The data contained in this product are based on the Census 2000 sample. The data are estimates of the actual figures that would have been obtained from a complete count. Estimates derived from a sample are expected to be different from the 100-percent figures because they are subject to sampling and nonsampling errors. Sampling error in data arises from the selection of people and housing units included in the sample. Nonsampling error affects both sample and 100-percent data and is introduced as a result of errors that may occur during the data collection and processing phases of the census. This chapter provides a detailed discussion of both types of errors and a description of the estimation procedures.

The majority of addresses in the country are in what is known for census purposes as Mailout/Mailback areas, which generally consist of city-style addresses. The original source of addresses on the Master Address File (MAF) for the Mailout/Mailback areas was the 1990 Census Address Control File (ACF). The first update to the ACF addresses is a United States Postal Service (USPS) Delivery Sequence File (DSF) of addresses. The November 1997, September 1998, November 1999, and April 2000 DSFs were incorporated into the MAF.

Until shortly before the census, the ACF addresses and the November 1997 and September 1998 residential DSF addresses constituted the MAF. These addresses were tested against Census Bureau geographic information to determine their location at the census block level. The geographic information is maintained in the Census Bureau's Topologically Integrated Geographic Encoding Referencing (TIGER) system. When an address on the MAF can be uniquely matched to the address range in TIGER for a street segment that forms one of the boundaries of a particular block, the address is said to be geocoded to that block. Valid and geocoded addresses appeared on each address list used for a field operation.

The Block Canvass operation was the next major address list operation in the Mailout/Mailback areas for Census 2000. Between January and May 1999, there was a 100-percent canvass of every block in these areas. Every geocoded address was printed in a block-by-block address register. Block Canvassing listers identified each address as one of the following: a verified housing unit; a unit with corrections to the street name or directional; a delete; a duplicate, implying the unit exists elsewhere on the list with a different, unmatchable designation, such as a different street name or building name; uninhabitable; or nonresidential. Also, units that were deleted from one block and matched an added unit in another block were called a move.

A cooperative address list check with local governmental units throughout the country, called Local Update of Census Addresses (LUCA) 98, occurred in approximately the same time frame as Block Canvassing. In LUCA 98, the participating governmental units received an address list and were asked for input mostly on added units but also on deleted units and corrected street names or directionals. The outcome of this operation was similar to that of Block Canvassing; units were added to and deleted from blocks, and address corrections were made.

The Decennial Master Address File (DMAF), created in July 1999, was the file used for the main printing of the Census 2000 questionnaires. In Mailout/Mailback areas, the operations that had yielded housing units and their status before this initial printing stage were the ACF, the November 1997 DSF, the September 1998 DSF, LUCA 98, and Block Canvassing.

Updates to the DMAF followed the creation of the initial DMAF. Addresses were added by the November 1999, February 2000, and April 2000 DSFs. The LUCA 98 field verification and appeal processes were address update operations that occurred subsequent to the creation of the initial DMAF. Units receiving a conflicting status from Block Canvassing and the LUCA 98 operation were sent for field verification by the Census Bureau; the results of the field verification were sent to the governmental units. The governmental unit could appeal the Census Bureau's findings for particular units at this stage. At an appeal, the Census Bureau and the governmental unit submit their evidence of the status of a housing unit for independent review. The Census Address List Appeals Office, a temporary Federal office established outside the Department of Commerce, administered the appeal process. The Director of the Appeals Office (or their designee) was responsible for issuing a written determination that was considered final. Both the field verification and the appeal process had the potential to change the status of a housing unit.

The New Construction operation was another cooperative effort with participating governmental units that added addresses before Census Day. This was a final operation in Mailout/Mailback areas that used governmental units local knowledge to identify new housing units in February and March of 2000.

After Mailout/Mailback, the second most common method of questionnaire delivery was Update/Leave. Rather than obtaining addresses from the ACF and DSF, the address list for Update/Leave areas was constructed during a Census Bureau field operation called Address Listing. This was due to the fact that addresses in Update/Leave areas were primarily noncity-style. Census employees were sent to the field with maps of their assignment areas and were instructed to record the city-style address, noncity-style address or location description, or possibly some combination of the above, for every housing unit. In addition, the location of the unit was noted on the census map with what is known as a map spot . This operation took place in the fall of 1998.

