American Community Survey

Errors in Narrative Profiles on American FactFinder

The Census Bureau has identified errors in the 2011 American Community Survey 1-year Narrative Profiles. In some cases, the charts and graphs in the profiles displayed incorrect estimates. We have removed these profiles from American FactFinder and plan to re-release these profiles with corrected estimates on October 25, 2012. We suggest using the ACS Data Profiles to access the estimates until the Narrative Profiles are corrected and restored.

Minor errors in data collection and processing of the 2011 ACS

The Census Bureau recently discovered several small errors in the ACS Failed Edit Follow-up (FEFU) data collection instrument and processing system. In FEFU, households who have returned a mail form but whose data failed a completeness check are called to collect the missing data. This operation is designed to improve the final quality of mail-returned questionnaires. The errors occurred in late 2011 when changes were made to correct the issues noted in Errata Note 75. The errors had minimal impact on the published estimates. The affected households are spread throughout the United States, and we have no reason to believe that the errors had a notable impact on published estimates for any specific geographic area.

There are two types of small errors reported in this errata note. First, the wrong person was deleted for a small number of households reporting fewer household members during the FEFU interview than was reported on the household's mail return. Most interviewers were able to catch and correct this error when the final roster reconciliation was done in the interview. Second, during some FEFU interviews, the interviewer was not prompted to ask some of the questions regarding missing employment status, marital history, and home value on the household's mail return. Most interviewers were able to detect this error and account for it during the interview; however, it is possible that the error could result in more missing data for some items and lead to higher item allocation rates.

Based on the analysis of the impact of these errors and the risks associated with correcting them, there are no plans to correct the dataset and re-release the affected data products. Errors in the FEFU operation are now corrected.

Correction to the 2010 ACS 1-year Selected Social Characteristics in the United States (CP02) and Selected Social Characteristics in Puerto Rico (CP02PR) Comparison Profiles

On March 8, 2012, the Census Bureau corrected and re-released the 2010 ACS 1-year Selected Social Characteristics in the United States (CP02) and Selected Social Characteristics in Puerto Rico (CP02PR) Comparison Profiles.

Previous note about this error was posted January 26, 2012

Correction to Selected Economic Characteristics (DP03) and Selected Housing Characteristics (DP04) Data Profiles for the following ACS products: 2010 ACS 1-year, 2008-2010 ACS 3-year and 2006-2010 ACS 5-year.

On January 26, 2012, the Census Bureau corrected and re-released the Selected Economic Characteristics (DP03) and Selected Housing Characteristics (DP04) Data Profiles for the 2010 ACS 1-year, 2008-2010 ACS 3-year and 2006-2010 ACS 5-year.

Previous note about this error was posted January 9, 2012

Correction to 2008-2010 ACS 3-year Detailed Table B20004 and Subject Tables S1501, S2001, & S2002 for Estimates of Educational Attainment

On January 19, 2012, the Census Bureau corrected and re-released four tables in the 2008-2010 ACS 3-year estimates in American FactFinder. We have also released corrected estimates in the 2008-2010 ACS 3-year Summary File. The affected tables were:

• B20004, Median Earnings In The Past 12 Months By Sex By Educational Attainment For The Population 25 Years And Over
• S1501, Educational Attainment
• S2001, Earnings In The Past 12 Months
• S2002, Median Earnings In The Past 12 Months Of Workers By Sex And Women's Earnings As A Percentage Of Men's Earnings By Selected Characteristics

Previous note about this error was posted December 8, 2011

Minor errors in data processing from 2005 through 2011

The Census Bureau recently discovered several small errors in the ACS processing system. The erroneous code was identified and corrected in July, 2011, and data processed since then have been handled appropriately. This note briefly describes the errors and the impacted variables.

The code errors existed in the processing of paper questionnaires included in the ACS Failed Edit Follow-Up (FEFU) operation, causing a very small amount of self-reported data from the mail questionnaires to be omitted. The FEFU operation involves follow-up phone calls for paper forms that contain incomplete responses. In many instances, the original responses were re-captured in the follow-up interview.

The following items were affected by the errors, but had no notable impact on the published estimates at the national level: Minutes to work, Class of worker, Available to start work last week, Worked 50 or more weeks during the past 12 months, Relationship, Year moved in, and Year built. The errors appear to be spread throughout the United States, and we have no reason to believe they had a notable impact on published estimates for any specific geographic area below the national level.

The errors also affected two items that we estimate have small, but notable, impacts on national level estimates. The errors resulted in a small underestimate (approximately 1.5 percentage points) of grandparents responsible for caring for the their grandchildren for 5 or more years. In addition, the errors affected the service connected disability rating item, resulting in a small overestimate of the "Not Reported" category (approx. 1.9 percentage points) and small underestimates for the "30 or 40 percent" (approx. 0.7 percentage points), "50 or 60 percent" (approx. 0.4 percentage points), and "70 or more percent" (approx. 0.8 percentage points) categories.

Based on the analysis of the impact of these errors and the risks associated with correcting them, there are no plans to correct the dataset and re-release the affected data products.

Errors in 2009 ACS 1-year Subject Table S2701, "Health Insurance Coverage Status"

In the 2009 ACS 1-year Subject Table S2701 "Health Insurance Coverage Status", in the section "Work Experience" the "civilian noninstitutionalized population 18 years and older" and "worked full-time, year-round in the past 12 months" display incorrect estimates of the "number uninsured" and "percent uninsured". There are no plans for correcting these estimates in this table.

For the correct estimates for these groups, we suggest using Detailed Table "B27012 - Health Insurance Coverage Status and Type by Work Experience By Sex by Age" or "C27012 - Health Insurance Coverage Status and Type by Work Experience by Age".

