Abstractions in Defense Analysis

Abstractions are important for thinking about the world, and particularly for STEM types, taking a thing, putting it in a box and thinking about the box is basically what we do. If we had to reckon with the full complexity of everything at all times, nothing would ever get built. But as Joel Spolsky famously said, all abstractions are leaky, and the leakage can easily mess up any analysis if the person doing it doesn't know when things are likely to leak and how to deal with it.

I bring this up because outsiders attempting to analyze defense problems often make this mistake, particularly in the tech-adjacent sectors that I tend to see a lot of stuff from. Ideas go in the box, and then the box is folded, spindled and mutilated in complete ignorance of the gusher of complexity this produces. An excellent example of this is Austin Vernon's post on shipbuilding, where he attempts to diagnose the problems the USN is facing from a position of knowing very little about warships.¹

The first couple sections are a mess, mixing up changes in shipbuilding that US warship yards have kept up with with those they very much have not and being extremely unclear about how many shipyards the US has and if we should have more or fewer. But then he starts pointing to Korea and Japan as examples of countries that procure ships more cheaply than we do. But cost is also an abstraction. In everyday life, it's a remarkably good and non-leaky one. You can easily forget that money is a consensus fiction adopted for reasons of convenience and treat it as an underlying fact about the world, like physics. But this is far less true in the defense world for a lot of reasons, and once you start looking at international comparisons, abolishing money becomes a surprisingly attractive alternative to actually figuring out what is going on. Ships are big and complex, and very rarely bought from a single vendor. As a result, the number you see is as much a matter of accounting standards as anything. For instance, if one nation does its best to put all costs attributable to a ship under a single heading, while another buys components by department, so the hull is under the Bureau of Ships while radar and weapons come from the Bureau of Ordnance, the vast majority of sources will likely report the first ship as being considerably more expensive even if the combined cost for each ship is the same.²

Actually, things can get even weirder, as it isn't always consistent within a country. For instance, the budget book cost of a Burke doesn't include engines, which are procured in an entirely separate budget book, under a line item which also includes a lot of support work. This is because gas turbines are maintained by swapping them out, and it makes sense to just draw the engines for a new ship from the the pool of LM2500s maintained to support all of the Burkes and Ticos. However, the budget item for the Zumwalts does include the engines, as they use a different turbine, shared only with the Freedom class LCS, so there's a much smaller pool to draw on. Both ways of doing the accounting make complete sense given the operational realities of supporting the ships, but the fact that they are different for the same country in the same year illustrates the degree of complexity involved here.

But then we get to his section on "Rethinking Ship Requirements" and his leaky abstractions turn into a conceptual oil gusher. His main theory is that a lot of the problem we have today comes from trying to do too much with each ship, and that smaller single-purpose ships would be a better option. This is the sort of thing that sounds great at a high level of abstraction, but it's not a theory which can survive close contact with knowledge of naval practice. The problem is that for basic engineering reasons, anything capable of passing for an oceanic warship has to be at least a thousand tons or so, even if it's armed only with a .50 cal. And once you've cleared that threshold, it turns out that economies of scale kick in, and the marginal cost of adding the systems for an extra role is a lot lower than the cost of an entirely new hull. The best example of this is probably the Type 82/Type 42 debacle, where a proposal to combine two systems in one ship to save overall cost was shot down to keep ship size and unit cost down because that was more politically salient.