After processing the Address Listing data, the Census Bureau could tabulate the number of housing units in each block. Because the housing units in these areas may have nonstandard mailing addresses and may be recorded in census files solely with a location description, the governmental units participating in the local review operation in these areas were sent lists of housing unit counts by block. This operation was called LUCA 99. When a LUCA 99 participant disagreed with a Census block count, the contested block was sent out for LUCA 99 recanvassing. Census employees were redeployed to make updates to the address list. In addition, there was a LUCA 99 appeal process for settling housing unit status discrepancies that could potentially add units to the address list. The LUCA 99 recanvassing and LUCA 99 appeal process took place at various times during the DMAF updating process. Although most of the LUCA 99 entities had their recanvassing results processed before creation of the initial DMAF, many did not. There were DMAF updates designed specifically for obtaining late recanvassing and appeal results. These updates to the census files occurred in time for USPS delivery of a questionnaire.

The last address list-building operation in the Update/Leave areas was the Update/Leave operation itself. This operation was responsible for having a census questionnaire hand-delivered at every housing unit. The MAF and the maps were updated during this process.

In the most remote regions of the country, housing units were listed at the same time people within them were enumerated for Census 2000. These operations, called List/Enumerate and Remote Alaska enumeration, were the only source of addresses in these regions. All housing units were map spotted at the time of enumeration.

In some other regions of the country where an address list had already been created, the Census Bureau determined that direct enumeration of the population would be more successful than mailback of the forms. This operation was called Update/Enumerate. There were two types of Update/Enumerate areas - urban areas that were formerly Mailout/Mailback and rural areas that were formerly Update/Leave. The urban areas had passed through all the Mailout/Mailback operations up through the point of the creation of the initial DMAF, and the rural areas had passed through Address Listing, and sometimes LUCA 99, by the time of the creation of the initial DMAF. Because of these distinct paths, it was necessary to distinguish between the urban and rural Update/Enumerate areas.

Urban Update/Leave is another special enumeration that took place in areas where mail delivery was considered to be problematic. The addresses had passed through all the operations of the Mailout/Mailback areas up through the creation of the initial DMAF, but enumerators visited the area during the census. As a result, additions, deletions and corrections to the address list were made.

People who do not receive a questionnaire at their house could submit a Be Counted Form, or they could call Telephone Questionnaire Assistance and have their information collected over the telephone. Addresses from these operations that did not match those already on the DMAF and that were geocoded to a census collection block in an area where census enumeration did not take place were visited in a Field Verification operation to determine if they existed. Verified addresses were added to the address list.

Follow-up operations provided additional information about housing units listed on the DMAF. In Nonresponse Followup (NRFU), enumerators followed up on units that had not returned a preaddressed census form. These units could be enumerated, deemed vacant, or possibly deleted. At the same time, units that did not appear on the address list could be added and enumerated concurrently. Coverage Improvement Follow Up was designated for enumeration at addresses added by New Construction and the later Delivery Sequence Files, as well as a second check on NRFU vacant and deleted units. Adds were also possible. These operations occurred in the Mailout/Mailback, Update/Leave, and Urban Update/Leave areas.

Service Based Enumeration was designed to account for people without a usual residence who use service facilities (i.e., shelters, soup kitchens and mobile food vans). Only people using the service facility on the interview day were enumerated. In addition, people enumerated in Targeted Non-Shelter Outdoor Locations (TNSOLS) and people without a usual residence that filed Be Counted Forms (BCF) augmented the count. This component of the enumeration should not be interpreted as a complete count of the population without a usual residence.

Every person and housing unit in the United States was asked basic demographic and housing questions (for example, race, age, and relationship to householder). A sample of these people and housing units was asked more detailed questions about items, such as income, occupation, and housing costs. The sampling unit for Census 2000 was the housing unit, including all occupants. There were four different housing unit sampling rates: 1-in-8, 1-in-6, 1-in-4, and 1-in-2 (designed for an overall average of about 1-in-6). The Census Bureau assigned these varying rates based on precensus occupied housing unit estimates of various geographic and statistical entities, such as incorporated places and interim census tracts. For people living in group quarters or enumerated at long form eligible service sites (shelters and soup kitchens), the sampling unit was the person and the sampling rate was 1-in-6.

