Friday, December 10, 2010

Frequently Asked Questions, Part 2

For Part 1, click here.

6. What is a “multi-year period estimate” and how does it differ from numbers I got from the long form in 2000?

A. In 2000, a sample of housing units received a long form, which contained the same demographic and housing questions asked at every housing unit plus more than 50 questions on social, economic, and housing characteristics, such as language spoken, education, income last year, and number of rooms in a house. The ACS questionnaire is very similar to this long form, but the data are collected over a period of months and then “rolled-up” on an annual basis. For some places (with at least 65,000 population), 12 months of data are sufficient to create estimates. For smaller places, more months of sample are needed to produce estimates: 36 months (three years) of data for places of 20,000 to 65,000 and 60 months (five years) of data for places with fewer than 20,000 people.

Decennial census data were collected over a fairly short time period—about four months—and are usually interpreted as presenting the characteristics of population and housing around April 1. The ACS one-year estimates are based on data collected over a full 12 months, and they should be interpreted as describing the average characteristics for that year. Similarly, ACS three-year and five-year estimates should be interpreted as representing the average characteristics for the three- or five-year period.

7. Sometimes geographic areas change their boundaries. What boundaries are used in the ACS when the data are for multi-year periods?

A. For multi-year periods, the geographic boundaries are taken from the latest year of the period. So, if data are presented for 2005–2009 for a city or town, the place’s boundaries as of 2009 are used for the estimate.

8. I am interested in the census tract and block group data. I know that the Census Bureau sometimes changes the boundaries of these areas. How does this affect the 2005–2009 estimates for census tracts and block groups from the ACS?

A. The 2005–2009 ACS five-year estimates use census tract and block group boundaries from 2000. Starting with the release of the 2010 Census data in 2011 and for all decennial census and ACS products in the coming decade, census tracts and block groups will change to new 2010 boundaries. Therefore, the 2005–2009 ACS five-year estimates for census tracts and block groups will be the last estimates to make use of 2000 boundaries. Information on these changes will be available from the Census Bureau at the Redistricting web page:

9. What does MOE stand for?

A. MOE stands for margin of error. Both Census 2000 long-form estimates and ACS estimates are derived from a sample, and the estimates therefore can vary, plus or minus the MOE. This is frequently referred to as “sampling error” or the variability in a number based on the fact that it is taken from a sample and not an entire population. The ACS sample is smaller than the Census 2000 sample, so margins of error tend to be larger than those from the 2000 Census long-form sample.

10. How do I know if data are reliable?

A. There is no hard-and-fast rule regarding reliability, since it depends on how the data are being used. A general rule is to calculate the upper limit (number plus the MOE) and the lower limit (number minus the MOE) and ask whether the number is useful for the purpose at hand, given that range. For some purposes, a broad range is acceptable, for example, to obtain a general sense of the population with difficulties speaking English. In other cases, the range may be unacceptable and other data sources should be consulted, as when an exact count of persons of limited English proficiency is needed for estimating budget expenditures or space for an instructional program.

11. My town has 42,000 residents. I have numbers for a three-year period (2007–2009) and for a five-year period (2005–2009) available. Which should I use?

A. It depends on what the numbers are being used for. The three-year numbers are more current, but the sample is limited to three years; conversely, the five-year numbers have a larger sample but are not as current. If you are examining employment or other economic characteristic, the three-year estimates are likely better because they exclude the boom years prior to the recent recession. On the other hand, if you are looking to get a figure for something that may not be as time-sensitive, such as the percent of the adult population that completed high school, sample size will likely trump the need to be current, so the five-year numbers may better serve your purpose. A good rule-of-thumb is to start with the three-year estimates and assess the margins of error for your application; if you think you need estimates with smaller margins of error, move to the five-year data.

12. Are data for census tracts and block groups available on American FactFinder (AFF)?

A. All ACS data products can be accessed from the AFF. The full complement of ACS tabulations is available for census tracts as tables in the American FactFinder. This includes Profiles as well as more detailed tables. Block group tables are more limited and are only available in summary file format for download in the Download Center of AFF.

13. I have heard that the Census Bureau suppresses tables in an effort to minimize the release of unreliable estimates. How does this affect the release of the five-year estimates for census tracts and block groups from the ACS?

A. The Census Bureau has decided not to subject any of the five-year estimates to suppression based on their reliability. This is being done to encourage data users to use census tracts and block groups as building blocks for aggregation to larger customized geographic areas. Aggregation translates into larger samples and more reliable estimates and often results in areas that better suit specific applications. Keep in mind, however, that the one-year and three-year ACS data are subject to reliability thresholds, which may result in parts of tables or whole tables being suppressed.

14. Can the numbers from the 2005–2009 ACS be seen as representing the midpoint of that period – 2007?

A. No. The estimates represent an average for the entire period and should not be associated with any single-year time point.

15. What are the Public Use Microdata Samples (PUMS) and when should they be used?

A. PUMS files are samples of individual records of census or ACS responses for households and people, stripped of identifying information (for example, names and detailed geographic codes). Unlike the tables reported for geographic areas in AFF, which are fixed, PUMS records can be used to create custom tabulations to answer more detailed questions, with the only major limitation being sample size. To ensure confidentiality, records in the PUMS files cannot be identified for geographic areas of fewer than 100,000 persons. Thus, the PUMS cannot be used to create custom tables for smaller geographic areas.

16. What is the PUMA Summary Level and why is it important?

A. PUMAs are Public Use Microdata Areas. They are nonoverlapping geographic areas, composed of census tracts that partition a state, each containing at least 100,000 persons. State governments drew the PUMA boundaries at the time of Census 2000. Since full implementation of the ACS in 2005, PUMAs have appeared as a summary level for the creation of tabulations online through American FactFinder (AFF). Thus, ACS data have been made available for the one-year and three-year period estimates at the PUMA level through AFF. In December of 2010, the first five-year PUMA estimates are being made available through AFF. One-year and three-year PUMA estimates are useful when data are needed for substate areas (or subcounty or subcity areas in counties and cities with enough people to have two or more PUMAs) that are more up to date than the five-year data for census tracts and block groups.

17. Where can I find out more about the ACS and changes in the Census in 2010?

The CQ Press ACS primer Web site provides key information on the new American Community Survey, and the forthcoming encyclopedia will provide a wide knowledge base for the entirety of the Census and ACS. Moreover, both provide bibliographies for further exploration of individual topics.