Calgary Public Transit II

[MalcolmTucker]

Have to model choice somehow If I modeled it in arbitrary cupcake units would that make people happy (but in that case, would walking consume cupcakes or generate cupcakes).

[fusili]

Originally Posted by MalcolmTucker

Have to model choice somehow If I modeled it in arbitrary cupcake units would that make people happy (but in that case, would walking consume cupcakes or generate cupcakes).

Agreed, but let's not model behavior on outdated theories of decision making that have no basis in actual psychology or behavior.

[DizzyEdge]

Originally Posted by fusili

Agreed. Those places that do have them, usually have multiple jurisdictions that they cross over. BART in the Bay Area (San Francisco, Oakland, Berkely, Hayward, Freemont, etc) and Barcelona's Renfe and FGC systems (Barcelona, Terrassa, Sabadell, Besos Valley, Llobregat Valley) are two examples I can think of. I know Berlin has a zone system, but not sure about the metropolitan configuration.

Well and Vancouver of course, with Burnaby,N Vancouver, Richmond, Surrey, Coquitlam etc being in different zones than Vancouver proper.

[DizzyEdge]

I wonder how much the ticket would need to be to ride the LRT from Cochrane to 100% cover its cost. In town I think fare only covers 50%, so $6 right there, plus paying off the capital cost of extending to Cochrane, plus added maintenance. Maybe with a bit of a discount for less road wear from Cochrane vehicular traffic.

[Bassic Lab]

Originally Posted by DizzyEdge

I wonder how much the ticket would need to be to ride the LRT from Cochrane to 100% cover its cost. In town I think fare only covers 50%, so $6 right there, plus paying off the capital cost of extending to Cochrane, plus added maintenance. Maybe with a bit of a discount for less road wear from Cochrane vehicular traffic.

Our fare recovery rate is about 50% in general. That includes every empty bus trudging along at ten PM. The recovery ratio for just the LRT would be much higher, probably positive. If provincial money built the expansion and Cochrane provided the feeder bus network, then Calgary Transit would likely make money at $3.00 a rider (at least as long as there were enough of them to justify the service).

[MalcolmTucker]

It would be much cheaper to upgrade the CPR mainline and run a couple Sumitomo DMUs with additional stops in south Tuscany and Bowness. ~$500 million. a few years ago.

[mersar]

Originally Posted by Bassic Lab

Our fare recovery rate is about 50% in general. That includes every empty bus trudging along at ten PM. The recovery ratio for just the LRT would be much higher, probably positive. If provincial money built the expansion and Cochrane provided the feeder bus network, then Calgary Transit would likely make money at $3.00 a rider (at least as long as there were enough of them to justify the service).

From what I recall LRT fare recovery is actually closer to 150%. Its the feeder bus network that drags it down to around 50%. Aside from the top few routes (301/3/20/4/5, etc) most bus routes are absolutely horrendous in terms of fare recovery vs cost.

[sim]

Originally Posted by MalcolmTucker

Well, if you want to get into the transportation economics of it you have to start with an assumption, in this case I will assume a time value of 25 ¢ a minute. I'll assume the centre of downtown is Knox United Church.

You have to operate at 5 minute frequencies before you can attract half of the walk/transit trips between the centres at close to current fares.

Based on my memory of a 3rd year transportation economics class more than a few years ago.

Given that an hour of transit service has the operational cost of $116.40, and for those three barebones routes you would need 14 buses on the road (with very little room for contingency, turnarounds, breaks), you would need to generate more than 1000 boardings an hour across all service hours to get to 50% farebox ratio. For 7 am - 7 pm it would be $3,5 million a year if ridership goals were met. (this assumes you serve the 14th street and 8th street routes as different routes)

Well I'd be pretty darn hesitant to present such a crude modelling exercise as even being a little bit conclusive. I know this is just a forum and all and probably just meant as an exercise, but you never know where/who this resonates with.

The time value assumption might be somewhat reasonable. Equating the disutility of the two times, however, not so much. This can easily be 2 or more times more arduous - hence a 1 for 1 time trade off is, well, not so good an assumption. Plus, transit users walk. Nor does this encapsulate the full breadth of modes, and it sure does not account for the other factors that contribute to mode choice, but anyway let's leave it at that. And since I'm already being pedantic, unless you don't give your drivers breaks and have ultra reliability you'll need more than 14 buses.