The sample designation method for housing units depended on the data collection procedures. Approximately 95 percent of the population was enumerated by the mailback procedure. In these areas, the Census Bureau used the Decennial Master Address File (DMAF) to select electronically a probability sample. The questionnaires were either mailed or hand-delivered to selected addresses with instructions to complete and mail back the form.

The housing unit sampling rate varied by census block. Long Form Sampling Entities (LFSEs) were used to determine sampling rates in Census 2000 similarly to the way governmental units were used in the 1990 census sample design. LFSEs were:

• Counties and county equivalents (such as parishes in Louisiana).
• Cities.
• Incorporated places (including consolidated cities).
• Census designated places in Hawaii only.
• Minor civil divisions in certain states only (Connecticut, Maine, Massachusetts, Michigan, Minnesota, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, and Wisconsin).
• School districts (based on the 1995-1996 school year).
• American Indian reservations.
• Tribal jurisdiction statistical areas (replaced for Census 2000 by entities called Oklahoma Tribal Statistical Areas).
• Alaska Native village statistical areas.

Size estimates for LFSEs were based on housing unit counts from the DMAF and occupancy rates from the 1990 census. If the smallest LFSE that included all or any part of a block had an estimated housing unit count of less than 800, the housing units in the block were sampled at a 1-in-2 rate. If the smallest LFSE that included all or any part of a block had an estimated housing unit count of 800 or more but less than 1,200, housing units in the block were sampled at a 1-in-4 rate. If a block was not in either of the two previous sampling rate categories, and was part of an interim census tract with 2,000 or more estimated housing units, the housing units in the block were sampled at a 1-in-8 rate. Housing units in all remaining blocks (those not assigned to 1-in-2, 1-in-4, or 1-in-8 rates) were sampled at a 1-in-6 rate.

In List/Enumerate areas (accounting for less than 0.5 percent of the housing units), each enumerator was given a blank address register with designated sample lines. Beginning about Census Day, the enumerator systematically canvassed an Assignment Area (AA) and listed all housing units in the address register in the order they were encountered. Completed questionnaires, including sample information for any housing unit listed on a designated sample line, were collected. If an AA contained any blocks that would qualify as above for a 1-in-2 or 1-in-4 rate, all households in the AA were sampled at 1-in-2. Housing units in all other AAs were sampled at 1-in-6.

Housing units in American Indian reservations, tribal jurisdiction statistical areas (replaced for Census 2000 by entities called Oklahoma Tribal Statistical Areas), and Alaska Native villages were sampled according to the same criteria as other LFSEs, except the sampling rates were based on the size of the American Indian and Alaska Native population in those areas as measured in the 1990 census. Trust lands were sampled at the highest rate of any part of their associated American Indian reservations. If the associated American Indian reservation was entirely outside the state containing the trust land, then the trust land was sampled at a 1-in-2 rate. All Remote Alaska assignment areas were sampled at a rate of 1-in-2. Housing units in Puerto Rico were sampled at a constant 1-in-6 rate in all blocks.

Variable sampling rates provide relatively more reliable estimates for small areas and decrease respondent burden in more densely populated areas while maintaining data reliability. When all sampling rates were taken into account across the Nation, approximately 1 out of every 6 housing units was included in the Census 2000 sample.

The Census Bureau has modified or suppressed some data in this data release to protect confidentiality. Title 13 United States Code, Section 9, prohibits the Census Bureau from publishing results in which an individual can be identified. The Census Bureau's internal Disclosure Review Board sets the confidentiality rules for all data releases. A checklist approach is used to ensure that all potential risks to the confidentiality of the data are considered and addressed.

• Title 13, United States Code

Title 13 of the United States Code authorizes the Census Bureau to conduct censuses and surveys. Section 9 of the same Title requires that any information collected from the public under the authority of Title 13 be maintained as confidential. Section 214 of Title 13 and Sections 3559 and 3571 of Title 18 of the United States Code provide for the imposition of penalties of up to 5 years in prison and up to $250,000 in fines for wrongful disclosure of confidential census information.

