African American voters, 142, 153, 154, 156, 192n9
Alvarez, R. Michael, 149
American National Election Studies (ANES): economy and, 192n10; pre- and post-election interviews, 139, 143–44, 145, 146, 154–57, 192n3; turnover in successive elections and, 140–42; variance compression and, 157, 160
April polls (ED-200), 2, 4, 61–63, 165
autoregressive (AR1) process, 46–49, 60, 63. See also random walk (integrated series); stationary time series
Bartels, Larry, 41
beginning of election year, 1, 4, 166
bounces, 44–46, 166; conventions and, 76–77, 107, 188n23; intensification effect and, 55; of stationary time series, 47, 49, 50, 52. See also late campaign effects; shocks
bumps, 44–46, 166; conventions and, 76–77, 107, 166; of stationary time series, 47, 50, 52. See also shocks
Business Retrospections, 112, 116, 117–18
campaign effects, 1–2; downward trend over the years, 144; events perspective on, 11–13 (see also convention season; debates); fundamentals and, 2, 5–7, 13, 15, 41–42, 57, 177–78; models of, 13–14, 41–42; overestimated by observers, 176–77; partisan predispositions and, 55–57, 148–49; predicting specific elections and, 169; seeming contradictions of, 165; time-series properties and, 49, 50, 51–52; voter decision-making and, 10–11; voters’ relative utilities and, 53–55, 56. See also enlightenment (learning); fundamentals; late campaign effects; persuasion; priming; shocks
Campbell, James, 7
candidate attractiveness, partisan effects and, 57, 160. See also presidential approval; relative utility (relative liking)
candidate competence, 2; persuasion about, 10
candidate positions, 7. See also platform ideology
Catholic voters, 154, 156, 192n9
change in voter preferences, 15. See also campaign effects; volatility of voter preferences
Comparative Manifesto Project, 128
conservatism. See platform ideology; policy mood
convention season, 4, 12, 72–79, 102–7; bump associated with, 76–77, 107, 166; changes in scheduling of, 37, 60–61, 72; decline in poll variance over, 37–38; difference between Election Day vote and polls of, 90; economic factors independent of, 115; fundamentals and, 8, 167; learning in, 107, 131–32, 188n21; political model and, 130, 131–32; predicting the Election Day vote and, 97, 99, 100, 101–7, 188n19; volatility of voter preferences in, 31, 32, 67, 72–75, 79, 166
cross-sectional analysis, 60–72; biweekly-by-biweekly, 67; conventions and, 72–79; long-term equilibrium in, 69–72; mid-April vs. election eve, 61–63; of time series, 60–61; week-by-week, 63–69
cross-sectional variance, 32–38. See also variance compression, cross-sectional
crystallization of voter preferences, 2, 148–54, 155
daily analysis of vote division, 35–38; vs. Election Day vote, 88–90; reliability in, 183n15
decision-making by poll responders, 10–11
defections, partisan, 150–51
demographic variables, 140, 148; campaign evolution and, 151–54; Election Day vote and, 154–57
design effects on poll accuracy, 30
dropouts. See nonvoters
dynamics of vote intentions, 2, 3, 41–42; conclusions of initial analysis, 57–58; individual voter and, 42–44; models of, 46–52; shifting over time, 52–57; types of shocks contributing to, 44–46. See also random walk (integrated series); shocks; stationary time series; time series
early polls. See April polls (ED-200)
economic model, 109, 110–25; compared to political model, 133–36; disparate views of, 110; objective factors in, 110–11, 112–15, 117, 121–22, 123–25, 188n3; predicting Election Day vote from, 110, 121–25; subjective factors in, 111–13, 115–19, 122, 123–25, 189n12; summary of, 125
Economic News Heard, 189n13
Economic Perceptions, 117–18, 122, 123, 124, 125, 189n13; omitted from combined model, 134
