xkcd

Critics of transit investment – especially rail investment – frequently cite a failure to achieve a budgeted ridership estimate as evidence of the ineptitude or corruption of the agencies planning the lines in question.  While I never wish to discourage due criticism, ridership estimates are constructed via theoretical models, and critiquing a model for being wrong is tautological, akin to critiquing a human for being mortal. Frustrated at popular confusion over the nature of modeling, I thought I’d write a post on the limitations and capabilities of models.  To my mind there are four main points:

More after the jump…

(1. Ridership models attempt to quantify what would otherwise be qualitative phenomena; they translate dynamic human behaviors into static numerical inputs. Any number of input factors (endogenous variables) may be included in the construction of the model, such as tolerable walking distance, seasonal weather patterns, political affiliations, other transit connections, demographics, while other inputs are excluded (exogenous variables), such as unforeseen economic recessions.  In short, modeling is all about throwing numbers at sentiments in search of rigor.  (For philosophical problems with this approach, see Nancy Cartwright’s “The Vanity of Rigour in Economics”)

(2.  Models aggregate particular behaviors upwards to create generalized assumptions. Such “bottom-up” science is inductive, and it has long been recognized that such reasoning is useful but tricky.  You may see 1,000 black cats and conclude that all cats are black, only to see a white one.  Similarly, all transit riders are willing to walk ½ a mile, until they aren’t.  Etc etc…

(3.  Simpler models are better.  One tricky feature of models is that they are better at sketching a picture than painting it.   The more variables one includes, the more sensitive (i.e. touchy and error-prone) the model becomes.  Thus modelers try to select only the core components driving behavior.  Never let anyone impress you by boasting about how complex their model is.

(4.  Models don’t do any real work; rather, they merely actualize the assumptions they contain. The incredible ability of models to describe system dynamics far exceeds human capacity, but the inputs the models contain are usually entirely human-derived.  A model is thus like a psychiatrist; you litter it with anecdotes and it tells you what they mean.  But they don’t give you any information about the veracity of the anecdotes themselves.

Despite these many flaws, models nevertheless tend to outperform human judgment alone.  While they may never be right, they’re almost always close.  Models do a wonderful job of establishing baseline approximations from which to tweak, experiment, and innovate.  When we use predictive models to aid the formation of transportation policy, we should only do so with a clear understanding of their limitations. Variances of 10-20% from modeled estimates are par for the course, and patience should be afforded to transit agencies when their models either under- or (more often) overestimate ridership demand.  The real success of a transit service is its relative and ongoing performance once its baseline is empirically (rather than theoretically) established, and by this criterion Link is doing very well.

115 Replies to “Ridership Modeling and Fallibility”

  1. Zach’s last sentence pretty much sums up a models usefulness. Managers are held to their predictions of actual performance as compared to their predictions(model)when selling an idea to their boss, the board, or the voters.
    A good comparison to public transit was my experience running an assembly line. If you wanted to buy a machine or re-arrange the flow of work, the capital investment had to gain efficiency or total output over what you were currently doing, say over X years to have a payback. You were measured by how close you met your prediction.
    I don’t know why, but public transit investment seems to overlook this basic approach to spending when deciding to build things that fail the test – either prediction or outcome.
    I’m not up to a cat-fight this morning over Link, BRT, or bus, but examples abound over costs and benefits. Oran did the math on the Mercer Is. parking spaces. $96,000 each? C,mon, that’s not a good investment. Managers and supporters then scramble to create the ‘soft’ benefit list, like urban renewal opportunities, congestion relief, or TOD, or even worse, the anecdotal picture of … “last week I saw umpteen people doing…”, or “were building a system to last 100 years so….”. Money has present value, and must be spent wisely, otherwise you run out of it, or the support of those willing to give you more.
    That’s were the model is useful, and if you missed your prediction, re-calibrate the model and do better next time – IF YOU STILL HAVE A JOB.

  2. Thanks for the well thought out post. I’m sure the folks who modeled Link grit their teeth anytime they read or hear comparisons of projected to actual ridership. Frequently these comparisons leave out the economic conditions or other input variables (which are uncontrollable) which have resulted in slightly lower ridership than projected. As the system (and ridership) grows this criticism about projected ridership will evaporate.

  3. Also, ST’s ridership forecast for light rail does not incorporate tolling, which can boost the ridership dramactially with right rates, so the future ridership is likely to be much higher than the forecast. I expect the regionwide tolling will happen in the near future, starting with 520 bridge and HW99 downtown tunnel (already done-deal). On the horizon, it’s inevitable to toll the other bridge (I-90) to pay for the $2b funding shortage of 520 bridge replacement, and the WSDOT is studying partial tolling (HOT) on the entire I-405, possibly in 2014. They could significantly increase ridership of the East Link.

    1. You think in an area with libertarian leanings and a notorious initiative process that “excessive” tolls are going to be tolerated by taxpayers? Tolls as you seem to be suggesting that are used as “punitive” measures to force people out of cars will have the unintended consequence of bringing the wrath of taxpaying car drivers down on transit projects. Tolls should only really be for the purpose of recouping bond debt obligations expended for the construction or re-furbishing of that structure.

      I think it is enough that natural economic forces will make car ownership and operation untenable for significant numbers of people. That coupled with the highway infrastructure only capable of supporting a fraction of current population and that the area’s population is forecast to continue growing will further create incentive for people to seek alternatives to driving.

      My strong advice to transit enthusiasts is to not make war on the car because people will make choices towards transit without coercive intrusion by government. It isn’t the governments fault that oil will spike into triple digit pricing. The market will do that on its own. And oil that is $150/bbl translates to nearly $5/gal gas. Which translates into unaffordable driving. That will affect the desirability of suburbs, the demand for walkable neighborhoods, and the value of cities and neighborhoods with excellent transit infrastructure.