But to the whole modelling thing:

Originally Posted by fusili

Yeah, people definitely don't think or behave this way. Treating travel behavior through an rationale choice theory model is stupid for several reasons. First, rationale choice theory was thoroughly trashed by Daniel Kahneman and Amos Tversky in their studies of decision making behavior. Second, travel decisions are about a lot more than time and money. Let's just stop using completely failed economic models to try and explain, let alone predict anything.

Here is a start: Peter Jones, et al "Understanding Travel Behavior", Aldershot Press (IIRC) 1980.

Seems someone read The Black Swan?

The value of time used above is actually, with some reasonable approximations, not bad. 0.25 cents/min or $15/hour which is give or take about 50% of prevailing wage. I do believe this is what Transport Canada recommends and numerous other authorities when doing a Cost-Benefit analysis on a project.

I'm not sure I 100% understand what you are criticizing here, but Tversky and Kahneman are very often referenced in transportation demand modelling (I did a search of the name and had about 20 papers come up that I actually have alone). They are also some of the founders of behavioral economics, which is preciously what "good" models use to come up with values of time (although there are other methods). While the above analysis was very parsimonious, you better believe that a similar rationale is indeed used for a lot of analysis, albeit at the aggregate level in most practical applications (of mode choice). The use of this on a disaggregated scale remains largely within academia.

To my understanding they didn't thrash rational choice modelling per se, but showed the irrational and uncertain elements exist. And incidentally, there are extreme value distributions that incorporate individual idiosyncrasies in the known (or rational) attributes. Heck, there's even decision making models to account for variability in unobserved attributes.

So if Malcom's analysis were to use this (but this is a rather large exercise) with that basis, it could account for the gal that walks 1 hour one way when she owns a vehicle and has "free" parking, or the guy that drives 2 blocks to work. In the end, this gets re-aggregated anyway for real decision level data - and that, given that a sample size is large enough quickly causing "convergence to the Gaussian" is what you would plan infrastructure or services around. You don't build HSR for the 6 people willing to pay 500 for a one way ticket.

Transportation is well within a physical realm. Rationality does exist within choices people make pertaining to aspects of it, especially because it's so often repeated and a derived demand anyway. However, what's rational or what isn't is in this case rather subjective and this CAN be accounted for in discrete choice theory. The models are by no means right, and they are not even necessarily useful. In some more extreme cases them being wrong might be impactful.

In most cases they are better than simply estimating a catchment population and applying some a priori mode share to it and hope for the best. I am picking on something that I saw about a month or so ago.

I guess I'm a bit surprised that people don't know the extent of, hmmm, well science that goes on, or at least could behind transportation modelling and therefore policy and infrastructure decisions.

Originally Posted by fusili

Agreed, but let's not model behavior on outdated theories of decision making that have no basis in actual psychology or behavior.

This even pervades into books such as Human Transit (of which I know there are some fans of on here).

I apologize for the story.

[sim]

Originally Posted by Bassic Lab

A trip that only lasts a couple of minutes has less value for the rider than a long commute but it often has a higher marginal cost to Calgary Transit than a long suburban commute. Having built the C-Train, it costs next to nothing to add another rider from Somerset-Bridlewood. Alternatively, an off peak bus driving a handful of people a couple of blocks would cost much more to introduce.

This seems to come up here on a recurring basis. I'll reiterate my stance, that this is actually a rather short-sighted way of looking at the issue. If you extend this supposition, you might as well throw city structure and accessibility out the window. Ideally, you'd have two highly dense nodes that a transit line would serve with a station at each end. Fill it up one end, drop em all off at the other. It's not the underlying point of a transit system, nor should it be used as rationalization post-construction [of the LRT], because a fare-structure may just have long-run effects on built-form. You've also taken away resources to provide a system that is perhaps more effective for long distance commutes, at the cost of providing one that is more effective in centralized areas where it could then serve as something (in conjunction with other modes) that serves more trip purposes.

