Home > Data Services > Census Support > Tools . 2010 . Topics . Experts

Search Data Services

Meta Search
search across all the following databases:

Data Catalog
Data and documentation

Common questions and answers.

Entire collection of data resources.

Latest Data News

RSS Feed icon

Altmetrics: What are they and/or should I care?

The Undercover Historian

Weighting Makes a Difference

Creating Cartograms

Big Data is not about the Data

Expect the Unexpected: Apportionment Projections based on Census Bureau Estimates

The 2010 Census has finished its data collection phase. Its next newsworthy milestone is the release of state-level population counts to Congress (December 31, 2010). This will tell us the winners and losers in the apportionment of power in Congress. Some analysts have already made predictions of the winners and losers based on the Census Bureau's most recent population estimates.

The predictions using an apportionment calculator and July 1, 2009 estimates seem reasonable:

3 seats (Texas)
1 seat (Arizona, Florida, Georgia, Nevada, South Carolina, Utah, Washington)

2 seats (Ohio)
1 seat (Illinois, Iowa, Louisiana, Massachusetts, Michigan, New Jersey, New York, Pennsylvania)

Links to just missed out and just made it states.

But, how well did the Census Bureau's population estimates work in predicting the 2000 apportionment results?

While most of the above projections will hold, one should recall that there were unexpected discrepancies in the projections based on the July 1, 1999 postcensal estimates and the final apportionment results based on the 2000 Census. And the deviations were not really foreseeable, e.g., they did not just involve the states who were projected to barely retain or lose a seat. Montana was the only state really on the brink in a published projection based on 1999 data; it ended up not getting the seat. Michigan ended up losing a seat and yet had a seemingly safe margin of approximately 200,000. Likewise, North Carolina was supposed to be well out of the running needing almost 150,000 more in population and in 447th place and yet ended up with the 435th seat.

Below is a quick summary of projections made on the basis of July 1, 1999 data and the actual apportionment results for 2000. Links to the data for all states is provided below.

Predicted Apportionment Results (based on July 1, 1999 postcensal1 estimates):

2 seats (Arizona, Texas)
1 seat (California, Colorado, Florida, Georgia, Montana, Nevada)

2 seats (New York, Pennsylvania)
1 seat (Connecticut, Illinois, Mississippi, Ohio, Oklahoma, Wisconsin)

Actual Apportionment Results (based on Census 2000 apportionment population)

2 seats (Arizona, Florida, Georgia, Texas)
1 seat (California, Colorado, Nevada, North Carolina)

2 seats (New York, Pennsylvania)
1 seat (Connecticut, Illinois, Indiana, Michigan, Mississippi, Ohio, Oklahoma, Wisconsin)

Suprises: Population Estimates Did Not Predict This

Went from 0 to +1 or +1 to +2
North Carolina, Florida, Georgia

Went from +1 to 0 or 0 to -1
Montana, Indiana, Michigan

One obvious weakness in using the 2009 or 1999 Population Estimates to project apportionment for 2010 or 2000 is that states vary in their population growth. The 2009 Population Estimates are leaving out the last 9 months of growth (July 1, 2009 – April 1, 2010). Is this how a seemingly safe state loses a seat (Michigan) or how a state vaults over many other states to gain a seat (North Carolina)? Or was there something more fundamental that led to the surprises in 2000?

While projecting the July 1 estimates forward to April 1, 2010 might improve our apportionment projections - particularly for the states on the bubble, one can still be left with surprises if the underlying estimates have any problems with them. It might help to review how these numbers are generated. The Census Bureau's Population Estimates program uses demographic methods to calculate state-based estimates of the US population. The weakest part of the equation is its migration estimates. The United States does not have a population registry that allows the federal government to keep track of interstate moves and international moves, so the Census Bureau has to make its best approximation of population movement based on proxies. It should be noted that the proxies have changed over the past decade with the advent of the American Community Survey. This source did not exist in the 1990s. The methodology for the postcensal estimates were very much the same back in 1999; it just used different proxies.

There has been a dramatic slowdown in both international and internal migration with the economic downturn and housing collapse. [Frey, 2010; Silver, 2009]. This could have a positive benefit to states that have been donor states (e.g., the Northeast and Midwest) and a negative affect on the Sunbelt or the West. Likewise, according to a study from Pew Hispanic, the undocumented population coming to the US has declined in the past year - mostly related to the collapse of the economy and the construction industry in particular. This has hit selected states - Florida, Virginia, and Nevada the most. Will this help Ohio, which is projected to lose 2 seats?

Related to the collapse in the housing market is an increase in foreclosures. Census residence rules are quite clear on how families (or individuals) living temporarily with others should be enumerated. However, most respondents filling out the census are not familiar with these rules, skipped the instructions, or did not answer the screening question (Question 2) on the census form, which might have prompted a follow-up by the Census Bureau. So, states with high foreclosure rates (Arizona, Nevada, Florida, and Michigan) could be adversely affected if doubled-up households were left out of the count. An indication of how serious an issue this might be is recent results from the Current Population Survey (CPS). The most recent poverty release (September 2010) showed that there has been an 11.6 percent increase in the number of doubled up families. The CPS is a survey that uses professional interviewers. It is more likely that the CPS will pick up doubled-up families than the self-enumerated census will.

