Analyses of ORG variables should use EARNWT. Because earner study participants are a smaller group than those in the BMS, using WTFINL would underestimate weighted population counts. If you are using ORG and BMS variables together (e.g., tabulating mean hours worked (BMS variable UHRSWORKT) by earnings (ORG variable EARNWEEK2)), applying EARNWT restricts your results to earner study participants and appropriately adjusts your estimates to be representative of this target population.