Data generated for this tool was created using polling data from the University of Texas, Yale Project on Climate Change Communication, Pew Research Center, and additional polling commissioned by ClearPath conducted by trusted Republican pollsters. Congressional district and state level projections were modeled using multilevel regression and poststratification (MRP), a technique for localizing national polling data by modeling demographic and geographic predictors associated with polling responses. Projections reflect opinions of registered voters.

Geographic predictors used for these models included Cooks Partisan Voting Index, state and PADD monthly retail gas prices, state annual oil production, and power plant capacity data from the U.S. Energy Information Administration, district level opinion projections from the Yale Project on Climate Change Communication, annual district level employment in extraction industries from the American Community Survey (ACS) conducted by the US Census Bureau, and congressional district religiosity measures modeled by Mark Setzler of High Point University.

Demographics associated with political party affiliation and voter registration were taken from the Pew Research Center’s political surveys and congressional district demographics used for poststratification were taken from the ACS. In case where 2015 district level demographic and geographic predictors were not yet available, 2015 estimates were imputed based on historical trends from 2011-2014. All 2016 projections were created using historical opinion trends and imputed demographic and geographic predictors.

MRP was chosen since the technique has been shown to provide more accurate locality estimates than disaggregation, the technique for calculating state or congressional district estimates by simply taking the mean for the subset of respondents in each locality. As with all other polling data, the wording of questions affects the responses and there is a margin of error associated with the responses. Projections from questions in University of Texas Energy Poll are based on an initial sample of ~20,000 respondents collected between 2011 and 2015, projections from questions in the ClearPath Poll are based on an initial sample of ~1,300 respondents collected in 2015, Pew Research data is aggregated from multiple surveys, creating a sample of ~30,000 respondents collected in 2014. Howe, Mildenberger, Marlon, and Leiserowitz (2015) found MRP results based on survey aggregates totaling ~12,000 respondents provided projections typically within ±5% of additional region specific polling done for external validation. See below for more information about MRP.

Support for MRP

“Bayesian Multilevel Estimation with Poststratification: State-Level Estimates from National Polls” http://www.stat.columbia.edu/~gelman/research/published/parkgelmanbafumi.pdf

“How Should We Measure District-Level Public Opinion on Individual Issues?” Author(s): Christopher Warshaw and Jonathan Rodden Source: The Journal of Politics, Vol. 74, No. 1 (Jan., 2012), pp. 203-219 Published by: University of Chicago Press on behalf of the Southern Political Science Association Stable URL: http://www.jstor.org/stable/10.1017/s0022381611001204

Howe, Peter D., Matto Mildenberger, Jennifer R. Marlon, and Anthony Leiserowitz (2015). “Geographic variation in opinions on climate change at state and local scales in the USA.” Nature Climate Change, doi:10.1038/nclimate2583.

“Forecasting elections with non-representative polls” http://www.stat.columbia.edu/~gelman/research/published/forecasting-with-nonrepresentative-polls.pdf

Data Sources Used

Congressional District Cooks Partisan Index http://cookpolitical.com/file/2013-04-47.pdf

2011 – 2014 1 year estimates from the American Community Survey (ACS), US Census Bureau. https://www.census.gov/data/developers/data-sets.html

Setzler, Mark. Forthcoming. “Religious Differences among Congressional Districts and the Success of Women Candidates” Politics & Gender. Accepted for publication in October 2014.http://acme.highpoint.edu/~msetzler/SetzlerPubsData/PAG2015/ReligEstimatesByCDw00thru12.xlsx

Retail gasoline prices from the U.S. Energy Information Administration. https://www.eia.gov/dnav/pet/pet_pri_gnd_dcus_nus_m.htm

2011-2015 Political Survey and Polarizations data from the Pew Research Center. http://www.people-press.org/category/datasets/The Pew Research Center bears no responsibility for the analyses or interpretations of the data presented here.

The University of Texas at Austin Energy Poll. http://www.utenergypoll.com/

Howe, Peter D., Matto Mildenberger, Jennifer R. Marlon, and Anthony Leiserowitz (2015). “Geographic variation in opinions on climate change at state and local scales in the USA.” Nature Climate Change, doi:10.1038/nclimate2583.