Errors in HINCP and FINCP in the 2004 and 2005 ACS 1-Year and 2005-2007 3-Year Public Use Microdata Samples (PUMS) in DataFerrett

Three of the PUMS datasets available in DataFerrett have errors in the variables HINCP (Household Income) and FINCP (Family Income). The affected datasets are the 2004 1-year PUMS, the 2005 1-year PUMS, and the 2005-2007 3-year PUMS. For these datasets, all values for HINCP and FINCP are incorrect in DataFerrett. We plan to correct the variables and re-release the files in DataFerrett.

The PUMS files accessed through the FTP site do not contain these errors. Data users should obtain these files through the FTP site until the variables have been corrected in DataFerrett.

Errors in the 2008 and 2009 ACS 1-year Subject Table S1811, "Selected Economic Characteristics for the Civilian Noninstitutionalized Population By Disability Status"

In the 2008 and 2009 ACS 1-year Subject Table S1811, two lines under the "EARNINGS IN THE PAST 12 MONTHS" heading are labeled incorrectly. The lines in error describe the percentage of people with earnings of "$1 to $4,999 or loss" and "$5,000 to $14,999". The estimates released were produced using different earnings cut-offs. The estimates on those two lines actually show the percentages with earnings of "$1 to $9,999 or loss" and "$10,000 to $14,999", respectively.

The error affects the 2008 and 2009 ACS 1-year estimates available on American FactFinder and as comma separated value (csv) files via the Census Bureau's FTP site. There are no plans to re-release these tables.

Corrections to documentation for ACS Public Use Microdata Samples (PUMS) in DataFerrett

On July 29, 2011, several errors were corrected in the variable descriptions and value labels associated with the ACS PUMS files in DataFerrett. The problems affected variable descriptions within the application, as well as variable labels and value labels attached to customized tables and downloadable data files.

Correction to the PUMS variable descriptions and value labels were made to the following variables: ELEP, FULP, GASP, GRPIP, INTP, NOC, NP, NRC, PERNP, PINCP, RETP, SSIP, VALP, WATP, and WKHP. The corrections were made in DataFerrett to the 2004, 2005, 2006, 2007, 2008, and 2009 1-year PUMS; and the 2005-2007 3-year PUMS.

Previous note about this error was posted June 20, 2011

Errors in SMOCP in the 2005-2009 ACS 5-year Public Use Microdata Sample (PUMS) Files

The 2005-2009 ACS 5-year PUMS has errors in the housing unit variable SMOCP (selected monthly owner costs). SMOCP is created by summing several variables associated with the monthly costs of owning a housing unit. For cases collected in 2005 and 2006, SMOCP was incorrectly calculated to include both annual mobile home costs (MHP) and monthly mobile home costs (MHP/12). As a result, SMOCP values were overestimated for all mobile homes.

This error affects cases on the housing unit records that were collected in 2005 and 2006. Cases collected in 2007-2009 were not affected. We are reviewing feasibility of re-releasing the 2005-2009 5-year housing files in the future.

To correct this error, use the SERIALNO variable to identify the data collection year, and then subtract MHP from SMOCP for all 2005-2006 cases where BLD=1 (mobile homes). Each case's data collection year is indicated in the first 4 digits of SERIALNO.

Re-release of 2009 ACS 1-year Subject Tables S2702 "Selected Characteristics of the Uninsured in the United States" and S2702-PR "Selected Characteristics of the Uninsured in Puerto Rico"

Revised: 5/13/2011

Subject Tables S2702 and S2702-PR were re-released on AmericanFactfinder and on the Census Bureau's FTP site on May 13, 2011. These tables were removed on March 30, 2011, due to several data errors.

The revised tables contain corrected information for the following rows:

Rows for "Language Spoken at Home" have been removed and will not be re-released.

Previous notes about this error were posted on:
May 2, 2011

March 30, 2011

Re-release of the 2007-2009 ACS 3-year Public Use Microdata Sample (PUMS) Files to correct for errors in ADJINC and ADJHSG

Prior to April 7, 2011, the 2007-2009 ACS 3-year PUMS had errors in the variables ADJINC and ADJHSG on the housing unit files. Corrected housing unit files were released on April 7, 2011.

Previous note about this error was posted March 17, 2011

The Re-release of Selected Sequences from the 2009 ACS 1-Year Summary File

Processing errors were discovered in the 2009 ACS 1-Year Summary File that was released on September 28, 2010. The errors affected fifteen sequences: 1, 2, 107, 108, 113, 114, 164, and 169-176. These errors occurred in both the estimate (e) and margin of error (m) files.

Sequences 164, 169, 170, and 176 were replaced for Puerto Rico only. The other erroneous sequences were replaced for most or all areas.

The corrected sequences were released on January 26, 2011.

The Re-release of two of the 2008 ACS 1-year Public Use Microdata Sample (PUMS) Files

Between November 9, 2010 and November 29, 2010, the 2008 ACS 1-year PUMS files named "csv_pus.zip" and "csv_hus.zip" erroneously contained 2009 ACS 1-Year PUMS data. These two 2008 ACS 1-Year PUMS data files were reissued on November 29, 2010. The remaining 2008 ACS 1-Year PUMS files – including the state-level files and SAS format files – were unaffected by this problem.

The 2003, 2004, 2005, and 2006 ACS 1-year PUMS and the 2005-2007 3-year PUMS files were revised in November 2010 to correct an error in the AGE variable and to make a number of other smaller improvements. The table below is a complete list of changes made to the revised files. The original version of this note posted in November 2010 did not describe the complete list of changes made to the revised files.

Previous notes about this error were posted November 18, 2010

A small number of errors were detected in the 2008 ACS internal microdata file, leading to errors in the 2008 ACS 1-year PUMS files and the 2006-2008 ACS 3-year PUMS files. The 2008 ACS 1-year PUMS and the 2006-2008 ACS 3-year PUMS files were corrected to the extent they could be without causing a risk of disclosure.

