U.S. CENSUS DATA: Background & Choosing a Dataset

U.S. Decennial Census of Population & Housing ("The Census")

• Collected every 10 years, starting in 1790.
• Goal is to count every single person living in the United States on April 1 of the Census year. 
• The questions have changed over time. 
  • Visualization: https://pudding.cool/2020/03/census-history/
  • Starting in 2010, the Census form is very brief and collects only basic housing and demographic information. Questionnaire for Census 2020: https://2020census.gov/en/about-questions.html
  • Prior to 2010, the Census had a "long form," sent to 1 in 6 households, that included more detailed questions. The long form was replaced by the American Community Survey (see below). 

American Community Survey (ACS)

• Complimentary survey to the Decennial Census; earliest available data is from 2005.
• Sample of ~300,000 households per month on an ongoing basis. 
• Data are released in two kinds of datasets:
  • 1-year estimates (includes data from 12 months of surveys) and
  • 5-year estimates (includes data from 60 months of surveys)
• Questions are more detailed than those in the Decennial Census. Questionnaire for 2020 ACS: https://www2.census.gov/programs-surveys/acs/methodology/questionnaires/2020/quest20.pdf
• Because the ACS is a sample survey, it has margins of error. There are also smaller areas that are not included in the datasets because of disclosure risk.  

Evolution of the American Census Data Viz

• Interactive digital timeline documenting historical changes to Census questions and organization 
• Text and data viz by Alec Barrett and TWO-N
• Published at The Pudding

Which dataset/survey should I use? 

There are pros and cons to all of them, but here are some things to think about as you decide:

U.S. Decennial Census

Example: "Census 2010"; also sometimes labeled "DEC Summary File 1" or "DEC S.F. 1"

• Most accurate (+)
• Includes all geographies (+)
• Only every 10 years, so is out-of-date quickly (-)
• Limited scope of question topics (-)

ACS 5-year Estimates

Example: "ACS 2018 (5-Year Estimates)"; also sometimes labeled "ACS 2014-2018 5-year Estimates"

• Broad scope of question topics (+)
• Released every year, so usually more current than the Decennial Census (+)
• Sample survey, so has a margin of error and does not include data for geographies where the population is smaller than 20,000 [citation]. (-)
• Utilizes 5 years' worth of surveys, so more accurate than 1-year estimates (+)

ACS 1-year Estimates

Example: "ACS 2019"; also sometimes labeled "ACS 2019 1-year Estimates"

• Broad scope of question topics (+)
• Released every year, so usually more current than the Decennial Census (+)
• Sample survey, so has a margin of error and does not include data for geographies where the population is smaller than 65,000 [citation]. (-)
• Utilizes 1 year worth of surveys, so less accurate (with larger margins of error) than 5-year estimates (-)