All Things Census
The Census Bureau began a gigantic release of 2010 Census data today, publishing detailed race, Hispanic and population totals down to the block level for Louisiana, Mississippi, New Jersey and Virginia. The bureau will release data for the remaining states (and District of Columbia) over the next two months, with the final state numbers due out before April 1.
The legal purpose of this data is that it will be used as the basis for redrawing the boundaries of political districts within states, from congressional districts to local government boards. But the numbers also will have a fascinating story to tell about racial change, dispersion of the Hispanic population nationwide, population growth or decline and other themes.
A few of the earliest news stories include: Hispanics largest minority in New Jersey in the Star-Ledger; Asian and Hispanic populations skyrocket in Virginia in The Washington Post; New Orleans population drops since 2000 in the Times-Picayune; and Mississippi coast population loss not as great as feared in the Biloxi Sun Herald.
How well did the Census Bureau’s population estimates for the first decade of the 21st century match the actual counts from the 2010 Census? The short answer: Pretty well for the nation, and for all but a handful of states.
The accuracy of these population estimates is important because the numbers, which are released each year in between the once-a-decade census counts, are the basis for distributing billions of dollars in federal funds and are the denominators for rates used in some federal surveys. Unlike the census, which counts people directly, the estimates are assembled using government data, including birth and death certificates, immigration estimates and tax-return statistics on people who changed residences.
As Census Bureau Director Robert M. Groves pointed out in a news conference today, it greatly increases confidence in the census count if population numbers that are derived using different methods are similar. Groves said it reflects well on 2010 Census accuracy that the 2010 Census count for the nation, 308,745,538, was close to the bureau’s national population estimate of 308,977,944 on Census Day (April 1, 2010). The 2010 national count also matched up well with population estimates the bureau released last month using demographic analysis, an alternative measurement technique that uses a method somewhat similar to that used in the population estimates.
Digging into the estimates data, Pew Research Center demographer Jeffrey Passel has analyzed how the bureau’s population estimates for states compare with the official 2010 Census counts for states. He began with state estimates for July 1, 2009 (the latest available), and projected them forward to Census Day based on average growth rates for each state for 2007-2008 and 2008-2009. (This method produces a slightly different total for the nation than the Census Bureau computed, 309,081,328.) Read more
A decade ago, the apportionment counts from the 2000 Census showed that North Carolina was the luckiest state in the country. Based on its population gains, it won the last seat in Congress, the 435th. In 2010, though, North Carolina’s fortunes were not as good.
When the 2010 Census apportionment counts were announced last month, they showed that North Carolina fell short of winning the 435th or last seat. This time, the 435th went to Minnesota. (If it is any comfort to North Carolinians, the state would have gained the theoretical 436th seat in Congress, if one existed.)
Census numbers are used to apportion seats in the U.S. House of Representatives using a formula that assigns each state one seat, and, one at a time, allocates the remaining 385 places. The allocation is based on the size of each state’s resident population, plus any overseas military and civilian federal workers (and dependents) from that state.
In 2010, North Carolina fell nearly 15,800 people short of the number it would have needed to win the last seat instead of having it go to Minnesota, according to calculations by Pew Research Center senior demographer Jeffrey Passel. Read more
When the Census Bureau announced the first population totals from the 2010 Census for the nation (308.7 million) and for states on Dec. 21, the numbers did not include ethnic or race breakdowns. Those will be released later this year. A new report from the Pew Hispanic Center, using other Census Bureau data, concludes that Latinos played a key role in growth of the nation’s population and growth in states that gained congressional seats because of reapportionment, which is based on census numbers.
Using 2009 population estimates from the American Community Survey, Hispanics accounted for 51% of the nation’s population growth since the 2000 Census, which counted 281 million U.S. residents. From 2000 to 2010, the nation’s population grew 9.7%. From 2000 to 2009 (the last year available), the Hispanic population grew 37%.
Looking at population change for each state and the District of Columbia, Hispanics represent a greater share of eligible voter and resident populations in states that will gain congressional seats than in those states that will lose congressional seats, the report concludes. Hispanics represent 15.2% of eligible voters in the eight states that gained congressional seats, compared with 5.4% of eligible voters in the 10 states that lost congressional seats, according to estimates based on the 2009 American Community Survey. (Eligible voters are defined as U.S. citizens who are 18 years or older.) Latinos also represent 23.6% of the resident populations in states that gained seats, compared with 8.4% in states that lost seats.
The report makes the point that because many Latinos are either too young to vote or are not U.S. citizens, their population growth has not translated fully into electoral strength. However, the report states, “the electoral strength of the nation’s largest minority group will continue to grow in the coming decades.”
The first numbers from the 2010 Census, to be released tomorrow, are the state population totals that have been the basis of the proportional division of seats in the House of Representatives since the nation’s early days. The number of House seats has been fixed at 435 since 1913, but there have been numerous tweaks in the methodology used to divide them up—and debate continues today.