economy, 2, 3, 5, 6, 7; as driver of vote intentions, 167, 168; election of 2000 and, 8, 168; political variables and, 134; predicting specific elections and, 169; presidential approval and, 109, 119–24, 133, 190n15; priming about, 9, 115, 168; vote switchers and, 192n10. See also fundamentals
ED-200 (April) polls, 2, 4, 61–63, 165
educational level of voter, 153, 155, 156
election markets, 176–77, 181n4, 193n5–6
election of 2000: economy and, 8, 168; intangibles in, 168
electoral (true) preferences, 27
enlightenment (learning), 8; in convention season, 107, 131–32, 188n21; about economy, 115, 119; fundamentals and, 6, 8, 41, 148, 168; individual-level relative utility and, 158; of least interested voters, 11; moving equilibrium and, 14; in primary season, 107; priming effect on, 9
equilibrium: bounces or bumps and, 45; campaign effects and, 13–14; of stationary time series, 47–48, 49, 52, 58
equilibrium, moving, 13–14, 50–51, 165–66; convention period and, 103; fundamentals as driver of, 45–46, 167; regression analysis of, 69–72, 87, 98, 100–102; year-specific, 69, 82
Erikson, Robert S. See Macro Polity
error. See sampling error; survey error
error correction model, 50
error variance, 27–28, 60. See also sampling error
events. See shocks
external fundamentals, 6, 7; increasing variance of, 159. See also economic model; economy; fundamentals; political factors; political model; presidential approval
final campaign weeks: graphs of polls by year, 26; stability of vote division in, 66, 67–69, 166; two-stage least squares analysis, 69–72; variance of polls during, 26, 28–30, 31, 32. See also late campaign effects
final week polls: convention effects and, 76–77; election outcome and, 83–88, 90; incumbent party vote and, 83–85; shrinking lead and, 4, 69, 84, 87–88, 91; uncertainties of prediction and, 167
Finkel, Steven F., 148–49
first-time voters, 142
floating voters, 5. See also independent voters
forecasting. See predicting Election Day vote
fractionally integrated time series, 184n45
framing, 181n6
fundamentals, 167–68; campaign effects and, 2, 5–7, 13, 15, 41–42, 57, 177–78; defined, 5–6, 181n3; enlightenment and, 6, 8, 41, 148, 168; equilibrium representing, 45–46, 167; external vs. internal, 6–7; market predictions and, 177; poll-based predictions and, 177; post-convention slopes and, 103; predicting specific elections and, 169, 170–71, 193n1; in stock market, 181n4; in time-series model, 49, 50, 51; typical outcome and, 3, 11. See also economy; external fundamentals; internal fundamentals
Gallup polls, 17; demographics in, 152, 154; individual voters from, 139, 149; party identification in, 126, 150, 191n24; of presidential approval, 119; variance compression and, 157–58
Gallup Report, 19
gambling. See election markets
GDP growth, 110
gender gap, 154
gender of voter, 154, 155, 156; priming based on, 9
group characteristics, 6. See also demographic variables
Hibbs, Douglas A., 111
Hispanic voters, 153, 154, 155
house effects on poll accuracy, 30, 183n10
ideological proximity of voters to parties, 2, 109, 126–27, 129, 134, 167, 191n23. See also platform ideology; policy mood
Income Growth, 111; political variables and, 134; predicting specific elections and, 170; presidential approval and, 121–22, 123–24; subjective measure correlated with, 117; vote and, 113–15
income growth, 110–11
income of voter, 152, 153, 154, 155, 156
incumbent party vote: over convention season, 75–77, 105–7; final week’s polls and, 83–85
incumbent president. See presidential approval
independent voters, 5, 55, 56–57, 149. See also least politically involved voters
Index of Consumer Sentiment, 111, 116
index of leading economic indicators, 189n8
individual voter: factors influencing, 5; hypothetical time series of, 42–44; intensification effect and, 53–55; partisan effect and, 56; survey data sources, 139. See also internal fundamentals; relative utility (relative liking); vote shifts by individual voters
