Normalizing Citations – Beyond the H-index

The proper metric for an academic’s influence on the academic world of academic publishing is academic citations. An academic might make many (say 100) small contributions, each cited a small number (say 10) of times, or one contribution cited widely (say 1000) times. Neither is inherently superior, despite claims to the contrary, a

Citation needed. Source: Unknown.
Citation needed. Source: Unknown.

nd for the academic in question, it was probably easier to write one widely cited piece than 100 smaller ones, but that was unpredictable at the time.

Academic citations are cumulative distribution function, they can never go down (they can with retractions, but we will neglect that). So by this measure on average senior academics appear more influential than younger academics, which they of course are. But this is not a useful measure for filtering prospective candidates for hiring and promotion, which is why these metrics exist, to sort people based on productivity and establish a social hierarchy.

So to begin, we have two corrections to make. First, senior academics have more opportunities to write papers. A junior academic simply has not had the cumulative time to author 100 papers. Second, the senior academic’s papers have had more time to accumulate citations. So I suggest dividing total citations by Years^2 to account for these two temporal accumulating factors.

But which “Years”? Years since terminal degree? — This favors the young who start publishing before their degree. Years since they began their degree? Almost no one has any paper in year 1 of their graduate career. So we can estimate and split the difference and say years since graduation with terminal degree +2, on the theory that by the time you graduate you should have had at least 3 papers, and that means you started about 2 years before graduation. Still this is highly sensitive to assumptions for younger academics, it will wash out for the older academics. Domains will vary of course in terms of publishing culture.

There are other problems, for instance, co-authorship. At the extreme, all 108 billion people who ever lived have contributed fractionally to every paper, but they don’t all get co-authorship (except on experimental physics papers). But someone who puts all of their PhDs on all of their group’s papers is gaming the system to the detriment of those who assign more individually authored papers. So each citation should be divided by the fraction of authorship that the academic in question deserves. While this is impossible to assess, (promotion files sometimes ask for percentages on co-authored papers, but this is never systematically estimated or consistent). Computing an average dividing by the number of authors on the paper is a good surrogate.

I am not in this business of bibliometrics, I will leave that to others. But hopefully someone in the industry (Scopus, Web of Science, Google Scholar) can run the proposed corrections on these databases and produce a normalized citation measure as a standard output.

How traffic signals work: Some terms

TrafficCycleDiagram.001

Traffic engineers have developed terminology to aid in communication

The ‘approach’ is the set of lanes that are coming into a particular intersection, from a given direction. So, there might be an eastbound approach of traffic that is moving in the easterly direction.  A ‘cycle’ is the complete amount of time that it takes to go from a red light to a red light. We think of it as a clock. ‘Cycle length’  is the amount of time it takes to complete a cycle, measured in seconds.  A ‘phase’ is part of a cycle that is allocated to a particular movement, which receives the right-of-way. There might be multiple movements that receive right-of-way simultaneously, as long as they are not conflicting.  The northbound and southbound movements might both get the green light at the same time. They’re on the same phase, and they’re not conflicting.

“What do you do with right (left – in right hand drive countries) turns?” Do you give them a separate phase? Or do they share the phase? If they share the phase, then it becomes more complicated. There are many possible patterns, from which traffic engineers aim to select the ‘optimal,’ but that depends on the objectives and conditions.

There are ‘movements’. ‘Protected’ movements have right-of-way, and don’t have to yield to any other conflicting movements,  opposing vehicles, or to pedestrians. The ‘permitted’ movement is most common for right turns (in left-hand drive countries like Australia), for instance when making a turn without a green arrow, the driver has the permission to make that movement, so long as it is safe, but is not protected by a red light in the conflicting direction. Left turns are also permitted if there are no conflicting pedestrians or bicyclists.

‘Lost time’ occurs at the start of the phase because the first car has to accelerate from a dead stop, which takes some time: drivers first perceive the green signal, then check to make sure the intersection is clear, and then accelerate from a stop. So the speed at which that first (and second, and third) car goes through the intersection is slower than subsequent vehicles. There is also lost time at the end of the phase as some drivers are reluctant to go through on an amber (yellow) signal. There is also an ‘all red’ phase in some places to make sure the intersection is fully cleared of vehicles and pedestrians.