Why shouldn't users be more responsible for capital costs of a system if they require those capital costs in the first place? The contra-argument always seems to be, well it's better than them driving. True.. if you forget transportation (and land-use) work as one giant system. What if the NW leg carried the same amount of passengers but only went to Capitol Hill for example... ohh imagine the farebox recovery!! You'd even approach something that resembles the ideal (above) with shorter distances. Ahh, but that would require more people living closer and therein the dreaded word, density! Don't like it? Want to continue working downtown? No problem, but PAY for it.

Besides, for anyone still worried about marginal costs in relation to usage - I can confidentially say that not a single agency doesn't already charge some form of marginal cost just for stepping on the train/bus/vehicle. I.e. The fare isn't proportional to distance alone (you'd have a y-intercept to speak graphically), as the majority of the fare is charged regardless of distance. A base fare, like Taxis if you will.

Put me in the pro zone or distance-based faring camp. Incidentally, whether overlapping or regional agencies exist or not, most of the world's more successful transit systems do instate one or the other kind of graduated faring.

[Socguy]

Originally Posted by technomad

Came across this this morning, wonder how well it could work for a Centre St line?

http://youtu.be/_n47ExxV6QY

• Video Link

This seems interesting but ultimately I don't think it would work for the center street line, at least the whole way. I could see it employed in the more northerly parts of the line where there is a little more room to play with and people decided it was worth it to hide the line. But as you move closer to the core I suspect that your only two real options are an elevated guideway or a full underground tunnel, at least if you want to follow the center street alignment.

I say this because in my opinion any line that goes into the core from now on will need to be underground. Once you have to go to the expense of tunneling some parts, you might as well just stick with it across the river and up Center st. Then it's just a matter of picking where to come out of the ground, which will be a trick in itself.

Coming out of the ground, I can't think of any great places until you at least get to 16th and even that may be too short. Don't get me wrong, I believe that an LRT line up Center street would be very popular once built but a surface line or a trench line would either take lanes out of the road or force the demolition of some houses and I can't imagine much political will to do that at the moment. That is why I say that the only realistic options are a tunnel or elevated track.

Of course all of this is simply my own opinion on the matter. I'm sure that there is somebody around here who is more well versed than I in these particular matters. I do think the technology should prove useful in the right application though!

[MalcolmTucker]

Modeling long distance transportation modal choice is much easier than short trips of course. The purpose of the exercise wasn't to judge demand in numbers, but to work backwards from what would be a useful level of service to see roughly what it would cost, and how much demand would be needed to justify that level of service. It was also an exercise in 'how much service has to be provided to actually save people time'. Turns out it is a lot.

I think a beltline collector much like a neighborhood LRT collector for rush hours would be very useful, but outside of those hours it would be pretty hard to justify, especially since the walk times are google's estimates and seem long to me, and that the area has many routes already that would not be overburdened outside of rush. I walked from 17th Ave to 6th Ave almost everyday for 3 years and never found it arduous.

I understand that our rough thoughts on here can lead to wider opinions outside of direct contributors on here. I hope for example that on the main thread thoughts about inclusionairy zoning spread. I wouldn't want an overly ambitious push into Beltline mobility setting back long term sustainability of Beltline mobility by having to be cut after ridership didn't materialize either.

[Allan83]

Originally Posted by MalcolmTucker

Well, if you want to get into the transportation economics of it you have to start with an assumption, in this case I will assume a time value of 25 ¢ a minute. I'll assume the centre of downtown is Knox United Church.

Destinations:
17th Ave and 8th Street SW
26th Ave and 4th St SW
17th Ave and 14th St SW

At what frequencies could you operate if you charged nothing, and still have 50% of people of the transit/walk modal share choose transit?

Capturefree by Odarwin1, on Flickr

Fares for 15 minute frequency

Capture15 by Odarwin1, on Flickr

Fares for 10 minute frequency

Capture10 by Odarwin1, on Flickr

Fares for 5 minute frequency

Capture5 by Odarwin1, on Flickr

You have to operate at 5 minute frequencies before you can attract half of the walk/transit trips between the centres at close to current fares.

Based on my memory of a 3rd year transportation economics class more than a few years ago.

