Central Link failing to meet, meeting, or beating projections is a common topic when it comes to discussing the success (or failure) of the initial segment.  This discussion will only intensify as we move towards a possible 2016 Sound Transit 3 ballot measure.  In order to help the discussions be as fact based as possible I’ve put together a couple posts discussing the different projections for Central Link over time compared to reality, and thrown in some unscientific projections of my own just for visualization.  This first post will focus on short term projections gathered from Service Implementation Plans.

The original projections come from the 2009 and 2010 SIPs, both produced before the full Initial Segment was open.  As discussed multiple times due to the Great Recession (80,000 less jobs downtown than projected) and other factors Central Link did not meet its initial ridership projections.  After having a full years worth of data to analyze, in 2011 Sound Transit lowered their projections.  In response to the downward revision and Sound Transit’s initial numbers being off in the first place Martin had a great piece with Sound Transit planner Bob Harvey on the process for creating projections and why the projections were off.  It’s well worth a refresher read.  A more detailed explanation of individual factors was given in the Central Link – Initial Segment Before and After Study discussed last week.

The continued strong growth in the years since opening caused Sound Transit to revise their numbers back upwards in 2012.  And this year Link has beaten even those higher projections.  In August Link reached a yearly moving average of 27,925 weekday riders compared to Sound Transit’s projection of 27,900 at years end.  If Link can maintain its 10.7% year over year growth rate in the four months from September through December it should finish the year with an average of around 28,832 weekday boardings.  In the 2013 Service Implementation Plan Sound Transit projected 29,600 weekday boardings for 2014, meaning that on January 1st 2014 Link will already be over a third of the way to their year end goal of 1,700 additional riders.

Because I like data and graphs I put this all down into my ridership excel spreadsheet and created the below. Projections come from 2009, 2010, 2011, 2012, 2013 and 2014 Service Implementation Plans.  The 2009 SIP had the same projections as 2010 so it was removed.  The ridership data comes from Sound Transit’s monthly ridership reports covered here.


It’s not shown on the page but if Central Link can maintain the 10% growth its experienced since opening then in 2018 actual ridership will cross the 2009 and 2010 SIP projections.  That is of course completely hypothetical as basing 5 years of growth on 3 years of data is extremely shaky, not to mention the fact that in 2016 with the opening of the University Link Extension it all becomes academic anyway.

38 Replies to “Link Ridership v The Short Term Projections”

    1. Given how competitive this is with taxis from the airport, I suspect that at least the taxi traffic has gone down a bit.

      Not everyone is within reasonable distance to a link station though, so people will continue to get rides to and from the airport.

    2. Also have to consider how much worse traffic would be without it. Imagine if the Link disappeared overnight, some portion of that ridership would likely hop onto I-5 and then things would get really ugly.

    3. Airport ridership is spread throughout the day, unlike peak commute traffic. TIBS might have made a little dent in peak hour traffic, but park & rides have to be truly massive to make more than a dent. And then new people would just move further away to cheap housing because of the lack of traffic plus convenient free parking (induced demand). This is why some of us fight against park-and-rides and long distance expansions of Link.

    4. Some might say it’s something their PR dept will never admit, but they want freeway traffic to continue to be bad. It seems counterintuitive, but it’s bad for their business model if too many people give up their cars and take the train or bus into work. Why do you think they design lines that don’t take people out of their cars and get cars off the freeway? Some say they design lines that serve airport travelers, former bus riders, carless cap hill residents, carless uw students, etc. They people who take Link weren’t driving anyway. Some say they don’t want to solve a problem. Sure, they sold Central Link on the idea that it will take cars off the road, but that’s never what it was intended to do, some say.

    5. It doesn’t matter what the traffic is like coming and going to Sea-Tac. You can take the light rail and avoid the traffic.

  1. So….

    Despite that each and every curve shows an initial high rise, and then a leveling off, you create a new function, based on 25 additional riders above projected for one month, and project it straight up outward for several years.

    Is that what you want us to believe?

    1. What the hell are you talking about? Where are you getting this “extra 25 riders” nonsense?

      So far, link ridership has grown about 10% a year. They’re projecting it to continue to do so. That may well be wrong, but it’s hardly insane.

      1. More of a visualization than an actual projection, I have no training in these things at all, I’m just going on past performance.

