<result> <column_name>Education: Bachelor+</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.85</confidence> <requires_user_confirmation>false</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, min, max, median, distribution</valid_reducers> <excluded_reducers>sum (summing percentages across entities is meaningless), product (not a ratio measure)</excluded_reducers> <average_semantics>arithmetic_mean</average_semantics> </reducers> <inner_outer_measure> <inner_measure_reducer>min</inner_measure_reducer> <outer_measure_reducer>max</outer_measure_reducer> </inner_outer_measure> <multiplicative_partner_derivable>false</multiplicative_partner_derivable> <reasoning> Table-type gate: DIMENSION with no sequence axis makes multiplicative highly implausible. The column contains percentage values (1.02%, 1.12%, etc.) representing the proportion of population with Bachelor+ education at each entity. This is an intensive property—a demographic rate or level—not an extensive quantity that accumulates. Summing percentages across different geographic entities is meaningless; the natural rollup is averaging. The dtype is object (likely formatted strings with %), preventing numeric statistics, but the sample values clearly show small percentages representing demographic composition levels, consistent with intensive_additive. </reasoning> <table_type_conflict>false</table_type_conflict> </result>