The purpose of this paper is to provide a robust process to statistically analyze reflective cracking field performance data. There is often a lack of consistency and transparency in statistical analysis of pavement field performance data, which may not satisfy ANOVA or regression modeling assumptions. Twelve full-scale asphalt concrete (AC) overlay pavement test sections located at the MnROAD test facility are used to demonstrate the statistical framework. The percentage of cracking reported at joint locations (%RC) is used to represent reflective cracking performance, and its relationships to pre-overlay load transfer efficiency (LTE), truck traffic, overlay thickness, and common performance indices determined from laboratory tests are investigated. The three laboratory tests considered in this study are the disk-shaped compact tension (DCT), semi-circular bend (SCB) and overlay tester (OT). Logistic regression models are used for estimation. Predictive abilities of various models are compared in relation to the percentage odds (%odds) of reflective cracking. Model output is binary: it estimates not the relative amount of reflective cracking but instead the probability of a given pavement structure cracking. Varying ability to perform asphalt mixture laboratory performance testing is assumed. One such model, where no laboratory performance testing variables are included, shows that a one-unit increase (1-in.) in AC overlay thickness may result in approximately a third decrease in the %odds of reflective cracking. A logistic regression model developed that considers laboratory performance data from DCT, SCB, and OT results in the optimal model balancing best fit and best prediction properties without overfitting.