      1. Subsidies removed, oil is already that expensive and more. The problem is that auto-oriented suburban development, you’ve already created an economy upon heavily subsidized industries and facilities. People will still demand cars because of this. It’s not government “intrusion” we support, but merely policies that go the other way.

      2. Tolling in excess of the actual bond debt will be perceived as Government intrusion whether you like it or not.

        Just as we transit enthusiast get all bent out of shape when transit is attacked by politicians, so too will politicians who cannot fathom life without dependency on a car or who have been bought by cement, road builders, and other development types will gleefully scuttle transit plans.

        Like I said, let sleeping dogs lie (in more ways than one). As was shown in the last peak oil price spike, there is not much the market can do to further mask the real cost of oil and as the price of gas rose, people rapidly shifted their transportation habits. We don’t have to do anything further and we should not be seen as doing anything active to punish car users. Instead, we need to be ready to welcome the throngs of new users into the post car era.

      3. Charles, you can classify lots of things as government intrusion on that basis. And I think whether it is intrusion or not is irrelevant. People will care if it’s credible and necessary. Neocons like government intrusion in the form of military spending. Liberals like government intrusion in the form of social spending. Excessive tolling can be justified outside of “forcing people out of their cars.” For one, gas tax funding will soon become insolvent.

      4. Sherwin, my argument is that any taxing scheme or any fee that is perceived as a taxing scheme (such as a toll) had better be perceived as having relevance and value to the payer of said tax or fee or there will be consequences. The citizens of this state are notorious for shirking taxes even for valuable and needed things. At one point license fees on cars were quite substantial and during that time it could be said that such fees more than covered the WSDOT budget and probably even contributed to the general fund. A taxpayer revolt rolled those fees back to very modest levels. I remember paying $400+ to renew my license tabs on a 6 year old car years ago.

        My argument is to anticipate political blow back. For example, Chicago is busy privatizing its public assets in a desperate bid to shore up it’s finances. The latest was selling the on street parking rights to a private company. For a one time billion dollar payment, this company now gets the rights to the revenue streams from all the parking meters in the city for the next 20+ years. And immediately the parking fees have quadrupled and will continue to rise. The citizens of Chicago are furious and the supposedly invincible Mayor Daley could be in for a very rude political reckoning for this. It’s not that their isn’t justification for raising parking rates but the manner in which it was done, the shocking increases and the feeling of having no say in the matter have converged to the slow boil among the citizenry.

        Such actions have political consequences. How do you propose to “sell” the benefit of system wide tolling on highways when users feel they pay enough in gas taxes, license fees, tire taxes, environmental mitigation fees, tonnage taxes, and the occasional parking ticket or speeding ticket to warrant their use of the roads?

        Let’s not make war on the car with no expectation of push back.

      5. Mayor Daley happily retired recently.

        No political blowback, even though the parking meter sale had a lot of bad aspects to it.

  4. “The real success of a transit service is its relative and ongoing performance once its baseline is empirically (rather than theoretically) established, and by this criterion Link is doing very well.”

    This is ridiculous. The success or failure of a transit service is how much it costs for what it provides, compared to other modes of transportation.

    Compared to BRT or even cars on highways, LINK is a financial disaster and a failure by any cost-effectiveness measure.

    In other words, LINK is stupdily expensive for the tiny number of people it moves.

    Ridership projections are irrelevant.

    But, this article does bring up one decent point: ST has no idea how many people will ride Link in 2016, let alone 2030. Those ridership projections for years out in the future are useless. I always find it amusing when posters here quote ST’s future ridership projections as gospel — as if they are absolute fact, and not just guesses. As in “In 2016, when the U Link opens, Link will average 117,000 boardings per weekday!” Uh, huh. You think that projection will not change at all between now and 2016?

    1. One of the simplest of models is that things will continue in the way they have in the past. I believe there was a NYT article on this a year or two ago, and that happens with 95 percent confidence.

      So, given, LINK built a route which was designed to be the most heavily used in the whole system. This is not some “test route” to demo the technology, but the breadwinner. If it failed to meet projections, then you can probably say that any expansion of the system will result in it failing by the same amount.

      Unfortunately for LINK designers by choosing to build the most heavily trafficked route first, they can no longer use the argument about “critical mass” or having a “full system”.

      Another thing I find bewildering is the inability of LINK planners (or STB readers) to use BRT ridership as input to their models, and also as competitive routing. For example, the 577/578 bus that goes direct from Federal Way to Seattle and takes 35 minutes, versus the LINK system that, when built, will take people through a tour of South Seattle and take 50 minutes or more!

      So, it’s not important what the models said, but that they should now be rebuilt based on real life data and the rest of the system reprojected.

      1. So, given, LINK built a route which was designed to be the most heavily used in the whole system.

        What?

        Northgate to downtown is projected to be the most heavily used in the system. The RV was built first for technical and financial reasons, not because it had the highest ridership.

      4. JB,

        The most heavily travelled route in the Link system will be the downtown Seattle to UW/Northgate route – ST did not build the most heavily travelled route first.

        Your post is totally inaccurate.

    2. Norman, there is a difference between “useless” and “wrong”. Of course ULink projections are wrong, but they are closer than could be expected by human intuition alone. All good projections are both useful and wrong. The figures give us a shorthand from which to frame debates, and no one in their right mind can think that ULink won’t dramatically or even exponentially increase ridership.