The following housing variables were corrected: telephone service (TEL), number of rooms (RMSP), number of bedrooms (BDSP), kitchen facilities (KIT), refrigerator (REFR), and their respective allocation flags. For more detail on these errors, see Errata Notes #53 and #54. The "speaking a language other than English at home" (HHL) variable was also corrected due to the editing error referred to in Errata Note #61.

In addition, on the 2006-2008 3-year PUMS housing unit-level files, the variable FMVYP has been renamed FMVP so that it is consistent with the documentation and previous files.

For more details on how the corrections were made, see the User Note titled "The Correction and Re-release of the 2008 ACS 1-year and 2006-2008 ACS 3-year Public Use Microdata Sample (PUMS) Files."

Previous notes about this error were posted November 18, 2010

A column heading is in error in single-year data from 2005, 2006, 2007, and 2008, and multi-year files from 2005-2007 and 2006-2008. The column "Private not-for-profit wage and salary workers" should have read "Employee of private company workers." The column is labeled correctly in the version of this table produced using the 2009 ACS.

The universe description of these tables is incorrect in single-year data from 2005, 2006, 2007, and 2008, and multi-year files from 2005-2007 and 2006-2008. These tables erroneously identify the universe as the "civilian employed population 16 years and over with earnings." The universe for the total, male, and female columns is the civilian employed population 16 years and over. This includes persons who have earnings as well as those who do not have earnings. The universe for median earnings (dollars), median earnings (dollars) for male, and median earnings (dollars) for female is the civilian employed population 16 years and over with earnings. Starting in 2009, "with earnings" was removed from the universe description to more accurately describe the frequency distributions.

Several minor errors were discovered in the 2008 ACS edit programs for a number of variables including age, citizenship status, employment status, household language, income, industry and occupation, race and Hispanic origin, real estate taxes included in mortgage payment, and second mortgage. In each case, the edit programs were not completely updated to reflect variable changes resulting from the addition of three new questions and the revision of twelve other questions in 2008. This error affects almost every table but the effect is very small. Corrected 2008 data will appear in the 2009 ACS 1-year Comparison Profiles. Corrected 2008 data will also be used to produce the 2005-2009 ACS 5-year and 2007-2009 ACS 3-year data products as well as future multiyear products. We do not plan to rerelease any of the existing 2008 ACS 1-year or 2006-2008 ACS 3-year tables on American FactFinder, or the 2008 or 2006-2008 ACS PUMS files.

An error was discovered in the processing of the ACS that has produced an overestimate of the number of veterans that served between World War II and the Korean War in the ACS from 2000-2008. Answers of "don't know" or "refused" to the period of service question were incorrectly recorded as served between World War II and the Korean War for a small number of people. We estimate that the national totals for that population are overestimated by about 8,000 veterans (about 5 percent of the veteran population) in the 2008 ACS tables, and by a similar amount in the 2000-2007 ACS tables. However, the estimated percentage of veterans that served between World War II and the Korean War did not change. Corrected 2008 data will appear in the 2009 ACS 1-year Comparison Profiles. Corrected 2008 data will also be used to produce the 2005-2009 ACS 5-year and 2007-2009 ACS 3-year data products as well as future multiyear products. We do not plan to rerelease any of the existing 1-year or 3-year ACS tables on American FactFinder, or any of the ACS PUMS files.

An error was discovered in the processing of the ACS that has produced an overestimate of the Vietnamese population in the United States in the ACS from 2000-2008. Answers of "don't know" or "refused" to the race question were incorrectly recorded as Vietnamese for a small number of people collected via mail questionnaires. We estimate that the national totals for the Vietnamese population are overestimated by approximately 170,000 people (about 10 percent of the Vietnamese population) in the 2008 ACS tables, and by a similar amount in the 2000-2007 ACS tables. We will produce tables of the Vietnamese population from the corrected 2008 ACS data and update this errata notice with the link to the tables when they are available. Corrected 2008 data will appear in the 2009 ACS 1-year Comparison Profiles. Corrected 2008 data will also be used to produce the 2005-2009 ACS 5-year and 2007-2009 ACS 3-year data products as well as future multiyear products. We do not plan to rerelease any of the existing 1-year or 3-year ACS tables on American FactFinder, or any of the ACS PUMS files.

An error was discovered in the processing of the race data on some ACS mail forms from 2000 to 2008. Responses on the mail questionnaire that checked only the "Samoan" box for the race question (with no other race reported) were erroneously changed to "don't know." In these cases, race was imputed as if the respondent had not provided an answer to the race question. We estimate that the national total for the Samoan population is underestimated by approximately 23,000 people (about 15 percent of the Samoan population) in the 2008 ACS tables and by a similar amount in the 2000-2007 ACS tables. Data collected via the telephone and personal visit modes are not affected. We will produce tables of the Samoan population from the corrected 2008 ACS data and update this errata notice with the link to the tables when they are available. Corrected 2008 data will appear in the 2009 ACS 1-year Comparison Profiles. Corrected 2008 data will also be used to produce the 2005-2009 ACS 5-year and 2007-2009 ACS 3-year data products as well as future multiyear products. We do not plan to rerelease any of the existing 1-year or 3-year ACS tables on American FactFinder, or any of the ACS PUMS files.