The U.S. Constitution requires that a census be taken every 10 years in order to divide the House of Representatives “among the several States according to their respective numbers, counting the whole number of persons in each State,” except for slaves, who, until the late 1800s, were counted as three-fifths of a person, and certain Indians. Under federal law, the Census Bureau must deliver population totals to the president nine months after Census Day, which now means the deadline is Dec. 31. The reapportioned Congress will convene in 2013.
The first numbers for each state will include a total of residents, as well as an apportionment count that will also include any members of the military or federal employees overseas (and any dependents living with them) who can be allocated back to a particular state. In 2000, more than half a million Americans overseas were included in state apportionment counts. These overseas Americans also were included in the apportionment counts from the 1970 and 1990 censuses. (They won’t be included in the set of totals that are used to redistrict within states.) Read more
The Census Bureau did a better job in 2010 than it had in 2000 reaching out to “hard-to-count” groups, such as minorities and renters, who are more likely to be missed by census-takers than other Americans, according to a report released this week by the Government Accountability Office. The bureau’s outreach included a massive targeted advertising campaign, partnerships with thousands of community organizations, and special counts at institutions such as soup kitchens.
The report, one of three census evaluations released by GAO the week before the Census Bureau is scheduled to publish the first numbers from the 2010 Census, had general praise for the bureau’s outreach efforts (especially the advertising campaign), but noted some faults. It cited a balky and hard-to-use database used by the partnership staff. The report also quoted some census officials who said there was poor coordination between the local offices managing the count in their area and the outreach staff who were assigned to their localities. The Be Counted/Questionnaire Assistance Center program, which provided questionnaires at 38,000 staffed and unstaffed locations, returned and checked in an average of only 20 forms per site, indicating an operation that “was very resource intensive relative to the number of forms that were returned.”
The GAO noted that the Census Bureau is itself evaluating various aspects of its outreach campaign, with results expected in 2012, so the full impact of its efforts is not yet known. But the report stated, “the Bureau’s rigorous effort to raise awareness, encourage participation, and enumerate HTC (hard-to-count) populations likely played a key role in holding mail participation rates steady in 2010 for the overall population, a significant achievement given the various factors that were acting against an acceptable mail response in 2010.”
A newly released Government Accountability Office review of Census Bureau follow-up efforts to reduce errors in the 2010 Census raises an issue that is familiar to survey researchers: The problem of reaching the growing share of Americans who only have cell phones and not landlines. Although the 2010 Census was mainly a mailback operation, the Census Bureau also telephoned nearly 7.4 million households to seek clarification of confusing or potentially incorrect information on their census forms.
This operation, called Coverage Follow-up (CFU), was completed on schedule and under budget, GAO concluded in a report that was one of three evaluations released by the agency the week before the Census Bureau is scheduled to publish the first counts from the 2010 Census. The GAO evaluation also found the bureau completed another follow-up task, Field Verification, on time and under budget. But the agency’s evaluations raised questions about some aspects of the design of both operations, which had as their goal a reduction in the number of Americans who were missed, counted more than once or counted in the wrong place by the 2010 Census.
The problems encountered with telephone follow-up were among a number of issues discussed in the GAO report, but they are the focus of this posting because they are relevant to any research that relies on reaching respondents by telephone. As GAO points out, the National Center for Health Statistics has found that more than 24% of U.S. households only use wireless telephones, not landlines, and work by the Pew Research Center also has found that when survey-takers only call landlines, it can skew their results. Including cell phones in a survey sample adds significantly to the cost and complexity of conducting research.
In the Coverage Follow-up operation, the Census Bureau contacted households with census form problems by telephone, which enabled it to reach more households than it would have been able to with more expensive in-person visits. It also used a commercial database to identify landline phone numbers that were missing or not valid, so as to target its calls more effectively. The Census Bureau only used landlines from the commercial database “due to concerns about not knowing the geographic location of wireless phones it might dial and the possible financial burden on wireless customers of unsolicited calls,” the GAO report said. Read more
A few days before the Census Bureau is scheduled to release the first population totals from the 2010 Census, the Government Accountability Office published three reports evaluating key operations of the decennial count. In a letter to Congress summarizing its findings, the agency pronounced the 2010 Census “operationally successful,” but also suggested that the basic census design “is no longer capable of delivering a cost-effective headcount.”
One of GAO’s reports on the 2010 Census evaluates the the non-response followup operation and other key field operations, and looks ahead to reforms for the 2020 Census. A second report examines several follow-up operations intended to reduce errors from the original count. A third GAO report looks at outreach and enumeration programs for “hard-to-count” populations such as immigrants and the homeless. This posting describes the first report; subsequent postings will explore the other two.
