Polyvariate regression outperforms difference scores for affective polarization

What the study found: The study found that polyvariate regression was superior to using difference score measures for examining affective polarization, based on a comparison in which both approaches were applied to the same example.

Why the authors say this matters: The authors suggest this matters because affective polarization research has often relied on difference scores, which the study says are problematic in general and may hide useful findings. They also conclude that polyvariate regression can help researchers study congruence as an outcome without those difference-score issues.

What the researchers tested: The paper reviews conceptual, statistical, and theoretical problems with difference scores, then gives a tutorial on polyvariate regression, an analytic approach that tests how a predictor relates to affective polarization without creating a difference score. The demonstration used people’s stances on government defense spending and their feeling thermometer ratings of the Republican and Democratic parties in the United States.

What worked and what didn't: In the demonstration, polyvariate regression was used to test the association between defense-spending stance and party ratings, and the same association was also tested with a difference score outcome. The comparison showed better performance for polyvariate regression, and the abstract says difference scores may obscure potentially interesting findings.

What to keep in mind: The abstract notes that limitations of polyvariate regression are discussed, including its use in multiparty contexts and under other conditions. It also says polynomial regression is recommended when polarization is the predictor rather than the outcome, but no additional limitations are described in the available summary.

Key points

• The paper argues that difference scores are a problematic way to measure affective polarization.
• Polyvariate regression is presented as an alternative that does not require a difference score.
• An example using U.S. party feeling ratings and views on government defense spending was used to compare the two approaches.
• The comparison favored polyvariate regression over difference score measures.
• The abstract also recommends polynomial regression when polarization is the predictor rather than the outcome.