Yes, the census is worth more than one Big Mac a year

The census is the biggest political issue that most people don't care about.

Much of what we know about our nation's people — their ages, their races, their incomes and housing — comes from Census Bureau workers patiently designing and tabulating its surveys. Their numbers frequently set the terms of major political debates, starting with political redistricting. And of course, there is the original purpose for which the census was established: counting the number of people in these United States, and apportioning representatives to each state according to those findings.

The census, in other words, matters a great deal. So it matters a great deal that Commerce Secretary Wilbur Ross is urgently petitioning Congress to give him more money so that we don't screw it up.

The worries over the census distill the essence of a typical Washington story: The things that sound big ultimately don't matter much. Small details and dull statistics turn out to be mighty levers that can shift the whole government.

Start with the big thing that doesn't matter: the amount of money that Wilbur Ross is asking for. That's $15.6 billion, roughly a fourth more than Obama-era projections that he dismissed at a congressional hearing today as "overly optimistic."

This is, of course, objectively a great deal of money. If you had that much money, you would be the 64th-richest person in the world. However, that's not the right way to think about government expenditures. What matters is the cost to individual Americans, and this works out to about $48 apiece, for a procedure we perform only once every 10 years. That's roughly the cost of one Big Mac Meal a year — an eminently reasonable price to pay to ensure that political representation matches as closely as possible the number of people in each state.

But why is the cost rising?

But then, it's tempting to argue that the government ought to be able to do more with less. We did, after all, have a technological revolution that has enabled companies to slash costs, and manipulating this sort of data is exactly what technology is best at. Why can't the Census Bureau follow where the private sector has led?

Secretary Ross was asked approximately this during the hearing, and seemed slightly bemused by the question. While he was critical at several junctures of the sometimes absurd ways in which the government violates corporate best practices, he understands what the congressman didn't: The Census Bureau simply cannot replicate what companies do with IT.

For one thing, it is hemmed in by government procurement rules. Above a certain size, all institutions start choking on the managerial controls that are required to keep money from leaking out of the treasury like a sieve. And the federal government, which is the biggest institution in the country, has the biggest red tape problem.

Add in a bevy of rules designed to ensure that no one ever abuses their power to reroute government funds to friends, family or their own pocket, and you have a recipe for a procurement process that is ponderous and dysfunctional. That sort of system rarely produces cutting-edge technology. It's more likely to produce, well, the disastrous initial launch of the federal Obamacare exchange.

But the Census Bureau has its own peculiar problem on top of the general issues with federal government IT, because it does the census only once every 10 years. Even if it managed to design and build a cutting-edge census taking system, that system would be obsolete by the time the next census rolled around. And since it's only used once, you can't build the way Silicon Valley does: starting small, and then improving through iteration, and gradually bringing it up to scale. This kind of bespoke, single-use system is an expensive endeavor that will never stay around long enough to generate the kind of savings that corporations wring out of technology over time. And it's unlikely to reach their level of quality, either.

Private-sector researchers deal with these challenges by resorting to a variety of statistical techniques to guess at the data that falling response rates are costing them. But that's not a choice open to the census, which is required to count everyone. Unfortunately, nonresponders are wildly expensive compared to the millions of dutiful citizens who mail off their forms the week they get them. Most people, Ross said, will be pretty cheap to reach, but the last few million will require expensive man-hours tracking them down and cajoling information out of them. The more of those folks there are, the more expensive the census will become. A private company would probably write those folks off.