CoChat
Sign Up
PreprintCited by 242Misinformation and Its ImpactsElectoral Systems and Political ParticipationMedia Influence and Politics

Partisan Gaps in Political Information and Information-Seeking Behavior: Motivated Reasoning or Cheerleading?

Open access
Peterson, Erik, Iyengar, Shanto·2019·American Journal of Political Science
DOI 10.31235/osf.io/pmn7h
Open this paper in CoChat — free accountOpen PDF

Abstract

Do partisan disagreements over politically-relevant facts, and preferences for the information sources from which to obtain them, represent genuine differences of opinion or insincere cheerleading? The answer to this question is crucial for understanding the scope of partisan polarization. We test between these alternatives with experiments that offer incentives for correct survey responses and allow respondents to search for information before answering each question. We find that partisan cheerleading inflates divides in factual information, but only modestly. Incentives have no impact on partisan divides in information search; these divides are no different from those that occur outside the survey context when we examine web browsing data from the same respondents. Overall, our findings support the motivated reasoning interpretation of misinformation; partisans seek out information with congenial slant and sincerely adopt inaccurate beliefs that cast their party in a favorable light.

AI snapshot

Written from the paper’s full text.

TL;DR

This paper tests whether partisan gaps in political facts and news-source preferences reflect sincere motivated reasoning or insincere cheerleading. Across two incentivized survey experiments plus linked web-browsing data, it finds incentives modestly reduce factual divides but do not affect information search, supporting motivated reasoning.

Key findings

  • Incentives reduced the partisan information divide but did not eliminate it: the baseline divide was .32, the Democrat × incentive interaction was -0.10, and a .22 divide remained among incentivized respondents; roughly 60–70% of the initial divide persisted.
  • Among respondents certain of their answers, the remaining incentivized divide was .27 (95% CI [0.24, 0.30]); among confidently misinformed respondents, the remaining gap was .13 (95% CI [0.11, 0.15]).
  • Incentives had no effect on news choice: the estimated effect on selecting a co-partisan source was -0.003 (95% CI [-0.03, 0.02]), with similarly null effects for expert, mainstream, and out-partisan sources.
  • Survey news choices corresponded with real-world browsing: the correlation between survey and behavioral media slant was .58, and partisan divides in media choice were similar across unincentivized survey (1.24 SD), incentivized survey (1.30 SD), and real-world media diet (1.36 SD).

Methods

  • Two YouGov surveys: September 2018 (N = 1,000) and March 2019 (N = 1,700), pooled; 954 respondents completed both, and 855 had linked desktop web-browsing data.
  • Respondents were randomly assigned to receive $0.50 per correct answer on a five-item political information battery; before each question they chose one news source from left-leaning, right-leaning, expert, and two mainstream options, then answered and reported confidence.
  • Main analysis used pooled models with item fixed effects, Democrat × incentive treatment interactions, and standard errors clustered by respondent; news-choice models regressed source selection on incentive treatment and item fixed effects; web-browsing comparison used Bakshy et al. (2015) alignment scores.

Limitations

  • The paper states it cannot distinguish motivated reasoning from differential beliefs about source credibility or evidence quality, cannot rule out that psychic rewards for cheerleading exceed the monetary incentives, and cannot make claims about the realism of its incentive manipulations.
  • The paper notes the web-tracking measure captures only desktop browsing, missing mobile web and cable television, and that the unweighted sample is older and more highly educated than a nationally representative sample, though weighted robustness checks are similar.
  • The text also makes evident that the information items were selected for intense partisan disagreement, that respondents had no “Don’t know” option, and that demand effects or post-treatment concerns around response certainty are possible, though the authors report no detectable incentive effect on certainty.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.31235/osf.io/pmn7h

Cites

74 of 74 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

8 of 30 cited sources checked · 5 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Bullock et al. (2015, 521) find that only 40% to 45% of the initial partisan gaps remain when incentives are made available.

