Replication code for "Attribute Associations of Municipal Green Bond Yield Spreads: A Demand Perspective." Includes R scripts for constructing the green bond yield spread dataset by merging bond characteristics with longitudinal yield observations, performing descriptive and distributional trend analysis of municipal green bond spreads, and applying machine learning-based attribute screening via the Apriori Association Rule Learning (ARL) algorithm to identify bond attribute combinations associated with yield spread levels. Also includes ARL applied specifically to extreme spread observations, temporal stability analysis of discovered association rules, and ANOVA-based significance testing of attribute–spread associations across maturity subgroups. The demand-side analysis constructs ARL-screened bond groups based on second-, third-, and fourth-order association rules for both positive and negative spread classifications, matches each bond at issuance to contemporaneous Japanese and U.S. Treasury auction bid-to-cover ratios as proxies for global investor demand, and incorporates a set of macroeconomic control variables. Regression models are then estimated separately for each ARL-screened group to examine whether primary market demand conditions at issuance are systematically associated with the direction and magnitude of green bond yield spreads