The MAC layer of the IEEE 1901 standard for power line communications employs a CSMA/CA method similar to, but more complex than, this of IEEE 802.11 for wireless communications. The differences between these protocols raise questions such as which one performs better and under what conditions. We study the fairness of the 1901 MAC protocol in single contention domain networks, where all stations hear each other. We examine fairness at the packet level: a MAC layer protocol is fair if all stations share equitably the medium during a fixed time interval. We focus on short-term fairness, that is, over short time intervals. Short-term fairness directly impacts end-user experience, because unfair protocols are susceptible to introduce substantial packet delays. We evaluate short-term fairness with two metrics: Jain's fairness index and the number of inter-transmissions. We present test-bed results of both protocols and compare them with simulations. Both simulation and test-bed results indicate that 802.11 is fairer in the short-term when the number of stations N is between 2 and 5. However, simulation results reveal that 1901 is fairer in the short-term for N >= 15. Importantly, our test-bed measurements indicate that 1901 unfairness can cause significant additional delay when N = 2. Finally, we confirm these results by showing analytically that 1901 is short-term unfair for N = 2.