        Over the past 12 months (ending in August) Link grew at an average rate of 10.7%. The 12 months before that it was 10.3. In other words Link’s growth rate is actually INCREASING.

        No, that obviously cannot continue forever, but considering the factors involved in the initial slow start (discussed here) I don’t think it impossible for the 10% to be sustained over the next couple of years.

        I think it more likely to slowly decrease to about 6 or 7% by the time University Link opens, but that is 100% completely and totally just a guess.

      2. In August Link reached a yearly moving average of 27,925 weekday riders compared to Sound Transit’s projection of 27,900 at years end. If Link can maintain…

        – 27,900

      3. It is supposed to average 27,900 by the end of the year, it’s already (as of August) 27,925. There are still 4 more months of potential growth to go. As mentioned in piece: If Link can maintain its 10.7% year over year growth rate in the four months from September through December it should finish the year with an average of around 28,832 weekday boardings.

      4. 27,925
        – 27,900

        John, you looked at this chart and accused its creator of “creating an new function” based on this particular month to generate future growth. This makes no sense, and doesn’t explain the projection they come up with. The projection has nothing to do with the 25 rider differential between previous projections, it is simply based on continued stable growth, as we’ve seen the last three years.

        Your interpretation of the process behind the projections was arbitrary, nonsensical, and innumerate.

  2. OK, maybe I am missing something here, but why does the line for the 2012 SIP start in mid-2010?

      1. totally bogus. We’d get thrown out of the room for presenting “historical” data as “projections”.

      2. Who’s doing that? The projections are the part of the line after the SIP was published. It’s not that complicated.

      3. So lazarus, are you arguing that the SIP and the ST budget should not adjust from year to year based upon historical data?

    1. Probably because that’s when the draft 2012 SIP was published. That said, why does the 2013 SIP line start at the same place?

    2. The SIPs contain the prior years actual ridership along with that year and future years estimates. Keep in mind that the 2014 SIP is being worked on now and if the pattern holds the final version will likely be released in December, so the ‘actual’ numbers it contains will be 2012.

      See page 111 of last years SIP to see what I’m talking about:

  3. Ridership projections after a project is finished are not nearly as important as the ridership projections that got a project built. The cost-benefit projections actually take money from taxpayers’ pockets or from other projects and direct it toward a particular project. Not as much turns on the projections done after a project is built, some adjustments to the operations maybe.

    1. It does allow you to tune your ridership model with quicker turnaround to real world feedback. I imagine the same general model and process is being used to generate ridership estimates for alternatives analysis in ST2 and will eventually be used in planning for ST3. It makes sense to continue to refine the process in an iterative way with comparison to actual results instead of waiting until ST2 lines open to see if your methodology from 2011 is close to reality.

    2. Well if you’re really looking for the ridership projections that got the project built, the answer would be 107,000 per day for a line from the U-District to Angle Lake (http://www.soundtransit.org/Documents/pdf/news/reports/Sound%20Move/199605_SoundMove_AppendixC_Benefits.pdf page C6). However, this isn’t comparable because the line was shortened after it was approved by voters, because Sound Transit grossly underestimated costs (the line from U-District to Angle Lake was projected to be $1.7 billion in 1995 dollars, but it is actually going to cost ~$5 billion). Once Link is extended to UW and Angle Lake by 2016, these ridership projections will probably be matched (although at a much higher cost than was anticipated originally), although we can’t really do accurate comparisons before that.

      As for just the initial segment between DT Seattle and SeaTac, I seem to remember that the initial projection (that was made before it was built), assuming no further expansion, was 47,500 for the year 2020. However this projection is largely irrelevant as Link will be expanded before then.

      1. Josh – We’ll never know on that 107k/2010 number because it assumed that the line would be built earlier. If we shift everything back to when its actually built and build a buffer for adoption — 107k/day for UW to Angle Lake is looking fairly realistic by 2020.

        Also $1.7B 1995 dollars are $2.53B current dollars for a closer comparable (which should include past and future dollars if we were doing year of expenditure.) That means the political guestimate came in at about 50% of costs… pretty much on par for big projects. That said — ST no longer deals in political guestimates — it deals in preliminary engineering estimates. Thus the likelyhood of current/future projects coming in on (or even *gasp* below) budget.