Given that an hour of transit service has the operational cost of $116.40, and for those three barebones routes you would need 14 buses on the road (with very little room for contingency, turnarounds, breaks), you would need to generate more than 1000 boardings an hour across all service hours to get to 50% farebox ratio. For 7 am - 7 pm it would be $3,5 million a year if ridership goals were met. (this assumes you serve the 14th street and 8th street routes as different routes)

It seems that the consensus is that this is not the best model, but it did spark some good discussion, so good post in any event. Let me add something else to the equation which I think would be important. I think it would be a big plus to be able to accurately track the buses and know when your bus was going to be at your stop. I think a 10 minute frequency would be enough if you knew within a minute or two when the bus was going to arrive at your stop. It’s typically not that difficult for people to adjust their schedules by 5 or 10 minutes one way or the other, and if you know when you bus is coming you can either leave 5 minutes early or work 5 minutes late, and you haven’t wasted any time. Going to 10 minutes cuts the number of buses in half, and doubles the number of riders on each bus, and the economics start looking better in a hurry no matter what model you use.

[Allan83]

Here’s a secondary question. How much money could you generate from the advertising on the buses on these routes? These buses would be circulating all day long through the downtown and Beltline, so the advertising space should have a much higher value than the space on buses that spend most of their time travelling through the burbs.

[MalcolmTucker]

Originally Posted by Allan83

Here’s a secondary question. How much money could you generate from the advertising on the buses on these routes? These buses would be circulating all day long through the downtown and Beltline, so the advertising space should have a much higher value than the space on buses that spend most of their time travelling through the burbs.

The ad contract comes up for tender in 2016. I can't find specific numbers for Calgary, since it is rolled into revenue in the transit budget, likely for competitive reasons. For the TTC it is $27-million-per-year. If my memory serves at one point CT reported the contract was between $7 and $12 million-per-year.

[Allan83]

Originally Posted by MalcolmTucker

The ad contract comes up for tender in 2016. I can't find specific numbers for Calgary, since it is rolled into revenue in the transit budget, likely for competitive reasons. For the TTC it is $27-million-per-year. If my memory serves at one point CT reported the contract was between $7 and $12 million-per-year.

Thanks. Just by rough approximation you could probably offset the costs of these routes a bit with advertising revenues but not a huge amount.

[milomilo]

Fascinating stuff on the last few pages! One thought though - once you get to the point of <5min frequencies with possibly converging bus routes that stop extremely frequently I'd imagine the system would be close to breaking down, with the buses themselves causing congestion, bunching and overtaking each other. That in turn would reduce the desirability of the route.

[fusili]

Originally Posted by sim

Well I'd be pretty darn hesitant to present such a crude modelling exercise as even being a little bit conclusive. I know this is just a forum and all and probably just meant as an exercise, but you never know where/who this resonates with.

The time value assumption might be somewhat reasonable. Equating the disutility of the two times, however, not so much. This can easily be 2 or more times more arduous - hence a 1 for 1 time trade off is, well, not so good an assumption. Plus, transit users walk. Nor does this encapsulate the full breadth of modes, and it sure does not account for the other factors that contribute to mode choice, but anyway let's leave it at that. And since I'm already being pedantic, unless you don't give your drivers breaks and have ultra reliability you'll need more than 14 buses.

But to the whole modelling thing:



Seems someone read The Black Swan?

The value of time used above is actually, with some reasonable approximations, not bad. 0.25 cents/min or $15/hour which is give or take about 50% of prevailing wage. I do believe this is what Transport Canada recommends and numerous other authorities when doing a Cost-Benefit analysis on a project.

I'm not sure I 100% understand what you are criticizing here, but Tversky and Kahneman are very often referenced in transportation demand modelling (I did a search of the name and had about 20 papers come up that I actually have alone). They are also some of the founders of behavioral economics, which is preciously what "good" models use to come up with values of time (although there are other methods). While the above analysis was very parsimonious, you better believe that a similar rationale is indeed used for a lot of analysis, albeit at the aggregate level in most practical applications (of mode choice). The use of this on a disaggregated scale remains largely within academia.

To my understanding they didn't thrash rational choice modelling per se, but showed the irrational and uncertain elements exist. And incidentally, there are extreme value distributions that incorporate individual idiosyncrasies in the known (or rational) attributes. Heck, there's even decision making models to account for variability in unobserved attributes.

