Current Navy plans call for construction on the first of 12 Columbia-class nuclear-armed submarines to begin in 2021, with an overall price tag of $128 billion. But the GAO’s latest audit of the program found that the Navy both failed to account for typical shipbuilding cost increases, and the complexity of actual “hands-on” labor hours it will take to build the boats…
In a letter dated Feb. 1, Kevin Fahey, the Pentagon’s assistant secretary of defense for acquisition, wrote in response to the GAO that the Navy’s cost estimate for the Columbia will be updated in 2019. He also said that the Navy disagrees that labor and cost estimates are faulty, but instead, the “supplier base remains the highest risk to construction readiness” and the Navy is investing in supporting the industrial base in order to shore up some of those risks before they impact the program.
That was a good report from Breaking Defense. This is what Fahey was probably pointing to in that last part (from the GAO study):
… to achieve Columbia’s aggressive construction schedule, while simultaneously building Virginia class submarines, the shipbuilder is working to ensure that it has sufficient shipyard capacity—including new facilities, additional suppliers, and an increased workforce.
The Virginia-class is the fast-attack submarine that has a Program Average Unit Cost (PAUC) of $3.4 billion (then year). Recently, Electric Boat went on a “hiring frenzy” to staff up personnel. In late 2018, it looked to expand nearly 2,000 jobs. Here’s the GAO:
To support growing workload from both the Columbia and Virginia submarine programs, the shipbuilder plans to increase workforce at its two facilities over the next decade: by 66 percent at Quonset Point, Rhode Island—where the components and individual submarine modules will be constructed—and 174 percent at Groton, Connecticut—where the super modules will undergo final outfitting and assembly.
The problem is that the government’s plan to disburse money may be faster than the contractors can ramp up operations. This seems to be the case for the B-21, among many other programs. One issue is that when existing programs stretch-out, they continue absorbing the labor that could otherwise have moved to the replacement program, creating this temporary surge requirement.
It is interesting to note that Admiral Rickover took only 3 years (1951-1954) to design and construct the first nuclear submarine ever, the Nautilus. The detailed design of the Columbia-class started in 2011, and lead-ship delivery is planned for 2027, a whopping 16 years.
Here are some issues from the GAO on cost estimating:
The Navy anticipates that it will need 12 million labor hours to directly construct the lead submarine—referred to as touch labor. This represents 17 percent fewer labor hours than what was needed for the lead Virginia class submarine, when adjusted for weight differences.
… the Navy’s independent assessment of the cost estimate calculated that production costs could increase by $3.59 billion in constant year 2010 dollars if a learning curve of 93.9 percent was realized, rather than the 88.9 percent rate estimate…
In addition, the Navy assumes a higher rate for Columbia multiyear procurement savings than what has been typically achieved for other programs…
A reliable cost estimate is critical to program success. It provides the basis for informed investment decision making, realistic budget formulation and program funding, meaningful progress measurement, proactive course correction when warranted, and accountability for results.
The Columbia-class cost estimate is seen as unrealistic because, for the lead ship, it estimated 17% fewer touch labor hours-per-pound for a system that is in many respects more advanced than the Virginia-class analogy (even though it looks like the stern area, and not the nuclear launch, has the lowest technology readiness level). This is the essence of supposedly “realistic” cost estimates. Find historical analogies, make some adjustments and normalizations, then see where the numbers come out.
Similarly, in production, the Navy estimated 15- to 20-percent savings on multi-year contracting (like a bulk buy discount), when it had only estimated 10- to 15-percent savings for the Virginia-class. From these historical data, with a few more assumptions, you can see the basis for an upward adjustment to the cost estimate.
The problem with such “realistic” cost estimating, also called “will cost” analysis, is that it takes as sacrosanct existing cost figures and may perpetuate gross inefficiencies or neglect new opportunities. Here is Ernest Fitzgerald discussing the problem:
… over-dependence on the probable cost estimating techniques has had a bad effect in other areas. To begin with, since the techniques used do not recognize inefficiencies in the bases used for projections, the approach tends to build excess costs into future estimates. For example, the cost estimates for the new generation of fighter aircraft, the F-14 and F-15, are heavily influenced by cost experience on the F-111, which is highly suspect to say the least.
Fitzgerald preferred the “should cost” analysis, in which consultants walked the contractor’s line and reviewed operations to ensure the lowest possible price through recommendations to cost accounting and industrial engineering.
However, that kind of analysis only seems to work in high-rate manufacturing areas, not R&D like lead-ship design and construction.
