It was on the streets of her Harlem neighborhood in the 1940s that teenager Althea Gibson began working on the tennis skills that would take her all the way to winning Wimbledon.
But according to the 1940 census, the trailblazing athlete didn't even exist.
There's no record of Gibson and her family in the decennial census, the records of which were released online to the public April 2 by the U.S. National Archives after a 72-year confidentiality period lapsed.
She and her family aren't the only ones - more than a million black people weren't accounted for in 1940, an undercount that had ramifications at the time on everything from the political map to the distribution of resources.
It also had an impact on the Census Bureau itself, the agency said, leading to efforts that continue to this day as it counts people every decade, to assess how well it managed to count people and to determine what could be done to improve. An analysis of the 2010 Census' efficacy is being released May 22.
The undercount estimate has generally gone down, but it's always been disproportionately higher for blacks than nonblacks.
There are a variety of reasons for undercounts - people move around; people may not know or be reluctant to answer government questions; address lists may be inaccurate; extremely crowded areas can be difficult to count, as can extremely isolated areas. Experts believe some of those factors weigh more heavily on minority undercounts, particularly the challenges of counting in urban areas.
The 1940 census was long known to have a black undercount. Evidence of it was found within a decade in a demographic study of young children and another of draft-age men. But modern-day genealogists digging into the newly released 1940 census records may be rediscovering it when they cannot locate their relatives or friends.
The absence of Gibson and her family in the available records points toward an omission.
Celedonia "Cal" Jones knows that Gibson lived in Harlem at the time, because the Manhattan borough historian emeritus grew up on the same block as her and remembers playing with her as a child.
"I know she lived on the block, because she used to dominate the paddle tennis," Jones said. "Her nickname was `tomboy."'
It can be difficult to find entries in the 1940 census, since there's no complete name index for the records currently available and won't be for a few months longer. But Lillian Chisholm, Gibson's sole surviving sister who was born in August 1940, confirmed the family lived at 135 W. 143rd St. at that time, making it possible to look up the census ledger.
An enumerator visited the building on at least five occasions in April 1940, according to the census records. An Associated Press review of the records found no listing of Gibson, who was 12 at the time, or her parents, at that address, though other building residents were counted.
There had been anecdotal information of population undercounts in previous censuses, but it was the data from the 1940 effort that really made it clear, said Phil Sparks, former associate director of the bureau and now co-director of The Census Project, which advocates for an accurate count.
Government officials were able to see that the count was off, particularly in the count of black men of a certain age group in the South, because they were using census data to plan for how many would be registering to fight in World War II, Sparks said. More signed up than were expected.
"From the standpoint of the war effort, it was a good thing to have happened," he said, "but suppose it had been the other way around?"
According to census reports, the black undercount was estimated at 8.4 percent in 1940, meaning that a population counted at 12.9 million was actually more like 14.1 million. The undercount for the nonblack population was 5 percent, or about 6.3 million people. The total undercount for all races was 7.5 million.
The U.S. Census Bureau tried to reach out to the black community as it prepared to undertake the 1940 census. Documents obtained by The Associated Press at the Franklin D. Roosevelt Presidential Library and Museum in Hyde Park, N.Y., show that the agency was particularly concerned about counting blacks in America at that time.
"The Census of 1940 will answer two questions of primary importance," bureau officials wrote in a statement titled "The Negro and the 1940 Census" made available to teachers, speakers and writers at the time. Those questions included "how many Negroes are there now in the United States?" and "has their proportion decreased ... or has it taken an unexpected - and unprecedented - upswing?"
These facts "may have a tremendously important bearing upon the determination of the Negro's place in American life," the officials wrote.
The statement was part of a nationwide publicity campaign to "impress upon the Negro citizen of the United States the importance of full and honest cooperation." Letters were sent out to black trade associations, YMCAs and social and civic organizations. Edward Lawson, a managing editor of Opportunity: The Journal of Negro Life, was hired to supervise the campaign.
However, there is evidence that the campaign to count the country's blacks was uneven.
In one case, the New Orleans chapter of the Urban League Inc., sent a letter to President Franklin D. Roosevelt's wife alleging "willful omission of negroes from staff of census enumerators in New Orleans," according to a copy of the document obtained by the AP from the FDR library.
The U.S. Census Bureau said it would have to check into the situation when asked about Gibson and her family not being part of the 1940 count, but didn't respond with an answer.
Jones isn't surprised that his childhood friend and others somehow got left out.
"It's part and parcel of being written out of history, that's the first step," he said. "You don't count."
The importance of an accurate count is vital, since the data is used in a number of ways. That includes the main purpose, written into the U.S. Constitution, that Congressional districts are apportioned by the census population counts. But it also matters because federal dollars flow to states and localities based on that effort, meaning a wrong count in a census year can impact a whole decade.
"It literally can mean the difference of tens of millions, hundreds of millions, of dollars," Sparks said.