How much time is spent at traffic signals?

The 30-Minute City by David M. Levinson
The 30-Minute City by David M. Levinson 

While working on another piece, I came upon the question of how much time is spent at traffic lights, for which there is not a well-sourced answer. I posted to Twitter and got some useful replies.

With that and some additional digging, I attempt to answer the question.

As the saying goes: Your Mileage May Vary. This depends on your origin and destination and path and mode and time of day and local traffic signal policies and street design. Tom VanVuren notes: “Much of the impact is in slow moving queues, rather than waiting for the signal cycle to complete. I expect you can make this number smaller than 10% (time at the stop line) or larger than 50% (time affected by traffic lights).” For simplicity, I am considering vehicles that would be stopped if they could either move at the desired speed or must stop (i.e. they are subject to “vertical” or “stacking” queues), but clearly measurement will depend on assumption. Still, there must be a system average. I had heard the number 20% bandied about, which feels right, but let’s first begin with some thought experiment, then look for some empirical results. We take different modes in turn.

A signalized but porkchop-islanded crosswalk at a Free Left (Free Right for those in the right-side drive countries). Notice the pedestrian light is red (don't walk) but the pedestrians cross anyway. If the free left is not eliminated in a more comprehensive redesign, it could easily be de-signaled and the crosswalk raised, so pedestrians dominate, and cars travel when they can.
Pedestrian Crossing at Broadway and City Road, Sydney. Pedestrians crossing against the light.

Motor Vehicles

Thought Experiments

Thought Experiment 1 A

Imagine an urban grid.

  • Assume 10 signalized intersections per km.
  • Assume a travel speed of 60 km/h when in motion. (This is probably too high with so many intersections and no platooning, but we are imagining here that you would not be stopping.)
  • Time to traverse 1 km=1 minute + signal delay.  (Some of the distance traversal time overlaps some of the signal delay time, but we will imagine a stacking queue, rather than one that has physical distance for simplicity, we can correct this later if it matters.)
  • Assume each intersection has only 2 phases.
  • Assume fixed time signals at each intersection evenly distributing green time between N/S and E/W directions.  So red time = 1/2 cycle length.
  • Assume 1 minute cycle length
  • If a vehicle stops, it waits 1/2 red time.
  • Vehicles obey traffic signals.
  • Assume no platooning.

This means that the average vehicle will stop at 5 intersections for 15 seconds each = 75 seconds (or 1.25  minutes) (vs. 1  minute in motion time). In this case, 1.25/2.25 minutes (55.5%) is spent waiting at signals.

Thought Experiment 1 B

In contrast.

  • Assume near perfect platooning.

In this case, the vehicle will stop at 1 intersection per km, for 15 seconds = 15 seconds. In this case 0.25/1.25 = 20% of the time is spent waiting at signals.

Discussion

Now, not all travel takes place on an urban grid.

  • Assume 25% of travel is on limited access roads (this is approximately true in the US),  75% on non-limited access roads.

With perfect platooning on the grid, and 25% off-grid, then 15% of travel time is intersection delay with near perfect platooning.

Clearly in practice platooning is far from perfect. My guess is the green wave breaks down after one or two intersections during peak times, but can survive well in the off-peak. As a rule of thumb, about ~10% of travel is in the peak hour, ~30% peak period. ~60% AM + PM Peak.

Data

GPS Studies

Eric Fischer of MapBox was kind enough to offer to run this question on their open traffic data. The results are not yet in. I will update when they are.

Arterial Travel Time Studies

There are a variety of Arterial Travel Time studies for specific corridors, but nothing that is universally generalizable.  (And logically where people do arterial travel time studies, there is a congestion problem, otherwise why study it.)