The usual perspective is that “should cost” analysis is good for negotiating contracts with suppliers, to get the very best price, but that “will cost” is preferred to make budgeting decisions to ensure adequate funds are available if contractors can’t make the “should cost” figure and also so that program approval and tradeoff decisions are not biased.
Independent oversight offices prefer to recommend higher cost estimates and more funding so that programs avoid cost growth, which leads to public scrutiny and so forth. It is obvious that if you’re performance is judged on cost growth, then you want a higher baseline cost estimate so you have more “wiggle room.”
It’s like budgeting $100,000 for a Toyota Camry, and when you come in at $100,000, you claim acquisition success! Clearly that is not success. In fact, the program only gets anywhere close to “efficient,” “fair,” or “reasonable,” when you get a cost savings of 75% to the baseline.
Here’s the other way of looking at this problem from former CAIG director Don Srull, this time when we consider overly optimistic “buy-in” cost estimates:
“Cost growth”, in fact, was a misnomer. The main culprit was not poor management, sloppy or lazy contractors, or lack of clever contract details that led to costs growing above the original estimate. It was simply unreasonably low beginning estimates. One could not after all buy the new Cadillac automobile for $5,000. When it turned out to cost $25,000 it was not “cost growth”—it was an unrealistic, poor initial underestimate!
The problem is that with defense equipment, we simply have no benchmark for understanding price level, so we control programs using measures of cost growth to baseline. The baselines are set from the experience of past programs, grounding them in reality. But when you’ve had poor performance in the past, that inefficiency is built into the future baseline. Then, it gets adjusted up even further because of new capabilities, inflation, and so forth.
Usually, line managers want to get a program approved and get it accomplished for the lowest overall cost, even if that tends towards a bias. However, officials not responsible for executing the program often want higher program cost estimates out of a sense of accountability dominated not by judgments of best value, but by minimizing future changes to the plan — it could throw the whole portfolio out of balance.
Acquisition success is not determined by execution to plan. Oftentimes, our biggest gains come as surprises that we opportunistically follow up on. Acquisition success is determined by the subjective valuations we give to the equipment actually developed, and that depends on numerous considerations about capabilities and the price-level, not necessarily the degree of cost growth.
Returning to the Columbia-class program, the GAO also cited some Earned Value Management (EVM) data over the past year. The Cost Performance Index (CPI) fell to 0.86 by the end of August 2018. That means that it costed $1.00 to accomplish 86 cents worth of planned work. If the current performance continues, the total contract could see 16% cost growth. Not too bad, but the trend is downward, and no one believes EVM data anyway. It is simply too difficult to know whether it reflects reality.
Usually the contractors “accomplish” work, or “earn value,” by finishing drawings at this stage. But that doesn’t mean that the drawings were correct. It doesn’t mean future drawings, integration issues, or new requirements won’t result in rework. It doesn’t mean that there won’t be an increase in the total number of drawings required. We still have more than two years planned for doing designs, so we’ll see how that all changes. (Note: construction is planned to start in 2021 when 85% of drawings are complete, but when uncertainty rears its head and things change, it may be that construction starts with far less of the final drawings complete.)
It is interesting, however, that Columbia has admitted such a rapid decline in cost performance so early in the contract. Usually, in my analyses of EVM, you see reasonably good performance until 50-60% in, then they admit rapid declines in cost performance (Schedule Performance is often less useful because it can be gamed even more easily). But I don’t think Columbia is in danger of cancellation if they admit problems early, it is simply too important and the only option. The US taxpayers will have to fork over whatever it takes, because we’ve already gotten ourselves into it and the clock is ticking on the lifespan of the current fleet.
The real question is, if the Navy went through Milestone B with a figure 10-, 50-, or even 100-percent higher for the Columbia-class, would it have been cancelled? Delayed? What other options were there? What good would come of more studies to justify higher cost targets?
I suppose it was mostly about “not lying to ourselves,” that a much bigger portion of the budget would get tied up in this thing. It’s not like the public has much of a gauge for what is a good deal or bad deal on nuclear submarines (let alone for everyday healthcare items). If costs were 100% higher than “fair” or “efficient” prices, which is unseen by the public, but the public is told cost growth was only 10%, then things appear just fine. Private construction costs often grow by more than 10%, after all.
Then there’s Hirschman’s Hiding Hand principle, that if we knew the true cost of programs we never would have undertaken them, and after-the-fact, we usually are happy that we did even at the higher expense.
Leave a Reply