      1. Anyone that thinks a model can be “wrong” doesn’t understand models. Models are always “right” give your inputs and assumptions. They are always “wrong” when it comes to reality but as far as the math is concerned they are right.

      2. Maybe not necessarily “closer than could be expected by human intuition alone” but more likely to be closer to correct than human intuition alone.

    3. Huh?

      LR is cheaper than BRT when O&M costs are considered. The lower total cost of LR is one of the reasons LR gets built – along with other benefits like reliability, capacity, speed, etc.

      Yes, if you built LR in purely rural areas with very low ridership then LR might not be as cheap as a bus, but that’s not when ST is doing. Nor are they proposing to do that.

      1. “if you built LR in purely rural areas with very low ridership then LR might not be as cheap as a bus, but that’s not when ST is doing. Nor are they proposing to do that.”

        I’m a big fan of light rail and a daily rider. But ST certainly plans to put LR in places that it doesn’t belong and can’t possibly be justified by ridership and benefits. Maybe not in purely rural areas, but durn close. I’m of the distict opinion that LR investments north of Northgate (to Mill Creek P&R), south of Sea-Tac (to N. Federal Way), and east of Bellevue (Overlake area) are a terrible waste of money that – but for political deals necessitated to get the urban areas served – should never have happened.

      2. There’s significant growth expected in the Bel-Red corridor; I’d hesitate to call sending Link towards MS that way a waste.

        If we had feasible commuter rail in Snohomish, perhaps we wouldn’t need to send light rail that way. But the ROW doesn’t exist to serve the existing commuter patterns.

      3. If it were only going to Ash Way P&R (what I assume you mean by Mill Creek P&R) and North Federal Way, you’d be correct, those would be bad places for light rail. But it’s going to Shoreline, Mountlake Terrace, and Lynnwood P&R in ST2 (it’s not even getting to Ash Way in this phase) to the north of Northgate, not just Mill Creek. Lynnwood P&R has a pretty enormous number of commuters coming through it every day, it looks like it has up to 20 express buses to Downtown Seattle per hour coming through at the peak hour in the morning, plus tons of express buses to the U District and tons of local buses to all parts of Snohomish County, so it certainly looks like a great place for light rail. Shoreline and Mountlake Terrace, while not bringing nearly as many people as Lynnwood TC, will still bring plenty. Assuming it will be extended north beyond there, it will stop at Alderwood Mall on the way to Ash Way, and stop at Ash Way on the way to the huge destination of Everett.
        To the south, it is stopping at North Federal Way because it is on its way to the huge destinations of central Federal Way and eventually Tacoma in the future.
        To the east, are you kidding about Overlake? It is a huge jobs center with tens of thousands of Microsoft and other tech company employees, definitely not close to rural. And its on its way to a future extension to Downtown Redmond, another major urban center.

      4. I readily agree there are large numbers of commuter trips between Seattle and places like Lynnwood, Overlake/Microsoft, etc. But surely you don’t suggest that the massive – we’re talking many billions of dollars here, folks – capital costs and seriously high O&M associated with all day 10 minute headways, 20 hours, 7 days a week are justified to serve even large-ish numbers of commuters. The commutes to/from Lynnwood and Overlake are very concentrated in one-way peak periods.

      5. This is not true. BRT operating costs are no higher than Link light rail operating costs. And the capital costs of BRT are a fraction of what LINK light rail cost, and is costing, to build.

      6. Norman,

        Honest question: How do you feel about megaprojects in general? I’m thinking of things like the Alaskan Way Viaduct replacement here, the Big Dig in Boston, etc.

        I’m interested to hear both whether you think megaprojects are ever justified, and if you do, what criteria you think they should be judged on.

      7. I think the bored tunnel is a huge waste of money. They should just rebulid the viaduct, or build a new viaduct, which would give more capacity for less money: cost-effectiveness.

    4. The success or failure of a transit service is how much it costs for what it provides, compared to other modes of transportation.

      No, because transportation options don’t exist in a vacuum. You need to take the transportation infrastructure as a whole. The comparison isn’t between light rail and other modes, it’s between a system including light rail and a system without.

      The point of building Link isn’t to compete with cars. It’s to enhance our regional transportation infrastructure.

      1. You have to consider the cost and the effectiveness. Link’s cost is so insanely high, that it is the very definition of boondoggle. We could have moved the same number of poeple on buses for a fraction of what Link light rail cost.

      2. Norman, you’re still fallaciously pitting trains one-on-one against other forms of transportation. The proper analysis can only be done on the transportation system as a whole, including all modes in whatever form present: car, commuter rail, light rail, vanpool, bus, walk, bike, unicycle… You can’t single out Link vs. buses without studying the entire system as it would exist with Link versus an entire system with additional bus service. That means you need to define what that hypothetical bus-oriented system would actually look like. After all, we didn’t vote to “build some trains,” we voted on specific transportation packages.

        I argue that if you were to analyze the possible transportation systems we could have built that use buses to address the same needs we have chosen to address with Link, you will find that the bus-oriented systems have no cost or convenience advantage, and are on the whole inferior options to systems that have a light or metro rail component. I’m not arguing that our current Link option is the most efficient system we could have possibly built.

      3. That is just wrong. I have many times written that a few SWIFT-type bus routes, one from downtown to SeaTac, like the 194; a couple down MLK Jr Way, one an express and one a local route, with some buses continuing on to SeaTac; and continuing the old 174 route with new buses, and perhaps some express 174 buses, would have done everything Lihk does at a fraction of the cost.

        SWIFT cost $30 million for 17 miles. Link cost $2.6 billion for 15.7 miles. So, on a per-mile basis, Link cost about $166M/ mile, and SWIFT cost about $1.76M /mile, or Link cost about 94 times as much per mile as SWIFT.