Some 2008 ACS race responses of Japanese, Korean, Vietnamese, and "Other Asian" checkboxes were processed incorrectly due to processing changes made to accommodate the race question's new layout in the 2008 questionnaire. These errors were limited to mail forms. Data collected via the telephone and personal visit modes were not affected. We estimate that this error results in the following changes to the national 2008 ACS estimates: increases for the Japanese population (less than 5 percent) and the Asian Indian and Filipino populations (less than 1 percent); and decreases for the Chinese population (less than 3 percent) and the Korean population (less than 2 percent). The Vietnamese population was also impacted by the error described in Errata Note 59, with the net affect being an increase less than half of 1 percent. Additionally, the total for all of the other Asian groups alone or in combination decreases by about 4 percent. This error does not directly impact the national total for the Asian alone population. Given these errors, the 2008 estimates for these populations will not be published in the 2009 ACS Comparison Profiles, and we advise users against making 2009 to 2008 comparisons. We do not plan to rerelease any of the existing 2008 ACS 1-year estimates or the 2006-2008 ACS 3-year estimates on American FactFinder or in the 2008 ACS PUMS files. A very small number of mail returns in 2009 were also processed in error resulting in minor effects on the 2009 ACS 1-year tabulations. No revisions are planned.

An error was discovered in the processing of the 2008 ACS 1-year data and the 2006-2008 ACS 3-year data. About 0.5 percent of the people that should have been included in the survey in 2008 were omitted in error. These were a small percentage of cases whose data were collected via the mail mode. Data collected via the telephone and personal visit modes are not affected. The housing data and only the first person in the household were included in the processing, but all other household members were erroneously excluded. Corrected 2008 data will appear in the 2009 ACS 1-year Comparison Profiles. Corrected 2008 data will also be used to produce the 2005-2009 ACS 5-year and 2007-2009 ACS 3-year data products. We do not plan to rerelease any of the existing 2008 ACS 1-year or 2006-2008 ACS 3-year tables on American FactFinder, or the 2008 or 2006-2008 ACS PUMS files.

There was an error in the calculation of "Workers per car, truck, or van" in Subject Table S0801 for 2005 through 2008 ACS. Total number of workers was used for the numerator instead of total number of workers using car, truck, or van. The correct value can be calculated taking the total number of workers for car, truck, or van in Detailed Table B08006 or C08006 and dividing by the total aggregate number of vehicles (car, truck, or van) used in commuting in Detailed Table B08015. Table 1 shows the number of workers per car, truck, or van for the U.S. for the years affected.

An error was discovered in the 2008 ACS 1-year data and the 2006-2008 ACS 3-year data, which involved Question 7 on the housing section (number of rooms and number of bedrooms items) on the 2008 ACS questionnaire . For housing units where data were collected via a computer-assisted telephone interview or a computer-assisted personal interview, if the number of rooms was missing a value of one room and zero bedrooms was erroneously assigned (instead of being imputed from the entire distribution of responses). Approximately 3 percent of all housing units at the national level are affected. The impact is greater on vacant housing units (about 22 percent affected) than occupied housing units (about 0.3 percent affected). A headnote (including a link to this errata note) is included on all of the affected tables.

Two errors were recently found affecting the 2008 ACS 1-year data and the 2006-2008 ACS 3-year data for household kitchen facilities and telephone service. The errors involve the last two items in Question 8 on the housing section of the 2008 ACS questionnaire which asks whether the housing unit has a refrigerator and telephone service (including cell phones). The error involved the incorrect capture of the responses to those items. It affected the estimates of householders who reported no telephone service, resulting in an underestimate of "no" responses, especially for the telephone item, and an increased imputation rate for both items. The impact on the estimate of housing units with complete kitchen facilities (which is based in part on the refrigerator item) is negligible. For the telephone service item, however, the impact is more noticeable. For example, at the national level the percent of households reporting no telephone service in 2008 was 1.8 percent; however, after correcting the data capture error, the percent reporting no telephone service is approximately 2.8 percent. A headnote (including a link to this errata note) is included on all of the affected tables.

The 2002 codes for the Migration State or Foreign Country variable (MIGSP) was erroneously published in the 2003 ACS PUMS Data Dictionary. The Data Dictionary has been updated to reflect the correct codes and was re-released on December 28, 2009.

An error was discovered in the 2007 and 2008 ACS 1-year, 2005-2007 ACS 3-year, and 2006-2008 ACS 3-year Summary Files for Puerto Rico. The sequence number 146 that contains Detailed Table B98011 HOUSING UNIT COVERAGE RATE is not available for Puerto Rico and a comma should have been included to indicate the missing table. This error occurred in all three sets of data: estimate (e), margin of error (m) and standard error (s). These corrected Summary Files were released on December 28, 2009.

The 2006 ACS 1-year PUMS have been revised by applying a modified age perturbation edit to these files. Age perturbation is one of several disclosure avoidance techniques used in the creation of the ACS PUMS files. It disguises original data by randomly adjusting the reported ages for a subset of individuals. As discussed in Errata Note #47, this technique in some previous years of ACS PUMS processing resulted in inconsistent sex ratios for selected age groups, especially ages 65 and over. The modified age perturbation edit was applied to people aged 65 and over to reduce inconsistencies among this group. The revised file estimates for people 65 years and older align closer to the ACS full sample by single year of age for sex ratios, social security participation, labor force participation, person income and home ownership. The sex ratio inconsistencies for persons aged 65 and over will not be seen in the 2006-2008 ACS 3-year PUMS files since the 2006 ACS 1-year PUMS files have been revised. The spreadsheet [XLS 21KB] displays sex ratio estimates from the ACS full sample and from the PUMS files for the affected years.

The original 2006 PUMS files have been removed and replaced with the revised 2006 1-year PUMS files. To request the original 2006 PUMS files, please send an e-mail to acso.users.support@census.gov with "Request Original 2006 PUMS files" in the subject line.

