How important are party cues for public opinion? While the literature on party cues has grown rapidly following the ‘experimental revolution,' it offers no general answer: Party cue effects fluctuate widely across experiments, and they are highly sensitive to contextual features of the treatment material. For instance, effects sizes depend on whether the treatment material (typically news stories) includes policy information and interest group cues, on how salient or difficult the political issue is, and on the level of elite polarization on the issue. This contingency creates a problem of external validity common to experimental research: Not all versions of treatments are equally informative about the real-world effects of a phenomenon, but researchers seldom know which to trust more, making it impossible to draw general conclusions.

To address this problem of external validity, we present a novel experimental design testing the power of party cues embedded in real-world news. First, we randomly sample 80 news articles covering policy proposals from one of the world’s leading news agencies, the Associated Press. Second, we carefully edit the articles, creating two versions of each: One in which party cues are present and one in which party cues are masked. Third, we use the two versions as treatment material in a large-scale, factorial survey experiment with 8,000 nationally representative US participants. This experimental design allows us to estimate a party cue effect that averages across the different features of the news articles, but also to examine which features that enhance and reduce the effect.