Atlanta thought its big story for the last decade had been one of urban renaissance. The U.S. Census Bureau had told the city as much. In 2000, according to the decennial census count, Atlanta had 416,000 people. In the Bureau's estimate for July 1, 2009, it was up to 540,000. That figure represented the city's largest population ever-more than the 497,000 Atlantans the Census Bureau found in 1970. It marked an incredible comeback from 1990, when the city had fallen below 400,000 people.
That's why the Bureau shocked the city when it announced the results of the 2010 Census this March. Atlanta, it said, had only 420,000 people. In a decade, it had hardly grown at all. In effect, the Census Bureau was saying that its own estimate just months before had been off by more than 20 percent.
Atlanta provides merely the most dramatic example of one of the true puzzles of the 2010 Census. The Bureau found far fewer people in many major cities than its own estimates had found a year earlier. Now the Census Bureau and cities are debating which numbers are closer to the truth. No one knows for sure-and no one may ever know for sure.
While it won't be easy, Atlanta and other cities are talking about challenging the counts. That's because the discrepancy isn't just a question of civic pride. Atlanta and other cities will have far less state and federal political representation and state and federal aid in coming years than they would have if the Bureau had found more people.
How is it possible that one count for Atlanta's population could be 420,000 and another could be 540,000-that 120,000 people might or might not exist? The basic answer is that the decennial Census and the Bureau's annual estimates rely on very different approaches.
The once-a-decade Census tries to make contact with every person in the country, through mailed forms and door-to-door outreach. The annual estimates start with the most recent decennial count, then extrapolate how the population has changed using a variety of local administrative records.
For many cities this time around, the difference between the decennial count and the 2009 estimates was huge. Nationally, as you'd expect in a growing country, the Bureau's decennial count for April 1, 2010, was about 3 million people higher than the previous estimate, from July 1, 2009. For the 50 largest cities in the country, however, the number actually went down by 1.3 million.
Detroit's population was off by more than 20 percent, or nearly 200,000 people, from its 2009 estimate-an estimate that itself had reflected a drop in the city's population of around 40,000 in the first nine years of the decade. None of the other 50 largest cities had discrepancies anywhere near as large as Atlanta and Detroit did, but Omaha, Phoenix, Cleveland, Dallas, Miami, Houston, Mesa (Arizona) and Chicago all had counts that were at least 5 percent below their most recent estimate.
In the Census Bureau's view, it's clear which numbers are right: those that come from the decennial count. "Obviously, we're doing the best we can with the estimates," says Greg Harper, a Census Bureau demographer, "but they are just estimates."
In fact, the yearly counts for cities are estimates based on estimates. The Bureau uses tax return data, Medicare beneficiary numbers and statistics on births and deaths to estimate the population of each county each year. Then, it divvies up the county population among the county's cities using housing records, which aren't always an accurate guide to a place's population..
Harper notes something else about Atlanta and Detroit: Both cities successfully challenged the Bureau's mid-decade estimates to increase their population counts. When cities don't agree with these estimates, the Bureau offers them an alternative formula to come up with a new number. Among other things, this formula assumes that the housing vacancy rate has remained constant throughout the decade. That's an unlikely proposition in struggling Detroit or in Atlanta, which was hit by the housing bust. It isn't just the Census Bureau that's saying the once-a-decade count is the better guide. Ken Darga, Michigan's state demographer, agrees.
On the other hand, the Census Bureau wasn't the only organization saying Atlanta was growing quickly during the last decade. Virtually everyone was saying that. Jerome McKibben, a demographer who has studied Atlanta, says the lowest recent estimate for the city was 480,000, far more than the decennial count. "Just about every data vendor, demographer, city planner, just about every person that looked at Atlanta, everybody else had them above 500,000." McKibben says. "It's not like every demographer got collectively stupid."
McKibben's case is that a variety of records prove there are far more than 420,000 people living in Atlanta. He believes these records give a more accurate picture than old-fashioned enumeration. After all, doing it the traditional way isn't easy. The Bureau has to collect addresses for every place where people live. If it doesn't have an address on file, it won't know to send a form. Even with addresses, not everyone is easy to reach or eager to respond. Places with large transient populations or large immigrant populations-like many big cities-are particularly difficult. McKibben thinks Atlanta was undercounted by 80,000 to 100,000 people.
Yet making that case to the Bureau after the fact is exceptionally difficult. For the decennial count, the Bureau accepts challenges on only three grounds: if a housing unit was assigned to the wrong location; if boundary lines were in error; or if it omitted an address because of a processing mistake. In other words, cities have to find specific errors. A general case that a city was undercounted won't cut it.
McKibben thinks the most Atlanta could add in a challenge would be 30,000 to 40,000 people. Even that would require a painstaking process of hunting for addresses the Bureau didn't include. Plus, it wouldn't come in time to preserve the city's political power in the once-a-decade redistricting process, one of the major immediate uses of the new population numbers. Detroit and Atlanta both stand to lose multiple state legislators in redistricting because of the Census counts.
Despite the obstacles, officials in quite a few cities-including Detroit, Atlanta, New York City, Chicago, Houston, Cincinnati, Miami and Buffalo-have discussed challenges starting June 1, when the Census Bureau will begin accepting them. That's because the financial stakes are so large.
New York City, for example, estimates that it receives an additional $3,000 per year in federal aid for each additional person who is counted. In Michigan, local government aid from the state is tied to population in the decennial count, not the annual estimates. Millions of dollars are at stake for Detroit.
Census counts affect cities in lots of other subtle ways. In Michigan, state law grants cities with more than 750,000 people special powers to assess taxes on personal income and utilities. The laws were designed to apply to Detroit and only Detroit. Now city officials are pleading with state lawmakers to change the threshold because the city suddenly has fewer than 750,000 people-at least according to the Census Bureau's reckoning in 2010.