A formatting error was discovered in the 2008 ACS 1-year PUMS files released on October 30, 2009. The affected variables FINCP, HINCP , PERNP and PINCP were displayed in scientific notation. The incorrect format caused unintentional rounding of these values. This formatting error has been corrected and the revised 2008 ACS 1-year PUMS files were re-released on November 3, 2009. After correction, the variables WAGP, SEMP, INTP, SSP, SSIP, PAP, RETP, and OIP sum correctly to PINCP, variables WAGP and SEMP sum correctly to PERNP, and PINCP summed across family members or household members will equal FINCP and HINCP respectively.

Due to a tabulation error, the Margin of Error (MOE) for select geographies in Detailed Table B25109 is incorrect. The corrected tables will be available in American FactFinder on September 29, 2009. For a list of affected geographies and their correct MOE, go here.

UPDATE: The 2006 ACS 1-year PUMS files were re-released on December 18, 2009. See Errata Note #50 for details.

Disclosure avoidance techniques were applied to the 2003 through 2006 ACS 1-year PUMS files to protect the confidentiality of responses. One of these techniques resulted in some inconsistent ratios for the number of men to the number of women (sex ratio) for people age 65 and over. Table 1 below displays 2006 ACS data from Detailed Table B01001 (Sex by Age) for ages 60 and over for the nation and the corresponding data from the 2006 PUMS file. An example of the inconsistency is shown in the 65-66 age group, where the sex ratio estimate is 88, but the ACS PUMS data indicate a sex ratio estimate of 98. The disclosure avoidance technique has been modified to avoid this distortion of age data in future releases of the ACS PUMS files.

The sex ratio inconsistencies may continue to be seen in any multiyear PUMS files that include the 2005 or 2006 ACS 1-year PUMS, since for the time being, these files remain unmodified due to disclosure risks. This includes the 2005-2007, 2006-2008, 2005-2009, and 2006-2010 Multiyear ACS PUMS files. Table 2 below displays 2005-2007 ACS 3-year data from Detailed Table B01001 (Sex by Age) for ages 60 and over for the nation and the corresponding data from the 2005-2007 3-year ACS PUMS file. Again, looking at the 65-66 age group, the sex ratio estimate is 88, but the ACS PUMS data indicate a sex ratio estimate of 95.

Due to an editing error, veteran's period of service (VPS) was being incorrectly assigned for some individuals. The majority of the errors misclassified some people who reported only serving during the Vietnam Era as having served in the category "Gulf War and Vietnam Era." The remainder of the errors misclassified some people who reported only serving between the Vietnam Era and Gulf War as having served in the category "Gulf War." These errors have been resolved for the 2007 tabulations.

During the processing of data products for the urban areas we discovered the geographic boundaries were incorrect for two of these areas, Hanford CA (code 36703) and San Francisco-Oakland CA (code 78904). Therefore, we are unable to release data products for these urban areas at this time. We are correcting the geographic boundaries in our data files now, and will release the data products as soon as possible.

There was an unusually large amount of missing data for the group quarters populations in Gainesville city, FL and Clearfield county, PA.  The large amount of missing data coupled with the current edit and imputation procedures resulted in incorrect estimates for residence 1 year ago, earnings, and income for both the group quarters population and the total population.  This spreadsheet file [XLS 31KB] contains the full list of tables, table lines, and geographic areas that will not be released due to this problem.

There was an error in the phrasing for two categories in Table C02003 for the 2005 ACS and the 2006 ACS, and Table B02003 for the 2004 ACS. The data were not affected. The category "Two or more races including Some other race" was revised to read "Two races including Some other race." The category "Two or more races excluding Some other race" was revised to read "Two races excluding Some other race, and three or more races." Please note that the tables prior to 2006 will not be corrected.

The ACS Summary File was re-run and re-released because it was discovered that C06010PR was missing from the summary file and the accompanying documentation. In general all Puerto Rico based tables have a non-Puerto Rico counterpart with the exception of C06010PR, which resulted in the missing table. Table C06010PR was placed in sequence 0139, which breaks the ACS summary file rule of forcing collapsed tables to be in the same sequence as the base table (B06010PR is in 0140). The reason that this was done is that there wasn't enough space in sequence 0140 to include C06010PR and still maintain our cell width limit, so we placed it in the nearest logical sequence that fits all of our other rules so as to minimize the total affect on the files that have already been published.

Mobile home costs were omitted in the calculation of selected monthly owner costs (SMOCP) for mobile home owners. In 2003 through 2006 these mobile home costs included costs for personal property taxes, site rent, registration fees, and license fees; in addition to these costs, in 2000 through 2002, these mobile home costs also included installment loan payments. Therefore, the omission of mobile home costs from selected monthly owner costs in 2000, 2001, and 2002 resulted in a larger understatement of selected monthly owner costs for mobile home owners than the omission caused in 2003, 2004, 2005, and 2006. The user can calculate the selected monthly owner costs for mobile home owners by adding the variable MHP (mobile home costs) to SMOCP (selected monthly owner costs) for all cases on which BLD = '01' (mobile homes) to arrive at the selected monthly owner costs for mobile home owners.

The variable PAOC was designed to indicate for women 16 years and over, the presence of an own child in the household (a son or daughter under age 18 and never married). An error was found in the programming of this variable that resulted in the inconsistent identification of women in subfamilies having a child, producing an underestimate of the numbers of women with own children usually in the 1 percent to 2 percent range for each of the states.

This problem affected the PUMS file and several estimates in Detailed Table B23003. As a result, the PUMS files for 2005 and 2006 will be re-issued with the corrected PAOC variable, and the B23003 tables for 2005 and 2006 have been removed from the AFF along with necessary changes to ACS Summary File. These tables will not be re-issued, but the programming will be corrected for future releases. Please note that the PUMS files and data products for the years prior to 2005 will not be corrected.

