Driving Blind

BlogJun 1 2012

Imagine driving a car with no dashboard—that’s right, no speedometer, no gas gauge, no oil light, no odometer; strictly seat-of-the-pants.  Or, imagine trying to plan a party without knowing how many guests are coming, where they’re coming from, how long they’re staying (you have extra beds, right?), what they’re bringing (are kids OK?), what they don’t eat—or even who they are.

No responsible family, and certainly no developed country, attempts to function in the absence of basic information about its members.  The U.S. Census, mandated by the framers of the Constitution, supplies data that form the basis for literally thousands of critical decisions, including ensuring our “one person, one vote” system of political representation, as well as helping our schools, businesses, military forces, and retirement communities plan to meet the needs of a rapidly changing population.  Think “market research” at its most basic level.

While the decennial Census is a complete count of the population (tabulated by age, sex, race/ethnicity, and a few household characteristics), the American Community Survey (ACS) is the source for all the detailed socio-demographic information— marriage (status and history), residential mobility, veteran status, languages spoken, immigration and ancestry, disability, and numerous other topics—that was formerly collected on the much longer Census form.

But the key features of the ACS are its timeliness and its ability to produce updated estimates for even the smallest geographical areas (counties, cities and towns, and even neighborhoods).  The ACS accomplishes this by using a “continuous sampling method.”  Every month, households selected to be representative of the population are surveyed.

By accumulating their responses over a 12-month period, a sample is achieved that is sufficiently large to report estimates annually for the nation, all states, and all other jurisdictions (for instance, most counties and cities) with populations of at least 65,000.  Three years’ of data are necessary to report estimates for places with populations less than 65,0000 but at least 20,000; and five years’ of data are required for the smallest communities.  These multi-year estimates are updated every year by dropping the oldest year’s data, and adding the most recent year’s (so that 2006-2010 becomes 2007-2011).  That way, while wholly new data are available for some regions only every few years, estimates are still refreshed yearly.

At Child Trends, we know first-hand the value of this resource.  ACS data are among the components used to produce the Census Bureau’s population estimates.  These estimates, in turn, form the basis for calculating important indicators, such as rates of teen births, child abuse and neglect, and youth voting.

Because so much important planning occurs at the community level (think schools, transportation systems, housing), having more-frequent data has been a huge boon to people who work in smaller rural or urban communities.  Prior to coming to Child Trends, I worked in state government, and my tenure there corresponded with the advent of the ACS. In Vermont, the prospect of having, within a few years, annually updated data for even the smallest communities was not only enthusiastically welcomed, but was critical to our new focus on data-driven decision-making.  We could, of course, count the number of teen births; but that didn’t provide the information necessary to compare rates over time or with states like New York.  The ACS provides the population data needed to calculate rates.

Why does the ACS use sampling?  Because a complete count is very expensive.  We can reduce the burden (and cost) of a “count everybody” approach by selecting much smaller subsets of the population that will provide valid estimates, if carefully chosen.   This is no haphazard process, but one based on science that is widely used in the business sector, by opinion polling firms, and by governments around the world.

Making the survey explicitly “voluntary,” as some have suggested, would not only add greatly to the work (and cost) of surveying, but could produce data that are less representative of the whole population.

The American Community Survey is the Census updated for the 21st century.  The ACS now delivers on an annual basis the information our communities used to get just once a decade.  Does anyone seriously believe that in today’s world we want to base decisions on data that are ten or more years old?  Few enterprises would be content with, or survive long, with that kind of information; as a nation, we deserve at least as much.