Fig 1.
General ideological self-placement (at survey start) on the political ideology spectrum (x axis) plotted against three other measures to assess consistency and relative bias. a) Comparison to average of self-placement on 13 prominent political issues. b) Comparison to average policy-stance agreement ideology. We see a systematic centralizing and liberalizing effect of the agreement-based metric (best-fit slope of 0.52, intercept −8.0), along with a weaker, but still substantial fit (R2 = 0.6133) based on respondents’ general ideology. c) For “null” comparison, general ideology is seen to be very consistent with itself across the length of the survey, despite potential re-contextualization of the political environment from the intervening stimuli—the average absolute deviation between measures of general self-placement was 5.25 on the 100-point scale, serving as an upper bound on individuals’ inherent response deviation.
Fig 2.
Full response data for ideological self-placement on each “major issue,” plotted against each respondent’s general ideology. Black trend curves are Gaussian-weighted moving average (). Issues are ordered by decreasing polarization (far-left/far-right mean difference). Dots colored by respondent party affiliation, with additional hue variation to aid distinguishing despite dot transparency.
Fig 3.
Ideology histograms: Comparison of overall self-placement ideology (left panels) and policy-agreement average ideology (right panels), to illustrate the distribution of ideology for each party-identity group (top panels) and the overall ideology distribution of the population (bottom panels) by these two measures.
Top panels are visually normalized with differing y axes to show the shape and position of each party-identity group, rather than their magnitude, while the bottom panels aggregate the bars to show the overall envelope of the sampled population while preserving party breakdown at each ideology level. a) General self-placement results in a nearly symmetric set of distributions, distinctly ordered by party identity, evenly filling the ideology space. b) Agreement ideology skews distributions central and liberal, though the distributions are still well-ordered by party. c) General ideology shows a more spread-out distribution which fills the whole ideological domain. d) Agreement ideology shows a more condensed distribution, slightly left of center.
Fig 4.
Policy-stance agreement: Agreement ideology responses by policy area, ordered from most polarized to least polarized (difference between far-left/far-right means) and plotted along with their Gaussian-weighted moving average—compare to Fig 2.
Fig 5.
(a) Average ideological estimation and (b) acceptance of statements as a function of general ideology, divided by statement pool (left/center/right panels). Gaussian-weighted moving average trendlines () are shown for unmarked (solid) and marked (dot-dashed) treatment conditions. (a) We see overall “flat” trends for the three pools’ ideological positions, indicating a general universality of the abstract ideological scale (there remains differences in interpretation between individuals for any given statement—see Fig A5 in the SI—but there does not appear to be systematic observer-ideological bias in those differences). (b) We see clear “linear” trends in agreement for the liberal and conservative statement pools, and a centrally-peaked trend for the centrist statement pool. Comparing trend lines: We see little, if any, impact of speaker party-identity information in most cases. A slight liberalizing effect on liberal-statement estimation was not replicated in the other sample (see Figs. B5 and B10 in the SI).
Fig 6.
Mean responses to four prompts about the two main political parties (party in question indicated by column shading), broken down by respondent partisanship (dot color).
Proceeding from the leftmost column pair, respondents provided (1) their estimation of the ideological position of an average voter for each party, (2) their positive/negative sentiment toward such a person, (3) their agreement with each political party’s policy platform, and finally (4) agreement with each party’s actions. Vertical scale’s meaning is relative to each question (e.g., + 32 means “very conservative” ideologically, “very positive” sentiment, and “strongly agree” with party platform or actions). In (1), along with the estimations of average partisan ideology, the actual ideological means of respondents each party in our sample (Strong and Lean combined) are provided in black: Democratic respondents skewed more liberal than all estimates, and Strong Democrat respondents significantly overestimated the conservatism of Republicans.
Fig 7.
Overall percentage rates of using the endpoints of the slider for external assessments, either for ideological position, agreement, or sentiment.
On the log-log plot, we see evidence of a power-law relationship, with an exponent of about −1.7.
Fig 8.
Extreme-value usage rates by ideology, with moving median and 25/75 percentiles (window width = 7) in black.
High-usage outliers exist throughout the ideological spectrum, but median rates for non-extreme ideologues are 3% or lower (moving-average rates, not shown, hover around 6%). Extreme ideologues, on the other hand, demonstrate a distinctly higher propensity for these extreme judgments.