PAOC is also a small part of the editing procedure used to create the employment status recode (ESR). A review concluded that this error affected ESR and some other variables

The variable PAOC was designed to indicate for women 16 years and over, the presence of an own child in the household (a son or daughter under age 18 and never married). An error was found in the programming of this variable that resulted in the inconsistent identification of women in subfamilies having a child, producing an underestimate of the numbers of women with own children.

This problem affects PUMS and may possibly affect other data products. It is currently under review and any updates will be posted as soon as they are available.

Errors were discovered with the median age estimate in four subject tables. However, each of the corrected estimates is within the 90% margin of error of the original estimate. These four subject tables have been corrected and updated data are available on American FactFinder (AFF.)

The 17 iterated Selected Population Profiles (SPP) [XLS 15KB] in Puerto Rico that were missing on American Factfinder (AFF) are now available.

Also, the error discovered with the medians in the SPPs for 9 new iteration groups [XLS 14KB] has been corrected and updated data are available on AFF.

We discovered that 17 iterated Selected Population Profiles (SPP) [XLS 15KB] ]in Puerto Rico are currently missing on American Factfinder (AFF). However, these tables are available on the ftp site. These tables will be available on AFF very soon.

Also, an error has been discovered with the medians in the SPPs for 9 new iteration groups [XLS 14KB]. This error will be corrected and updated on AFF very soon.

We discovered an error in the calculation of the median age in the column of this table for 'Other Central Americans'. Currently, an "X" is displayed in this cell of the table. The S0506 table with the correct value for this cell will appear in the third ACS2006 data release on September 27, 2007.

For the 2004 and 2005 ACS, the percent of "Self-employed in own not incorporated business workers and unpaid family workers" in "Sales and office occupations" was incorrectly calculated in Table S2406, resulting in an underestimation of this percentage. No other data products were affected by this error. This error has been corrected for future releases of the ACS data files.

To calculate this percentage in the 2004 and 2005 Table S2406, divide the number of persons who are in "Sales and office occupations" within the "Self-employed in own not incorporated business workers and unpaid family workers" category as shown in Detailed Table B24060 by the total number of persons in the "Sales and office occupations" category as shown in Detailed Table B24060, then multiply that number by 100.

For guidance in calculating the standard error around this percentage, see page 19 of the ACS Accuracy of the Data (2006) document [PDF 185KB]

Some of our data products for 2004 and 2005 contained an error which caused median family income's to be understated when they are over $90,000. The 2004 data will not be corrected. The data for 2005 has been corrected and is now available on AFF and on the ftp site. No states were affected by this error; but a small number of counties, townships, cities, and other sub-state geographic entities across the U.S. were affected.

Some of the median gross rent estimates shown in tables B25111 and B25113 were incorrectly calculated. For these two tables, the error caused the median gross rent estimates in high rent areas to be understated. These tables have been replaced on the American FactFinder with the corrected median estimates and associated standard errors. No other data products were affected.

This note is related to errata note 30. After further investigation, we have determined the same problem in the collection of the 2005 travel time to work data affected travel time estimates for Cumberland County, New Jersey and Penobscot County, Maine. No other characteristics appear to have been affected. The data tables containing travel time estimates for Cumberland County, New Jersey and Penobscot County, Maine have been removed or altered on the American FactFinder. For each table, an explanatory note is provided for the users. The spreadsheet file referenced in Note 30 below contains the list of all affected tables along with a description of the action taken for each table. For further information, please contact the Journey-to-Work and Migration Statistics Branch at 301-763-2454.

We have discovered a problem in the collection of the 2005 travel time to work data for Coweta County, Georgia. No other characteristics appear to have been affected. The data tables containing travel time estimates for Coweta County have been removed or altered on the American FactFinder. For each table, an explanatory note is provided for the users. The spreadsheet file [XLS 28KB] contains the list of all affected tables along with a description of the action taken for each table.

This problem is currently under review and users will be notified as soon as corrected data are available.

A minor error was recently discovered in the 2005 ACS sample data. The error affected the data for only 13 sample housing units and 25 sample persons. These 13 sample housing units are in 13 different states (none are in Puerto Rico).

Due to the nature of the data products tabulations, many tables from the August 29 ACS income and poverty release were affected. However, each of the corrected estimates is within the margin of error of the original estimate. Furthermore, almost all of the corrected estimates represent less than a 1% change from the original estimate. The data products from the August 29 data release that are affected will be re-issued on October 19. All data products included in the October 3rd data release have been corrected in time for that release. There was no effect on the data products released on Aug. 15. And the data products to be released on Nov. 14 (the final ACS data release for this year) are being corrected and verified now. These changes will affect 101 detailed tables (and certain other data products derived from these tables) across 145 geographic entities, about 2% of the roughly 7,000 geographic areas for which the 2005 ACS data products are published. However, many of these geographic areas will have changes in only a few data tables. The spreadsheet file of release 2 corrections [XLS 334KB] contains detailed information on the geographic areas and the data tables affected by this error.

The conditional notes for the following subject tables are incorrect: S0102, S0102PR, S0103, S0103PR, S0502, S0502PR, S0503, S0504, S0602, S0803, S0803PR, S0901, S0902, S1001, S1703.

The note: "Table is not available for the selected geography if the population for the selected subject falls below a threshold of 65,000.," for S0102,S0103, S0502, S0503, S0504, S0602, S0803, S0901, S0902, S1001, and S1703 should read:

Table is not available for the selected geography if the population for the selected subject falls below a threshold of 1,000,000. Additionally, the population within a geographic area for the selected subject must meet a threshold of 65,000.

The note: "Table is not available for a selected Puerto Rico geography if the population for the selected subject falls below a threshold of 65,000." for S0102PR,S0103PR, S0502PR, and S0803PR should read:

Table is not available for a selected Puerto Rico geography if the population for the selected subject falls below a threshold of 1,000,000. Additionally, the population within a geographic area for the selected subject must meet a threshold of 65,000.

