This paper studies the performance of two traditional schedulers, Proportional Fair (PF) and Round Robin (RR), in the context of relay-enhanced LTE-A networks. These two schedulers are indeed natural candidates for implementation in relays nodes (RN) and, following the results obtained in singlehop networks, mobile operators could be tempted to adopt PF because of the good trade-off it offers between cell capacity and fairness. Based on a statistical throughput evaluation model, we show that this is not necessarily the right option. The number of RNs, their locations in the cell, and the backhaul link quality have indeed a decisive influence on the scheduler choice. In some scenarios, it is even not desirable to deploy relays as they degrade the network performance compared to the no relay case. For the purpose of performance evaluation, we develop a realistic and computationally tractable statistical network model that takes into account fast fading, multiple interferers, cell range expansion bias, backhaul link quality, and traffic load. We also propose an optimization of the radio frame structure and a sub-optimal RN placement scheme in order to fairly compare RR and PF.