Subnational units increasingly require detailed public opinion estimates, yet traditional methods like Multilevel Regression and Poststratification (MrP) face data limitations. This article introduces a new approach—MrsP—which relaxes these strict requirements by utilizing only marginal distributions instead of full censuses or detailed surveys.

Data & Methods: By leveraging marginal distribution information rather than complete datasets, MrsP significantly broadens applicability across countries without access to census-level details while improving prediction accuracy. The method involves using Monte Carlo simulations and empirical analyses (United States; Switzerland).

Key Findings: Results show that with the same predictors, MrsP performs comparably or better in standard applications than MrP. Notably, when additional relevant predictors are incorporated, MrsP surpasses traditional methods.

Why It Matters: This advancement promises more robust and accessible tools for scholars studying descriptive representation at subnational scales globally.