But that's an example from 50 years ago, when computers were expensive and missile launchers could only handle one type of missile. Today, there's a second problem with the thesis that becomes apparent when we boil away Austin's abstractions: "Navy ships like Burke-class destroyers, Littoral Combat Ships, or Constellation-class frigates take on many tasks. They include conventional air defense, ballistic missile defense, anti-submarine warfare, land attack, anti-ship surface warfare, convoy escort, patrol, scouting, etc."³ Now, this is an impressive array of roles, but it's worth asking what concrete systems do we need to accomplish each one. Air defense is going to require a missile launcher (probably VLS), a decent radar and a combat system (some moderately fancy computers and software to run on them). BMD means a VLS, a really good radar and a combat system. ASW needs a sonar, a helicopter, a combat system and maybe some sort of weapon we can launch from the ship itself (these days, that probably comes out of a VLS). Land attack is VLS and/or a gun, depending on exactly what you want to do. Anti-ship surface warfare is a combat system, and some combination of a helicopter and anti-ship missiles, which either come from simple deck launchers or out of a VLS. What you need to escort convoys depends on what you're escorting them from, but it's going to be some combination of air defense, ASW and patrol. Patrol is only a bit better-defined, but minimally you're going to want a gun to put shots across people's bows and probably a helicopter, maybe with a self-defense missile system in case you decide to patrol somewhere like the Bab-el-Mandeb. And scouting is mostly a job for other people, like the ship's helicopter.

Now, an astute reader will notice that there's a great deal of overlap between the systems required for the listed missions, which in turn suggests that "four specialized ships replacing one multi-role ship" is going to work a lot less well than you would think even leaving aside the economies of scale you find in hulls and in military procurement. Once you remember that you would really like to have things for your high-end warships to do when they're not hunting subs or shooting down planes, the obvious solution is to equip them for secondary missions as well (which, practically speaking, means giving them a gun), and then look at lower-end ships if you need extra hulls for peacetime presence missions. This is what almost every Navy on earth has done, except for the few rich enough to carry out presence missions with higher-end hulls, notably the USN and PLAN.⁴

There's another major abstraction leakage in this section, when he starts to get into a cost breakdown for a Burke, pointing out that a ULCC is 7 times the size of a Burke and costs 1/6th of what the structure does. Now, in fairness to Austin, he cites someone else and does report the numbers accurately. And to his credit, he indexes for inflation, avoiding an extremely common mistake when doing this kind of stuff. But he assumes that the cost labeled "Basic Construction" is just for the hull, and I don't think this is right. Unfortunately, the underlying source doesn't define exactly what "Basic Construction" covers, but I am extremely doubtful that is the cost for the basic hull steel alone. For one thing, it almost certainly includes the cost of installing all of the various other items (weapons, etc) onto the ship, as well as things like habitation facilities that are not called out elsewhere.⁵

And then we come to other ideas he has for improving the fleet. First, reduce speed, because "power required is proportional to the cube of speed". I have no idea where this came from, but it's simply not true and the reality is much more complicated. He pins the modern speed standards on the ability of the faster Japanese fleet to win at Tsushima, but suggests that it matters a lot less in an era when the main threat is airplanes and missiles. There is a bit of truth here, but he seems to be assuming that naval architects are idiots and haven't thought about this. I have read more than enough Friedman to assure him that they have.⁶ Speed is expensive, and figuring out how much you actually need is a major task of early design work.⁷ The main reason that a Burke can make 30 kts is because it needs to keep up with the carriers. The carriers can make 30 kts because that lets them generate more wind over deck to fly planes, and trying to get much past 30 kts gets really expensive.⁸ He also proposes using hybrid electric propulsion, calling on the azipods that cruise ships use. Hybrid propulsion is absolutely a thing, and is becoming more common, although the use of commercial azipods is unlikely because of concerns about power density and damage resistance.

There are similar problems with his analysis of radar. It is true that power requirements scale with the fourth power of range, but the increased use of sea-skimming missiles doesn't mean that long-range radars are now pointless. A more powerful radar is also going to detect stealthy targets at longer range, and it is better able to see through jamming. These are both quite important, but disappear if you treat radar as a single thing with a fixed range number. He then goes on to argue for armor: "Missiles aren't too destructive because the Houthis have fired anti-ship missiles at dozens of cargo ships in the Red Sea, most sailing away with minor damage." Again, the details become extremely relevant. Most anti-ship missiles are intended to mission-kill 4,000-ton frigates, and lobbing a weapon intended for that against what is basically a metal box full of random stuff (often rocks) 100 times that size will not give a good idea of how effective it will be against its intended target.⁹ And it would be fairly easy to develop a bigger missile capable of killing an armored warship. Those don't exist today because of a lack of armored warships to use them on, but it's a lot easier to develop a new missile than a new ship. He also discusses underwater protection, which would be helped by bigger hulls, but given his source for that section, I can be quite confident in pointing out several errors. Most notably, the limit on hull depth had nothing to do with speed (in fact, a deeper hull is likely to be slightly faster) and a great deal to do with being able to use ports and docking facilities, so his proposal wouldn't allow a TDS on the bottom of the ship anyway (leaving aside the fact that there are secondary effects which make underbottom explosions particularly dangerous regardless of the space available).