The variables WGTP1...WGTP80 and PWGTP1....PWGTP80 were inadvertently left off of the August 29, 2006 release of the data files and the data dictionary. Replicate weights are being provided with the 2005 PUMS for the first time. Please reference the PUMS Accuracy of the Data (2005) [PDF 160KB] for more information on replicate weights. The corrected files were released on September 11, 2006.

Some of our data products for 2005 contain an error that causes median family incomes to be understated when they are over $90,000. We are correcting this problem now, and we will let you know when the corrected data are available on AFF and on the ftp site. No states were affected by this error; but, a small number of counties, townships, cities, and other substate geographic entities across the U.S. were affected.

"Workers per car, truck, or van" is the derived measure used to indicate carpooling or vehicle efficiency. It is calculated by dividing the total number of workers who reported using a "car, truck, or van" by the number of "Total vehicles used." The values of number of vehicles for individual workers (DRIVESP) are derived from the JWRIP (ridership variable) as follows:

IF JWRIP = 1 (Drove alone) THEN DRIVESP = 1.000 vehicles;
ELSE IF JWRIP = 2 (In a 2-person carpool) then DRIVESP = 0.500 vehicles;
ELSE IF JWRIP = 3 (In a 3-person carpool) THEN DRIVESP = 0.333 vehicles;
ELSE IF JWRIP = 4 (In a 4-person carpool) THEN DRIVESP = 0.250 vehicles;
ELSE IF JWRIP = 5 or 6 (In a 5- or 6- person carpool) THEN DRIVESP = 0.200 vehicles;
ELSE IF JWRIP = 7 - 10 (In a 7-or-more person carpool) THEN DRIVESP = 0.143 vehicles;
ELSE DRIVESP = blank (nonworker or worker who does not drive to work).

The sum of DRIVESP is the estimated number of "Total vehicles used" by those workers who used a "car, truck, or van." Thus, "workers per car, truck or van" is the sum of all workers who reported using a "car, truck, or van" (JWTR = 1) divided by the sum of DRIVESP for all such workers.

The Class of Worker status "unpaid family workers" may have earnings. Earnings reflect any earnings during the 12 months prior to the ACS interview. The Class of Worker status reflects the job or business held the week prior to the ACS interview, or the last job held by the respondent.

Current residence, residence 1 year ago (migration), and place of work tabulations use the November 2004 Statistical Area Definitions for Metropolitan Statistical Areas, Metropolitan Divisions and Principal Cities, Micropolitan Statistical Areas, Combined Statistical Areas, New England City and Town Areas (NECTA), Combined NECTAs, and NECTA Divisions and Principal Cities. For more information on Statistical Area Definitions, see http://www.census.gov/population/metro/data/metrodef.html.

These tabulations are produced to provide estimates of workers at the location of their workplace. Estimates of counts of workers at the workplace may differ from those of other programs because of variations in definitions, coverage, methods of collection, reference periods, and estimation procedures. The ACS is a household survey that provides data pertaining to individuals, families, and households.

These workplace-based tabulations are only available for States; Counties; Places; County Subdivisions in selected states (CT, ME, MA, MI, MN, NH, NJ, NY, PA, RI, VT, and WI); Combined Statistical Areas; Metropolitan and Micropolitan Statistical Areas, and their associated Metropolitan Divisions and Principal Cities; Combined New England City and Town Areas; New England City and Town Areas, and their associated Divisions and Principal Cities.

The data dictionary shows incorrect category labels for the educational attainment variable (SCHL). The educational attainment question changed on the 1999 ACS questionnaire, which changed the response categories and eliminated the choice of "Vocational, technical, or business school degree." The PUMS data dictionary incorrectly shows labels for categories 10 and 11 as "Some college, but no degree" and "Vo/Tech/Bus school degree," respectively. The correct data dictionary labels for categories 10 and 11 are "Some college, but less than 1 year" and "1 or more years of college, no degree," respectively.

There was an error in the designation of the categories "Income in the past 12 months below poverty level" and "Income in the past 12 months at or above poverty level." Data for the category "below poverty" were shown for the category "at or above poverty" and vice versa. The corrected table was released on February 23, 2006.

There was a calculation error in the data for the category "With children under 18 years," in Subject Table S2201. Because of the error, the incorrect percent was shown for this category. The table has been corrected as of February 23, 2006 and only the corrected version is now available.

There was a calculation error in the data for the median and mean earnings categories for full-time, year-round workers 16 years and over with earnings, in Subject Table S2303. Because of the error, these categories displayed earnings data for all workers 16 years and over with earnings, rather than for the intended subset of workers who worked full time, year round. The table will be corrected on February 23, 2006.

There was a calculation error in the data for the categories "Both husband and wife in labor force," and "Husband in labor force, wife not in labor force," in Subject Table S2302. Because of the error, data for households in which the husband was unemployed were excluded from the calculations for each of these categories. The table will be corrected on February 23, 2006.

There is an error in the way the imputation flag for citizenship was set. This error led to an understatement of the number of people for whom citizenship was imputed. People who had their place of birth (i.e., country of birth) imputed as the United States, Puerto Rico, Guam, the U.S. Virgin Islands, or the Northern Marianas should also have had their citizenship flag set to imputed. However, this did not occur. This error will be corrected in future releases of the ACS data files.

Re-release of 2003 ACS PUMS:

1. The file has been re-run to correct inconsistencies in the data for veterans.
2. Ancestry code 593 (Other Sub-Saharan African) was changed to 587 (Other Sub-Saharan African).

Re-release of 2004 ACS PUMS:

1. The file has been re-run to correct inconsistencies in the data for veterans.

The "Other Pacific Islander" estimate in the 2004 single year and multi-year profiles has been removed from the American Factfinder website. Once revisions are complete, this estimate will be available again on American Factfinder.