    “In the closest comparison, Bullock et al. (2015, 521) find that only 40% to 45% of the initial partisan gaps remain when incentives are made available.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Bullock et al. (2015, 521) source
  • Interpretationcitation

    The treatment effects of incentives across these studies are of similar magnitude to those observed in Bullock et al. (2015).

    “While the treatment e↵ects of incentives across these studies are of similar magnitude to those observed in Bullock et al. (2015), the initial partisan gap on these items is much larger than previously obtained in incentive-based designs.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Bullock et al. (2015) source
  • Interpretationown data

    The persistence of partisan divides in factual beliefs among incentivized groups is noticeably higher than in previous work.

    “The persistence of partisan divides in factual beliefs among the groups provided with incentives is noticeably higher than in previous work.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Bullock et al. (2015, 521) source

      Experiment 1 reports that about 45% of the partisan gap remained when participants were incentivized to answer correctly. The Theory and Prior Evidence section says earlier incentive research did not examine partisan patterns in responses, so the paper does not establish that this persistence was higher than in previous work. Page numbers are unavailable in our copy. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Interpretationown data

    The initial partisan gap on these items is much larger than previously obtained in incentive-based designs.

    “While the treatment e↵ects of incentives across these studies are of similar magnitude to those observed in Bullock et al. (2015), the initial partisan gap on these items is much larger than previously obtained in incentive-based designs.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Bullock et al. (2015) source
  • Premisecitation

    Bullock et al. (2015) found that the partisan difference in information fell by half among respondents informed that their chances of winning a 200 dollar gift certificate would increase with correct answers.

    “The partisan di↵erence in information, which was substantial among respondents o↵ered no incentive, fell by half among respondents informed that their chances of winning a 200 dollar gift certificate would increase with correct answers.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites Bullock et al., 2015 source
  • Classified as probably not a substantial claim. 29
  • May combine two claims. 27

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    In the motivated reasoning account, polarization has strengthened partisans' willingness to ignore or counter-argue with credible facts that cast aspersions on their party or party leaders and to uncritically accept false information that reflects favorably on their side.

    “In this motivated reasoning account, polarization has strengthened partisans' willingness to ignore or counter-argue with credible facts that cast aspersions on their party or party leaders and, conversely, to uncritically accept false information that reflects favorably on their side (Gaines et al., 2007;Shapiro and Block-Elkon, 2008;Kraft et al., 2015).”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites Gaines et al., 2007 source
    • cites Shapiro and Block-Elkon, 2008 source

      Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites Kraft et al., 2015 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The amount of $0.50 per question falls between incentive levels used in previous studies.

    “In a point we later revisit when discussing the implications of our findings, this amount falls in between the incentive levels used in previous studies.”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites Prior et al. (2015) source
    • cites Bullock et al. (2015) source

  • Probably not a substantial claim: may not pass S1 (asserted). May combine two claims.

    Premisecitation

    In the cheerleading account, partisans are in fact well informed, but prefer to act misinformed in surveys.

    “In this view, partisans are in fact well informed, but prefer to act misinformed in surveys.”
    Quote in the paperNot a substantial claim1 of 3 sources checked
    • cites Prior et al., 2015 source
    • cites Bullock et al., 2015 source

      The Abstract says partisan survey answers may reflect expressive cheerleading rather than sincere beliefs. But the Conclusion says cheerleading can be either informed or uninformed, so the claim overstates the account by saying partisans are well informed. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites Scha↵ner and Luks, 2018

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Findingown data

    The point estimate for the effect of incentives on the probability of selecting a co-partisan news source is -0.003 (95% CI [-0.03, 0.02]), negative but substantively small and indistinguishable from zero.

    “For instance, in Column 1 the point estimate for the e↵ect of incentives on the probability of selecting a co-partisan news source is negative, but substantively small and indistinguishable from zero (-0.003, 95% Confidence Interval [-0.03, 0.02]).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The partisan gap in information closes due to incentives but is not eliminated; large partisan divisions remain in the incentivized condition.

