Path flow estimator (PFE) is a one-stage network observer that can esti- mate path flows and path travel times from traffic counts in a transporta- tion network. Because a unique set of path flows is readily available from the PFE, a trip table can be estimated by simply adding up flows on all the paths connecting individual origin–destination (O-D) pairs. In this paper, the effects of the number and locations of traffic counts on the quality of the O-D trip table estimated by PFE are examined. The set-covering model, studied in the location theory, is applied to determine the minimum number of traffic counts and their corresponding locations required to observe the total demand of the study network. Next, the effects of the error bounds used in PFE to handle the inconsistency prob- lem of traffic counts are examined, and a heuristic using the Lagrange multipliers to facilitate the adjustment of such error bounds is provided. Numerical results show that PFE can correctly estimate the total demand of the study area if a sufficient number of traffic counts collected at appro- priate locations is provided. The results further indicate that improper specification of the error bounds could lead to biased estimation of total demand utilizing the network. The origin–destination (O-D) trip table depicts the spatial distrib- ution of trips among traffic analysis zones and the total number of trips in a transportation network. It is one of the important inputs required in traffic assignment and traffic simulation. During the past several decades, a significant research effort—namely, the studies of Van Zuylen and Willumsen (1), Cascetta (2), Spiess (3), Fisk (4), Yang et al. (5), Ashok and Ben-Akiva (6 ), Sherali et al. (7 ), Bell and Shield (8), Hazelton (9), and Maher et al. (10)—has been devoted to developing efficient methodologies that can pro- duce a reasonable estimate of the O-D trip table based on readily available and inexpensive data such as traffic counts. The O-D estimation from traffic counts can be broadly classified based on network configuration to simple networks with no route choice, networks with route choice but no congestion, and networks with both route choice and congestion. Examples of simple networks without route choice include finding turning fractions at an inter- section (11, 12) and determining split ratios for a freeway system with several on- and off-ramps (6, 13). For uncongested networks with route choice, a proportional assignment, which is exoge- nously determined, could be used when estimating the O-D trip table. The statistical methods used in this class of problem include entropy maximization (1), maximum likelihood (3), generalized least squares (9, 14, 2), and Bayesian inference (15), which are used to reduce the degree of freedom of the estimation problem. For general networks with both route choice and congestion, the assumption of proportional assignment is no longer valid because route choice proportion and the O-D trip table are interdependent (16 ). It is necessary to incorporate a route choice model explicitly into the O-D estimation problem. An approach based on the bilevel program by endogenously determining route choice proportion while estimating the O-D trip table is one of the possible solutions to ensure consistency (4, 5, 10). In the bilevel programming approach, the upper-level problem uses one of the statistical techniques mentioned earlier (e.g., generalized least squares) to estimate the O-D trip table, and the lower-level problem endogenously determines the route choice proportion [e.g., deterministic user equilibrium or stochastic user equilibrium (SUE)] consistent with the estimated O-D trip table. Although the inconsistency issue is resolved, the bilevel programming approach could pose a computational difficulty when applied to large-scale networks. Among the efficient methods developed thus far, path flow estimator (PFE), originally developed by Bell and Shield (8), does not directly estimate O-D flows but estimates path flows from traffic counts. The attractiveness of PFE lies in the fact that it is a single-level mathematical program in which the interdependency between O-D trip table and route choice proportion (i.e., conges- tion effect) is taken into account without using a bilevel mathe- matical program. Network users are assumed to follow the SUE principle, which allows the selection of nonequal travel time paths due to imperfect knowledge of network travel times and yields the unique path flow estimate. Extensive studies on this topic during these past several years demonstrate that the quality of O-D trip table estimate is highly depen- dent on the underlying route choice models, the quality of observa- tions (17 ), and the number and locations of traffic counts (18). The quality of traffic data, which is usually defined by the error inherited from the data collection and processing, not only contributes to the accuracy of the O-D trip table estimate but also creates some oper- ational problems for some O-D estimation models, especially those that require exact preservation of flow continuity (i.e., nodal con- servation). The inconsistencies among traffic counts somehow need to be resolved either before or during the estimation process. In this study, the quality of O-D trip tables estimated by the nonlinear PFE with different sets of traffic counts is investigated. The quality of the O-D trip table estimate is examined in terms of two aspects: the capability to estimate the total demand in the network and the capa- bility to replicate the spatial pattern of the known (true) O-D trip table. In addition, a heuristic procedure is also proposed using the Improved Path Flow Estimator for Origin–Destination Trip Tables Piya Chootinan, Anthony Chen, and Will Recker P. Chootinan and A. Chen, Department of Civil and Environmental Engineering, Utah State University, Logan, UT 84322-4110. W. Recker, Department of Civil Engineering, University of California at Irvine, Irvine, CA 92697-3600. 9 Transportation Research Record: Journal of the Transportation Research Board, No. 1923, Transportation Research Board of the National Academies, Washington, D.C., 2005, pp. 9–17.