        Let’s say you need twice as many buses on my proposed routes as SWIFT is using, due to higher ridership on those routes compared to the current SWIFT route, and to provide Link-type headways. So, assume $3.5M/mile for those SWIFT-type routes instead of Link, and 5-minute headways in peak hours.

        Putting in 4 SWIFT-Type bus routes with 5-minute headways on each, instead of Link, would have cost around $200 million. That is about 7.7% of what Link cost. Or, a saving of about $2.4 BILLION for taxpayers.

        Operating costs for BRT and LRT are similar.

      4. “SWIFT cost $30 million for 17 miles. Link cost $2.6 billion for 15.7 miles. So, on a per-mile basis, Link cost about $166M/ mile, and SWIFT cost about $1.76M /mile, or Link cost about 94 times as much per mile as SWIFT.”

        So what? The return on the investment isn’t measured in platform-miles, it’s measured on the system as a whole.

        “Let’s say you need twice as many buses on my proposed routes as SWIFT is using, due to higher ridership on those routes compared to the current SWIFT route, and to provide Link-type headways. So, assume $3.5M/mile for those SWIFT-type routes instead of Link, and 5-minute headways in peak hours.”

        You can’t simply say “let’s double the number of buses.” You are completely ignoring the characteristics of SWIFT-style BRT as a transit mode vs. rail.

        “Putting in 4 SWIFT-Type bus routes with 5-minute headways on each, instead of Link, would have cost around $200 million. That is about 7.7% of what Link cost. Or, a saving of about $2.4 BILLION for taxpayers.”

        What’s the expansion cost of SWIFT versus Link? How do they complement each other? You are trying to ignore very important questions and turn a complicated analysis into a simple multiplication. Your first fallacy is assuming a linear return on investment for BRT. How do you know SWIFT-style BRT would scale linearly to fit the growth patterns we’re expecting for the central Puget Sound for the next 20 to 50 years?

        “Operating costs for BRT and LRT are similar.”

        Please back up this assertion. Start by defining what you mean by “operating costs.” How are they measured? Cost per passenger mile? Are you including the amortization of fixed costs? Are you accounting for potential future expansion, including the much lower marginal cost of adding capacity to Link than to SWIFT?

        Details, Norman. They are important.

      5. No, I am not including any amortization of fixed costs. If you did that light rail would be vastly more expensive than BRT, obviously.

      6. No, if you include amortization of road costs, highways astronomically more expensive than practically anything.

      7. SWIFT is simply not in the same corridor as Link. Different number of people, different density, different levels of congestion, et cetera.

    5. Compared to Link, BRT and cars and highways are financial disasters and a failure by any cost-effectiveness measures.

      In other words, highways are stupidly expensive for the tiny number of people they move.

      Sorry, Norman, facts against you again.

  5. Michael Ennis says voters should get what they were promised. And yet, Mr. Ennis is trying to deny voters what they were promised by trying to stop light rail construction from moving forward.

    The article is just disingenuous on its face.

    1. Voters were promised traffic reduction.

      Most people want mass transit because they think it will get other people off the road and out of their way.

      Does LINK do this?

      1. Of course, it doesn’t. And neither do buses, HOV lanes, more freeways, tunnels, or commuter rail. It’s called “induced demand”.

        We’ve replaced the carrying-capacity of the viaduct, but more cars have appeared to take the place of those people using other modes. Enabling the same number of cars to bypass downtown is a public safety emergency, dontcha know? (whereas removing the viaduct is not)

        BTW, Can you show where such a ridiculous promise was made?

      2. Buses absolutely take cars off the road. Whay do you think park-and-ride lots are for? Park you car, and get on a bus, instead of driving your car into Seattle.

      3. Yes, it is true that in order to use a bus one must leave one’s car. Nothing about that fact implies that increased bus service will lead to fewer total cars on the road. The bus service might attract primarily existing bus riders who simply shift routes, or more people might elect to work in the office rather than work from home, or unemployed people without cars might now have a transportation option to get to their jobs…

        You simply cannot extrapolate from one person shifting mode from car to bus to an entire region.

      4. What do you mean? Every car in a park and ride is a car that is not on the road. lol Take the Bellevue park and ride, for example. The cars in that lot never go over the I-90 bridge — they stay in that parking lot during the day. All those cars are cars that are taken of I-90!

        You dispute this?

      5. “What do you mean? Every car in a park and ride is a car that is not on the road.”

        The magic of large-scale transportation networks means that just because a car is in a park and ride does NOT mean that there’s one fewer car on the road.

        Say ST added a new Express bus across I-90, and it was a hit. 1,000 people a day left their cars at park and rides on the Eastside to take this new bus to work. Other people who aren’t served by this new bus see the roads clear up and adjust their commutes to take advantage of the 1,000 car deficit on I-90. Perhaps they shift from 520, they stop telecommuting as often, or they stop timeshifting their commutes because that made their daily routines very inconvenient. In response, demand shifts in other parts of the network, including SR520, where more people start taking their cars.

        “You dispute this?”

        I dispute your illogical extrapolation from individuals to populations. That’s just not how transportation planning works.

      6. I ride across the I-90 floating bridge round-tip twice a week in mid-day, and there is never any congestion. By your theory, people should see this, and immediately “take advantage” of the lack of congestion, and start driving across the I-90 floating bridge in mid day.

        But this never happens. Why not? What is wrong with your theory?

      8. What a waste! — all that empty space. Someone should spend all their day riding back and forth on that freeway, taking cellphone pics, and show the world how the voters didn’t get what they voted for.

