2018 Pilot Study
Updates & Errata
Current Dataset Release: January 29, 2019 Version
September 26, 2019
A supplemental data file with commercially sourced voter registration and turnout status for 2016 and 2018 has been made available. We thank Jennifer Jerit of Stony Brook University for making this possible. ANES will add these voter data to a new version of the Pilot Study dataset in a future release, but until that has been done, users can separately download the new voter data file and append the new data to the previously released Pilot Study dataset.
The commercial voter data were provided to YouGov by Target Smart. ANES has not evaluated the accuracy of the supplemental voter data; the data are provided “as is,” for researchers to evaluate the data’s fitness for any particular purpose.
These data were produced by matching the YouGov respondents to the TargetSmart voter database. ANES has no information about how this matching was performed or about how accurate it may be. Cases that were deemed matched are coded as registered to vote, and variables indicate the case’s turnout status in the 2016 and 2018 primary and general elections and, in some cases, the type of party ballot the voter cast in the primary elections.
Voting status categories are not necessarily mutually exclusive, and reflect differences in reporting across states.
Cases that were not considered matched to any TargetSmart record may be unmatched for several reasons, including: the survey panelist did not report their name or address; the panelist reported a false name or address; there was an error in the official voting record; there was an error in the process of matching with the official voting record; the panelist was registered to vote at a different address than the one reported; or the panelist was not registered to vote.
The primary party status may also be missing for several reasons, including: the panelist was not matched to a voting record, the state does not report this information, or the state does not hold partisan primaries.
Value labels for the turnout status variables (tsvote2018g, tsvote2018p, tsvote2016g, and tsvote2016p) are as follows:
A. Voter cast an absentee ballot
B. Absentee/mail-in, non-SOS, applied via voter recognition
E. Voter cast an early vote ballot
F. Early, non-SOS, applied via voter recognition
M. Voter cast a ballot by mail
P. Voter cast a ballot at the polls
Q. Voter cast a provisional/questionable ballot
R. Poll, non-SOS, applied via voter recognition
S. Questioned/provisional, non-SOS, applied via voter recognition
Y. Individual voted in the election
Z. Voted, no vote method available, non-SOS, applied via voter recognition
Value labels for the primary party variables (tsvote2018party and tsvote2016party) are as follows:
N. No party preference
O. Other party
February 8, 2019
On the initial data release, the variable inputstate has erroneous codes
for respondents from Colorado and Connecticut. To correct the variable,
cases coded -1 should be changed to 9 for Colorado and cases coded -7
should be changed to 8 for Connecticut, and the labels should be updated
accordingly. The following (SPSS) code will make the corrections.
recode inputstate (-7=8) (-1=9) (else=copy).
val lab inputstate
11 'District of Columbia'
33 'New Hampshire'
34 'New Jersey'
35 'New Mexico'
36 'New York'
37 'North Carolina'
38 'North Dakota'
44 'Rhode Island'
45 'South Carolina'
46 'South Dakota'
54 'West Virginia'
We thank Charles Stewart from MIT for noting this error.