This series on the key states for the 2016 presidential election is based on historical data and future projections provided to Next America by States of Change: Demographics and Democracy. The States of Change project is a collaboration supported by the Hewlett Foundation that brings together the Center for American Progress, the American Enterprise Institute, and demographer William Frey of the Brookings Institution. The project is directed by Ruy Teixeira of the Center for American Progress and Karlyn Bowman of the American Enterprise institute.
In a statement to Next America, the States of Change Project explained its program as follows:
The goals of the project are: (1) to document and analyze the challenges to democracy posed by the rapid demographic evolution of the United States, from the 1970’s to the year 2060; and (2) to promote a wide-ranging and bipartisan discussion of America’s demographic future and what it portends for the political parties and the policy challenges they (and the country) face.
Results from the first year of the project include:
“¢ Trend analysis of 40 years of demographic change in United States, 1974-2014, nationally and in every state, particularly as it has affected the pool of eligible voters
“¢ Projections of the racial composition of every state to the year 2060, both overall and by eligible voters
These findings are being released in several ways, first through this collaboration with National Journal/Next America. There will also be a detailed report discussing our national and state results over the entire 1974-2060 period and an interactive web feature that will allow users to trace the demographic evolution of any state’s electorate by a variety of different characteristics. The report and interactive feature will be released on or around February 24, when a public conference will be held at AEI to present and discuss the project’s findings.
All the historical data used in these are taken from the Census Bureau’s monthly Current Population Survey. States of Change analyzed the data for each relevant year from the Voter Supplement of the November survey. The stories examine two sets of results. The first is eligible voters-that is, the share of the state population that is over 18 and a citizen in each demographic category. The second is actual voters, the share of the state population in each demographic category that actually voted, according to the Census survey.
In analyzing the historic results, the States of Change project subjected the data to a statistical process known as LOWESS. The project explains that widely used statistical process this way:
The estimates presented here for eligible populations were produced using LOWESS, a statistical procedure that ‘smooths’ our CPS and projection data in order to generate more stable and accurate results. As with all surveys, results from the CPS are subject to random sampling error and any particular year’s estimates are likely to deviate from the true value. ‘Smoothing’ over these data points for characteristics whose rates of change should be fairly consistent across elections ultimately creates better estimates for any particular year.
The CPS survey generally paints a similar picture of the electorate each year as the exit polls conducted on Election Day for a consortium of media organizations. The CPS recently has found a slightly higher white share of the vote than exit polls, though both have followed a similar track of a declining percentage for whites. While the two data sources converged on the white share of the national vote in 1996 (83 percent) and 2000 (81 percent), in each of the past three presidential elections the Census has found the white share two percentage points higher than the exit polls did; in 2012, the Census put whites at 74 percent of the national vote, compared to 72 percent for the exit.
One difference between the two data sets remains larger. Compared to the CPS results, the exit polls consistently find that college-educated voters represent a greater share of the electorate. For instance, in Ohio in 2008, the exit poll found that college-educated whites equaled 33 percent of the electorate; the CPS put their share at 26 percent. In Virginia, that year the exit poll found that college-educated whites equaled 40 percent of voters; the CPS put their share at 33 percent.
The States of Change project has chosen to use the CPS data, so that is what Next America is using in these reports on the project’s work. (The Census data is the only available source on the eligible voter population; the exit polls only produce an assessment of the actual voters.) Next America has used both measurements of the electorate’s composition in the past, and will continue to do so in the future.
While the Census does ask about voting participation, it does not ask about voting preferences. All the historic data referenced in these stories about the voting preferences of the different groups in the various states comes from the Election Day exit polls. The name of the institution conducting the exit poll has changed over time; in 2012 it was conducted by Edison Research.
In examining the results by education in these stories, Next America is reporting the trends solely for whites with and without at least a four-year college degree, because that line of educational attainment has proven a significant fault-line in voting behavior among whites. The gap is not nearly as significant in the voting behavior of minorities with and without four-year degrees.
The States of Change project is unique because it has not only collected retrospective data but uses a demographic model constructed by William Frey of the Brookings Institution to forecast the future shape of the electorate, by state, through 2060. The project explained its methodology for making its projections of each state’s overall population in this statement:
The projections employ a multistate cohort component methodology which begins with the 2010 census and projects ahead in five year intervals, race and age-specific populations for each state to 2060 based on the components of domestic migration, international migration, fertility and mortality following from modeling put forth in Andrei Rogers,Introduction to Multiregional Mathematical Demography (New York: Wiley, 1975) These projections are performed separately for racial groups wherein each state’s domestic migration flows are projected between that state and the remainder of the four census regions (Northeast, Midwest, South West). International migration to the US for each interval is allocated to states and regions. In both cases, these migration flows and immigraton allocations are based on patterns recorded in the 2007-2012 multiyear American Community Survey. Race specific fertility and mortality rates for each state assume national rates specific to age and race. These as well as national immigration levels are broadly consistent with those used in US Census Bureau national projections.
In addition, the Project explains its methodology for forecasting the eligible voter population this way:
Like all of the data presented after 2014, the eligibility rates for these different populations are projections. These were generated by taking data from the American Community Survey and dividing up the American population into groups based on their state, race, and age. Multilevel statistical models were then used to estimate the eligibility rates and the naturalization rates for each state/race/age group. These groups were then tracked forward in time and had those unique naturalization rates applied to them. In addition, these estimates account for immigration into the state and the resulting effect it had on different groups’ overall eligibility rates. The end result is a procedure that is sensitive to the different rates of naturalization experienced by each of these groups and the immigration each state is predicted to experience in the future.