On the last three days of March, teams of temporary Census Bureau workers visited the types of places, including what the bureau calls "targeted non-sheltered outdoor locations" (TNSOL), where homeless people are known to congregate. These workers were carrying out the "Service-Based Enumeration" (SBE) phase of the Census, which counts the nation's homeless population.
The bureau gave these workers two instructions that seemed peculiar: When they counted a homeless person, the workers did not need to take the person's name or date of birth, and if a presumed homeless person insisted he or she had already been counted by the Census, the workers were supposed to count that person anyway.
These orders raise an obvious question: Is the 2010 Census counting some homeless people twice?
The issue has not escaped the notice of the Office of the Inspector General of the Department of Commerce, which oversees the Census Bureau. During the effort to count the homeless, officials from the IG's office visited 13 local Census offices to observe first-hand how the count was conducted. On May 5, the IG published a report for Congress on Census operations during the first quarter (January-March). It included the IG's observations on the homeless count.
The report indicated that the Census manual for Group Quarters Enumeration (GQE) -- of which the SBE count of the homeless is a part -- specifically instructed workers to recount people who said they had already been counted. The IG report also said the workers counting the homeless were not required to collect the names and birth dates of these people.
"Unique to this operation, enumerators were allowed to create an individual Census record based on their direct observation of the race, gender and ethnicity of the respondent," the IG reported. "Enumerators were not required to obtain names or dates of birth from such respondents. Additionally, the Census Bureau's GQE manual indicates that enumerators should recount any individual who asserts that he/she has already been counted."
The IG's office reported that some workers were naturally disinclined to follow the bureau's instructions to recount a person who claimed he had already been counted. Sometimes a person who said he had been counted already was counted again, sometimes not. The same happened with people who said they had an address.
"We identified concerns with ... inconsistent handling of individuals who either (1) stated that they had already been counted, or (2) stated that they had an address," the IG reported. "We observed 83 enumerations -- at shelters, soup kitchens, food vans and TNSOL sites -- carried out by 13 local offices. In over half of our observations, enumerators were inconsistent in deciding whether or not to recount individuals who stated that they had already been counted. We also identified inconsistent practices when respondents indicated that they had an actual residential address. In particular, some of these individuals were counted during SBE, while other individuals were told that they could not be counted because they were not homeless. The enumerators' natural inclination to avoid duplication often contradicted the procedures in the Census GQE manual."
The IG's report concluded there is a great risk that Census created duplicate records of some homeless. It also revealed that the IG's office had not examined how the bureau planned to correct its results to remove people it had counted twice.
"When deviating from established procedures, enumerators appeared to follow a more common-sense approach to reducing the risk of duplicate records," said the IG report. "However, this risk remains great for individual records created during SBE. We have not reviewed the process Census will use to remove duplicate records for enumerations that were simply based on direct observation of race, gender, age or ethnicity, and in which no birth date or name was provided."
In written testimony presented in February to the House Oversight and Government Reform Subcommittee on the Census, Robert Goldenkoff of the Government Accountability Office explained why Census accuracy is important. "Data from the census -- a constitutionally mandated effort -- are used to apportion seats in the Congress, redraw congressional districts, help allocate more than $400 billion in federal aid to state and local governments, and redraw local political boundaries," he said.
"Precision is critical," he said, "because, in some cases, small differences in population totals could potentially impact apportionment, redistricting decisions or both."
On Wednesday, I asked the Census Bureau why it told enumerators to count homeless people who said they had already been counted, why did it not record the names and dates of birth for the homeless people it counted, and what process it is using "to remove duplicate records for enumerations that were simply based on direct observation of race, gender, age or ethnicity, and in which no birth date or name was provided?"
Census spokesman Michael G. Cook responded by email. "We have a process for dealing with duplicate responses to the 2010 Census to which the Government Accountability Office (GAO) and Inspector General (IG) are very familiar," he said.