    “While the partisan gap in information closes due to incentives, it is by no means eliminated.Large partisan divisions remain in the incentivized condition.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Among respondents certain of their answers, the availability of incentives reduces partisan gaps on information items, but large divides remain even in the presence of incentives.

    “While the availability of incentives reduces partisan gaps on the information items, large divides remain even in the presence of incentives.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The pattern of no sizeable or statistically significant effects of incentives on information source selection continues for expert sources, mainstream news sources, and out-partisan news sources.

    “This pattern of no sizeable or statistically significant e↵ects of incentives on information source selection continues when we consider the probability of selecting an expert source (Column 2), one of the two mainstream news sources that was available (Column 3) or an out-partisan news source (Column 4).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The effect of incentives on partisan divides remains similar and statistically significant across strong partisans, information items where Democrats were motivated to answer incorrectly, and items where the incorrect answer was party-congenial for Republicans.

    “There is some minor variation in the initial magnitude of these partisan divides across these di↵erent specifications, but the e↵ect of incentives remains similar and statistically significant e↵ect across all three.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The interaction between partisanship and the incentive treatment is negative and statistically significant (-0.10, 95% CI [-.12, -.08]).

    “the interaction between partisanship and the incentive treatment is negative and statistically significant (-0.10, 95% Confidence Interval [-.12, -.08]).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Across the news selections made in the control condition of the surveys, 29% of selections went to a co-partisan source, 26% went to the expert source, 38% to one of the two mainstream news sources, and only 7% to an out-party source.

    “Across the news selections made in the control condition of the surveys (i.e., those where respondents were not assigned incentives for correct answers to the information questions), 29% of selections went to a co-partisan source, 26% went to the expert source, 38% to one of the two mainstream news sources and only 7% to an out-party source.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    When isolating the information items about unemployment, the pattern of findings is similar to previous work that considers this topic.

    “We also note that, when isolating the information items about unemployment in Appendix Table B.1, we observe a similar pattern of findings to previous work that considers this topic.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The remaining gap of .27 (95% CI [0.24, 0.30]) is both substantial and statistically significant.

    “This remaining gap of .27(95% Confidence Interval [0.24, 0.30]) is both substantial and statistically significant.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In the certainty analysis, the initial magnitude of partisan divides is smaller across specifications, but the pattern is largely consistent with the earlier analysis.

    “While the initial magnitude of these partisan divides is smaller across the various specifications in this more stringent coding, the pattern is largely consistent with our earlier analysis.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    In general, respondents expressed a relatively high degree of certainty in their answers to the information questions.

    “In general, respondents expressed a relatively high degree of certainty in their answers to the information questions.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    In the more stringent coding isolating confident and misinformed responses, the initial magnitude of partisan divides is smaller across specifications, but the pattern is largely consistent with the earlier analysis.

    “While the initial magnitude of these partisan divides is smaller across the various specifications in this more stringent coding, the pattern is largely consistent with our earlier analysis.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Using the 10-pt outcome incorporating certainty, the pattern is similar to the main text findings.

    “Using this alternative outcome, we observe a similar pattern to the findings reported in the main text.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    These divides prove durable to the presence of incentives and when examined outside the survey environment.

    “These divides prove durable to the presence of incentives and when examined outside the survey environment.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    This result does not hold in any of the other pooled or subset analyses considered, or if the effects are estimated in Study 1 without including sampling weights.

    “We note this result does not hold in any of the other pooled or subset analyses we consider (or if we estimate the e↵ects of these features in Study 1 without including sampling weights)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Table B7 replicates the analysis for Republicans, again finding no substantively significant effect of incentives on news choice.

    “Table B7 replicates this analysis for Republicans, again finding no substantively significant e↵ect of incentives on news choice.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    When isolating the unemployment items, the findings resemble earlier studies.

    “We note that when we isolate the unemployment items, our findings resemble earlier studies.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The availability of incentives reduced, but did not eliminate, the partisan divide in information.