So if Malcom's analysis were to use this (but this is a rather large exercise) with that basis, it could account for the gal that walks 1 hour one way when she owns a vehicle and has "free" parking, or the guy that drives 2 blocks to work. In the end, this gets re-aggregated anyway for real decision level data - and that, given that a sample size is large enough quickly causing "convergence to the Gaussian" is what you would plan infrastructure or services around. You don't build HSR for the 6 people willing to pay 500 for a one way ticket.

Transportation is well within a physical realm. Rationality does exist within choices people make pertaining to aspects of it, especially because it's so often repeated and a derived demand anyway. However, what's rational or what isn't is in this case rather subjective and this CAN be accounted for in discrete choice theory. The models are by no means right, and they are not even necessarily useful. In some more extreme cases them being wrong might be impactful.

In most cases they are better than simply estimating a catchment population and applying some a priori mode share to it and hope for the best. I am picking on something that I saw about a month or so ago.

I guess I'm a bit surprised that people don't know the extent of, hmmm, well science that goes on, or at least could behind transportation modelling and therefore policy and infrastructure decisions.

This even pervades into books such as Human Transit (of which I know there are some fans of on here).

I apologize for the story.

Yup, read "Black Swan" by Taleb and "Thinking Fast and Slow" by Tversky and Kahneman. Both great, great books.

When it comes to transportation behavior, people don't wake up and think "I am going to make a cost/benefit analysis of each of my trips today and make the optimal choice." (This is what the four-step model used often in transportation modelling assumes). Transportation engineers have this belief that transportation decisions come down primarily to time and cost, and have no idea about anything else that affects a travel decision. In fact, the activity characteristics (time, time flexibility, duration, duration flexibility, regularity, frequency, location, location flexibility, necessity, mode availability, activity cost) are just as important to a travel decision as cost and time (what I call trip duration), but are never considered.

Here is the problem with every single transportation model: it is never tested. Take any TIA ever done. Do people go back and test out any of the assumptions put into the model? No. Do they count peak trips post-development? No. Do they do so for 20 year timelines? No. It is pseudo-science.

Travel decisions seem irrational to transportation engineers, because all they consider is time and cost. They have no consideration when it comes to the actual activity that the trip is accessing. The woman who takes the bus to Walmart, and walks home with arms full of groceries is completely inexplicable to a transportation engineer, but only because they don't understand the full decision making process behind travel.

Why this is relevant to this conversation: frequency matters less to the girl who takes a streetcar to go get groceries, because she can control the time when she gets groceries, or can move the activity to a different day. She can work the activity schedule around the transit schedule. I agree frequency is important, but it is more important for certain trips than others. And we shouldn't assign a standard dollar value of time to all trips. They vary greatly in their necessity.

[sim]

Originally Posted by fusili

Yup, read "Black Swan" by Taleb and "Thinking Fast and Slow" by Tversky and Kahneman. Both great, great books.

When it comes to transportation behavior, people don't wake up and think "I am going to make a cost/benefit analysis of each of my trips today and make the optimal choice." (This is what the four-step model used often in transportation modelling assumes). Transportation engineers have this belief that transportation decisions come down primarily to time and cost, and have no idea about anything else that affects a travel decision. In fact, the activity characteristics (time, time flexibility, duration, duration flexibility, regularity, frequency, location, location flexibility, necessity, mode availability, activity cost) are just as important to a travel decision as cost and time (what I call trip duration), but are never considered.

Here is the problem with every single transportation model: it is never tested. Take any TIA ever done. Do people go back and test out any of the assumptions put into the model? No. Do they count peak trips post-development? No. Do they do so for 20 year timelines? No. It is pseudo-science.

Travel decisions seem irrational to transportation engineers, because all they consider is time and cost. They have no consideration when it comes to the actual activity that the trip is accessing. The woman who takes the bus to Walmart, and walks home with arms full of groceries is completely inexplicable to a transportation engineer, but only because they don't understand the full decision making process behind travel.

Why this is relevant to this conversation: frequency matters less to the girl who takes a streetcar to go get groceries, because she can control the time when she gets groceries, or can move the activity to a different day. She can work the activity schedule around the transit schedule. I agree frequency is important, but it is more important for certain trips than others. And we shouldn't assign a standard dollar value of time to all trips. They vary greatly in their necessity.