      2. kylek: 107 by 2020 is grossly unrealistic to expect by 2020.
        Central/Airport Link was pegged at 45,000 for 2020. Adding Angle Lake’s 3,000 and U-Links 2030 ridership projection of 39,000 still leaves you 20,000 Short. Even the unusual 10% gains of late still leave the current system 20% short of projection, so another 9,000 to makeup. In all, even if all the 2030 ridership folks start riding 10 years earlier, you’re down 29,000 – so let’s not get all giddy about some nice ‘catch up’ gains of late.

  4. My understanding, after reading the article by Martin to which you link, is that the predictions are based on a model that understates the ridership contribution of sporting events and aireport journeys. Am I misunderstanding something?

    If my understanding of the models is correct, it seems that the projections aren’t really directly comparable to the actual numbers. In particular, if I’m understanding things correctly, the fact that the numbers are more or less in sync right now is a result of cancelling bugs. Is there any reason to believe that these effects will continue to cancel each other?

    In any case, while it’s good to see that ST is getting better at projections, as yvrlutyens suggests, the more interesting question is whether this ability to make short term projections translates into better LT projections. Those are the numbers that drive the big decisions about capital expenditure.

  5. The Link projections originally assumed bus routes like the 101 and 150 would be truncated to connect to Link, only for such restructurings to be scrapped when political reality kicked in.

    I’ve been thinking about ways to overcome the political obstacles. First, in order for a transfer at Ranier Beach to get to Renton to be palatable, the connection needs to be quick, with minimal waiting, and a fast bus, once it comes. In particular, this means:
    1) Reliability. A guaranteed timed connection at Ranier Beach in the outbound direction. This means a Renton shuttle bus waiting for you right next to the Link station. Instead of leaving at arbitrary times on the clock, the shuttle would wait until a southbound Link train actually arrives and everyone who exited the train has had a chance to walk over to the bus, which would be clearly visible from the train.
    2) Frequency. This means a bus for every train all day, of if that’s not fiscally feasible, a bus for every train during the day and every other train during the evening. Of course, there would be a schedule and the guaranteed connection for every other train would still apply. For example, Metro would promise to you that if you are in Westlake Station by 10:03 PM and take the first Link train that comes, you would get a no-wait connection to Renton.
    3) Ease of connection. Since the whole point of the Renton shuttle bus would be to connect to Link, proximity of the connection is crucial. The bus must be situated as close to the train station as possible, yet facing in a direction so it doesn’t have to do a big loop-de-loop to maneuver out of the stop. Also, with the train station at Ranier Beach in the median, the stop in the inbound direction should be nearside to minimize the number of stoplights that people transferring to the train need to wait for. (Farside stops are better for thru-riders, but here, there are no thru-riders. Since near-side stops are better for people exiting the bus at that stop, the nearside stop should win).
    4) Speed. The Ranier Beach->Renton shuttle bus must be fast. At a minimum, this means following the 101’s routing. I would start with the 101’s routing, but straighten the route a bit within Renton – like eliminating the around the block detour into South Renton P&R, when it already serves another stop on a street, a mere 200 or so feet away. At any rate, taking the 106 route with stops every hundred feet is absolutely not acceptable.

    However, even if Metro promised to do all of the people, people would naturally be skeptical. After all, given Metro’s history of botching transfers, most people would assume that promises like these would be broken in favor of the same untimed-connection-to-half-hourly-bus that connections generally mean today. Such people would, very understandably, fight to maintain their one-seat ride.

    One way I can think of for Metro to prove to people that it is taking the idea of guaranteed, timed, fast connections seriously, is to try it out. Pick one day to operate the timed Renton Shuttle, but do it alongside the existing 101 so that people who don’t want to break with routine aren’t screwed and political opposition to losing the one-seat ride doesn’t scuttle whole effort. Yes, I know it’s expensive and somewhat wasteful, but politically, it’s the only way a Renton Shuttle trial like this could ever happen, and if it works, the ongoing savings would far outweigh the hours wasted on redundant service for one day. To encourage people to ride the shuttle, Metro could even offer free fares to the shuttle on the trial day so that people who don’t have Orca cards wouldn’t have to pay twice. They could also conduct on-board surveys regarding the satisfcation of people with the experience. If they can pull off the timed connection well, and let people see first-hand, how much better the bus-for-every-train experience is than the half-hourly 101 they currently get, a 101 truncation will be much easier politically to push through.

