Work zones and lane closures on urban arterials can cause significant disruptions to the traveling public, and methods are increasingly needed to estimate the reductions to saturation flow rates that result from work zones at signalized intersections. A set of statistical models that estimate saturation headways as a function of the presence and configuration of the work zone on signalized arterial streets is presented. More than 10,000 individual vehicular headway observations were collected from video observations in and after work zones at six study sites in North Carolina. Conventional multiple-regression and path-based-regression models (structural equation model) were used to develop the saturation headway models. Three models are provided at different aggregation levels of the collected data with identical work zone configurations. The models developed at cycle-length, 15-min, and full aggregation produced adjusted R-squared values of .3259, .7209, and .895, respectively. The proposed model incorporates the effects of lane configuration, pavement condition, turning percentage from shared lanes, work intensity, and number of closed exclusive turning lanes. Based on path analysis, the structural equation model satisfies all the rule-of-thumb criteria for goodness-of-fit indices. The model uses Highway Capacity Manual default values for turning-vehicle headway effect as its intercept coefficient value.