By Terri Ann Lowenthal

Sometimes, words escape me. (At least, words that are printable in a respectable, philanthropy-funded blog about the sacred foundation of America’s democratic system of governance, still the envy of the modern world, imperfect though it is.)

So let me just say this: Really, Congress?

The very first task the founding fathers gave you in the U.S. Constitution—to direct the taking of a census once every 10 years—and you kick the can down the road? With the decennial clock ticking and the window of opportunity to figure out how to make it all work for less money closing fast? Words are failing me.

Lawmakers are trying to wrap up a broad spending bill for fiscal year 2015, which started on Oct. 1, before a short-term funding measure runs out Thursday night. The draft bill, unveiled Tuesday, allocates $840 million for the account covering the 2020 Census, $123 million less than the budget request. Congress essentially is cutting the proposed ramp-up for decennial census planning by almost half. The Obama Administration’s proposed 28 percent funding boost might sound like a lot, but as Arloc Sherman of the Center for Budget and Policy Priorities noted in a recent Huffington Post blog, mid-decade ramp-ups for the 2000 and 2010 Censuses were 30 percent or more.

Most of the increase the bureau requested relates to 2020 Census planning. 2015 is a pivotal year: the Census Bureau will conduct three major field tests to inform its selection of the 2020 Census design by next fall. A fourth test, scheduled for late summer, will evaluate revised questions on race, ethnicity and household relationship, as well as strategies for boosting Internet response and for helping language minorities participate.

Congress doesn’t want to pay more for the 2020 Census than it did for the 2010 count. The Census Bureau has to meet that goal while maintaining accuracy and trying to reduce the historic, disproportionate undercount of people of color, low-income households, rural residents and young children. It will take a big change in census methods to pull this off, as well as careful research, testing and preparation to be sure those reforms work. The payoff for investing in the groundwork now is significant: $5 billion in potential savings from automating response options and field work and from tapping government and commercial databases to update the address list and reduce costly door-to-door visits. All promising ideas, but we won’t know if they can produce a lower-cost and equally or more accurate census until we see and weigh the evidence.

Now the Census Bureau is really in a bind. It is wrapping up the first test, which focused on administrative records, aerial imagery and other governmental and commercial sources to update the master address list and digital mapping system. Preparations are underway for two tests—one in Maricopa County, Ariz.; the other in the Savannah, Ga., media market—with a “Census Day” of April 1. These are crucial research opportunities in census-like environments: the bureau will evaluate the use of administrative records to streamline and reduce the cost of door-to-door follow-up visits; targeted digital advertising to boost self-response among hard-to-count demographic subgroups; ways for people to respond via the Internet without a pre-assigned identification number that links them to a specific address; and new contact and notification strategies to cut down on paper communications and encourage prompt participation.

These initiatives aren’t incremental improvements on traditional census methods. They are significant departures from the tried-and-true mail and door-knocking design. They might work. They might not. But the Census Bureau can’t wait another two or three years to figure that out. It has one year to decide which methods hold enough promise for saving money without sacrificing the accuracy of the count and the quality of the data, in order to move ahead with IT systems and operational development. The decision is already a year overdue, thanks to previous budget cuts and sequestration.

Delaying or streamlining the 2015 tests would put effective 2020 Census reform in serious jeopardy. If the bureau pushes ahead with the full testing schedule, something else has to give. The Census Bureau can’t put off systems development; the risk of failure is too great. Other vital components of a successful census—the Partnership Program and advertising campaign— could be put on the back burner.

Other programs funded through the same account might take the brunt of the budget cut. The bureau could trim American Community Survey coverage of group facilities such as college dorms, military barracks and nursing homes, or cut out data products; it could slow down planning for the 2017 Economic Census. It could ditch its new initiative to build an enterprise system for data collection and processing, which it hopes will replace numerous (and costly) survey-specific systems.

I don’t know what hard choices the Census Bureau will make in the coming weeks and months. But here’s what I do know: Congress is responsible for a fair and accurate decennial census. The Constitution says so. And right now, it is really blowing it.