September 2nd

9/2/2010 — Mark Partridge

Census in Detail

Many censuses have 2 different questionnaires (previous censuses, not 2010)

  • Long form: goes to a sample of households (could be large, several million…)
    • short form questions, disabilities, income, veteran status, citizenship status, migration, place of birth, etc…Goes to about 1/6 of households
    • US 2000 Long Form
  • “100% Data” [in theory] comes from the short form
  • “Sample Data” comes from the long form
  • “SF1” and “SF2” are 100% data (SF=Summary File)
  • “SF3” and “SF4” are sample data

Censuses, in theory, are

  • Universal (100% coverage
  • Simultaneous (single “snapshot” where no person is counted twice)
  • Individual (counting and recording who lives in households, not only their numbers or characteristics)
  • Private information for the US census is made publicly available after 70 years. (Names of respondents, respondent characteristics…)

Census residence definition can be:

  • de jure (legal place of residence, “usual”) — used by US
  • de facto (where are you sleeping on census day?) — used by other countries such as Australia

US Census Geographic Coverage

  • 4 census regions: Northeast, South Midwest, and West.
  • 9 census divisions: Pacific, Mountain, West North Central, West South Central, East North Central, East South Central, South Atlantic, Middle Atlantic, New England.
cendivco.gif
  • Small Area Geography. Commonly, states are divided into counties, towns, cities, blocks, census tracts.
census_small_area_geography.jpg

A Census is a big administrative and logistical undertaking for collecting data:

  • setup administrative system: supervisors, statisticians, enumerators, software developers,…
  • developing questionnaire
  • cartography
  • pretest
  • design data processing system
  • follow up for non-response
  • consistency checks
  • imputation and “correction”

imputation flag, if flag=1 this variable was synthetically generated, not an item that was completed by respondent or was completed incorrectly by respondent