The cost of producing public transit service is not uniform, but varies by trip type (such as local or express), trip length, time of travel, and direction of travel, among other factors. Yet the models employed by public transit operators to estimate costs generally do not account for this variation. The exclusion of cost variability in most public transit cost allocation models has long been noted in the literature, particularly with respect to time-of-day variations in costs. This analysis addresses many of the limitations of cost allocation models typically used in practice by developing a set of models that account for marginal variations in vehicle passenger capacity, capital costs, and time-of-day costs using FY 1994 capital and operating data for the Los Angeles MTA. This analysis is unique in that it combines a number of previously and separately proposed improvements to cost allocation models. In comparison to the model currently used by the MTA, we find that the models developed for this analysis estimate: (1) higher peak period costs and lower base period costs, (2) significant variations in costs by mode, and (3) substantially lower costs for incremental additions in bus service. While this study uses Los Angeles MTA data, the focus of this analysis is not on the MTA per se, nor is this work intended as a critique of MTA practice. Rather, the focus of this study is on the limitations of the rudimentary, average cost allocation models employed by most transit operators. Toward that end, this analysis shows quite clearly that an array of factors addressed separately in the cost allocation model literature can be simultaneously and practically incorporated into a usable transit cost allocation model to provide transit systems with far better information on the highly variable costs of producing transit service.