National Neighborhood

Data Archive (NaNDA):

Voter Registration,
Turnout, and
Partisanship by County, United States,
2004-2018
openICPSR-125781
nanda_voting_county_2004-2018_01P.dta
nanda_voting_county_2004-2018_01P.csv
nanda_voting0418C_01P.sas7bdat

Overview and Data Dictionary

Documentation Version: 1.1
Last updated: 11/4/2020

Dataset Overview

Description
This dataset contains counts of voter registration and voter turnout for all counties in the United
States for the years 2004-2018. It also contains measures of each county’s Democratic and
Republican partisanship, including six-year longitudinal partisan indices for 2006-2016.

Principal Investigators
● Megan M. Chenoweth, University of Michigan Institute for Social Research
● Mao Li, University of Michigan Institute for Social Research
● Iris N. Gomez-Lopez, University of Michigan Institute for Social Research
● Ken Kollman, University of Michigan Institute for Social Research
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

Funding Sources
● United States Department of Health and Human Services. Administration for Community
Living. National Institute on Disability, Independent Living, and Rehabilitation Research
(90RTHF0001)
● United States Department of Health and Human Services. National Institutes of Health.
National Institute on Aging (RF1-AG-057540)

Data Sources
Data on voter registration and turnout was taken from the Election Administration and Voting
Survey (EAVS) datasets (United States Election Assistance Commission, 2004-2018). The
EAVS is conducted every two years following a federal election by the United States Election
Assistance Commission (USEAC). Information on voting, voter registration, and election
administration is collected from local election officials at the state and county levels (USEAC,
2020).

Citizen voting age population (CVAP), which is used to calculate voter turnout, was taken from
United States Census Bureau data sources, specifically the 2000 decennial census and the
2012 and 2017 American Community Survey five-year estimates. (More information about how
we selected CVAP for each year is available in the methodology section of this codebook.)

Partisanship indices are constructed from county-level presidential election results and precinct-
level senate election results. All presidential race data comes from the MIT Election Data and
Science Lab (2018a). Senate race data for 2000-2014 comes from the Harvard Election Data
Archive (Ansolabehere et al., 2014, Ansolabehere et al., 2018). Senate race data from 2016
come from the MIT Election Data and Science Lab (2018b).

Coverage
The dataset contains one observation per county in the United States. Alaska is excluded, as
are the U.S. island territories.

Methodology
Our research team created this dataset in order to explore the relationship between voter
engagement (as expressed through registration and turnout rates), partisan political leanings,
community health, and public policy.

Voter turnout is a contextual predictor of social trust, which contributes to health and happiness
of a neighborhood’s residents (Rahn et al., 2009, Kawachi et al., 1997). Prior research has
found that voter turnout is low in disadvantaged neighborhoods with high levels of immigration
(Levine et al., 2017) and that voter turnout is lower in neighborhoods with certain design
features such as car-oriented development (Hopkins & Williamson, 2012). Rates of voter

Page 2
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

participation, as well as partisan leanings, are correlated with self-rated health (Pacheco &
Fletcher, 2015) and affect attitudes toward health care and policy (Bye et al., 2016). The
COVID-19 pandemic in particular has shown that partisanship affects health behavior such as
mask wearing (Gadarian et al., 2020, Pew Research Center, 2020), which has the potential to
affect health outcomes as well.

Voter Registration and Turnout

To construct measures of voter registration and turnout, we began by calculating three figures
for each county in the United States.

Registered voters: the total number of people registered to vote in the county. We obtained
these values from variable A1a (total registered voters) of the 2004-2018 Election
Administration and Voting Survey (EAVS, 2004-2018). (In 2004, this variable was called
“Reported Total Registration.” In 2006, it was called q022006total.)

