Not only was George Washington 6 foot 20, had a wig for his wig, and a brain for his heart, but his mental prowess conceived of some Statistical Issues that arise from census operations.
“… by which it appears that we shall hardly reach four millions; but one thing is certain our real numbers will exceed greatly, the official returns of them…”
Washington’s quote was one of the opening slides of Dr. Stephen E. Fienberg’s Morris Hansen lecture discussing possible avenues of the 2030 Census.
Even with today’s modern technology and know-how, the huge difficulties in Census taking is compounded as our population grows. The key problem, is “undercounting” a person.
Dr. Steve’s wonderful talk covered a large scope and a one line summary is to pursue a bottom up approach instead of (the current) top down approach.
Top Down: Geographies –> Locations –> Households –> Families –> Individuals
Bottom Up: Individuals –> Families –> Households –> Locations –> Geographies
By combining modern tools such as:
- Statistical linkage (infering if two people from different datasets are one in the same)
- Error Propogation by modeling every component involved in a Census operation
- Spatial models to assign people to locations
By sketching out the process, an apparition of this road is slightly visible.
One “humane” issue I see is during the Statistical Linkage Step. From The Snowden Files, are the general public warm to having a register of people by combining “Administrative Records?”
Some thoughts, taking neither sides:
- People are giving up their information to private business’ such as Facebook and Google
- These business’ profit from the information
- Shoe-phone Agencies (NSA) ransack and Deebo the information
- “Humane” Agencies (Census) using puny Dell’s in place of Shoe-phones care about the products’ sensitivity.
- Would people be willing to “lend” their info for a more “humane” cause?
- By “humane” i mean, public Census estimates are summarized and censored
Most importantly, the Census estimates are publicly accessible and useful for a wide range of downstream applications (check out this mobile app http://www.census.gov/mobile/ )