I’m happy to announce that I’m an Emergent Ventures fellow! I start with the Mercatus Center at George Mason University on October 22.
My venture is reforming the weapon systems acquisition process in the Department of Defense. Not that there is a lack of reform activity. I want to shift its focus and historical narrative. My views are pretty well spelled out on the History, Technology, and Competition of this area. I find that the reason acquisition reform fails is because it doesn’t address the flow of funds through the program budget.
The purpose of this post isn’t to rehash these ideas. I’m constantly looking for arguments that force me to revise my position rather than confirm it. As Tyler Cowen recommends, we should always “steel-man” the other side’s case. From here on, I will present what I consider a good case from the other side.
__________________________
__________________________
Defense acquisition cannot be effectively decentralized. Generally, 70% of lifecycle costs are in operations and sustainment activities. If we let a bunch of competing systems get developed and produced, the logistical tail will be unaffordable. Tough decisions will have to be made. Resources will be wasted. Force structure will be in disarray.
 |
| Generic system life-cycle cost. Generally, 70% is spent in O&S. |
Some say that rivalrous competition will only exist in R&D. Production decisions will be centrally controlled. Those organizations which fail to get cost-effective systems into production will then be punished.
This is a fantasy. The argument rests on the idea that redundant R&D projects will provide not only more types of systems, but better systems. Yet it isn’t clear that a “diversity” of R&D projects will get you combat effective systems at all. The result is endless tinkering on pet technologies, not fully operable systems. I’ll explain.
Redundancy used to be the buzz word of reliability engineering ever since John von Neumann’s 1956 paper. He found that a system (e.g., a computer) could be more reliable than its constituent parts (e.g., vacuum tubes) through redundancy. When one part fails, back-ups are available. The idea caught on to theories of administration as part of a backlash to Weberian straight-line hierarchies of zero redundancy. The market appeared to be an example of a redundant system, more reliable than any firm.
Redundancy, however, doesn’t provide the insurance that is often assumed. Systems are only reliable if the redundant components fail independently of one another. Redundant engines on an airplane will not save you from volcanic ash, for example. They will fail simultaneously.
 |
| NASA tests a volcanic ash detection system for airline engines. |
Similarly, when R&D organizations come from similar cultures, they will pursue the same military requirements with similar engineering designs. They will then fail in similar ways. More cash will be burned than required.
Redundancy may then lead to excessive risk-taking. Perhaps each team will believe there is another one working toward the most feasible solution. They will feel justified in taking more radical shots with high expected payoffs but low probability of success. This increases the “diversity” after all. But most will fail. There’s no inevitability of success, particularly because the scale of defense projects is so large that the numbers of things that must go right is massive.
We could wait many decades to collect just one of these “black swans.” All the while, affordable systems—the unglamorous—went unpursued.
More likely, such R&D gambles will only work toward proving out one component or subsystem. Even if we do collect several successes there, they still need to be integrated into a system that is producible, affordable, and more lethal than the alternatives.
Because redundancy leads to technological tinkering of components and subsystems, additional management systems are required to tie them together. Building blocks, no matter how revolutionary, still need to be integrated. Unless they gain central direction, management systems are destined to fail.
Just as seemed to happen before World War II, decentralized in-house R&D resulted in endless tinkering. There is a sort of social shirking that occurs. Good-enough systems are already designed. They’re working on the next big thing. But these technologies were rarely connected to the soldier’s needs. They rarely end up providing operational systems.
Oh, by the way, we already have decentralized innovation going on at DARPA. They have achieved much success. But there is a reason DARPA only proves out emerging technologies. They would require much more cash and central coordination to integrate numerous technologies into fully developed systems. Failure at such scales is not an option.
The idea that weapon systems will arrive faster and cheaper due to decentralized R&D is false. If anything, such an arrangement will lengthen the “roundaboutness” of the stages of production. Organizations will focus on specializing and perfecting their technologies. How they integrate into a cost-effective system is never made clear. There is no price mechanism in the DOD acquisition system that allows an “invisible hand” to work. Coordination must be centrally directed like it is in the firm. It is dangerous to assume that weapon systems and coherent strategies will order themselves spontaneously.
____________________________
____________________________
I think that’s a good argument. Ultimately, I don’t buy it. Stay tuned to hear why decentralization, particularly for R&D, can get the DOD out of the mess its in.
Notice that the anti-manifesto above made a good argument without falling back on traditional claims of the central planner resting on the efficiency of portfolio optimizations.
Notice that the anti-manifesto above made a good argument without falling back on traditional claims of the central planner resting on the efficiency of portfolio optimizations.
I forget where I found this story, but when Markowitz (founder of asset portfolio theory) came to RAND, he was asked whether he used his theory to allocate his retirement. No, was the response, he just split it 50/50 to stocks and bonds.
And yet because of RAND, portfolio theory became the foundation of all defense resource decision-making. More to come.
Good post, I agree.
Please cover this in a future post, the propriety of hiring a lobbyist while in government service: https://www.realclearpolicy.com/articles/2018/10/12/congress_strikes_back_at_cronyism.html [T]he process of crafting JEDI bears all the hallmarks of the swamp that Trump has vowed to drain. Though there has long been talk about the Defense Department joining the cloud, the current call for bids was put together only after Defense Secretary James Mattis hired a D.C. lobbyist who had previously consulted for Amazon.
Regarding MPT: I once set up a Markowitz portfolio using Excel (using "Solver" to find the best mix of stocks and solve the Linear Optimization equations; wow I'm ambitious). The trouble is, I found out this analysis is perfect for backward looking analysis but useless for the future, since the covariance matrix between stocks often changes quickly, making the optimal mix obsolete. Wikipedia says: "The risk, return, and correlation measures used by MPT are based on expected values, which means that they are mathematical statements about the future (the expected value of returns is explicit in the above equations, and implicit in the definitions of variance and covariance). In practice, investors must substitute predictions based on historical measurements of asset return and volatility for these values in the equations. Very often such expected values fail to take account of new circumstances that did not exist when the historical data were generated."
Nevertheless, the MPT portfolio I did set up has outperformed the S&P500 index in the last few years. I have good positions in Macy's, Walmart (I don't believe the 'drone delivery' hype) and QCOM, all of which did well, go figure! Bonus trivia: I'm getting into Fannie Mae and Freddie Mac penny stocks now, and so should you! They'll be privatized within 10 years, you watch!