        But did we ever actually vote for it?

      11. Nope, they’re just being “underutilized”, as the Magnolia Community Car Club likes to describe the BAT lanes.

      12. Our freeway system doesn’t have a capacity problem. In fact we have a huge amount of excess capacity. What we have is a peak capacity problem. Light rail sounds like a good solution until you realize that off peak you’ve created an extraordinarily expensive system that has the same flaw.

      13. Our freeway system doesn’t have a capacity problem. In fact we have a huge amount of excess capacity. What we have is a peak capacity problem. Light rail sounds like a good solution until you realize that off peak you’ve created an extraordinarily expensive system that has the same flaw.

      14. Actually, light rail is much cheaper than extra freeway lanes after the first (approximately) four lanes.

        Which is why it’s such a good solution for high peak demand. The trouble is that you’re not using full cost accounting on the fantastically expensive roads, because “somebody else” is paying for them.

      15. Voters were promised a travel option other than being stuck in traffic. What most people “think” is often wrong.

        ST is right to stick to their mandate of providing a travel option.

      16. Most people want mass transit because they think it will get other people off the road and out of their way.

        BS. From where did you divine this assessment?

        I want mass transit because I don’t want to pay the cost of owning a car. So far not owning a car has worked out pretty well for me.

      17. “I don’t want to pay the cost of owning a car.” I bet you would not pay the full cost of using transit, either, if you had to pay the actual cost, including capital cost, of your trips. You like transit because taxpayers are paying almost all the cost of your trip, while you would have to pay the full cost of buying and operating a car, yourself.

      18. Norman,

        Your statement that “you would have to pay the full cost of buying and operating a car” is misleading, since it neglects the cost of roads and parking.

        Here’s a copy of the WSDOT 16-year operating budget (1). From 2009 to 2011, WSDOT will spend $494.4 million in operating costs, and $3,873.8 in capital costs on highways. During the same period, they will spend $158.9 million in operating costs and $98.4 in capital costs on public transportation and rail.

        According to (2), Washington has about 4.4 million licensed drivers. Thus, the WSDOT costs represent a subsidy per driver of about $993/year, or $82/month.

        Note that I am making no claims about public transit here. All I’m attempting to show is that, if you believe these numbers, then it’s incorrect to claim that drivers are unsubsidized.

        Do you disagree with any of these numbers? If you do, please say which ones.

        (1) http://www.wsdot.wa.gov/NR/rdonlyres/D63C71DA-C88F-4D9A-A577-CB3D2824B89D/0/Proformasforenactedbudget6109.pdf

        (2) http://www.statemaster.com/graph/trn_lic_dri_tot_num-transportation-licensed-drivers-total-number

      19. Just to emphasize: I am not making any claims, positive or negative, about public transit. I’m not claiming that public transit is unsubsidized (which clearly isn’t true), nor am I claiming that the only subsidies for public transportation come from the WSDOT (which clearly isn’t true), nor am I claiming that driving is more heavily subsidized on a per-user basis than public transit (which may be true, but I don’t actually know.)

        A possible conclusion from this data is that people simply aren’t willing to pay the full cost of transportation by any means. This would mean that, without subsidies, people would simply travel a lot less. Thus, if we think that mobility is a social good, then the most appropriate public policy would be to subsidize some form of transportation.

      20. Norman, my tax dollars are already paying to subsidize everyone else’s car journeys. I’m also paying in terms of degraded environment from everyone else’s use of their cars.

      21. WSDOT gets the revenue from the state and federal gas tax, which drivers pay. If a driver pays that money to WSDOT, how is spending that money “subsidizing” that driver? WSDOT gets their revenues from drivers!

        How is the cost of buying a car subsidized? Other than “cash for clunkers”, which I thought was a stupid idea. I paid for my car — taxpayers did not help pay for my car.

        But taxpayers paid the entire cost of Link light rail cars. People who ride in Link cars don’t pay even a fraction of the cost of those cars, with their fares — their fares pay only a small portion of the operating cost of Link. Fares pay zero part of the cost of buying those Link cars, and building the Link system.

      22. “WSDOT gets the revenue from the state and federal gas tax, which drivers pay.”

        As has been shown repeatedly, gas taxes are insufficient to pay for roads. In 2008, $9 billion was transferred from the general fund to the Highway Trust Fund. That right there blows away your gas tax argument.

        “How is the cost of buying a car subsidized?”

        Nobody said it was. But if you want to do a thorough analysis, everyone paid for the impact the manufacture of that car made on the environment. That’s true of pretty much anything our planet makes.

        “But taxpayers paid the entire cost of Link light rail cars.”

        And taxpayers are deriving the benefits—even the ones who aren’t taking Link. That’s the point of infrastructure. It increases mobility and makes the local economy more efficient.

        “People who ride in Link cars don’t pay even a fraction of the cost of those cars, with their fares — their fares pay only a small portion of the operating cost of Link.”

        Ah, the farebox recovery dead horse.

        It’s as unjust to require Link to recover its entire capital and operating costs through farebox recovery as it is to require highways to be paid for entirely by user fee when we all directly or indirectly benefit from these projects.

        Without a decent highway network, this region wouldn’t be attractive to live or work in, and there would be no revenue generated here. With an inefficient highway network, we’re just throwing money down the drain. The same can be said about all infrastructure.

        At least we got to vote on Link.

      23. “As has been shown repeatedly, gas taxes are insufficient to pay for roads. In 2008, $9 billion was transferred from the general fund to the Highway Trust Fund. That right there blows away your gas tax argument.”

