U.S. Census Bureau
Census 2000 Gateway  |  Glossary
Link to Census 2000 Gateway

Special Tabulations

Disclosure Review Board Rules/Requirements

These rules were approved by the Disclosure Review Board (DRB) on January 29, 2004. They will continue to be scrutinized and possibly changed as the DRB reviews more requests for special tabulations.

  1. All Census 2000 special tabulations must be reviewed by the Disclosure Review Board.

  2. All cells in any Census 2000 special tabulation must be rounded. The rounding schematic for most tables is:
    • 0 remains 0
    • 1-7 rounds to 4
    • 8 or greater rounds to nearest multiple of 5 (i.e., 864 rounds to 865, 982 rounds to 980)
    • Any number that already ends in 5 or 0 stays as is.

    In some circumstances, Census 2000 special tabulations must be rounded to 10’s. In particular, any special tabulations presenting data on the population in households or the population in group quarters must be rounded to 10’s. These data could be found in the universe of a table, presented as a variable, or obtained by subtraction when comparing datasets (e.g., comparing the variables and universes of the special tabulation with those in Summary File 3).

    The exact rounding scheme for rounding to 10’s is:

    • 0 remains 0
    • 1-4 rounds to 0
    • 5-14 rounds to 10
    • 15-24 rounds to 20, etc.

    This rounding to the nearest 10 also applies to those special tabulations pertaining to poverty status and disability where there is a possibility of obtaining group quarters or household data by subtraction.

    This rounding applies to all special tabulations that pertain to the population in households or the population in group quarters -- those done under reimbursable agreement, those done for working papers, tables, professional papers, etc. Examples of tabulations that must be rounded to 10’s can be obtained from the Decennial Programs Coordination Branch of the Population Division (301-763-2429).

    Any totals or subtotals needed should be constructed before rounding. This assures that universes remain the same from table to table, and it is recognized that cells in a table will no longer be additive after rounding.

  3. Medians or other quantiles may be calculated as
    1. an interpolation from a frequency distribution of unrounded data (these are not subject to additional rounding), or

    2. as a point quantile. These must be rounded to two significant digits: 12,345 would round to 12,000; 167,452 would round to 170,000. There must be at least 5 cases on either side of the quantile point. It is recognized that the interpolated quantile may indeed be some individual’s response, but it is coincidental, not by design.

  4. Thresholds on universes will normally be applied to avoid showing data for very small geographic areas or for very small population groups (often 50 unweighted cases for sample data). Tables may normally not have more than 3 or 4 dimensions, and mean cell size lower limits may also be required (mean cell size of each table is 3 for 100% data, or 20 weighted for sample data).

  5. Percents, rates, etc., should be calculated after rounding, but the DRB has granted exceptions to this rule when the numerator and/or denominator of the percent or rate is not shown.

  6. Means and aggregates must be based on at least 3 values.

  7. The finest level of detail shown for Groups Quarters data will be Institutional/Noninstitutional.

  8. For Demographic Profiles from user-defined geographic areas (neighborhoods), all areas must have at least 300 people in them. Using a computer program, the user-defined areas will be compared with standard Census Bureau areas to make sure users cannot obtain data from very small geographic areas by subtraction. If such small areas are found, the boundaries of the user-defined areas must be changed.