    “The availability of incentives reduced, but did not eliminate, the partisan divide in information.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The initial partisan divide over these items is slightly smaller in the second survey relative to the first.

    “The initial partisan divide over these items is slightly smaller in the second survey relative to the first.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    We observe web tracking data for 46% of the pooled sample across the two surveys.

    “We observe web tracking data for 46% of our pooled sample across the two surveys.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    This analysis shows a largely similar pattern of findings across the two surveys as in the pooled analysis in the main text.

    “This analysis shows a largely similar pattern of findings across the two surveys as we find in our pooled analysis in the main text.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The results remain similar when reanalyzing the experiments in each survey individually using the sampling weights provided by YouGov.

    “although we note our results remain similar when we reanalyze the experiments in each survey individually using the sampling weights provided by YouGov (See Appendix B)”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The total number of unique respondents is less than the sum of respondents across the two survey waves because 954 respondents were re-interviewed and appeared in both.

    “Because there was a panel structure to the survey data, in which 954 respondents were re-interviewed and appeared in both, the total number of unique respondents is less than the sum of respondents across the two survey waves.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The pattern of reduction but persistence of the partisan divide was observed for both Democrats and Republicans, depending on the partisan valence of the topic.

    “We observed this pattern for both Democrats and Republicans, depending on the partisan valence of the topic.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    When the outcome variable is correct answers, the point estimate for the effect of incentives among independents falls in between those for Democrats and Republicans, although it does not reach statistical significance.

    “When the outcome variable is operationalized in this way, the point estimate for this group falls in between those for Democrats and Republicans, although it does not reach statistical significance.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    In Survey 1 only with sampling weights, the study observes a decline in co-partisan news choice selection and an increase in the probability of selecting neutral news sources.

    “It is important to note that in this subset of the data and approach to analysis we observe a decline in co-partisan news choice selection and an increase in the probability of selecting neutral news sources.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Individuals accept misinformation and go on to incorporate it into their future political decisions.

    “Individuals accept misinformation and go on to incorporate it into their future political decisions (see e.g., Thorson, 2016;Swire et al., 2017;Flynn and Krupnikov, 2019).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites Thorson, 2016

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Swire et al., 2017

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites Flynn and Krupnikov, 2019

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    The surveys use YouGov's standard procedure of drawing a sample matched to voting-age population on key demographic characteristics.

    “These surveys use the firm's standard procedure of drawing a sample from its panel that is matched to the voting-age population on key demographic characteristics (e.g., Vavreck and Iyengar, 2011).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Vavreck and Iyengar, 2011 source

  • Classified as probably not a substantial claim.

    Study design and conductown data

    Respondents were randomly assigned to either receive or not receive incentives for correctly answering political knowledge questions.

    “First, following earlier studies, we randomly assign survey respondents to either receive or not receive incentives for correctly answering political knowledge questions.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    From the motivated reasoning perspective, the divides in political information that result from motivated reasoning are real and troubling.

    “Importantly for what follows, from this perspective the divides in political information that result from motivated reasoning are real and troubling.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    Overall findings support the motivated reasoning interpretation of misinformation; partisans seek out information with congenial slant and sincerely adopt inaccurate beliefs that cast their party in a favorable light.

    “Overall, our findings support the motivated reasoning interpretation of misinformation; partisans seek out information with congenial slant and sincerely adopt inaccurate beliefs that cast their party in a favorable light.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Our design does not let us distinguish motivated reasoning from other processes such as differential beliefs regarding source credibility and evidence quality.

    “Our design does not let us distinguish motivated reasoning from these other processes.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The results are no different when the two surveys are considered separately.

    “Because of their similar formats, we pool the surveys together for analysis, although in Appendix B we note the results are no di↵erent when we consider the surveys separately.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    There is little evidence that the sample's age and education deviations contribute to the findings; robustness checks incorporating survey weights yield similar results.