• Disclosure limitation

Disclosure limitation is the process for protecting the confidentiality of data. A disclosure of data occurs when someone can use published statistical information to identify an individual who provided information under a pledge of confidentiality. Using disclosure limitation procedures, the Census Bureau modifies or removes the characteristics that put confidential information at risk for disclosure. Although it may appear that a table shows information about a specific individual, the Census Bureau has taken steps to disguise the original data while making sure the results are still useful. The techniques used by the Census Bureau to protect confidentiality in tabulations vary, depending on the type of data.

• Data swapping

Data swapping is a method of disclosure limitation designed to protect confidentiality in tables of frequency data (the number or percentage of the population with certain characteristics). Data swapping is done by editing the source data or exchanging records for a sample of cases when creating a table. A sample of households is selected and matched on a set of selected key variables with households in neighboring geographic areas that have similar characteristics (such as the same number of adults and the same number of children). Because the swap often occurs within a neighboring area, there is no effect on the marginal totals for the area or for totals that include data from multiple areas. Because of data swapping, users should not assume that tables with cells having a value of 1 or 2 reveal information about specific individuals. Data swapping procedures were first used in the 1990 census and were also used for Census 2000.

Statistics in this data product are based on a sample. Therefore, they may differ somewhat from 100-percent figures that would have been obtained if all housing units, people within those housing units, and people living in group quarters had been enumerated using the same questionnaires, instructions, enumerators, and so forth. The sample estimate also would differ from other samples of housing units, people within those housing units, and people living in group quarters. The deviation of a sample estimate from the average of all possible samples is called the sampling error . The standard error of a sample estimate is a measure of the variation among the estimates from all possible samples. Thus, it measures the precision with which an estimate from a particular sample approximates the average result of all possible samples. The sample estimate and its estimated standard error permit the construction of interval estimates with prescribed confidence that the interval includes the average result of all possible samples. The method of calculating standard errors and confidence intervals for the data in this product appears in the section called "Calculation of Standard Errors."

In addition to the variability that arises from the sampling procedures, both sample data and 100-percent data are subject to nonsampling error . Nonsampling error may be introduced during any of the various complex operations used to collect and process census data. For example, operations such as editing, reviewing, or handling questionnaires may introduce error into the data. A detailed discussion of the sources of nonsampling error is given in the section on "Nonsampling Error" in this chapter.

Nonsampling error may affect the data in two ways: errors that are introduced randomly will increase the variability of the data and, therefore, should be reflected in the standard error; errors that tend to be consistent in one direction will make both sample and 100-percent data biased in that direction. For example, if respondents consistently tend to underreport their incomes, then the resulting counts of households or families by income category will tend to be understated for the higher income categories and overstated for the lower income categories. Such biases are not reflected in the standard error.

The estimates that appear in this product were obtained from an iterative ratio estimation procedure (iterative proportional fitting) resulting in the assignment of a weight to each sample person or housing unit record. For any given tabulation area, a characteristic total was estimated by summing the weights assigned to the people or housing units possessing the characteristic in the tabulation area. Estimates of family or household characteristics were based on the weight assigned to the family member designated as householder. Each sample person or housing unit record was assigned exactly one weight to be used to produce estimates of all characteristics. For example, if the weight given to a sample person or housing unit had the value 6, all characteristics of that person or housing unit would be tabulated with a weight of 6. The estimation procedure, however, did assign weights varying from person to person or housing unit to housing unit.

The estimation procedure used to assign the weights was performed in geographically defined weighting areas. Generally, weighting areas were formed of contiguous geographic units within counties. Weighting areas were required to have a minimum sample of 400 people. Also, weighting areas never crossed state or county boundaries. In small counties with a sample count below 400 people, the minimum sample size condition was relaxed to permit the entire county to become a weighting area.

As described earlier, Census 2000 long form data were collected on a sample basis. Cities and incorporated places were used to determine sampling rates to support estimates for these areas. As a result, each city, incorporated place, school district, and county had addresses selected in the long form sample.

To produce estimates from the long form data, weighting was performed at the weighting area level. In forming weighting areas, trade-offs between reliability, consistency of the estimates, and complexity of the implementation were considered. The decision was made to form weighting areas consisting of small geographic areas with at least 400 sample persons (or about 200 or more completed long forms) that do not cross county boundaries. No other boundary constraints were imposed. Thus, total population estimates from the long form data will agree with census counts reported in SF 1 and SF 2 for the weighting area, county, and other higher geographic areas obtained by combining either weighting areas or counties. Differences between long form estimates of characteristics in the SF 3 and their corresponding values in the SF 1 or SF 2 are particularly noticeable for small places, tracts, and block groups. Examples of these characteristics are the total number of people, the number of people reporting specific racial categories, and the number of housing units. The official values for items reported on the short form come from SF 1 and SF 2.

