RELAP5/MOD3.2 is a thermal-hydraulic system analysis code used to predict the response of nuclear reactor coolant systems in the event of certain accident scenarios. It is important that RELAP and other system analysis codes are able to accurately predict various two-phase flow phenomena, particularly the interfacial transfers between the liquid and gas phases. It is also important to understand how much uncertainty exists in these predictions due to uncertainties in the constitutive relations used to close the two-fluid model. In this paper, the uncertainty in the interfacial drag calculated by RELAP5/MOD3.2 due to errors in the drift-flux models used to close the model is evaluated and compared to the correlation developed by Goda et al. (2003). The case of downward flow is considered due to the importance of co-current and counter-current downward flow for predicting behavior in the downcomer of reactor systems during small-break Loss of Coolant Accidents (LOCAs) in nuclear reactor systems. The overall uncertainty in the interfacial force calculations due to error in the distribution parameter models were found to have a bias of +8.1% and error of 20.1% for the models used in RELAP5, and a bias of -30.8% and error of 23.1% for the correlation of Goda et al. (2003). However this analysis neglects the effects of compensating errors in the drift-flux parameters, as the drift velocity is assumed to be perfectly accurate. More physically meaningful results could be obtained if the distribution parameter and drift velocity were calculated directly from local phase concentration and velocity measurements, however no studies were available which included all of this information.