The Hispanic population is roughly 15 percent of the U.S. population. The Hispanic population is a more difficult population to estimate using demographic methods than other groups partly because of the undocumented population. See [Passel] for a description of issues. This throws a wrench into all apportionment projections because a shift in the population of one state can have an impact on another state. And, there is reason to think that the Census Population Estimates could be off a bit for Hispanics.

Is the Hispanic population larger than the Census Estimates? Just a tiny bit larger (100,000) or substantially larger (1,000,000)?

The political posturing of whether or not to include non-citizens in the 2010 Census in the fall of 2009 and the new Arizona immigration law [SB 1070] might have an affect on the participation rate of Hispanics. This could hurt states with high proportion Hispanic - particularly a state like Arizona, which was at the center of the latter controversy. On the other hand, a Pew Hispanic study found that foreign born Hispanics were more positive and knowledgeable than their native born counterparts about the 2010 Census. So maybe, the distribution of states' by their nativity profile is critical to the final count of Hispanics. For instance, the Hispanic population is Texas is more native than foreign whereas the reverse is true for Florida, Georgia, and North Carolina.

What will the final count of Hispanics look like in the 2010 Census? There were many reasons for Hispanics to avoid the census. In the end, the effects could cancel each other out and have no net impact on the apportionment.

Finally, there may be some lingering affects of disasters: Hurricane Katrina residents moving back to Louisiana (or not) and Haitians coming to the US on the heels of the January 2010 earthquake. In the current projections, if Louisiana's population ends up being ~70,000 larger than the Census estimates, it will not lose its 7th Congressional seat. That is a fairly substantial population for a state of 4,000,000, but within reason considering the population loss Louisiana experienced from Hurricanes Katrina and Rita.

Haitian refugees have been settling in two very populous states (New York and Florida). Their numbers are unlikely to have an impact on apportionment for these states. In the end, the more interesting numbers will be the final counts and demographic characteristics for the population in Orleans parish (New Orleans).

Expect the unexpected. There should be some discrepancy in the projections based on the July 2009 Population Estimates data.2

Table Comparing Census 2000 results with July 1, 1999 Projections
[Excel] [pdf ]

1 Postcensal population estimates are calculated for the years following a census and before the next census is taken. These are based on demographic components of change (births, deaths, and migration). Here is a link to the source for the July 1, 1999 postcensal estimates.

2 The estimates by Kimball Brace at Election Data Services do more than just use the July 1, 2009 estimates. They have many variations - mostly based on previous patterns of growth for states. Likewise, Frey has several scenarios based on the abrupt drop in population movements (e.g., migration) in the latter part of this decade. His mid-decade projection was also heavily influenced by migration patterns.

Brace, Kimball. 2009.
New Population Estimates Show Additional Changes for 2009 Congressional Apportionment, With Many States Sitting Close to the Edge for 2010. Election Data Services.

Brace, Kimball. 1999. Congressional District Reapportionment Program. Election Data Services.

Benson, Clarke. 2009. Mid-Recession Migration in 2010: Apportionment in 2010. Polidata.

Census Bureau. 2010a. Residence Rules and Residence Situations for the 2010 Census. Census Bureau Web site: http://www.census.gov/population/www/cen2010/resid_rules/resid_rules.html.

Census Bureau. 2010b. 2010 Census Questionnaire (sample). Census Bureau website: http://2010.census.gov/2010census/pdf/2010_Questionnaire_Info.pdf.

Census Bureau. 2009. Methodology for the State and County Total Resident Population Estimates (Vintage 2009): April 1, 2000 to July 1, 2009.

Census Bureau. 1997. Methodology for U.S. Population Estimates by Age, Sex, Race, and Hispanic Origin: 1990 to 1996.

Frey, William. 2010. A Demographic Lull at Census Time.The Brookings Institution.

Frey, William. 2009. A Rollercoaster Decade for Migration. The Brookings Institution.

Frey, William. 2005. The Electoral College Moves to the Sunbelt. The Brookings Institution.

Johnson, David. 2010. Income, Poverty, and Health Insurance Coverage Online News Conference. Census Bureau.

Lopez, Mark and Paul Taylor. 2010. Latinos and the 2010 Census: The Foreign Born are More Positive.

Passel, Jeffrey. 2007. Unauthorized Migrants in the United States: Estimates, Methods, and Characteristics. OECD Social, Employment and Migration Working Papers.

Passel, Jeffrey and D'Vera Cohn. 2010. U.S. Unauthorized Immigration Flows are Down Sharply Since Mid-Decade. Pew Hispanic Center.

Ramirez, Andres. 2009. The New Contituents: How Latino Population Grown Will Shape Congressional Apportionment after the 2010 Census. American's Voice Education Fund.

Silver, Nate. 2009. Red States Gaining Ground, But Migration Slowing Down. FiveThirtyEight.com.