Due to an error, American Community Survey core tables PCT004 and PCT005 were not restricted to the correct population subgroup for the years 2000, 2001, 2002, and 2003. Therefore, these core tables have been removed from the American Factfinder website. Once revisions are complete, these tables will be available again on American Factfinder.

To ensure high quality data, we continuously review the estimates produced by the American Community Survey. Whenever anomalies are discovered, we investigate to learn and find out its causes and make methodological changes, as appropriate, to improve the ACS estimates. These changes in methodology may result in large differences in ACS estimates for a small number of geographic areas. Since ACS weighting is controlled at the county level, small changes in methodology are more likely to affect sub-county estimates. For example, a modification in the weighting methodology to improve sub-county estimates resulted in an unexpected large drop in the estimates of housing units for Anaheim, California. This modification was implemented in 2003. The large drop in the housing unit estimate occurred in 2003 from levels estimated in 2001 and 2002.

We discovered a calculation error that caused a gross underestimation of Average Family Size, Average Household Size of Owner-occupied Unit, and Average Household Size of Renter-occupied Unit in the 2000 to 2001 Supplementary Survey Margin of Error. This error only occurred in the following geographic areas:

• Houston City, Harris County pt., Texas
• Harris County, Texas
• Franklin County, Ohio
• Broward County, Florida
• Bronx County, New York

The profiles have been corrected as of December 3, 2002 and only the corrected versions of these profiles are available.

We discovered a calculation error which caused a gross underestimation of Median Family Income in the 2000 to 2001 Margin of Error. This error only occurred in the Primary Metropolitan Statistical Areas (PMSA). The profiles have been corrected as of November 13, 2002 and only the corrected versions of these profiles are available.

The American Community Survey is continuously introducing methodological improvements to maximize the quality of the estimates. Based on results from the survey evaluation program we continue to make changes and improve the estimation methodology. The difference in the estimates from previous releases is a result of revised edits, and reweighting the data using an improved estimation methodology.

The 109 core tables have been corrected as of November 7, 2002 and only the corrected versions of these tables are available.  Over the next few months, the noncore tables will be updated as well, but until then there may be inconsistencies within the Census 2000 Supplementary Survey data.

Users may find inconsistencies in estimates for areas which are in the American Community Survey as well as in the Supplementary Surveys. The 2000 American Community Survey estimates are currently being revised and will be replaced shortly.

The household and family tables [PDF 25KB] were tabulated using the housing unit weight instead of the person weight of the householder used in the previous release. As a result of this change, the number of households will equal the number of occupied housing units but will not equal the number of householders, as it was the case in the previous release. This change results in more consistent data. This change affects 31 summary tables, the tabular and narrative profiles, and the Median Household Income rank table.

In the previous release of table H031 and the Profile of Selected Housing Characteristics the label designations for the two most recent year-built categories were specified incorrectly. The correct label specifications are "1999 or later" and "1995 to 1998" rather than "2000 or later" and "1995 to 1999".

This current release of Median Room data has changed from the previous release, and table H021 and profile table 4, Profile of Selected Housing Characteristics, have been corrected as of March 25, 2002. Only the corrected versions of these tables are available.

In computing median rooms, the whole number is used as the midpoint of the interval; thus the category "3 rooms" is treated as an interval ranging from 2.5-3.5 rooms. Linear interpolation is used to estimate median rooms between the lower and upper bounds of a specified interval. Median rooms is rounded to the nearest tenth.

In the previous release of the American Community Survey data, the midpoint, rather than the lower value of the interval, was incorrectly used as the starting value for the derivation of median rooms. For the category "3 rooms," we previously used the interval 3.0-4.0 rooms.

The American Community Survey household and family tables [PDF 25KB] were reweighted using the person weight of the householder instead of the housing unit weight used in the previous release. Because of this change, the number of households will equal the number of householders, which was not previously the case. This change affects 31 summary tables, the tabular and narrative profiles, and the Median Household Income rank table.

This current release of Industry and Occupation data has changed from the previous release, and tables P066, P067 and P068, as well as profile table 3, Profile of Selected Economic Characteristics, and the narrative profiles, have been corrected as of March 25, 2002. Only the corrected versions of these tables are available. The American Community Survey Industry data follow the guidelines issued in Clarification Memorandum No. 2, "NAICS Alternate Aggregation Structure for Use by U.S. Statistical Agencies," issued by the Office of Management and Budget.

Narrative Profiles released in August 2001 contained an error. In the first section (POPULATION), the sentence beginning "For people reporting one race..." had percentages with the total population as the denominator. The profiles released in November 2001 have recomputed percentages with the population reporting one race as the denominator. The race and Hispanic origin percentages in the following sentences were correct.

The Agriculture, Forestry, Fishing and Hunting Sector in the previous release did not comply with 1997 North American Industry Classification System (NAICS). This current release of Industry and Occupation data has changed from the previous release, and tables P066, P067 and P068 have been corrected as of November 20, 2001. Only the corrected versions of these tables are available. However, the industry data in the current release do not strictly adhere to the coding structure guidelines recently issued by the Office of Management and Budget. Future industry products will follow NAICS classification structure as designated in the May 15, 2001, NAICS Clarification Memorandum No. 2, "Alternate Aggregation Structure for Use by U.S. Statistical Agencies."

Processing errors were discovered in the median tables H061 (Median Gross Rent), H074 (Median Value) and H085 (Median Selected Monthly Owner Costs). This also affects the Profile of Selected Housing Characteristics, and the last paragraph (Housing Costs) of the Narrative Profile. The tables have been corrected as of August 28, 2001 and only the corrected versions of these tables are available.