Congestion pricing schemes controlled by the gMFD: a comprehensive design and appraisal to bridge the engineering and economic perspective

Abstract

Charging drivers for the congestion they cause is a well-known concept among economists, traffic engineers and transport professionals. Many studies have been performed to explore and better address this issue during the history of transport economics and in recent years the first congestion pricing schemes have been finally implemented in some cities (Singapore, London, Stockholm). Nonetheless, several theoretical and practical problems still restrain this measure from being widely adopted. The lack of consistency with the traffic engineering perspective and limited degree of realism represent the main theoretical limitations of traditional congestion pricing models. The public resistance and fair distribution of gains and losses are the major practical barriers to the implementation of congestion pricing schemes. In this thesis, we discuss congestion pricing models controlled by the gMFD within agent-based simulation (MATSim). With this approach we seek efficient tolling schemes from an economic and traffic perspective. Furthermore, by means of investigations of the distributional effects of the proposed schemes we look for the strategy that maximizes the social welfare and with the highest potential acceptability. In the first phase of this study, through a review of theoretical studies and real experiences we identify the main reasons behind and factors influencing public resistance to the introduction of congestion pricing measures. Particularly, the question of equity is discussed in order to derive the necessary background for the evaluation of distributional effects. A thorough review of the major congestion pricing models proposed to date is made to construct our model. The survey goes from the traditional approaches like Marginal Cost Pricing and the bottleneck model to the most recent approaches based on macroscopic traffic models. During the review we discuss the main advantages and drawbacks of each approach. In the end, we describe the concept of Macroscopic Fundamental Diagram (MFD) and introduce important issues like the distribution of traffic and the phenomenon of hysteresis. The experiment is set up with MATSim to study the metropolitan area of Zurich. First, we test the consistency of simulation outputs with the traffic flow theory by playing with different parameters of the traffic model in order to obtain the most reliable result. Then, we derive and explore the main macroscopic traffic relations of the network defined by the cordon. This analysis shows the influence of distribution of traffic, expressed by the spatial spread of density, on the performance of the network. Furthermore, this property can be associated to the phenomenon of hysteresis loops. In order to account for this additional complexity, a generalized macroscopic fundamental diagram (gMFD) is experimentally derived to express the relationship between accumulation, production and spread of density. Finally, three tolling schemes differently controlled by the gMFD are designed: a uniform toll (Flat Toll) that allows the system working below the critical accumulation threshold; a time-varying toll (Step Toll) that charges drivers according to the total delay inside the cordon estimated through the gMFD; the Spread Toll that explicitly considers the question of spatial distribution of traffic inside the cordon. In particular, the Step Toll applies a charge in order to completely eliminate delay, whereas the Spread Toll penalizes users only for actual delays due to the overall increase of demand rather than clustering of congestion. By integrating different aspects like the traffic performance improvements, travel behavior responses and economic impacts we carry out a comprehensive evaluation of the schemes. In this broad appraisal framework the typical traffic enhancement measurements (heaviness of congestion, delays etc.) are combined with economic indicators such as the variation of agents’ utility. An analysis of travel behavior responses is carried out to identify the major trends associated with the tolling schemes and possible relationships with the traffic and economic impacts. Different socio- demographic groups are defined in order to evaluate the distributional impacts on inhabitants of different neighborhoods and areas, and on different trip purposes (work, home, education, shopping and leisure). Finally, considerations on the potential levels of public acceptance of the schemes are made on the basis of some indicators such as the share of winners, the average gain and loss, and the benefit-cost ratio. The proposed schemes have shown to perform differently from each other. While the Step Toll and the Flat Toll determine comparable traffic enhancements, the first one outperforms the second one in terms of economic impacts and distributional effects. The smoother adaptation of demand determined by the Step Toll is probably its strong point. The Spread Toll, which was conceived as the fairest scheme, does not produce any considerable traffic improvement and it is even detrimental from the economic perspective. As to the question of public acceptance, although the slightly lower share of winners, the Step Toll determines higher gains than the other two schemes. Hence, from a social welfare perspective (combination of efficiency and distributional impacts), the Step Toll is the best performing scheme, followed by the Flat Toll and Spread Toll. Besides these results, the investigations of macroscopic traffic properties of networks have shed light on important aspects of the traffic flow theory. First, the distribution of congestion expressed as spatial spread of density has considerable influence on the network performance. This property could be modeled in heterogeneous networks by introducing the spread as additional variable in the gMFD. Second, hysteresis loops in the MFD plane occur in presence of congestion and their pattern can be identified with the spread. Furthermore, the decrease of performance seems to be related to the frequency of loading-unloading cycles.