        Hate to point out the truth to you, but that was $8 billion, and it was just a repayment to the highway trust fund of the $8 billion that was transferred OUT OF the highway trust fund to the general fund in 1998 to reduce the federal deficit:

        http://www.house.gov/list/press/wa02_larsen/PR_072308_Restoring_Highway_Trust_Fund.shtml

        From Congressman Rick Larsen: “Today’s bill restores $8.017 billion to the Highway Trust Fund that was transferred from the Highway Trust Fund to the General Fund in 1998.”

        Does that clarify things at all?

        By the way, 25% of the highway trust fund revenues are now being spent on transit, and not on roads.

        People who don’t ride Link are not deriving any benefit from it — just ask those people, and they will tell you.

      24. Hate to point this out to you, but state highway budgets are largely not funded by the gas tax; the federal highway budget was largely not funded by the gas tax in the crucial prior-to-1950 era; and local highway budgets are rarely funded by the gas tax at all.

        And you can claim that people not riding Link are not deriving any benefit from it. You’d be wrong, of course. I could also claim that anyone not driving across a particular road is not deriving any benefit from it. Would I be right? Depends on the road, actually.

      25. “Voters were promised traffic reduction.”

        This always comes up in large transit projects. “The transit agency promised it would reduce traffic/congestion!” Transit agencies don’t promise this because it’s patently false. A few overenthusiastic people say it, and people get the impression the transit agency promised it. Sometimes the agency’s early brocures mention it, but the fallacy is quickly pointed out and the brochures are changed. Still, people think a minor comment on an early brochure (or by people not affiliated with the agency) is the same as a solemn promise on election day.

        The second I-90 bridge was uncongested for ONE WEEK while drivers adjusted their travel patterns, then it was back to congested as usual. If we add a third bridge, the same thing will happen. Only if you added three more bridges at once would demand be so saturated it couldn’t fill them.

        The purpose of adding rapid transit is not to reduce congestion but to provide a bypass through it. If you’re willing to accept the limitations of transit, you can avoid the congestion. You can’t reduce congestion except via tolls and such.

  6. Primary xkcd violation: posting a xkcd comic without a mouseover.

    Secondary violation: not linking back to their page (it’s all they ask in their license).

    2. My apologies for the oversight. I think I had the caption in an earlier draft but reuploaded the image later and forgot.

  7. Can we get an inline fact check going on in these comments? I think moderators should be well equipped to add annotations in the middle of posts that contain inaccurate information. I wouldn’t recommend deleting on-topic but incorrect content, just adding a highlighted annotation next to it, much like how editor comments are inserted in print media.

    1. We rely on commenters to help. We simply don’t have the time to constantly argue with some specific commenters that don’t put a good faith effort into their comments.

  8. All I can say is, I envy anybody who’s young enough to think a year and a month and a half is long enough to know if anything is a success or a failure. Beware: early preoccupation with statistics predicting transit failure is a sure predictor of incurable addiction to transit board Public Comment periods in later life!

    Mark Dublin

    1. We know what Central Link cost to build. We have ST’s ridership projections for up to 2030. Even if Central Link meets those projections, it will be an incredible waste of billions of tax dollars. So, you don’t have to go by what Link is carrying now. Even using ST’s most-optimistic projections, the ridership will be very low, oompared to the cost.

      2. It is ST who moves the goal posts. And ST who sets the targets.

        Why would anyone accept ST’s opinion of what is a “success” and what is a “failure”? ST is the one who is building the system! lol

        How absurd to let the agency that is building a transit system determine if that system is a success or failure. You don’t see the foolishness in this?

      3. “It is ST who moves the goal posts. And ST who sets the targets.”

        No, you moved the goalposts when you said “Even if Central Link meets those projectsion, it will be an incredible waste.” You yourself said that this goalposts were not good enough.

        “How absurd to let the agency that is building a transit system determine if that system is a success or failure. You don’t see the foolishness in this?”

        It’s not absurd at all. ST put forth a system and a set of acceptance criteria. We accepted their criteria and agreed to hold them to it by voting Yes. We have independent auditors to ensure ST is performing an accurate and candid analysis of their satisfaction of the agreed-upon criteria.

        This exact arrangement exists in the private sector when dealing with Service Level Agreements (SLAs). Typically, a vendor will make an offer to contract for services with a given guarantee of reliability or performance. Note that it is the vendor who is offering both the service and the SLA. If the parties contract, the onus is usually on the vendor to ensure they meet the SLA, and the customer to ensure the vendor is accurately meeting their SLA. The customer probably (and certainly, if they are a publicly traded company) has an independent auditor that will periodically investigate the business relationship and ensure that both sides are correctly upholding the terms of the agreement.

      4. Not true. What ST said would be built at the time Central Link was voted on is not at all what we got. You dispute that?

      6. We voted for Sound Transit packages 1 and 2, and “Transit Now”.

        That’s more than I can say about certain multi-billion-dollar freeway projects.

      7. Voters were told that one light rail line can carry as many people as a 12-lane highway.

        Will any segment of Link ever carry as many people as a 12-lane highway?

        Hint: I-5, which is 10 lanes between Seattle and SeaTac, carries 425,000 people per day past any point on that highway. Link is currently averaging about 24,000 boardings per day, and only about 12,000 people per day past the mid-point on the line. The projection is that by 2030, Central Link will carry about twice that, or around 24,000 people per day past the mid- point on the line.

        Are voters getting what they were promised — as many people moved as a 12-lane highway? I would say that Link is carrying only about 3% of what a 10-lane I-5 is carrying, or only about 3% of what the public was told.

        It is easy to win elections when you out-spend your opponents by many times, and make promises which are patently absurd, and which will never be met.