Ballots cast: the total number of votes cast in the November general election for each year.
The EAVS variable code and description used for this figure varies across years:
 2004: “Total Ballots Counted” (no code)
 2006: q34total (“Ballots Counted—Total”)
 2008-2014: F1a
 2016-2018: D1a (“Votes Cast: Total”)

Note that survey data from the EAVS was not cleaned or imputed in the process of creating
these measures. Some data is missing for one of the following reasons.
 Data is available only for even years. Elections do not occur in odd years in most United
States jurisdictions.
 Data is missing for all years and counties in Alaska, because elections are not
administered at the county level.
 Voter registration counts are zero or missing for most years for the state of North
Dakota, which does not require voter registration. (Voter registration figures for 2004
appear in the source data and are based on each county’s own CVAP estimate.)
 Responses from some counties may be missing for certain years due to survey
nonresponse on the part of local election officials.

Voting population: following McDonald and Popkin (2001) and McDonald (2020b), we
evaluated three possible measures for this component.
 Voting age population (VAP), the total number of people age 18+ in the county.
 Citizen voting age population (CVAP), which excludes noncitizens. (Per United States
law, non-citizens are not allowed to vote in federal, state, and most local elections
[Findlaw.com, 2020].)
 Voting eligible population (VEP), which excludes noncitizens and ineligible felons (where
disallowed by state law).

Page 3
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

While VEP is the most accurate measure, we were not able to obtain or estimate county-level
figures for ineligible felons. We selected CVAP because it is more accurate than VAP and is
readily available from the United States Census Bureau.

To maintain consistency with data sources that are commonly used in other NaNDA datasets,
we used the following Census Bureau sources to calculate CVAP:
 2004-2008: CVAP for each year is interpolated based on the 2000 decennial census and
the 2012 ACS 5-year estimate.
 2010-2012: CVAP is taken from the 2012 ACS five-year estimate (covering the period
2008-2012) with no interpolation.
 2014-2018: CVAP is taken from the 2017 ACS five-year estimate (covering 2013-2017,
carried forward to 2018).

For more information on why NaNDA uses these data sources and on interpolation methods,
see the codebook for National Neighborhood Data Archive (NaNDA): Socioeconomic Status
and Demographic Characteristics of Census Tracts, United States, 2008-2017 (Melendez et al.,
2020).

We then calculated three ratios using these components:

 Voter registration: registered voters / voting population.
 Voter turnout: ballots cast / voting population.
 Registered voter turnout: ballots cast / registered voters.

In some counties, we found that the number of registered voters is higher than CVAP. In fewer
counties, the total number of ballots cast also exceeds CVAP. One possible cause for this is
“deadwood” in voter registration rolls: voters who have moved or died but not yet been removed
from voter rolls in their former county by local election administrators (Shaw et al., 2015, Levitt,
2007). Another is that CVAP is survey-based rather than based on a true enumeration of
population (Nyhan et al., 2017). A third is that the EAVS, the source for our registered voters
and ballot count data, is a survey and is subject to responder error. To account for these issues,
we have top-coded all three ratios at .98 (98% voter registration and turnout).

Partisanship
For the years 2006-2016, we calculated Democratic and Republican partisanship indices for
each county based on its voting history in presidential and senate races over the current and
three prior elections. We built this index based on votes for president and senator because
these races occur consistently in even years all fifty U.S. states.

We extracted votes for Democratic and Republican presidential candidates from county-level
data on presidential election outcomes for the years 2000-2016. For senate races, we
summarized 2000-2016 precinct-level votes for Democratic and Republican candidates to the
county level, then joined them with the presidential election vote counts. This resulted in four
figures per county per year:

Page 4
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

 pres_dem_votes: Votes for Democratic presidential candidates

 pres_rep_votes: Votes for Republican presidential candidates
 sen_dem_votes: Votes for Democratic senate candidates
 sen_rep_votes: Votes for Republican senate candidates

For each county and year, we created four ratios:

 pres_dem_ratio: pres_dem_votes / (pres_dem_votes + pres_rep_votes)
 pres_rep_ratio: pres_rep_votes / (pres_dem_votes + pres_rep_votes)
 sen_dem_ratio: sen_dem_votes / (sen_dem_votes + sen_rep_votes)
 sen_rep_ratio: sen_rep_votes / (sen_dem_votes + sen_rep_votes)
Note that votes for third party and independent candidates are not included in the denominator
of these ratios.

