In recent years, several states including Missouri, Wyoming, California, and Texas experienced fracture failures of traffic signal mast arms. Almost all the failures are associated with the propagation of defects or cracks. It is therefore imperative to evaluate existing mast arms using a simple yet accurate procedure. A statistical methodology is proposed to predict the fatigue life of signal mast arm structures on the basis of field-measured strain data. The annual occurrence of various stress levels is determined using the historical wind speed data in the vicinity of a mast arm structure and the strain readings of the structure under specific wind gusts. For each stress level, the crack initiation and propagation lives are estimated with the strain-life approach and the Paris crack-growth-rate model. They are combined to account for variable stresses by means of Miner's rule and the root-mean-square model, respectively. The stress concentration factor around the arm-post connection is determined using a finite element model. The parameters in the life prediction models are determined with ASTM flat tension and compact tension tests. The proposed methodology was applied to a 12.8-m (42-ft) long octagonal mast arm and a 16.5-m (54-ft) long circular mast arm in Missouri. It is concluded that signal structures in perfect condition will not crack under natural wind gusts during their service life. However, the 16.5-m-long arm is likely to be vulnerable to tiny defects around the weld connection, but the 12.8-m-long arm is safe unless a visible crack exists.