    “At the same time, there is little evidence this contributes to our findings, particularly as we see similar results in robustness checks that incorporate survey weights to correct for these deviations between our realized sample and nationally-representative targets (See Appendix B).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    There is agreement that on many political information questions Democrats and Republicans gravitate to survey responses that cast their party in a favorable light.

    “In summary, there is agreement that on many political information questions Democrats and Republicans gravitate to survey responses that cast their party in a favorable light.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    The unweighted sample used in the primary analysis is older and more highly educated than a nationally-representative sample would be.

    “We note here that the unweighted sample used in our primary analysis is older and more highly educated than a nationally-representative sample would be (see Table A1 for demographic information).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The study's evidence cannot speak to the correspondence between survey-based and behavioral measures of media exposure in settings other than desktop web browsing.

    “We note our evidence cannot speak to the correspondence between survey-based and behavioral measures of media exposure in these other settings, which represent a setting for future work.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    This pattern is expected, as independents lack the directional pressures to engage in motivated reasoning or cheerlead for their party that face partisan respondents across the set of items focused on here.

    “This pattern is expected, as independents lack the directional pressures, either to engage in motivated reasoning or cheerlead for their party, that face partisan respondents across the set of items focused on here.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The study cannot address the possibility that for some partisans the psychic rewards for cheerleading exceed the monetary rewards available in the studies, which would mean the incentive conditions still contain some degree of insincere cheerleading.

    “One challenge to the validity of this motivated reasoning interpretation is that we cannot address the possibility that, for some partisans, the “psychic” rewards for cheerleading exceed the monetary rewards available in our studies.This would mean the incentive conditions in our design still contain some degree of insincere cheerleading.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The articles in the survey were edited to ensure they were of similar length (roughly 250-300 words) for each information source.

    “The articles that respondents encountered in the survey were edited to ensure they were of similar length (roughly 250-300 words) for each information source.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    The available sample for the analysis in C.3 is 840, slightly smaller than the entire web traffic sample because 15 respondents did not visit any of the political websites on the list.

    “The available sample here (840) is slightly smaller than the entire web tra c sample because 15 respondents in the web tra c sample did not visit any of the political websites on our list, meaning we are unable to categorize the partisan slant of their real-world political media exposure.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationown data

    This appears to be, in part, because including the sampling weights increases the amount of co-partisan source selection in the control group.

    “This appears to be, in part, because including the sampling weights increases the amount of co-partisan source selection in the control group.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationown data

    The study has interpreted the magnitude of the partisan information divide and the stability of partisans' preferences for congenial information as evidence of motivated reasoning stemming from polarization.

    “We have interpreted the magnitude of the partisan information divide and the stability of partisans' preferences for congenial information as evidence of motivated reasoning stemming from polarization --the interpretation commonly given in the misinformation literature.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    The articles were based on content that had originally appeared from the source it was attributed to in the survey, to ensure each article could be realistically viewed as coming from that outlet and to ensure the choice set reflected the real-world array of news coverage.

    “These articles were all based on content that had originally appeared from the source it was attributed to in the survey.This was done to ensure each article could be realistically viewed as coming from that outlet and to ensure the choice set in the study reflected the real-world array of news coverage available on that topic from media outlets with di↵ering ideological reputations.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Respondents were recruited using YouGov's standard approach of matching a non-representative online sample to a nationally representative target.

    “Respondents were recruited using YouGov's standard approach of matching a non-representative online sample to a nationally representative target.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    The study's measure only captures web browsing conducted from desktop computers and misses other forms of media use such as mobile web browsing and cable television consumption.

    “We emphasize here that our measure only captures web browsing conducted from desktop computers.This misses other forms of media use (e.g., mobile web browsing and cable television consumption).”
    Quote in the paperNot a substantial claim
49 of 49

  • The paper relies on this without support.

    The partisan leanings of the news sources (MSNBC/Huffington Post as left-leaning, Fox News/Breitbart as right-leaning, CNN/New York Times as mainstream) are as categorized.