We then calculated an annual Democratic and Republican partisanship indices and a six-year
aggregate Democratic and Republican partisanship indices for each county for 2006 through
2016. The Democratic index is the average of the presidential and senate Democratic vote
ratios over the current and previous four elections. The Republican index is calculated using
Republican vote ratios in the same manner. The two indices add up to one for each county and
year.

A chart illustrating a hypothetical example of how these indices are calculated is available in
Appendix A: Partisanship Index Calculation.

The partisanship index and its source data components are missing in some circumstances.
 Data is available only for even years. Elections do not occur in odd years in most United
States jurisdictions.
 Partisanship indices are not calculated for 2000-2004 because six years of voting data
are required to calculate them. (Although voter turnout data is not available for 2000 and
2002, we have included the presidential and senate vote counts and ratios so that users
may replicate our partisanship ratio calculations.)
 Precinct-level voting data is not available for 2018, so we could not construct
partisanship indices for that year.
 The District of Columbia does not vote for senators, so we did not construct partisan
indices for its one county.
 Data is expected to be missing for presidential races in all states in 2002, 2006, 2010,
and 2014 because presidential elections do not occur in those years.
 Senate vote counts are missing in years in which senate races do not occur. (Every
state elects two senators to six-year terms, usually in different years.)
 Precinct-level senate returns are unavailable for the following states and years. (Note
that this list has not been checked to confirm whether a senate race would have
occurred in all years.)
 Arizona 2000 and 2010
 California 2000 and 2012

Page 5
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

 Colorado 2000 and 2002

 Delaware 2000
 Florida 2000-2008 and 2012
 Georgia 2000 and 2010
 Hawaii 2000 and 2002
 Illinois 2000-2012
 Indiana 2000-2012
 Kentucky 2000-2006
 Maryland 2002
 Massachusetts 2000-2002
 Michigan 2012
 Missouri 2000 and 2012
 Mississippi 2000-2002
 Nebraska 2000-2002, 2006, and 2010
 New Jersey 2010-2012
 New Mexico 2002, 2006, 2010
 Nevada 2000-2008, 2012
 New York 2000-2006 and 2012
 North Dakota 2000
 Ohio 2000-2002, 2012
 Oklahoma 2000-2006
 Oregon 2000-2006, 2012
 South Carolina 2000-2002
 South Dakota 2000-2002
 Tennessee 2000
 Utah 2000-2006 and 2010-2012
 Virginia 2000-2004
 Washington 2000-2006
 West Virginia 2000-2010
 Wisconsin 2000 and 2012
 Whenever possible, we summarized precinct-level senate returns with county-level data
by FIPS code, or if FIPS code was not available, on a combination of state and county
name. Precinct-level source files that do not identify a county in one of these two ways
cannot be summarized to the county level. Because of this, senate vote data and
partisanship ratios are missing in 2000-2012 for the following states: Connecticut,
Delaware, Hawaii, Maryland, Massachusetts, Minnesota, Rhode Island, and Utah.

Usage Note
A number of factors besides voter engagement and motivation can affect rates of voter
registration and turnout. For example, the requirement to register tends to lower turnout
(Ansolabehere & Koninsky, 2006). Some states permit Election Day voter registration, which

Page 6
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

has been shown to increase turnout (Neiheisel & Burden, 2012). In some states, voters are
automatically registered to vote when they interact with certain state agencies (often those
issuing driver’s licenses), which increases registration rates without necessarily increasing
turnout (Rakich, 2019). Turnout is consistently higher in presidential election years (McDonald,
2020b). In addition, demographic factors that affect an individual’s propensity to vote (such as
education) may have a cumulative effect at the neighborhood level (Sondheimer & Green,
2010).