Here are some key sentences, from its letter to Congress, summarizing GAO’s conclusions in the first report: “Although some additional work and more data releases lie ahead, and information on the accuracy of the count is not scheduled to be available until early 2012, this much is clear: The Bureau generally completed the enumeration phase of the 2010 Census on schedule and consistent with its operational plans, and largely surmounted a series of risks that jeopardized the success of the headcount…an operationally successful census was no small accomplishment.”
However, the letter notes, “the 2010 Census required an unprecedented commitment of resources, including recruiting more than 3.8 million total applicants–roughly equivalent to the entire population of Oregon–for its temporary workforce; and it escalated in cost from an initial estimate of $11.3 billion in 2001 to around $13 billion, the most expensive population count in our nation’s history.
“Further, our oversight of the 1990, 2000, and now 2010 Censuses suggests that the fundamental design of the enumeration–in many ways unchanged since 1970–is no longer capable of delivering a cost-effective headcount given the nation’s increasing diversity and other sociodemographic trends.” Read more
The Census Bureau announced today that the first numbers from the 2010 Census will be released on Tuesday, Dec. 21. These include resident population totals for each state (and D.C. and the nation), as well as apportionment population totals that are used to divide the 435 seats in the U.S. House of Representatives among the states. The apportionment totals include the resident population for each state, as well the number of overseas members of the military and federal workforce (plus dependents living with them) who can be allocated back to a particular state.
The bureau also has a nifty interactive map showing population change and apportionment trends going back to 1910. Click on the reapportionment map for 1920, and there is no data shown. That’s because Congress refused to reapportion that year because officials from rural areas feared the loss of power to growing cities.
This year’s release is a week earlier than in 2000, when the reapportionment and resident population totals were released on Dec. 28, 2000. A dozen seats in the U.S. House changed hands in 2000, with Southern and Western states gaining at the expense of those in the Northeast and Midwest. The same general pattern is expected from the 2010 Census, reflecting a long-term pattern of national population change.
The Census Bureau today released five sets of population estimates for the nation as of April 1–but not from the soon-to-be-released 2010 Census count. The estimates are based on an alternative measurement technique, called demographic analysis, that agency officials say employs plausible assumptions about population change. The demographic analysis estimates range from a low of 305.7 million to a high of 312.7 million. The first numbers from the 2010 Census, which will include a national population total and state totals for reapportionment purposes, will be released later this month.
At a news conference today, Census Bureau Director Robert Groves said the five different estimates all make “plausible assumptions.” The analysis will be among three major tools used by the bureau to evaluate the quality of the 2010 Census. The other two include quality measures such as mail return rates from the census itself, and a post-census quality-check survey for which results will be released in 2012.
The demographic analysis methodology is based mainly on use of birth, death, immigration and Medicare records since the 2000 Census. For the population under age 65, bureau officials subtracted the number of deaths from the number of births, added in immigration and subtracted emigration. For the population ages 65 and older, the analysis relied mainly on Medicare enrollment data, supplemented by estimates from the Current Population Survey of older Americans not enrolled in Medicare.
The data exercise sounds deceptively simple, but as this technical paper demonstrates, Census Bureau experts had to make numerous choices based on sometimes-incomplete information. The release of five different sets of estimates reflects a lack of certainty about the data on which the estimates were based. Each of the five sets is based on a different estimate of net international migration over the past decade. The lowest and highest estimates also are based on different estimates of births, deaths and the 65-and-older population. The range of the estimates is 7 million people, of which 5 million is due to differing net immigration estimates, especially counts of unauthorized immigrants.
Bureau officials have used this alternative technique for decades to check the quality of the door-to-door census count, and to improve the accuracy of the annual population estimates between census counts. A decade ago, the bureau’s initial total derived by demographic analysis was markedly lower than its 2000 Census count–279.6 million, compared with 281.4 million. A major reason was that demographic analysis had underestimated unauthorized immigration; there also were issues with birth statistics and race classification. For 2010, bureau officials attempted to improve their immigration estimates by analyzing data from the American Community Survey, which went fully operational in mid-decade, and by consulting outside experts.
In addition to the national population totals, the five series of estimates also include numbers for men and women, for each single year of age, for race groups (only black and non-black because of data limitations), and for the Hispanic population under age 20 (again, because of data limitations, the full Hispanic population was not estimated). There are no estimates for states or other geographic units below the national level, again because of limited data.
Although the estimates released today are not 2010 Census numbers, bureau officials will use them to help evaluate the quality of the census count. In particular, after detailed race, age and gender numbers from the census come out in 2011, demographers will examine any patterns of difference between the census counts and demographic analysis estimates to see how both measurements can be improved.
At a news conference, Groves emphasized that the estimates had been developed independently of the 2010 Census counts; the demographic analysis team was “sequestered,” he said, from the work on 2010 Census totals. The demographic analysis released today also was not based on the annual population estimates that use a similar methodology; the last national estimate said that as of July 2009, the U.S. population was 307 million.