The data basis for the effective width expression used in cold-formed steel design is analyzed statistically. Several plausible effective width formulas are studied. It is concluded that the effective width is not significantly different for most test data on plates, beams, and columns in steel (including cold-formed and stainless steel), with the exception of specimens welded without stress relief. There is no significant dependence of the effective width on the flat width of partially effective flanges. A new reliability-based effective width formula is then proposed for design.