The measures available in this dataset may be useful to exploring these trends. However, we
caution users against comparing registration and turnout rates across states and over time
without considering the effects of these other factors.

Page 7
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

Variables
Variable Type Obs Unique Mean Min Max Label
stcofips string 29067 3114 . . . County FIPS code
year int 29067 10 2009.501 2000 2018 Year
reg_voters long 24817 18594 58870.92 0 6858459 # registered voters
ballots_cast long 24817 17087 34056.86 0 3551506 Ballots cast in general election, all races
cvap long 24914 10509 69477.43 60 6218280 Citizen voting age population
reg_voters_pct float 24812 19598 0.8309706 0 0.98 % eligible voters registered (reg_voters / cvap), top coded
% eligible voters casting ballots (ballots_cast / cvap), top
voter_turnout_pct float 24812 23465 0.5002885 0 0.98 coded
% registered voters casting ballots (ballots_cast /
reg_voter_turnout_pct float 24539 22840 0.5385218 -1 0.98 reg_voters), top coded
pres_dem_votes long 15570 9916 19902.93 4 2464364 # votes for Democratic presidential candidate
pres_rep_votes long 15570 11198 18938.44 54 1076225 # votes for Republican presidential candidate
sen_dem_votes float 10916 7319 15361.83 0 1940493 # votes for Democratic senate candidate
sen_rep_votes float 10916 8216 17056.07 0 917527 # votes for Republican senate candidate
pres_dem_ratio float 15570 15525 0.3903424 0.0324675 0.9569519 % votes for Democratic presidential candidate
pres_rep_ratio float 15570 15517 0.6096576 0.0430481 0.9675325 % votes for Republican presidential candidate
sen_dem_ratio float 10890 10633 0.3965579 0 1 % votes for Democratic senate candidate
sen_rep_ratio float 10890 10631 0.6034421 0 1 % votes for Republican senate candidate
Democratic partisanship index (% votes cast, past 6
partisan_index_dem float 21791 17500 0.4005668 0.024595 0.9261177 years)
Republican partisanship index (% votes cast, past 6
partisan_index_rep float 21791 17492 0.5994332 0.0738823 0.975405 years)

Page 8
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

References
Ansolabehere, S., & Koninsky, D. M. (2006). The introduction of voter registration and its effect on
turnout. Political Analysis 14(1), 83-100. doi: 10.1093/pan/mpi034

Ansolabehere, S., Palmer, M., & Lee, A. (2014). Precinct-Level Election Data (V1) [Data set]. Harvard
Dataverse. doi: 10.7910/DVN/YN4TLR

Ansolabehere, S., Ban, P., & Morse, M. (2018). Precinct-Level Election Data, 2014 (V1) [Data set].
Harvard Dataverse. doi: 10.7910/DVN/B51MPX

Bye, L., Ghirardelli, A., & Fontes, A. (2016). Final Report: The American Health Values Survey. Robert
Wood Johnson Foundation.
https://www.rwjf.org/content/dam/farm/reports/reports/2016/rwjf437263/subassets/rwjf437263_1

Findlaw.com (2020, October 16). Can noncitizens vote in the United States?
https://www.findlaw.com/voting/my-voting-guide/can-noncitizens-vote-in-the-united-states-.html

Gadarian, S., K., Goodman, S. W., & Pepinsky, T. W. (2020). Partisanship, Health Behavior, and Policy
Attitudes in the Early Stages of the COVID-19 Pandemic. Social Science Research Network. doi:
10.2139/ssrn.3562796

Hopkins, D. J., & Williamson, T. (2012). Inactive by design? Neighborhood design and political
participation. Political Behavior 34, 79-101. doi: 10.1007/s11109-010-9149-2

Kawachi, I., Kennedy, B. P., Lochner, K., & Prothrow-Stith, D. (1997) Social capital, income inequality,
and mortality. American Journal of Public Health 87(9), 1491–1498. doi: 10.2105/ajph.87.9.1491