information: intensification effect and, 53; vote stability and, 10, 146–48, 192n7. See also enlightenment (learning)
in-person polls, 18
instrumental variable, 60, 69, 74, 87, 101
integrated time series. See random walk (integrated series)
intensification effect, 53–55
interactive voice response (IVR) polls, 18
intercepts: of daily equations, 187n11; of vote intention by lagged vote intention, 63
interests of voters, 6, 148–49. See also fundamentals
internal fundamentals, 6, 148–54, 162, 168. See also demographic variables; fundamentals; partisan predispositions; party identification
Internet as information source, 144
Internet polls, 18
interpolation: of daily economic measures, 112, 189nn5–6; of daily vote division, 35–36, 88–89, 183n15, 187nn6–7
Inter-University Consortium for Political and Social Research (ICPSR), 139. See also American National Election Studies (ANES)
Intrade commercial election markets, 176, 193n5
Iowa Electronic Market, 176, 193n5
issues. See platform ideology; policy mood
late campaign effects, 6, 13–14, 15, 46, 166, 167, 178; in autoregressive model, 48, 49; as bounces, 44; regression coefficients and, 99–100; R-squared statistic and, 93. See also final campaign weeks
late-deciding walk-ins, 154, 160, 161, 162, 163, 167
late electoral shifting, 86–88
learning. See enlightenment (learning)
least politically involved voters, 10, 11, 56. See also independent voters
least squares. See OLS (ordinary least squares); TSLS (two-stage least squares)
Lenz, Gabriel, 181n7
liberal-conservative platform ideology. See platform ideology
liberal-conservative policy mood. See policy mood
likely voters, 19, 162, 167, 182nn5–6. See also late-deciding walk-ins; nonvoters
live-interviewer polls, 18, 19
long-term predispositions, 10–11, 182n8. See also internal fundamentals; partisan predispositions
lowess (locally weighted scatterplot smoothing), 192n4
MacKuen, Michael. See Macro Polity
macropartisanship, 109, 126, 127, 128, 190n16, 191n24; economy and, 134; Election Day vote and, 132–33, 134, 136, 191n25; major shifts in, 190n19; platform ideology and, 130. See also party identification
Macro Polity, 109, 126, 130, 191n24
marital status of voter, 156
markets, election prediction, 176–77, 181n4, 193nn5–6
memory-based processing, 11
“minimal effects” view of campaigns, 7, 13
minor-party candidates, 19, 142
mobilization efforts, 12
models of campaign timeline. See time series
models of campaign variables. See economic model; political model
mood. See policy mood
multistage random samples, 18, 182n3
national conventions. See convention season
near-integrated time series, 52, 184n3
newly eligible voters, 142
newspaper organizations, polls of, 17
nomination. See convention season; primary season
nonvoters, 88, 154, 160–61, 162–63, 167
normal vote, 57
OLS (ordinary least squares): conventions and, 74, 75, 77–79, 102–6; with lag-2 values, 185n6; late polls and, 87, 186n4; regression coefficients in, 98–100, 101–2; at various lag lengths, 71; of vote-by-last-poll relationship, 84; with weekly vote intentions, 63, 67–69
online processing, 10–11
partisan defections, 150–51
partisan predispositions: apparent rise in, 181n2; Bayesian models of, 184n9; campaign effects and, 55–57, 148–49; perceptions based on, 149; priming of, 9; pushing vote margin to the center, 55–57, 58, 159–60; stability of, 10, 11; state-level polls and, 192n12
party conventions. See convention season
party identification, 5, 6, 167; campaign evolution and, 149–52, 162; demographic variables and, 154; Election Day vote choice and, 155; policy mood and, 128; in political model, 109–10, 125–27; variance compression and, 159–60. See also macropartisanship
Personal Retrospections, 112, 189n11
Pew polls, 148
platform ideology, 109, 126, 128–30, 132, 133, 134, 167, 190n21