I recall that in my childhood, I did a study in Montgomery County, Maryland using such data (from 1987 traffic counts and a floating car study published by Douglas and Douglas), I did not actually compute the percentage, but fortunately I reported enough data that allows me to compute the percentage now. (The sample is of course biased to what is measured). For the average arterial link, the speed was

 Variable Inside the Beltway  Outside the Beltway
 Speed (km/h) 34.88  41.60
 Length (km)  0.46  0.72
 Time (min)  0.792  1.04
 Downstream Delay (min)  0.27  0.24
 Percentage of Signal Delay  25%  18.75%

Which is consistent with expectations that signals are more significant in more urbanized areas (inside the beltway is basically Bethesda and Silver Spring, MD), and with our general estimates. Now of course the speed here is impacted by downstream signals, and so is lower than the speed limit and certainly lower than the free-flow speed sans-signals. More details are in the paper.

Engine Idling Studies

Moaz Ahmed pointed me to a Vehicle Idling Study by Natural Resources Canada.

The percent of time of vehicle idling ranged from 20-25%. (Not all vehicle idling is at signalized intersections).

(Engine idling of course burns fuel without doing work, so if the engine is going to be idling for an extended period, it would save fuel (and reduce air pollution) to turn it off. Turning the engine on and off also has costs, so the estimate was if idling was going to be longer than 10 seconds, it uses more fuel, but considering other wear and tear costs, the recommended threshold is if idling is longer than 60 seconds, then turn off the engine.  But at a signalized intersection, how will vehicles know how long they will wait? Smart traffic signals with connected vehicles could provide this, but now they don’t. Eventually this will be moot with a full electric vehicle fleet. Until that time, it matters. I suspect given the longevity and sluggishness of the traffic control sector, smart signals informing trucks will not be widely or systematically deployed before trucks are electrified.)

Pedestrians

Now as noted above, Your Mileage May Vary. If you are a pedestrian, you are unlikely to hit a greenwave designed for cars, though of course your travel speed is slower is well. So redoing the Thought Experiment

Thought Experiment 2

Imagine an urban grid.

  • Assume 10 signalized intersections per km.
  • Assume a travel speed of 6 km/h when in motion. (this is a bit on the high side, average pedestrian speed is closer to 5 km/h)
  • Time to traverse 1 km=10 minutes + signal delay.  (Some of the distance traversal time overlaps some of the signal delay time, but we will imagine a vertical stacking queue, rather than one that has physical distance for simplicity, this is a much better assumption for pedestrians than vehicles.)
  • Assume each intersection has only 2 phases.
  • Assume fixed time signals at each intersection evenly distributing green time between N/S and E/W directions.  So red time = 1/2 cycle length.
  • Assume 1 minute cycle length
  • If a pedestrian stops, she waits 1/2 red time. (That is the “walk” phase for pedestrians is as long as the green phase for cars. Strictly speaking this is not true, it is more true in cities with narrow streets than it is in suburban environments with wide streets, as narrow streets can be crossed more quickly, so the amount of “walk” time allocated can be most of the phase. This is certainly not true in Sydney, where the “walk” phase is cut short so turning cars have fewer conflicts with late pedestrians.)
  • Pedestrians obey traffic lights.  (This is not as good an assumption as vehicles obey signals, pedestrian signal violation is probably higher. This is not a moral judgment one way or the other, people tend to obey authority, even when authority abuses power.)
  • Assume no platooning. (This is probably too severe, a quick pedestrian with some signal coordination can probably make a couple of lights in a row).

Here the average pedestrian will stop at 5 intersections for 15 seconds each = 2.5 minutes (vs. 10  minute in-motion time). In this case, 2.5/(2.5+10) minutes (or 20%) is spent waiting at signals. Now, this number is probably true for more pedestrians than the vehicle delay estimate is for vehicles, since pedestrians are more likely to be found on an urban grid and less in a suburban or limited access environment. (Self-selection at work).

Bicyclists

If you are a bicyclist, you are unlikely to hit a greenwave designed for cars unless you travel at exactly an integer fraction (1/1, 1/2, 1/3) of the green wave, as your travel speed is slower is well. So redoing the Thought Experiment

Thought Experiment 3

Imagine an urban grid.