    “In each case, respondents could choose to read information from one left-leaning source (i.e., MSNBC or Hu ngton Post), one right-leaning source (i.e., Fox News or Breitbart), one source with specialized expertise (e.g., excerpts from a Bureau of Labor Statistics press release in the case of a question about the unemployment rate), or from a pair of mainstream news sources (i.e., CNN, New York Tim”

  • The authors state this as an assumption.

    Comprehension of information provided by expert sources requires greater effort.

    “On the assumption that comprehension of information provided by expert sources requires greater e↵ort”

  • The authors state this as an assumption.

    In the main text comparison, visits to CNN and the New York Times are considered neutral from the perspective of partisans by giving them a similar code to the expert sources.

    “In the comparison between survey-based and real-world news choice presented in the main text, we consider visits to CNN and the New York Times to be neutral, from the perspective of partisans, by giving them a similar code to the expert sources.”

  • The paper relies on this without support.

    In public perceptions of perceived slant and audience partisanship, the mainstream outlets CNN and the New York Times appear similar to left-leaning news sources such as Huffington Post and MSNBC.

    “We recognize that, in public perceptions of their perceived slant and the partisanship of their audiences, the mainstream outlets we consider (i.e., CNN and the New York Times) appear similar to the left-leaning news sources (i.e., Hu ngton Post and MSNBC).”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The YouGov survey platform is a representative sampling platform that matches a randomly-drawn sampling frame of the U.S. population with opt-in respondents and has been widely validated.

    “We recruited 500 participants from the YouGov survey platform, a representative sampling platform that matches a randomly-drawn sampling frame of the U.S. population with members from their opt-in respondents based on a large set of variables38,39; the platform has been widely validated. 40”

    From Public opinion on types of voice systems for older adults (2024) · cites this paper as [39]

    The Data and Additional information sections (pages not given) describe YouGov matching an online panel sample to a nationally representative target on key demographics. They do not establish that the target frame is randomly drawn or that the platform has been widely validated; the Data section also notes that the study’s unweighted sample was not nationally representative. (LLM, conf=0.85)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Pooled sample size conflicts with wave sizes and overlap

    Minor

    Table A1 reports 1,017 and 1,754 respondents in the two waves and 1,852 unique pooled respondents. If 954 respondents appeared in both waves, those wave sizes imply 1,017 + 1,754 − 954 = 1,817 unique respondents, 35 fewer than reported.

1 of 1

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. Iyengar, Shanto, Gaurav Sood and Yphtach Lelkes. 2012. "A↵ect, Not Ideology: A Social Identity Perspective on Polarization." Public Opinion Quarterly 76(3): 405-431.

Cited by

Sorted by citations

Related

Authors
Peterson, E., & Iyengar, S.
Affiliations
  • Texas A&M University (US)
  • Stanford University (US)
Corresponding author
  • Peterson, Erik
  • Iyengar, Shanto
ORCID
Venue
American Journal of Political Science
Publisher
Center for Open Science
Published
2019-10-29
Volume
65
Issue
1
Pages
133-147
ISSN
0092-5853 (ISSN-L)
DOI
10.31235/osf.io/pmn7h
OpenAlex
W3036144303
Semantic Scholar
b411421cb65bc4fbfaf5893995d7b481907711da
Cited by
241
References
72
Citations by year
2026: 48 · 2025: 51 · 2024: 26 · 2023: 48 · 2022: 49 · 2021: 17 · 2020: 2
FWCI
38.04
Type
journal article
Language
en
Open access
Licence
Locations
  • American Journal of Political Science (publishedVersion)
  • SocArXiv (OSF Preprints) (submittedVersion, cc-by)
  • SocArXiv (OSF Preprints) (submittedVersion)
Funders
William and Flora Hewlett Foundation
Topics
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Social Sciences › Political Science and International Relations › Electoral Systems and Political Participation
  • Social Sciences › Sociology and Political Science › Media Influence and Politics
Keywords
motivated reasoning, partisan polarization, information seeking, misinformation, political information sources, survey incentives, web browsing data

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-23.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-23.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review