Because the weighting areas were formed at a smaller geographic level, any differential nonresponse to long form questionnaires by demographic groups or geographical areas included in a weighting area may introduce differences in complete counts (SF 1 and SF 2) and the SF 3 total population estimates. Also, an insufficient number of sample cases in the weighting matrix cells could lead to differences in SF 1, SF 2, and SF 3 population totals. Thus, differences between the census and SF 3 counts are typical and expected.

In 1990, separate tabulations were not prepared for small areas below a certain size. In contrast, Census 2000 tabulations are being prepared for all areas to maximize data availability. This approach may lead to a greater number of anomalous results than what may have been observed with tabulations released from the 1990 census. A similar phenomenon occurred in the 1990 census when weighting areas respected city and place boundaries. Census counts differed from the long form data estimates in small places. As expected, these differences were sometimes large.

The SF 1 tables provide the official census count of the number of people in an area. The SF 3 tables provide estimates of the proportion of people with specific characteristics, such as occupation, disability, or educational attainment. The total number of people in the SF 3 table is provided for use as the denominator, or base, for these proportions. Estimates in the SF 3 tables give the best estimates of the proportion of people with a particular characteristic, but the census count is the official count of how many people are in the area.

The SF 1 gives exact numbers even for very small groups and areas; whereas, SF 3 gives estimates for small groups and areas, such as tracts and small places, that are less exact. The goal of SF 3 is to identify large differences among areas or large changes over time. Estimates for small areas and small population groups often exhibit large changes from one census to the next, so having the capability to measure them is worthwhile.

In any large-scale statistical operation, such as Census 2000, human and processing errors occur. These errors are commonly referred to as nonsampling errors. Such errors include: not enumerating every household or every person in the population, failing to obtain all required information from the respondents, obtaining incorrect or inconsistent information, and recording information incorrectly. In addition, errors can occur during the field review of the enumerators work, during clerical handling of the census questionnaires, or during the electronic processing of the questionnaires.

While it is impossible to completely eliminate nonsampling error from an operation as large and complex as the decennial census, the Census Bureau attempts to control the sources of such error during the data collection and processing operations. The primary sources of nonsampling error and the programs instituted to control this error in Census 2000 are described below. The success of these programs, however, was contingent upon how well the instructions actually were carried out during the census. As part of the Census 2000 evaluation program, both the effects of these programs and the amount of error remaining after their application will be evaluated.

The objective of the processing operation was to produce a set of data that describes the population as accurately and clearly as possible. In a major change from past practice, the information on Census 2000 questionnaires generally was not edited for consistency, completeness, or acceptability during field data collection or data capture operations. Census crew leaders and local office clerks reviewed enumerator-filled questionnaires for adherence to specified procedures. Mail return questionnaires were not subjected to clerical review and households were not contacted, as in previous censuses, to collect data that were missing from census returns.

Most census questionnaires received by mail from respondents as well as those filled by enumerators were processed through a new contractor-built image scanning system that used optical mark and character recognition to convert the responses into computer files. The optical character recognition, or OCR, process used several pattern and context checks to estimate accuracy thresholds for each write-in field. The system also used "soft edits" on most interpreted numeric write-in responses to decide whether the field values read by the machine interpretation were acceptable. If the value read had a lower than acceptable accuracy threshold or was outside the soft edit range, the image of the item was displayed to a keyer who then entered the response.

To control the possible creation of erroneous people from questionnaires containing stray marks or completed incorrectly, the data capture system included an edit for the number of people indicated on each mail return and enumerator-filled questionnaire. If the edit failed, the questionnaire image was reviewed at a workstation by an operator. The operator identified erroneous person records and corrected OCR interpretation errors in the population count field.

At Census Bureau headquarters, the mail response data records were subjected to a computer edit that identified households exhibiting a possible coverage problem and those with more than six household members - the maximum number of people who could be enumerated on a mail questionnaire. Attempts were made to contact these households on the telephone to correct the count inconsistency and to collect census data for those people for whom there was no room on the questionnaire.

