Time Savings vs. Access-Based Benefit Assessment of New York’s Second Avenue Subway

Recently published:

  • Wang, Yadi and Levinson, D. (2022) Time savings vs Access-based benefit assessment of New York’s Second Avenue Subway. Journal of Benefit Cost Analysis. 13(1) 120 – 147. [doi]


Under the current practice of benefit-cost analysis, the direct economic benefits produced by a newly built transit facility are assessed based on how it affects travel time and various costs that are associated with transport needs and travel behavior. However, the time-saving-based benefit calculation approach has been questioned and criticized. Given the strong correlation between accessibility and land value, we propose the access-based land value benefit assessment as an alternative, and apply this assessment method to analyzing the Second Avenue Subway project in Manhattan, New York. The primary principle of the access-based method is that the economic value of a transport project’s intangible gains is largely capitalized by nearby properties’ value appreciation, which is directly caused by improved transport accessibility. We find that: (i) the actual travel time saving is lower than originally forecast; (ii) a strong positive correlation between residential property value and job accessibility by transit is observed; (iii) the appreciation in sold property value and rented property value both far exceed total project cost; and (iv) such results support the decision to approve and construct the Second Avenue Subway.

Dr. Ang Ji

Congratulations to Ang Ji for “satisfying the requirements for the award of the degree of Doctor of Philosophy at the University of Sydney.”

Thesis Title: Traffic programming: Aligning incentives for socially efficient lane changes among non-connected vehicles.

Lead Supervisor: Professor David Levinson.

Abstract: This dissertation explores the rationality of drivers’ risky and aggressive behaviors in lane-changing scenarios and discusses some feasible ways to hold selfish drivers accountable for their decisions. Regardless of potential congestion and crashes suffering by other road users, rational drivers prefer to maximize their gains and demand others’ yielding. However, when all of them have such thoughts, conflicts (dilemmas) are embedded in their interactions, leading to unexpected consequences for the whole traffic. This question is investigated analytically by exploiting the game theory concept. A simplified 2×2 non-cooperative game is built to model strategies executed by human drivers without communications. This research learns driver behavior in two predefined sub-phases: `Stay’ and `Execution’ from empirical data. This procedure examines the factors that impact drivers’ execution of lane changes. From the results, we understand that lane-changing is motivated by the urgency to change and the dissatisfaction with current circumstances. The analytical model is then established by integrating driver incentives into payoff functions. The `greed’ and `fear’ of drivers in this process are quantified by speed advantages and possible crash costs respectively, so they trade off these factors and make decisions based on their own and opponents’ estimated payoffs. Using a numerical case study, we find that social gaps exist between user-optimal and system-optimal strategies when drivers mostly engage in selfish behaviors, significantly deteriorating the total system benefit. Pricing can be a sufficient tool to incentivize users to cooperate with others and achieve win-win outcomes. It is posited that the designed pricing schemes may promote the negotiation between drivers, reducing collision risks and improving operational traffic efficiency. Several simulation experiments are then conducted to evaluate this dissertation’s hypotheses on the performance of pricing rules. Overall, the proposed framework develops a behavioral model and improvement schemes from the perspective of microscopic vehicular interactions. The conclusions will hopefully find their applications in autonomous vehicle-human interaction algorithms and future transportation systems.

Ang Ji
Ang Ji

Journal articles related to the dissertation include: 

Dr. Ji now has a position at Southwest Jiaotong University in Chengdu, one of China’s leading transport programmes.

The idea of Traffic Programming was first raised in this blog a while back (in 2016).

We recently were awarded a grant from the Australian Research Council to examine this question in further depth.

  • Design of micro-decisions in automated transport. Australian Research Council DP220100882 Professor David Levinson; Professor Michael Bell; Dr Mohsen Ramezani; Professor Dr Kay Axhausen; Professor Dr Hai Yang.

Ensemble Models of For-Hire Vehicle Trips

Recently published:

Wu, Hao and Levinson, David (2022) Ensemble Models of For-Hire Vehicle Trips. Frontiers in Future Transportation. 3 [DOI]

Ensemble forecasting is class of modeling approaches that combines different data sources, models of different types, with different assumptions, and/or pattern recognition methods. By comprehensively pooling information from multiple sources, analyzed with different techniques, ensemble models can be more accurate, and can better account for different sources of real-world uncertainties. The share of for-hire vehicle (FHV) trips increased rapidly in recent years. This paper applies ensemble models to predicting for-hire vehicle (FHV) trips in Chicago and New York City, showing that properly applied ensemble models can improve forecast accuracy beyond the best single model.

Transportist: April 2022

Welcome to the latest issue of The Transportist, especially to our new readers. As always you can follow along at the  transportist.org or on Twitter.


  • Sydneysiders will get  12 days of free travel across the public transport network after “the NSW Rail, Tram and Bus Union threatened to strike every Friday in June unless the government instituted a period of free travel for commuters as an apology for last month’s fiasco when trains shut down for 24 hours.” This occurs during the school holidays when demand is lower.

