A statistician helps planners get inside the Census numbers.
As MU’s Jonathan Bradley sees it, there are 400 billion reasons for using U.S. Census data more effectively.
The figure is the dollar amount expended each year by federal, state and local agencies for infrastructure, community development and social services projects. Decisions concerning how much gets spent where often boil down to information gleaned from the American Community Survey (ACS), the U.S. Census Bureau’s multi-year repository of wide-ranging data elicited from across the nation. “While the ACS data are quite helpful for those needing specific data, you can’t get to the data that you want in a lot of cases,” says Bradley, a postdoctoral fellow in statistics.
“For example, the Department of City Planning in New York may be interested in making policy decisions based on community districts, data that doesn’t exist within the ACS as multi-year estimates. Currently, the ACS can only provide annual data for populations of more than 65,000. Also, the ACS only publishes one-year, three-year, and five-year period estimates, so planners wishing to access data on different time periods are unable to acquire the information they need.”
Bradley and fellow MU researchers Christopher Wikle and Scott Holan, both MU professors of statistics, have come up with a fix using spatial and temporal correlations to produce precise estimates. The result, they say, will allow ACS users to estimate demographic variables at any time period and for any geographical location. For rural and isolated communities, their method could be particularly helpful.
“In our methodology we were able to make accurate estimates of income on Native American Reservations—data that normally is not easily accessible,” Bradley said. “We were motivated by the section of language assistance from the Voting Rights Act, to determine if regions that require language assistance also require other types of assistance based on their income status. Additionally, planners will be able to access the data and estimates at any time.”
Bradley’s research is part of the Missouri node of the NSF-Census Research Network, an interdisciplinary team addressing the methodological questions related to the Census’ shift from the 10-year long-form data to the ongoing survey that releases data annually.