Volatile organic compound (VOC) emissions from crude oil storage tanks are commonly estimated using emission factors derived from laboratory analyses and equation of state (EOS) models, but the reproducibility and uncertainty of these methods have not been well characterized, particularly for the waxy crude oil produced in the Uinta Basin of Utah, U.S.A. We collected replicate pressurized liquid and gas samples from three Uinta Basin oil wells, distributed them among three commercial laboratories, and used a subset of the laboratory results as input for two EOS models (ProMax and PVTsim). Significant differences existed among laboratories for most measured parameters, even though each laboratory generally followed the same standard methods. Emission factors calculated from the flash liberation data of two laboratories were not significantly different from one another, but this agreement resulted from compensating differences rather than from consistent underlying measurements. EOS models added a substantial additional layer of variability; uncertainties for model-derived emission factors exceeded 100% in almost all cases. Laboratory flash liberation analysis provided lower-uncertainty emission factors than EOS model-based approaches for Uinta Basin waxy crude, though EOS models with binary interaction parameters tuned specifically for waxy crude oils may provide more reliable results.