Evaluating Evaluation

We need a way of assessing strategic (i.e. UTPS-like ‘four-step’) transport models. 

My first real job was working as a transport modeler for the Montgomery County, Maryland Planning Department. At the time, models were about a third of a century old, dating from the mid-1950s, deriving from the work of Douglas Carroll et al. in Detroit and then Chicago. I worked on developing a new model for the Washington DC region, (as we didn’t really trust the model from MWCOG, and wanted our own), Travel/2, using then current data and ideas in transport modeling. It had some good features which are yet to be mainstream like ensuring travel time consistency between trip distribution and route assignment, and used then au courant logit mode choice models and detailed trip generation. It wasn’t an agent (activity)-based model, the world wasn’t quite ready, TranSims was just being developed, though we played with the idea some and I wrote not very efficient Fortran code for population synthesis. Route assignment was still static (though multi-class), DTA was not an off-the-shelf product. We tried to document everything, (which is probably sitting in WordPerfect files on my computer) though there was no internet or GitHub to host our code or macros, and of course the source code for the modeling platform (EMME/2) was and remains proprietary, not open access.

Models have not advanced as much in the past third of a century since I started as we would have hoped at the time. Certainly they have moved in the direction we anticipated, just very slowly. Ideally we would assess their accuracy, there is some literature on ex-post analysis of course, and we have done that, but that takes too long, years are required to know whether the forecast was accurate, and by then decisions have been made and the models have evolved, so anything that was wrong was obviously obsolete and cannot be used to criticise the new models.

It should however be possible to assess the models (and each component model) themselves on a number of criteria, e.g. (but not limited to):

  • Transparency,
  • Replicability,
  • Methodological Consistency,
  • Methodological Quality (without prescribing a particular modeling technique, but definitely proscribing those that are known inferior),
  • Internal Consistency (the input travel times to travel demand equal the outputs from route choice, e.g.),
  • Calibration and Validity (how does the base model compare with observed data of various kinds, has backcasting been done, are the results accurate, … is this done systematically, following standard documented procedures?),
  • Recency of Estimation Data,
  • etc.

And develop some kind of scorecard that individuals or teams can apply to strategic transport models, roughly similar to how the ITDP BRT scorecard works. The score should not be a weighted average, but a product of the scores of the components. A zero on any attribute would result in a 0 for the whole thing. Models could then be rated as Gold, Silver, Bronze, or Tin. This can be used to argue for model improvements by external comparison in some sort of systematic (rather than ad hoc) way, perhaps using league tables to compare models, as well as providing a means to discredit bad forecasts, which is what scares away modellers from this approach. 

The Zephyr Foundation aims to improve transport modeling, but it is largely organised by modelers themselves, and thus subject to political sensitivities.

Similarly, we need a way of assessing Business Cases (Benefit-Cost Analyses) (BCA) based on how good the methodology is.

Obviously many BCA in Australia often fail on the criteria of transparency and replicability, as they are tightly held as “Cabinet in Confidence” or “Commercial in Confidence”, so no one knows how the results were actually established. By the product rule, if they fail on that criteria, they fail on everything else, as none of the other criteria can be properly evaluated.

We can ask for some of the same kind of criteria as above. There are various industry committees (TRBATAP) that aim to improve and standardise methods, but I am not aware of anyone from outside the sector who is actually scoring the quality of the evaluation. Let me know of any.


  • Lahoorpoor, Bahman, and David Levinson. 2022. “In Search of Lost Trams: Comparing 1925 and 2020 Transit Isochrones in Sydney.” Findings, March. https://doi.org/10.32866/001c.33040. Has Sydney lost access by removing its extensive tram network? We compare the 1925 tram network with today’s bus network, and conclude that the access provided today exceeds what would have been provided by just trams. The Sydney CBD would have had better access if 1925’s central tram lines were still in operation.
  • Access by Trams and Trains in Sydney and Melbourne in the 1920s. Sydney map by Bahman Lahoorpoor inspired by map of MelbourneMap of Sydney and Suburbs showing railway lines including trams. (1925)Map of Sydney and Suburbs showing railway lines including trams, buffered. (1925)Metropolitan Town Planning Commission Map of Melbourne and Environs: Minimum Railway and Tramway Time Zones.[Cam Booth sells a restored version of the Melbourne map, suitable for framing]
  • The latter image is on the Cover of Elements of Access,

Elements of Access: Transport Planning for Engineers, Transport Engineering for Planners. By David M. Levinson, Wes Marshall, Kay Axhausen.

A Pattern Language.jpg

Research by Others

News and Opinion

Substack App

You can now read Transportist in the new Substack app for iPhone.

Now available for iOS

Get the app

With the app, you’ll have a dedicated Inbox for my Substack and any others you subscribe to. New posts will never get lost in your email filters, or stuck in spam. Longer posts will never cut-off by your email app. Comments and rich media will all work seamlessly. Overall, it’s a big upgrade to the reading experience.

The Substack app is currently available for iOS. If you don’t have an Apple device, you can join the Android waitlist here.