policy mood, 109–10, 126, 127–28, 130, 132, 133, 167, 190n20, 191n24
policy preferences, 5, 6; long-term dispositions on, 182n8; persuasion about, 10
political factors, 2, 5, 7, 167. See also fundamentals
political knowledge, 146–48
political model, 109–10, 125–33; compared to economic model, 133–36; predicting Election Day vote, 110, 130, 132–33; testing over the campaign, 130–32; variables of, 109, 125–30
PoliticsNow website, 19
poll of polls, 20–21
polls: decision-making by responders to, 10–11; historical development of, 17–18; vs. market predictions, 177; “silence” of loser’s supporters in, 88; wording of, 17, 19
polls included in analysis, 2, 3; of individual vote decisions, 139–40; selection of, 18–21; summarized graphically, 21–26
predictability of elections, 7–8, 175–77
“predictable campaign” perspective, 7, 8; priming and, 9
predicting Election Day vote: with combined model, 135–36, 191n27; conclusions about, 107–8; convention season and, 97, 99, 100, 101–7, 188n19; from economic variables, 110, 121–25; from final week polls, 83–88, 90; with political model, 110, 130, 132–33; from polls over campaign timeline, 88–90; presidential approval and, 119–22; regression coefficients and, 90, 97–102; R-squared statistic and, 90–97. See also economic model; political model
predicting specific elections, 168–77; compared to election markets, 175–77; graphical presentation of results, 171–75; model for, 169–71
predicting vote intentions, 136–37; with economic model, 121–22, 134–35; with political model, 130–32, 134–35; with presidential approval, 121–22
predictive factors. See campaign effects; fundamentals
presidential approval, 7, 119–22, 167; the economy and, 109, 119–24, 133, 190n15; with incumbent president running for reelection, 190n14; as indicator of performance, 167; omitted from combined model, 134, 191n26; predicting from political model, 130–31; predicting specific elections and, 170–75. See also candidate attractiveness
presidential performance, 3, 5, 6, 167. See also presidential approval
priming, 8–9, 41, 168, 181n7; about economy, 9, 115, 168; time-series models and, 53
print media, 166
probit equations: demographics and, 152–53, 155, 156, 192n9; party identification and, 149–50, 155; variance compression and, 158–59; Zaller information index and, 147
pseudo R-squared, 152–53, 155, 156
Public Opinion, 19
Public Perspective, 19
quota sampling, 18
race, 9, 148, 152, 154, 155, 156
random sampling, 18; departures from, 183n11; multistage, 18, 182n3. See also sampling error
random walk (integrated series), 47, 48; for bounded variable, 184n4; of individual preferences, 53–54; problems with model using, 52; R-squared statistic and, 91, 93; with stationary component, 50–51, 91; vs. stationary time series, 49–50; two-stage least squares analysis and, 70. See also near-integrated time series
registered voters, 19
regression. See OLS (ordinary least squares); TSLS (two-stage least squares)
regression coefficients, 90, 97–102; for convention period, 102–6
relative utility (relative liking), 43, 52; intensification effect on, 53–55; learning and, 158; partisan effect and, 55–57
reliability, statistical, 28, 30; in daily analysis, 35–36, 183n15; with lagged weekly vote intentions, 68–69
religion of voter, 152, 153, 154, 155, 156, 192n9
rolling cross-sections, 34–38
R-squared statistic, 90–97; adjusted, definition of, 187n16; over campaign timeline, 94–97; conventions and, 76, 102; in cross-sectional analysis, 64, 67, 68, 71; debates and, 81; economy and, 114; final week polls and, 86; theory of, 90–94. See also pseudo R-squared
sampling error, 3, 4, 12, 165–66; daily analysis and, 35–36; dwarfed in the cross-section, 35, 60; estimation of, 27–28; in fall campaign, 30; within-year variance adjusted for, 30–32
