopher) to the technical knowledge that's out there about strategy. But it seems like the stuff that's been out front on the web until now is skewed toward the thinking of the arms control community, and away from the thinking of the US defense pros. This is something James Wells was trying to impress upon me a while back, but I didn't realize the extent of it.
(Also very helpful to hear in the episode about the relationship between limited options and damage limitation; this was not something that was clear to me previously.)
Thanks for all the great thoughts about my question from the mailbag episode!
The point about optimum instability is interesting, I was just reading a piece that touched on that in the recent Brad Roberts counterforce collection (maybe it was yours?). One thing that occurs to me here: we now have a very strong conventional deterrent against Russia, compared with the Cold War situation. It seems like that would mean that the point of optimum instability is "lower" (more stable) than the point of optimum stability was back then?
Thanks for the follow up Austin. I noticed that you didn’t answer the question about counterforce and crisis instability in the podcast, so it’s good to see your response here. I remain unpersuaded, but at least it’s a serious response.
I also wanted to follow up on the P5 glossary with a link (https://2009-2017.state.gov/documents/organization/243293.pdf) and a comment. I was involved in the glossary effort and found it both time-consuming and mostly pointless. China led the effort and seemed to want to maximize length rather than adding value. That’s why there are many entries copied from IAEA glossaries. It would have been useful to spend more time discussing differences in thinking about terms like deterrence and stability rather than set aside terms on which there were significant divergences. For now, I think the greatest value of the P5 process is to discuss doctrines and identify actions to reduce escalation risks.
Mark- thanks for this. Since you are unpersuaded, let me ask you a question. If you were in the leader of a country involved in a conflict with the United States and you believed the United States had the ability to limit or eliminate your ability to retaliate but you did not have a similar ability- the position I posit the Soviets were in by the late 80s if not sooner- what would you order your military target in a large scale first strike in response to the use or lose pressure? Striking many U.S. cities would invite a massive retaliation. Targeting U.S. nuclear forces would be ineffective and actually help remaining U.S. forces finish disarming your forces, with plenty left to coerce war termination by threatening your assured destruction. Bear in mind this is in a conflict that is currently conventional only or conventional with only limited nuclear use.
So what's your move in that situation? Suicide or assisted disarmament? Or do you have some other targets in mind? I will note that for most leaders launch under attack/launch on warning (LUA/LOW) options appear to be the preferable response- but it has its own problems. I plan to post more on LUA/LOW later this year.
On P5- my own experience of the P5 (2018-2025) as a grumpy Joint Staff officer is summed up well by your "both time-consuming and mostly pointless" remark. The only useful doctrine exchanges with the Russians I participated in were the bilateral Strategic Stability/Security Dialogues.
Austin, my view of deterrence is that it is much more visceral and that the fine tuning is mostly beside the point. And nothing I have read or heard about the fine tuning has changed that view. One side’s damage limitation is the other side’s first strike.
"One side's damage limitation is the other side's first strike"
No argument- but it does not obviate my question. If you intend to use because you fear to lose- you still have to target something(s). That is not fine-tuning that is basic military operations, of which nuclear operations are an example. If your options for things to target result in either dynamic disarmament or suicide it seems highly irrational to use, which is why leaders typically prefer having LUA/LOW (or early warning counterstrike) options despite their drawbacks.
I don’t spend much time analyzing deterrence scenarios, and I’m deeply skeptical of the value of doing so. I think they misrepresent the fundamental nature of nuclear deterrence.
OK... but how do you know a crisis is stable or unstable without analyzing the crisis scenario (i.e. force structure on both sides)?
Again, I am skeptical that the scenarios we focus on capture the reality of a potential crisis. Presenting the problem in a particular way narrows the field of view and may blind us to dangers don't understand or haven't thought of. I say that as a person who generally thinks it's important to do the analysis.
Sorry, on this... is it really obvious that it's irrational to target US silos, though? You're not *just* disarming yourself, you're causing a decent amount of "collateral" civilian casualties, which the US cares more about than you the dictator do. And you might hope you could do that without provoking the US to go countervalue in response. So counterforce attacks could be used as a sort of de facto punishment.
This certainly doesn't mean the Russians would want to go after US weapons with an all-out attack, but maybe a medium-large one, also holding a good fraction of their arsenal in reserve for launch on warning.
If we are talking about the situation I posited in my post- which is that much of U.S. counterforce capability is resident on subs- launching an attack on U.S. ICBMs (leave aside bombers) would require a minimum of 400 warheads to target every silo (and not any NC2/NC3 facilities). If each Russian warhead was 95% reliable and had a 90% chance of landing sufficiently close to a U.S. silo to destroy it then each of those 400 warheads would have a .95x.9= .86 probability of destroying the silo it was aimed at. This is probably an unrealistically favorable set of assumptions for Russian systems- and roughly 50 U.S. silos would still survive (.86. x400).
This attack would consume at least a third of Russia's deployed ICBM warheads- probably its aging SS-18s mod 6 (~45 deployed) which can carry up to 10 warheads. So to get ~350 US silos (currently 1 warhead each) would consume at least 40 heavy ICBMs at 10 warheads each. A decent exchange on launchers (40 to get 350) but a crummy exchange on warheads (400 to get 350). If the U.S. MIRVs the ICBMs this might be more attractive.
Following this the United States would retain hundreds of warheads on its submarine force to target 40 fewer Russian ICBMs and 400 fewer warheads. And of course this assumes the United States does not launch its ICBMs under attack. This course of action seems... unwise.