  • Assume 10 signalized intersections per km.
  • Assume a travel speed of  20 km/h when in motion. (This is a typical for experienced riders). Time to traverse 1 km=3 minutes + signal delay. (Assume a stacking queue)
  • Assume each intersection has only 2 phases.
  • Assume fixed time signals at each intersection evenly distributing green time between N/S and E/W directions.  So red time = 1/2 cycle length.
  • Assume 1 minute cycle length
  • If a bicyclist stops, she waits 1/2 red time. (That is the ‘bike’ phase for bicyclists is as long as the green phase for cars.)
  • Bicyclists obey traffic lights.  (This is not as good an assumption as ‘motor vehicles obey signals’, bicyclists signal violation is probably higher.)
  • Assume no platooning. (This is probably too severe, a quick bicyclists with some signal coordination can probably make a couple of lights in a row).

In this case the average bicyclists will stop at 5 intersections for 15 seconds each = 2.5 minutes (vs. 3  minute in-motion time). In this case, 2.5/(3+2.5) minutes (or 45%) is spent waiting at signals in an urban environment.

Strava Data

Strava, an app for tracking bicyclists and runners can produce some useful data. Andrew Hsu, e.g., reports “28 mile bike commute. 1:30-ish moving time. 10-15 minutes waiting at lights.” From this, for him, we estimate 15 / (15+90) = 14%. To be clear, 1:30 is an extreme commute. I don’t have access to the full database, and obviously this is biased by the nature of the trip.

Buses

Alejandro Tirachini produced an estimate of travel time for buses finds delay at traffic signals (in suburban Blacktown, Sydney, NSW) is 10-13% of total time.

The Transportist: March 2018

Welcome to the March 2018 issue of The Transportist, especially to our new readers. As always you can follow along at the blog or on Twitter.

Thank you to all who purchased Elements of Access. Copies are still available.

Book: Metropolitan Transport and Land Use

As cities around the globe respond to rapid technological changes and political pressures, coordinated transport and land use planning is an often targeted aim.
Metropolitan Transport and Land Use, the second edition of Planning for Place and Plexus, provides unique and updated perspectives on metropolitan transport networks and land use planning, challenging current planning strategies, offering frameworks to understand and evaluate policy, and suggesting alternative solutions.
The book includes current and cutting-edge theory, findings, and recommendations which are cleverly illustrated throughout using international examples. This revised work continues to serve as a valuable resource for students, researchers, practitioners, and policy advisors working across transport, land use, and planning.

PURCHASE

Transportist Posts

Transport News

HPVs/Bikes/Pedestrians

Aviation

Transit

Roads

AVs

EVs

SVs/Taxis/Car Sharing

HGVs/Freight/Delivery/Retail

Intercity Rail

Land Use

Science

Economics

Telecommunications

Justice/Equity

Security

Research

Books

Now available:

Nothing in cities makes sense except in the light of accessibility. Transport cannot be understood without reference to the location of activities (land use), and vice versa. To understand one requires understanding the other. However, for a variety of historical reasons, transport and land use are quite divorced in practice. Typical transport engineers only touch land use planning courses once at most, and only then if they attend graduate school. Land use planners understand transport the way everyone does, from the perspective of the traveler, not of the system, and are seldom exposed to transport aside from, at best, a lone course in graduate school. This text aims to bridge the chasm, helping engineers understand the elements of access that are associated not only with traffic, but also with human behavior and activity location, and helping planners understand the technology underlying transport engineering, the processes, equations, and logic that make up the transport half of the accessibility measure. It aims to help both communicate accessibility to the public.

Purchase:

Still available …
In this book we propose the welcome notion that traffic—as most people have come to know it—is ending and why. We depict a transport context in most communities where new opportunities are created by the collision of slow, medium, and fast moving technologies. We then unfold a framework to think more broadly about concepts of transport and accessibility. In this framework, transport systems are being augmented with a range of information technologies; it invokes fresh flows of goods and information. We discuss large scale trends that are revolutionizing the transport landscape: electrification, automation, the sharing economy, and big data. Based on all of this, the final chapters offer strategies to shape the future of infrastructure needs and priorities.
Purchase