shocks, 44–46; content of, 109 (see also economic model; political model); individual voter and, 43; intensification effect and, 55; modeling effects of, 46–52; regression coefficients and, 98, 102, 103–5; R-squared statistic and, 93–94, 96, 97. See also bounces; bumps; campaign effects; late campaign effects
short-term influences, 6, 10–11, 15, 182n8. See also bounces; late campaign effects
shrinking leads: over campaign timeline, 99; final week polls and, 4, 69, 84, 87–88, 91
southern white voters, 153, 154, 156
speeches, 12
“spiral of silence” effect, 88
stability: of aggregate vote intentions, 140; of individual vote choice, 139–40, 144; of partisan predispositions, 10, 11; of vote division in final weeks, 66, 67–69, 166. See also volatility of voter preferences; vote shifts by individual voters
state-level polls, 192n12
stationary time series, 47–49; problems with model using, 51–52; vs. random walk, 49–50; random walk with component of, 50–51, 91; R-squared statistic and, 93
Stimson, James A., 126, 127–28, 167, 190n20, 191n24. See also Macro Polity
survey error, 12, 27, 30. See also sampling error
Survey of Consumers, University Of Michigan, 111, 115, 118
surveys. See polls
switching votes. See vote shifts by individual voters
telephone interviewing, transition to, 18
timeline, presidential campaign: crystallization of voter preferences over, 2, 148–54, 155; for fifteen specific elections, 171–75; four key points of, 139; most important periods of, 4, 166–67; pace of change over, 166–67; summary of, 165–68. See also campaign effects
time series: campaign effects and, 51–52, 93–94; difficulty of modeling poll data with, 34, 42, 59–61; of individual voter, 42–44; models of, 46–52; R-squared and, 90–94; vote intentions over the campaign as, 165–66. See also random walk (integrated series); stationary time series
tracking polls, 17
trial-heat polls, 2–3, 4, 14, 17. See also polls
true preferences, 27
true variance, 27–30; in different periods of campaign, 30–32
TSLS (two-stage least squares), 69–72, 74, 185n11; conventions and, 77–79, 103–6; late polls and, 87, 186n4; regression coefficients in, 100–102; of vote-by-last-poll relationship, 84
undecided voters, 87–88, 148, 167, 186n5
University Of Michigan Survey of Consumers, 111, 115, 118
utility. See relative utility (relative liking)
variance: of autoregressive process, 47; cross-sectional, 32–38; induced by party identification, 149–50; of random walk (integrated series), 47, 50, 51; of stationary series, 47, 48–49; within-year, 22, 25–26, 30–32. See also R-squared statistic
variance compression, cross-sectional, 32–34, 36–38, 39, 157–63; over campaign timeline, 158–60; conclusions about, 58, 162–63; convention effect on, 37, 75; from fall campaign to final vote, 160–62; intensification effect and, 53–55; partisan activation and, 55–57, 159–60; random walk model and, 50, 51, 53; R-squared statistic and, 93; two-stage least squares analysis and, 71
variance compression, within-year, 26
Vavreck, Lynn, 9
volatility of voter preferences: in convention season, 31, 32, 67, 72–75, 79, 166; in debate periods, 81; over the election year, 30–32, 38–39, 183n13; across election years, 25, 29; in fall polls, 29; overestimated by election markets, 176–77. See also stability; vote shifts by individual voters
vote. See predicting Election Day vote
vote intentions, 1, 2; aggregate stability of, 140; slow evolution of, 177; as a time series, 165–66. See also dynamics of vote intentions; polls; predicting vote intentions
voter. See individual voter
vote shifts by individual voters, 140–48; across elections, 140–42, 146; information and, 146–48; from pre-election intention to Election Day choice, 143–44, 147, 154, 192n5; timeline of, 145–46. See also internal fundamentals; volatility of voter preferences