For sure. But I'm imagining the real objective of the attack wouldn't be to get the ICBMs for its own sake. It would be to kill a million or so US civilians, cause a public panic, hopefully put the country into a mood of "We don't want to die for Estonia/Taiwan/wherever." And the hope would be to accomplish that without provoking the US to escalate to city-busting, the way you probably would if you killed that many people with a countervalue attack.
Or maybe a less risky way to try for the same goal would be to target conventional military bases in the continental US?
If you want to kill a million Americans, attacking ICBM fields is not such a great way to do it. They are almost literally in the middle of nowhere- take the 90th Missile Wing- its silos are closest to towns like Pine Bluffs WY (pop ~1100) and Harrisburg NE (pop ~80). If your goal is to kill Americans you are better off attacking the bases for ICBMs- Cheyenne WY where FE Warren AFB, the 90th MW HQ, is located has a population of... ~66,000 so still a long way from a million but does not require dynamic disarmament.
If you just want to kill a million Americans you are gonna have to hit some larger urban areas aka city busting. So sure target non-nuclear military bases that are near urban areas (Ft. Belvoir here in northern Virginia perhaps) And then the U.S. may go ahead and launch a damage limiting attack anyway to keep you from doing it again. Or may launch a reciprocal limited attack. Or quit. But your nuclear first use against the U.S. homeland has not really solved your use or lose problem unless the U.S. quits after you kill 1 million Americans. You like Russian roulette it seems.
Ah I was thinking of the old First Strike documentary, which predicted casualties in the millions for a Soviet counterforce attack. But yields were higher then and there were more silos in more different parts of the country.
This is a good point when it comes to the Soviets, but in the present day it seems entirely possible that if you're in a position Vladimir Putin could end up in--you believe that Russian defeat or capitulation would likely end with you dead in a coup--then you might well go for the countervalue first strike rather than risk being disarmed. At least then you'd probably survive in a bunker.
Edit: Or perhaps more realistically, you might launch a large-scale nuclear attack against US/NATO conventional forces to try to prevent Russian defeat without escalating to direct attacks against US territory?
It would be a weird world in which one believes you are more likely to survive an attack by the un-attrited nuclear forces of the United States than a coup, bunker or no. The United States has in the past made the destruction of enemy political leadership a key objective of targeting in large scale attack (see for example
https://nsarchive2.gwu.edu/NSAEBB/NSAEBB173/SIOP-25.pdf).
Therefore a counter-value first strike would likely mean your bunker was pretty heavily targeted in retaliation by all U.S. nuclear forces, including bombers. This attack could thus presumably include the B61-11 gravity bomb: https://www.nukestrat.com/us/afn/99-29_ACC120897.pdf
So this large scale counter value attack would likely be committing suicide for fear of death, or at least playing Russian roulette with several rounds in the cylinder- to mix Bismarck with the Deer Hunter, analogically speaking. In contrast, Putin has already rather handily put down one quasi-coup in 2023 (that was a fun time to be in Pentagon...). Hardly seems worth rebuilding the glory of Russia to see it all burned down, unless one really has no other choice. LUA/LOW seems preferable I think, despite its problems.
Your point about large scale use against NATO conventional forces is probably correct but would be unlikely in the first instance to involve Russian strategic forces. The U.S. has assessed for some time that Russia has up to 2,000 "non-strategic" (i.e. not New START accountable) nuclear warheads for such eventualities. See
https://www.dia.mil/Portals/110/Images/News/Military_Powers_Publications/Nuclear_Challenges_2024.pdf
Thank you, Austin... I had the impression from the Miller/Acton CSIS targeting debate that attempting to kill specific Russian leaders isn't a realistic objective. It seems like there's more disagreement among experts about these questions of fact than I thought.
Re the Russian tactical weapons, does the damage limitation approach target these, or does it disregard them because they're not a threat to the continental US?
I didn't see that debate, pretty sure I was washing my hair. Regardless I am not referring to targeting specific leaders (e.g. Putin), I am talking about targeting classes of targets like NC3 or leadership, which are located in facilities (e.g. bunkers, which was your starting point). Whether attacks on those facilities (and those that inhabit them) are successful depends on intelligence about them and weapons that can damage/destroy them. The B61-11 link I sent addresses the latter, the former is harder to say at the unclassified level but U.S. intelligence has a decent track record of finding underground facilities when it is looking. See for example:
https://www.cia.gov/readingroom/docs/CIA-RDP84T00171R000301360001-9.pdf
So one could decide "the U.S. does not know where my bunker is and/or does not have a weapon that can kill/entomb me in it" so I will load up for Russian roulette with nuclear weapons and go big countervalue first. But this seems unlikely at least in the conventional/early limited nuclear use phase, particularly when a LUA/LOW option is available.
On the question of "experts"- disagreement among them is normal and fine. I will note not all experts are created equal. More importantly, look at arguments and especially evidence experts employ. "In God we trust, all others must bring data" is a good maxim. This is why I include links to evidence in most of my posts.
Russian NSNW- damage limitation is about denying assured destruction so weapon that are of limited range/yield would not be a factor. Damage limitation would include, to the extent practicable, weapons that threaten allies with assured destruction but the principal logic is about U.S. willingness to run risks on behalf of allies- the core of extended deterrence.
That makes sense--the main point is to reduce the downside to the US sufficiently so that we have incentives to follow through on extended deterrence.
Thanks for putting all this information out there. I teach nuclear ethics at Michigan, and I thought I was pretty plugged in (for a philosopher) to the technical knowledge that's out there about strategy. But it seems like the stuff that's been out front on the web until now is skewed toward the thinking of the arms control community, and away from the thinking of the US defense pros. This is something James Wells was trying to impress upon me a while back, but I didn't realize the extent of it.
(Also very helpful to hear in the episode about the relationship between limited options and damage limitation; this was not something that was clear to me previously.)