Read on Twitter
There's a lot of discussion about how "The Latino Vote" falsely assumes a homogenous group. @betsylevyp & I show how this same faulty assumption is present in how psychologists try to understand/compare "group" beliefs through aggregate analyses https://psyarxiv.com/kfpjg
We analyzed racism judgments of anti-immigrant tweets in four samples of Latinx, white, and black participants using a variance component analysis whose main focus is to quantify sources of variance. We compared those results to traditional differences-in-means analyses.
The aggregate analyses show that Latinxs rated the anti-immigrant tweets on average as more racist than black and white participants. However, the variance analyses instead show substantial heterogeneity within all three racial categories.
A consistent source of disagreement came from differences in political ideologies which shaped the avg rating that different participants gave overall to the tweets, but another major source was stable idiosyncratic ways that specific participants rated specific tweets vs others
That reliable idiosyncratic form of disagreement can only be measured by collecting repeated judgments on all the stimuli - a task structure that is not very common in psychology. Without repeated judgments, this form of heterogeneity gets lost into the residuals as error/noise.
We examined many variables theorized to be important for racism perception to try to explain these disagreements. Some were significant in regressions, but ultimately, there was more unexplained than explained non-error variance.
Mapping variance > significance testing
Mapping variance > significance testing
I'll end with this conclusion from the paper. How we use race as an independent variable in aggregate analyses of beliefs is a form of racecraft, as @ruha9 pointed out in her review of the Fields' book. We transform avg diffs into races with diff beliefs
https://static1.squarespace.com/static/5bbd85f3809d8e6a1a3c5c9e/t/5bdc1ddc4d7a9c45193522d0/1541152220397/Racecraft_Review1.pdf
https://static1.squarespace.com/static/5bbd85f3809d8e6a1a3c5c9e/t/5bdc1ddc4d7a9c45193522d0/1541152220397/Racecraft_Review1.pdf
We need to pay careful attention to how our analyses and subsequent inferences flatten evaluative heterogeneity, treat categories as if groups by assuming psychological boundedness, and reify races.