Levine, J. R., Leenman, T. S., Gershenson, C., & Hureau, D. M. (2017). Political places: Neighborhood
social organization and the ecology of political behaviors. Social Science Quarterly 99(1), 201-215. doi:
10.1111/ssqu.12352

Levitt, J. (2007). The truth about voter fraud. Brennan Center for Justice.
https://www.brennancenter.org/sites/default/files/2019-08/Report_Truth-About-Voter-Fraud.pdf

McDonald, M. P., & Popkin, S. L. (2001). The myth of the vanishing voter. American Political Science
Review 95(4), 963-974. doi: 10.1017/S0003055400400134

McDonald, M. P. (2020a). National general election VEP turnout rates, 1789-present. United States
Election Project. http://www.electproject.org/national-1789-present

McDonald, M. P. (2020b). Overview of how the voting-eligible population (VEP) [is] constructed. United
States Election Project. http://www.electproject.org/home/voter-turnout/faq/sold

Page 9
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

MIT Election Data and Science Lab. (2018a). County Presidential Election Returns 2000-2016 (V6)
[Data set]. Harvard Dataverse. doi: 10.7910/DVN/VOQCHQ.

MIT Election Data and Science Lab. (2018b). U.S. Senate Precinct-Level Returns 2016 (V11) [Data
set]. Harvard Dataverse. doi: 10.7910/DVN/NLTQAD.

Neiheisel, J. R., & Burden, B. C. (2012). The impact of Election Day registration on voter turnout and
election outcomes. American Political Research 40(4), 636-664. doi: 10.1177/1532673X11432470

Nyhan, B., Skovron, C., & Titiunik, R. (2017). Differential registration bias in voter file data: A sensitivity
analysis approach. American Journal of Political Science 61(3), 744-760. doi: 10.1111/ajps.12288

Pacheco, J., & Fletcher, J. (2015). Incorporating Health into Studies of Political Behavior: Evidence for
Turnout and Partisanship. Political Research Quarterly, 68(1), 104-116. doi:
10.1177/1065912914563548

Pew Research Center. (2020, June 25). Republicans, Democrats move even further apart in
coronavirus concerns. https://www.pewresearch.org/politics/2020/06/25/republicans-democrats-move-
even-further-apart-in-coronavirus-concerns/

Rahn, W. M., Yoon, K. S., Garet, M., Lipson, L., & Loflin, K. (2009). Geographies of trust. American
Behavioral Scientist 52(12), 1646-1663. doi: 10.1177/0002764209331531

Rakich, N. (2019, October 10). What happened when 2.2 million people were automatically registered
to vote. FiveThirtyEight. https://fivethirtyeight.com/features/what-happened-when-2-2-million-people-
were-automatically-registered-to-vote/

Shaw, D., Ansolabehere, S., & Stewart, C. (2015). A brief yet practical guide to reforming U.S. voter
registration systems. Election Law Journal: Rules, Politics, and Policy 14(1), 26-31. doi:
10.1089/elj.2014.0273

Sondheimer, R. M., & Green, D. P. (2010). Using experiments to estimate the effects of education on
voter turnout. American Journal of Political Science 54(1), 174-189. doi: 10.1111/j.1540-
5907.2009.00425.x

United States Census Bureau (2000). Census 2000 Special Tabulation 76: Data for Citizenship by
Race and Hispanic Origin [Dataset]. https://www2.census.gov/programs-
surveys/decennial/rdo/datasets/2000/2000-cvap/stp_76.zip.

United States Census Bureau (2012). CVAP 2008-2012 5 Year ACS Data - CSV Format [Dataset].
https://www2.census.gov/programs-surveys/decennial/rdo/datasets/2012/2012-cvap/CVAP_2008-
2012_ACS_csv_files.zip.

