Loop detectors have been widely used to estimate travel time delay. One implicit assumption about the estimation is that traffic flow is stationary. Based on this assumption, the point data collected by loop detectors at different locations could be interpolated or extrapolated to estimate travel time delay for segments between adjacent loop detectors. This assumption is not quite valid when vehicle speeds fluctuate spatially and temporally. It is desirable to examine and understand the accuracy level of such travel time delay estimation methodology in real-world situations. The purpose of this article is to present an empirical analysis to explicitly illustrate the accuracy level of travel time delay estimated using loop detectors. Travel time delay obtained from eight freeway segments under various combinations of reference speeds, data aggregation intervals, and loop detector spacings are studied and compared with true delay determined using vehicle trajectory information. For all of the eight freeway segments evaluated, it was found that the mean absolute percentage errors of estimated delay are consistently less than 7%, when delay is defined using a reference speed higher than 60 mph (96 kmph). Higher mean absolute percentage errors were observed for lower reference speeds, longer detector spacings and turbulent traffic situations, such as traffic incidents. Adopting higher reference speed, say 60 mph (96 kmph), longer data aggregation interval, say 300-second, and shorter loop detector spacing, say 1,000 ft (305 m), helps produce lower mean absolute percentage error in loop detector-estimated travel time delay.