61 1 ethnics’ (Miller and Chaturvedi, 2018) and ‘Blacks’ (Whitby, 2007) alike. Shared racial/ ethnic identification matters in both experimental and non-experimental settings and drives voters’ choices in numerous ways. Studying shared identification in candidate experiments is a relatively new development in the US-literature. Whereas researchers used to study voter evaluations of candidates from the perspective of the general population, researchers increasingly study racial/ethnic subgroup assessments of descriptive candidates. In the past there might have been a streetlight effect: researchers are more prone to shine their light on questions that are relatively easy to answer with existing data, while questions that require greater data-gathering effort remain in the dark despite scientific and societal relevance. More recently, researchers have been going the extra mile to oversample racial/ethnic subgroups or to increase sample sizes and sampling methods are changing rapidly (Cassese et al., 2013; Coppock and Green, 2015; Coppock and McClellan, 2018; Peyton et al., 2022). Earlier experiments in the social sciences were almost all conducted amongst students. These samples comprised less racial/ ethnic diversity than the general population. Currently, researchers use online survey agencies such as Lucid, MTurk, TESS and YouGov. Despite some pitfalls in sampling methods, the advantage is that they are more diverse than student samples (Cheung et al., 2017: 349). This diversity is vital to avoid external validity problems. Due to high internal validity in experimental designs, it can be tempting to prefer this over external validity. But even when causal claims can be made, it does not mean that the particular causal path holds across groups. Until recently, many social science experiments were done on Western, Educated, Industrialized, Rich, and Democratic (i.e. WEIRD) samples, which distorts outcomes.
RkJQdWJsaXNoZXIy MjY0ODMw