Damn, I started reading the actual study and their methodology was pretty exhaustive. https://www.researchgate.net/publication/317252320_The_Visibility_of_Social_Class_From_Facial_Cues I think your answer is in Studies 5 and 6, although I can't be sure. This was somewhat interesting to me:I'd be curious how they did this.
Perceptions of targets’ Positivity (a composite of positive affect, empathy, warmth, and reversed dominance ratings) and Attractiveness (a composite of attractiveness, health, and intelligence ratings) served as utilized cues in perceivers’ social class categorizations. That is, the participants in Study 1 were more likely to categorize as rich those targets that participants in Study 3 rated higher on Positivity and Attractiveness. This suggests that perceivers use class-related stereotypes (e.g., of the rich being happier and more attractive) when categorizing people as rich or poor. Only Attractiveness validly cued targets’ actual social class, however. Unsurprisingly, then, not all wealth-related stereotypes are correct.