Now in fairness, a couple of Austin's ideas are halfway decent. Building bigger ships to increase damage resistance has been looked at occasionally and the idea of just building new ships instead of doing a midlife refit was actually implemented by the British for the Type 23 frigates. Of course, that program is an illustration of the potential problems with that approach, namely that when the time came to build new ships, it didn't happen and the vessels were refitted anyway. Buying something as expensive as a modern warship is inherently going to involve politics, which limits engineering options. And as for iterative ship classes, well, I agree we got that wrong with the Zumwalts and LCS, but that's basically the story of the Spruance/Tico/Burke extended family. As for modular weapons, I agree we should adopt StanFlex, but even that has limitations. As Samuel Johnson didn't say, what is good isn't new and what is new isn't good. This is a field where a lot of smart people have been thinking about things for a very long time. Sometimes abstractions work, but in that case, you're likely to find out that you're either reinventing the wheel or proposing something that's already been considered pretty seriously. When they don't, you end up making silly and obviously unsuitable suggestions.

And while I've spent most of this beating up on Austin Vernon, that's mostly because, full credit to him, he did an unusually good job of showing his work, which made it easier to take apart. But he's far from the only culprit. Another recent example is Ares Industries, a startup that got a fair bit of press a few months ago for their attempt to build a cheaper anti-ship missile. Unfortunately, while I am immensely sympathetic to their goal, a lot of that press made little sense: "Existing anti-ship cruise missiles are low-volume, multi-million dollar, 3,000 lb behemoths meant to take out large cruisers and destroyers. But Chinese shipbuilding capability is far superior to ours, and we can’t make enough anti-ship missiles to counter. In addition, huge missiles aren’t needed to take out the smaller corvettes and frigates that make up the People’s Liberation Army Navy." This is a case where the obvious abstraction ("frigates are smaller than destroyers") leaks a lot. The obvious adversary frigate, the Type 054A, is about 4,000 tons. Their numbers seem to hint at a 300 lb missile, which implies a warhead of 100 lb or so. A typical anti-frigate missile (something along the lines of Harpoon or Exocet) has a warhead of 400-500 lb. The obvious counter is that those are just intended to hit anywhere on the ship, and that you could do better by placing the warhead precisely where you want it. That much is true, but NSM has been pulling that trick for a decade, and it has a 260 lb warhead. As it is, I remain skeptical that 100 lb is going to be enough against a Type 054A, which has the size and toughness of a front-line warship, but it's an easy story to sell if you simply pull out the abstraction.¹⁰

Unfortunately, the practice of abstracting away a lot of detail seems to be spreading to the Pentagon, in particular with the adoption of the Navy's new Distributed Maritime Operations concept. The problem is that while a lot of the buzzwords involved sound fine when you're looking at a powerpoint, it doesn't work quite as well when you sit down and start to ask what ships are carrying which missiles and how they're getting targeting data. Despite this, the Navy appears to be investing heavily in the concept, presumably because former SecDef Mattis finally met an enemy he couldn't defeat: PowerPoint.

The key point to take away from all of this is that expertise is important when dealing with complex fields, and defense is exactly such a field. If you are not careful to make sure you have a good grasp of what can go into a box and when that box will leak, you are likely to end up saying very silly things. Unfortunately, learning about defense is hard, and I don't know of a good way to grasp all of this other than reading an absurd number of things over a lot of years. And because defense is full of specialized and complex niches, you should be very suspicious of anyone who tries to bulldoze it all and solve problems with a nice abstraction that isn't carefully grounded to the fundamental question of "what systems are we using here and what are they going to do?"