Finding Census numbers... and using them correctly
The U.S. Census Bureau's Web site has a huge amount of data that can be the basis for entire stories, or to enhance reports with a nugget or two of information.
There are more than 65 links on the census.gov Home page. Three are especially helpful in finding data quickly for news stories.
- Populations
- The Population Finder on the right side of the page provides quick and easy navigation to the latest population numbers for each city, county and state.
- Demographic profiles
- The Find an Area Profile selection box on the right is the quickest way to get demographic data on dozens of topics for each state, plus larger cities and counties.
- American FactFinder
- The link in the left rail is your gateway to detailed Census information, including (1) annual estimates since 2000; (2) 2000 Census data and (c) 1990 Census data.
The Census Home page features many helpful links. See a larger version of this screenshot for a quick look, or go directly to the Census Web site to begin exploring.
Decennial Census vs. American Community Survey
The annual estimates from the Census Bureau are not the same, or as complete, as the information produced from the Census every 10 years.Decennial Census
- The bureau's best, most reliable data comes from Census each 10 years.
- Every effort is made to reach each household to complete a Census form with basic information.
- One out of every six American households completes a longer form with more questions.
- Of note:
- Remember, several years have past since the 2000 Census. Before using the data in your story, consider whether the population makeup may have change significantly for the data you're citing.
American Community Survey
- The best resource for recent data.
- The bureau releases annual estimates for larger places based on a survey.
- Data is released for places of at least 25,000 people (Until 2007, the cutoff was 65,000).
- Of note:
- Like any survey of a sample of people, there is a margin of error. Take care to check the margin of error released with each number to judge whether conclusions should be drawn for rankings and other purposes. This is especially true where a small number of people fit a particular demographic.
- When it is not important to use data from the most recent year, consider instead getting data from the 3-year estimates. The latest 3-year estimates are for 2005-07. The data is a bit older, but the margins of error are much smaller because more people have been surveyed.
- Do not mix 1-year and 3-year American Community Survey information. The numbers usually will be different.
- Do not use the American Community Survey for population estimates. The survey understates the total population for most places because it is a snapshot of the people reached for the survey. The Census Bureau provides separate population estimates.
Population Estimates
The Census Bureau each year produces population estimates for each township, village, city, county and state. The best estimates are done for counties. Here's how it is done.
State populations
- State population estimates are simply the sum of each county's population.
County populations
- The Census Bureau takes into account a variety of information.
- Births, based on recorded births.
- Deaths, based on recorded deaths.
- Migration for those under age 65, based on federal income tax returns.
- Migration for those 65 and over, based on Medicare enrollees.
- Net international migration, based on information from Census Bureau surveys, Census 2000 and military movement records.
- The Census Bureau takes into account a variety of information.
City, village and township populations
- The bureau follows these steps, a less accurate process than what is done for the counties:
- Creates a new housing unit estimate for each place by taking into account:
- Number of housing units from Census 2000.
- Estimating new units constructed since 2000, based mainly on new housing permits.
- Estimating units lost based on previous trends. For 2008, the estimate was based on the trend found in the 1997-2000 American Housing Survey.
- Multiplies the new estimate of housing units for each place by the 2000 Census number of residents per household for each place.
- Multiplies the result by the Census 2000 occupancy rate for each place.
- This results in a preliminary estimate. An adjustment is made by following these steps:
- Sums the preliminary population estimate for each place in a county.
- Each city, village or township is then adjusted up or down at the same rate countywide so the population for all the places equals the previously released, and more precise, county population estimate.
- Creates a new housing unit estimate for each place by taking into account:
- The bureau follows these steps, a less accurate process than what is done for the counties:
Something to keep in mind about population estimates
This method, while very sound, potentially can be flawed for certain places.
What if housing vacancy increased sharply in one area of a county, while it was stable elsewhere?
This could result in an artificially high estimate for the city where the vacancy rate has increased, and an artificially low estimate for areas where the occupancy rate has been stable.
Remember, 2000 Census data is used for housing occupancy. It is the latest available.
Are the estimates right for Cuyahoga County cities?
A real example of this potential flaw may have occurred in Cuyahoga County, Ohio, where the latest estimate of 1,283,925 is a drop of 7.9 percent from 2000 to 2008.
The 2008 city estimates show a number of inner-ring suburbs of Cleveland incurring among the sharpest population declines in the state.
But Cleveland represents one-third of the county's population. What if Cleveland's vacant housing rate has increased far faster than in the suburbs?
If this is the case, Cleveland has lost more people than shown by the latest estimate (down 9.2 percent since 2000 to 433,748 in 2008).
And more people would live in the suburbs than the latest estimates show.
The answer will be produced with Census 2010.