Again, I'm not sure what you are criticizing and I by and large agree with you. The 4 step model doesn't really assume that at all though, because it isn't based on behavioral economics (the mode share on an aggregate scale might count). But to be clear, time and money are the two biggest factors in decision making when it comes to mode or route choice.

I guess you are having a go at transportation engineers and I'm with you, but I think you are also looking at a subset (albeit probably a majority especially on this continent). TIAs are usually based off of ITE trip generation rates or direct observations and then projected and have effectively zero to do with time or money - only linear statistical correalation. So I think you're confusing some things here. For the record I believe TIAs as are should go the way of the dinosaur.

Believe it or not, frequency can and often is modelled. Depending on your data, even comfort can be. Reliability as well. Heck, the difference in perceived time between stop and go traffic or on average slower traffic with an equal trip time can be evaluated. This largely comes from behavioural economics, and it precisely accounts for what you are saying isn't accounted for. The difference is that this largely remains academic and is admittedly hard (given that you do have these engineers still running the practical show, and that they are data intensive and take time) to see it be implemented in practice.

And yes, the next step is activity based modelling which is generally just an extension of what I've said but accounting also for the reason a trip is made to begin with.

People do wake up and do a quasi cost/benefit analysis for every essentially every decision they make. You do it, I do it, we all do it. They make it based on what is optimal for them not some prescribed optimum. From the analyst's point of view, its obviously impossible to account for all of these. Again, it takes data and there are different ways to obtain this - revealed or stated preference.

Values of time ARE determined for different types of trips, different modes, different whatever. The aggregate is generally used so some investment decision can reasonably be made. These can of course be aggregated to different levels, and that's why I've stated on here several times that walking and waiting are not to be given the same weight as in-vehicle times, or that the value of reliability is not to be overlooked.

Strangely, this is just another form of econometrics, which Taleb so adamantly disparages, even though it aligns with behavioural economics (or can, at least in the transportation world).

In sum, I agree with you that accounting for only time and money will lead you astray, however these remain the largest factors. The old guard isn't even really accounting for these and anyone familiar with the type of modelling I keep referring to is aware that the two factors aren't the only ones that are relevant - and to varying degrees, this can be accounted for. These can be used to find varying Values of Time.

[fusili]

Originally Posted by sim

Again, I'm not sure what you are criticizing and I by and large agree with you. The 4 step model doesn't really assume that at all though, because it isn't based on behavioral economics (the mode share on an aggregate scale might count). But to be clear, time and money are the two biggest factors in decision making when it comes to mode or route choice.


People do wake up and do a quasi cost/benefit analysis for every essentially every decision they make. You do it, I do it, we all do it. They make it based on what is optimal for them not some prescribed optimum. From the analyst's point of view, its obviously impossible to account for all of these. Again, it takes data and there are different ways to obtain this - revealed or stated preference.

Values of time ARE determined for different types of trips, different modes, different whatever. The aggregate is generally used so some investment decision can reasonably be made. These can of course be aggregated to different levels, and that's why I've stated on here several times that walking and waiting are not to be given the same weight as in-vehicle times, or that the value of reliability is not to be overlooked.

This is what I disagree with. Time and money are not the primary determinants of travel behavior. Activity pattern is. How a household delegates activities, distributes resources and establishes priorities is much more important than time or cost. The guy who drives to work every single day, spends crazy $$$$ on parking and spends more time in traffic than using transit can only be understood if we also understand he's got a kid that might get sick at school and needs a ride home at a moment's notice. Understanding how this pattern is created is fundamental. Someone will drive to an activity that transit is faster and cheaper to do, just because they don't do it very often and couldn't be bothered to look into transit availability.

And people don't make quasi-cost/benefit analysis for every decision they make. They generally create a activity and travel pattern early on, and then exhibit incredible inertia once that pattern is established. Nobody seriously wakes up each morning and calculates the time and cost of driving vs taking the bus. They just take the bus because they made that decision awhile ago. Housing changes and major job changes will shake up an activity pattern, but once it is set, it is difficult to change.

[MalcolmTucker]

Modeling according to fusili :


Source: http://coldcallmemaybe.tumblr.com/