The government cost analyst’s parable


Joe, a consulting OR/MS analyst, was still thinking about his conversation with his accountant, Art, about the distortions accounting methods can introduce into management decisions. On his next visit to Art’s office, he reopened the topic. “You’ve convinced me that accounting methods and the incentives they create have really messed up the U.S. healthcare system,” Joe began. “But I know that government foul-ups are rarely confined to one area. What other weird things are they doing?”

“Well,” Art replied, “consider the way they analyze the cost of a new program. One of my colleagues, Steve, used to do this for a living before he joined our firm – partly because he’d had more than he could stand of how these government analyses are done. You remember, don’t you, from basic economics, how in a pure competitive market, the individual firm’s marginal cost is constant, but the overall cost of additional production in the industry is increasing, right?”

Joe nodded, somewhat uncertainly.

“Remember why?” Art continued.

Joe looked unsure.

“At the industry level, firms are competing for resources that become scarce if enough buyers want them,” Art explained. “So that cost element gets bid up in price as volume of production increases, but the individual firm doesn’t see those effects.”

Joe nodded again, somewhat more confidently.

“OK,” Art went on, “so now think about government programs. The marginal cost of any one program might be just the direct cost – the payments or subsidies plus the salaries of the people needed to run the program. But if you start enough government initiatives all at once, you create a shortage of people with certain skills and background. You also create competition for borrowing money! So the true marginal cost of government programs is higher than the direct marginal cost – but most analyses pretty much ignore the difference. See?”

“Yeah,” Joe agreed.

“Of course,” Art added, “there’s a flip side. The people who oppose all government expansion assume, in effect, that the difference is infinite. So then the two sides argue. But if there’s one thing we know from practical economics, it’s that virtually nothing costs either zero or infinity. Wouldn’t the whole discussion make a lot more sense if people began with realistic cost estimates?”

“Sounds perfectly sensible to me,” Joe concurred. “So why isn’t it done?”

“Politics, of course, is part of the reason,” Art said. “Advocates who know better may still try to convince everyone that the cost is either negligible or intolerable, depending on how badly they want to enact or block the program under debate. But there’s another problem – accounting methods again. If a company offers a new product or service, the costs and revenues happen within that company. You can see, with proper accounting and auditing, how well the new activities did. But government agencies typically don’t work that way.

“If, for example, FAA changes the way it does air traffic control, the benefits and costs mostly affect the airlines, the traveling public, other businesses that benefit from faster or less expensive or more secure shipping, maybe even the economy as a whole. If the Food and Drug Administration changes the approval process for new drugs, the costs and benefits are to the patients, the healthcare providers and the payment sources, not just the agency. So how do you accurately capture costs and benefits if all you have are data from within the agency?”

Joe frowned, “This sounds like a serious problem and more widespread than I’d imagined. I’m starting to see just how difficult it must be to do truly evidence-based cost-benefits analysis.”

“It is,” Art affirmed. “And timing is an issue, too. I’ve heard that one of your colleagues, Gerry McNichols, used to explain that cost-benefit analysis for a new defense system is like telling you what a new car will cost in 10 years, when you won’t have the full specifications of the car until seven years from now, and even then, after that, the requirements may change on short notice.”

“I think I remember that,” Joe smiled, somewhat ruefully.

“And if you’re not depressed enough yet, there’s a bigger and deeper issue than cost-benefit analysis of programs,” Art asserted. “Think about overall measures of how well we’re doing as a country. Like Gross National Product, or GNP, one of the most important measures we use as a basis for all kinds of government policies. We base a lot of policy decisions on whether GNP is going up or down.

“But remember what I told you before about income measures versus asset measures? GNP is an income measure. So if I spend $400,000 to build a house then decide to spend another $50,000 to tear it down and throw the debris into the river, I’ve added $450,000 to GNP. But asset accounting says I’ve just turned some useful materials into trash and polluted the river.

“That’s why I rather liked Obama’s insistence that stimulus money should be concentrated on building or repairing infrastructure assets, like dams and roads and bridges – and information technology, if we can agree about what’s beneficial there. But of course he ran into some of the same accounting difficulties anyway, because they’re so deeply imbedded in the way all policy decisions get made.

“So you quant folks who work in policy have some important work to do,” Art urged. “Insist on better metrics and better data as a basis for policy decisions!”

Doug Samuelson ( is president and chief scientist of InfoLogix, Inc., in Annandale, Va., and a principal decision scientist with Great-Circle Technologies, Inc., in Chantilly, Va.