Incomplete or inconsistent information on the questionnaire data records was assigned acceptable values using imputation procedures during the final automated edit of the collected data. As in previous censuses, the general procedure for changing unacceptable entries was to assign an entry for a person that was consistent with entries for people with similar characteristics. Assigning acceptable codes in place of blanks or unacceptable entries enhances the usefulness of the data.

Another way in which corrections were made during the computer editing process was substitution. Substitution assigned a full set of characteristics for people in a household. If there was an indication that a household was occupied by a specified number of people but the questionnaire contained no information for people within the household, or the occupants were not listed on the questionnaire, the Census Bureau selected a previously accepted household of the same size and substituted its full set of characteristics for this household.

Table A. Unadjusted Standard Error for Estimated Totals [Based on a 1-in-6 simple random sample]
Estimated total1	Size of publication area2
	500	1,000	2,500	5,000	10,000	25,000	50,000	100,000	250,000	500,000	1,000,000	5,000,000	10,000,000	25,000,000
50	15	15	16	16	16	16	16	16	16	16	16	16	16	16
100	20	21	22	22	22	22	22	22	22	22	22	22	22	22
250	25	31	34	34	35	35	35	35	35	35	35	35	35	35
500	-	35	45	47	49	49	50	50	50	50	50	50	50	50
1,000	-	-	55	63	67	69	70	70	71	71	71	71	71	71
2,500	-	-	-	79	97	106	109	110	111	112	112	112	112	112
5,000	-	-	-	-	112	141	150	154	157	157	158	158	158	158
10,000	-	-	-	-	-	173	200	212	219	221	222	223	223	224
15,000	-	-	-	-	-	173	229	252	266	270	272	273	274	274
25,000	-	-	-	-	-	-	250	306	335	345	349	353	353	353
75,000	-	-	-	-	-	-	-	306	512	565	589	608	610	611
100,000	-	-	-	-	-	-	-	-	548	632	671	700	704	706
250,000	-	-	-	-	-	-	-	-	-	791	968	1090	1104	1112
500,000	-	-	-	-	-	-	-	-	-	-	1118	1500	1541	1565
1,000,000	-	-	-	-	-	--	-	-	-	-	-	2000	2121	2191
5,000,000	-	-	-	-	-	-	-	-	-	-	-	-	3536	4472
10,000,000	-	-	-	-	-	-	-	-	-	-	-	-	-	5477

¹For estimated totals larger than 10,000,000, the standard error is somewhat larger than the table values. Use the formula given below to calculate the standard error.

²The total count of people, housing units, households, or families in the area if the estimated total is a person, housing unit, household, or family characteristic, respectively.



The 5 in the above equation is based on a 1-in-6 sample and is derived from the inverse of the sampling rate minus one, i.e., 5 = 6 1 .

Table B. Unadjusted Standard Error in Percentage Points for Estimated Percentages [Based on a 1-in-6 simple random sample]
Estimated percentage	Base of estimated percentage1
	500	750	1,000	1,500	2,500	5,000	7,500	10,000	25,000	50,000	100,000	250,000	500,000
2 or 98	14	11	10	08	06	04	04	03	02	01	01	01	00
5 or 95	22	18	15	13	10	07	06	05	03	02	02	01	01
10 or 90	30	24	21	17	13	09	08	07	04	03	02	01	01
15 or 85	36	29	25	21	16	11	09	08	05	04	03	02	01
20 or 80	40	33	28	23	18	13	10	09	06	04	03	02	01
25 or 75	43	35	31	25	19	14	11	10	06	04	03	02	01
30 or 70	46	37	32	26	20	14	12	10	06	05	03	02	01
35 or 65	48	39	34	28	21	15	12	11	07	05	03	02	02
50	50	41	35	29	22	16	13	11	07	05	04	02	02

¹For a percentage and/or base of percentage not shown in the table, use the formula given below to calculate the standard error. Use this table only for proportions; that is, where the numerator is a subset of the denominator.



The 5 in the above equation is based on a 1-in-6 sample and is derived from the inverse of the sampling rate minus one, i.e., 5 = 6 1 .