      8. “Will any segment of Link ever carry as many people as a 12-lane highway?”

        It can, whether or not it will. (It probably will on special event days, or if Seattle gets the Olympics or something.) Do you dispute that? Or do you just want to move the goalposts?

      9. I’d say there is a damn good chance U-Link/North Link will approach that. I’m pretty sure that soon after opening it will need 4 car trains every 3 minutes for at least the morning and evening peaks.

  9. My Comp Sci Formal Models prof has a good quote:

    All models are wrong. Some models are useful.

  10. Aren’t ridership projections really beside the point? What’s important about Link are three things: TOD, giving those living within a half-mile of stations an alternative to commuting by car if their workplaces are pretty near stations, and environmental benefits. Based on those criteria Link’s a great success!

    1. That only makes some sense if cost is not a factor. We just have unlimited tax revenues, so how much any project costs is not important? This seems to be the theme on this site: “cost is no object!” “Don’t worry about the cost!”

      1. “That only makes some sense if cost is not a factor.”

        That’s not true. Quality of life improvements are not only valid if cost is no object. Like all goals, a value can be assigned and compared to the cost being paid, and a judgement made accordingly.

      3. “That only makes some sense if cost is not a factor.”

        ST’s costs are lower than average, so your argument lacks merit.

  11. Overall there’s another aspect of transit planning that I find ignorant in general. Let’s call it “reductio ad absurdum”.

    So, you take a highway full of cars and the transit planner looks at it and says — gee, if I put all those people on a train, it would be so much more efficient. Then he does his design (or more likely, gets people to actually build his scheme) and the ridership isn’t there. Does he reevaluate? No, he says, well, the problem is there isn’t enough density, so we’ll force people to move to where the stations are.

    You see the fallacies? A busy highway, looked at by the Reductionist, is seen as something “to be Fixed”. But if you realize that a full highway, is a Successful Highway, and the reason people are driving is because the like driving, that they want to be in low density sprawl, and have cars, and large homes, and a lawn, then trying to “optimize” it, by creating a failed rail system and a lot of onerous Government social agendas to make people live in condoes becomes pure folly!

    1. “full highway, is a Successful Highway, and the reason people are driving is because the like driving,”

      Doesn’t necessarily follow. Alternative modes may not exist, or be safe, or are too expensive. Ie. The 520 bridge is full of cars and yet it’s not possible to walk or bicycle, without going North around the lake, or crossing on the I-90 bridge.

      1. I think walking and bicycling are qualitatively different from Mass Transit.

        They are Personal Transit.

        If you can bicycle from Point A to Point B or walk as an alternative to a car, then it’s not an argument for Mass Transit.

        Take your case. Let’s say there were walking paths and biking paths across 520.

        All you’ve done then is offered new options in Personal Transit…and yes, you might see some car drivers convert to walkers, and bikers (more the later, I’d say).

        Those are 1 for 1 alternatives in my view.

      2. People cram onto the 520 bridge because it’s so successful it must be the fashionable place to be, not.

    2. The Successful Highway you mention doesn’t look so successful when you figure in all the externalities (e.g. land consumption, local air quality impacts, global co2 pollution, vehicle-related injuries and fatalities, noise pollution, water quality impacts, financial burden, physical inactivity).

      Ridership on the transit line that’s competing with the highway is an important indicator, but isn’t the end goal. Efficiency in transportation isn’t even necessarily the end goal. Highways are usually pretty efficient transportation-wise, at least until you get to a big enough city, but they also kill thousands of people and wreck the environment, so a holistic view points back to transit and nonmotorized as the efficient choice.

    3. Highways have incredible mobility benefits, but also incredible negative externalities. Yes, highways save a lot of time and facilitate quick movement from places within 10-300 miles. But on the flipside, vehicles as they are today generate incredible amounts of pollution and drive our country’s dependence on foreign oil. And even once we have electric cars, the automobile will still facilitate the development of sprawl (admit it, Kent would be much more pristine without all those arterials and subdivisions everywhere) and perhaps even continue to spur high rates of obesity.

    4. Strawmanning, Mr. Bailo? Or no, I see your error.

      I can think off the top of my head of many examples of planners considering a train, running a model and concluding that the ridership just isn’t there for a train.

      You’re suffering from a cherry-picked worldview because you aren’t a planning geek. You haven’t been reading all the plans and studies of agencies *including the ones which went nowhere*. Yes, indeed, there are plenty of studies which concluded that there simply wasn’t the ridership, but they aren’t very well known, are they?

    5. Fact is most people don’t give a damn about lawns. You can tell by how many people never USE their lawns.

      Some people do care, but there are plenty enough lawns for them; there’s an undersupply of “lawn-free” housing.

  12. Norman: “In 2016, when the U Link opens, Link will average 117,000 boardings per weekday!” Uh, huh. You think that projection will not change at all between now and 2016?

    Are you expecting the UW to close? Anything can happen, but it’s reasonable to assume the UW will still have 50,000 students and staff, and that the 71/72/73 express riders (who already chose a faster mode over the 70) will find Link an improvement. We can’t say exactly how many riders there will be, but we can make minimum and maximum estimates.

    Also, 2020 is a fairer comparison. 2016 is a severely incomplete construction phase, without the vital Brooklyn station for the north part of campus and the northern U-district. ST has said all along that Link won’t be at its best until it reaches Northgate.