Page 10
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

United States Census Bureau (2017). CVAP 2013-2017 5 Year ACS Data - CSV Format [Dataset].
https://www2.census.gov/programs-surveys/decennial/rdo/datasets/2017/2017-cvap/CVAP_2013-
2017_ACS_csv_files.zip.

United States Election Assistance Commission. (2004). NVRA, UOCAVA, and Election Day Survey
Datasets: Excel [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/6/State%20Data%20Tables%202004%20UOCAVA
%20Survey.zip

United States Election Assistance Commission. (2006). NVRA, UOCAVA, and Election Day Survey
Datasets: Excel [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/6/2006%20UOCAVA%20Survey%20All%20Data.zi
p

United States Election Assistance Commission. (2008a). NVRA, UOCAVA, and Election Day Survey
Dataset: Excel [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/6/2008%20eavs%20xls%20august%2011%202010
.zip

United States Election Assistance Commission. (2008b). NVRA, UOCAVA, and Election Day Survey
Dataset: Illinois [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/6/Illinois_2008_Data_All_Categories.xls

United States Election Assistance Commission. (2010a). NRVA Dataset: Excel [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/1/2012EAVS_NVRAData.xlsx.

United States Election Assistance Commission. (2010b). Election Day Survey Dataset: Excel [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/28/Final%20EAVS%20Data%20-
%20Sections%20C%20to%20F_EXCEL.zip

United States Election Assistance Commission. (2012a). Election Day Survey Dataset: Excel – Section
F [Dataset]. https://www.eac.gov/sites/default/files/eac_assets/1/6/Excel%20Files-Part%202.zip

United States Election Assistance Commission. (2012b). NVRA Dataset: Excel [Dataset].
https://www.eac.gov/sites/default/files/eac_assets/1/1/2012EAVS_NVRAData.xlsx.

United States Election Assistance Commission. (2014). Election Administration and Voting Survey:
Excel [Dataset]. https://www.eac.gov/sites/default/files/eac_assets/1/1/2014_EAVS_Excel_Files1.zip.

United States Election Assistance Commission. (2016). Election Administration and Voting Survey:
CSV (V1.2) [Dataset].
https://www.eac.gov/sites/default/files/Research/EAVS_2016_Final_Data_for_Public_Release_v2.csv.z
ip.

Page 11
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

United States Election Assistance Commission. (2018). Election Administration and Voting Survey:
CSV (V1.3) [Dataset].
https://www.eac.gov/sites/default/files/Research/EAVS_2018_for_Public_Release_Updates3.csv.

United States Election Assistance Commission. (2020). EAVS FAQs. https://www.eac.gov/research-

and-data/eavs-faqs.

Page 12
 National Neighborhood Data Archive (NaNDA)
Voter Registration, Turnout, and Partisanship by County, United States, 2004-2018

Appendix

Appendix A. Partisanship Index Calculation

This chart shows the calculation of a hypothetical example of the Democratic and Republican partisanship indices.

To calculate these indices:

1. Calculate ratios for the presidential and senate votes (pres_dem_ratio, pres_rep_ratio, sen_dem_ratio, sen_rep_ratio) for each
year and race that votes are present (highlighted in blue).
2. Calculate the average ratio for each year (annual average of Democratic ratios, annual average of Republican ratios) that races
are present (highlighted in orange). Note that these variables are calculated in the process of creating partisan_index_dem and
partisan_index_rep but are not included in the dataset.
3. Calculate the average of all ratios across all four years (highlighted in green). These are the partisanship index values for 2016.

Year pres_ pres_ sen_ sen_ pres_ pres_ sen_ sen_ Annual Annual partisan_ partisan_
dem_ rep_ dem_ rep_ dem_ rep_ dem_ rep_ average, average, index_dem index_rep
votes votes votes votes ratio ratio ratio ratio Democratic Republican
ratios ratios
2010 300 100 .75 .25 .75 .25
2012 900 600 800 200 .6 .4 .8 .2 .7 .3
2014
2016 800 800 600 400 .5 .5 .6 .4 .55 .45 .67 .33

Page 13