    Dente: “I’m of the distict opinion that LR investments north of Northgate (to Mill Creek P&R), south of Sea-Tac (to N. Federal Way), and east of Bellevue (Overlake area) are a terrible waste of money that – but for political deals necessitated to get the urban areas served”

    First, those regions are paying for it on their own. (Except Shoreline, which is married to Seattle for better or worse.) Second, those segments are the ones that can be most easily chopped off if the local residents change their mind, without destroying the core of the system. I’m 50/50 on Link between Federal Way-Tacoma and Lynnwood-Everett. Yes, it’s more efficient long term, but maybe Swift-like BRT is a suitable compromise until demand for rail becomes overwhelming.

    Bailo: “the 577/578 bus that goes direct from Federal Way to Seattle and takes 35 minutes, versus the LINK system that, when built, will take people through a tour of South Seattle and take 50 minutes or more!”

    Those buses will probably remain. The advantage of the 194 was only three or five minutes: within the margin of error, and Link gives more frequency and reliability. But the time difference is greater on the 57x and 59x. A limited-stop service can’t compete with an express service after a certain distance. Link’s main benefit in south KingCo will be in the mid-area, people going shorter distances from De Moines to Tacoma or Seattle or within the south county. But for even longer distances where you’d have to take two express buses (say from Des Moines to Northgate or Lynnwood or Bellevue), Link would probably be competitive again.

    Another purpose of Link is to encourage urban villages to grow at the stations. (Yes, buses could do this, but they didn’t, and Link is what we’re getting in any case.) That may seem far-fetched at Othello, Shoreline, Des Moines, but at least it provides the opportunity if the neighborhood is willing to take advantage of it. And the national trend is toward greater walkability and urban villages, not less, so maybe those neighborhoods will be grateful in 20-30 years that the stations were built now rather than later.

    1. ST has said all along that Link won’t be at its best until it reaches Northgate.

      Can we engrave that?

      So when the argument starts up again in 2016…

      2. You mean the argument is actually going to stop for 6 years? Hurray! We can get on to other topics.

      3. The Stadium-to-Northgate segment is clearly the backbone of the system, with many high-traffic pedestrian destinations, dense development going back a century, a tradition of transit ridership, and the tightest/most expensive parking. So yes, it’s the area that most needs rapid transit and will have the highest ridership. Both by people traveling within that corridor, and people going to those destinations from other parts of the line.

        The lament in 2016 will be the same as now. “Link is a little better, but it still covers only part of the highest-traffic corridor where it would be most useful.”

        So if that’s the primary ridership corridor, the second-level ones are Bellevue TC-Seattle, SeaTac-Seattle, and Lynnwood-Seattle. We can already see the SeaTac-Seattle segment, though it will get some increase when the north and east segments join onto it. (People who won’t make a 4-mile or 12-mile bus trip to the train, but will take the train if it goes all the way from their area to the airport, with at most one train-to-train transfer.)

  13. I’m amazed no one attempted to hijack the comment thread into a discussion of whether global warming is real, unless the moderators have been deleting them.

    1. I too was worried about that. There are parallels to be drawn regarding modeling and prediction, but to call that entire issue into question you’d have to doubt not only the models (which give themselves a pretty wide range of outcomes), but also a wealth of observational data that isn’t subject to the same limitations I’m talking about.

  14. I love blogs. You don’t have to be an expert on travel demand modeling but you can authoritatively lay out “four points of light” on the ambiguous notion of “modeling” that are the foundation for an argument that “it doesn’t matter how good a system is compared alternative investments, it only matters that it is better than yesterday.” Sounds like a motivational seminar for flunkies.

    1. The post wasn’t about the performance of rail vs. alternatives, but simply about the nature of models. The last point about Link was more of an aside. I’m not an expert on modeling, but I do have a master’s degree in economics that required extensive modeling projects, I’ve spoken at an academic conference about this very issue, and in the post I cited literature from a highly-respected philosopher of science. So we may not agree, and surely there’s much for me to learn, but I took this post seriously and did my level best to write responsibly about it. I do take offense to the insinuation that I’m merely a cheeto-chomping armchair amateur.

    2. What a low brow comment. Since you feel so free to criticize what are your transportation modeling credentials? I for one have taken a class on transportation modeling and I read this post before it went live and didn’t have any problems with it.

    3. For what it’s worth many of the same caveats about models Zach mentioned were covered in my undergraduate econometrics class.

  15. There are several problems with using models to estimate transit ridership. First, models produce projections, not forecasts. A projection is based on quantification and extrapolation of past behavior.

    If there is no experience with a particular transportation mode among a specific population, modelers either apply quantifications that have been seen in similar markets where that mode was introduced, or they try to infer how the population will react by estimating their sensitivity to new choices by surveying actual or expected reactions to changes in several variables, usually in-vehicle travel time, out-of-vehicle travel time and costs.

    Many variables like ride quality, reliability, land use changes, demographic changes etc. are difficult to quantify and are often left out of the models, which affects their accuracy.

    Also, often the planned system configuration used in the model is not what gets built and operated. For example, the modeler may have assumed that trains would run every 5 minutes, but the system operators actually decide to run trains every 15 minutes, instead. Or the planners may have assumed an operating speed of 50 mph on a particular segment, but due to cost overruns, the segment is built to allow only 25 mph speeds. Slower and more infrequent service can reduce ridership.

    Ridership projections also can be underestimated, but overestimation of ridership is a more common problem. There cases in which modelers have knowingly used assumptions that are unrealistic to increase projections, and some in which data may have been falsified. Most modelers, however, perform their work with integrity, are aware of the shortcomings of their work and are continually trying to improve the techniques.

    Project sponsors and supporters tend to use projections as forecasts and treat them as absolute truth, which gets them into trouble. In fact, the most appropriate use of models is to compare alternatives against a common set of assumptions.

Comments are closed.