    Of course, the danger of the Renton Shuttle trial day is that Metro has one chance to get the guaranteed, fast, timed connection business right. If they botch it, it will be decades or more before Metro gets another chance to try it again.

    As an additional aside, waiting until 2016 or 2021 might also strengthen Metro’s case for a 101 truncation, in that people commuting from Renton to the UW, rather than downtown, would gain increased frequency, but without the cost of an extra connection.

    1. Even if the existing 106 route was used to connect Renton to Link, but at a higher frequency and with timed connections, it would still have about the same travel time from DT Seattle to Renton TC:
      Link takes 24 min from University St to Rainier Beach. During the midday, the 106 takes ~25 min to get from Rainier Beach to Renton TC. With a 2-minute timed connection, this would be a total travel time of 51 minutes, compared to a travel time of ~44 min on the 101 (which increases to 50 at peak). Once average wait times are taken into account (15 min for the 101, vs. 5 min for Link) it’s basically a wash.

      Really the best solution (as has already been stated on this blog) would simply be to extend some of the routes that terminate at Renton TC such as the 105, 169, 148) to Rainier Beach Station in the place of the 106: they would be staggered to provide service between Rainier Beach and Renton every 10 minutes. The local part of the 101 would be served by a coverage route also going from Rainier Beach to Renton via MLK every 30 min, with timed connections to selected Link trains. This might require a few extra service hours, but it should be worth it.

    2. “a bus for every train all day”

      This is exactly what the Bellevue Vision recommends for some of its routes. Now that one government has aspired to it, it may be easier to get other governments to do the same.

      Did the Link projections really assume the 101 and 106 would be truncated, or is this a rumor that’s spreading? If it’s true, it would be interesting to know when and how Metro changed its mind.

      “The Ranier Beach->Renton shuttle bus must be fast. At a minimum, this means following the 101′s routing.”

      The other thing is, you can’t just delete the 101’s MLK segment without replacing it because there are a significant number of local riders and apartments, and the topography and lack of roads pretty much prevent them from walking to any other bus route.

  6. Mathew, Et All: Your last statement is troubling.
    “… after opening of the University Link Extension it all becomes academic anyway.” WHY?
    If we ignore history, are we destined to repeat it?
    Just because extentions are added to a system, or another station is added, it’s lunacy to think we suddenly stop counting, reset all the metrics, and start all over again. If that were the case, large, mature systems, would only report the total system ridership and never have any details to base good planning judgements on.
    Adding U-link is hardly reason quit looking south to 200th, as ‘acedemic’.
    Broadway and Univ Wash Stn’s add 14,000 and 25,000 daily riders, respectively in 2030. Yes, many of those will have a network effect by alighting along Central Link, but those riders are also countable and can easily be parsed back into their respective Origin/Destination pairs using ORCA data, to keep tabs on how Central Link is doing, well after 2016.
    Adding new extensions enable planners to see both cause and effect of individual decisions as well as network effects of having many more OD pairs for trip making.
    As ULink and all ST2 ridership numbers are quoted for the year 2030, it’s difficult to gage what growth curve rates apply in year 1, 2, etc, to get to that number. Central Link started off shaky and has done an impressive 10% growth in years 3 and 4. Is it sustainable according to your graph? Not very likely, as transit doesn’t grow that fast based on population alone (1-2%/yr). Just look at Metro – flat lined since 2008.
    What is ULink expected to do in the 2020, before we have to quit counting again because Northgate segment opened?

    1. It becomes academic b/c once U-Link opens then we will not have any more data from JUST the Initial Segment. Any estimates will be just that, meaning that any discussion will just be an argument over what assumptions should factor into the estimates, etc. Thus an ‘academic’ discussion.

      1. Not so Mathew. As I explained, ORCA data is available on a large percentage of all trips, showing origin and destination of each trip.
        The margin of error using that data would less than 1%, which is hardly an ‘estimate’, as you put it.
        It would be quite simple to continue to report Central Link ridership numbers internally, and I’d be flabbergasted if Management didn’t want that kind of detail.
        Public consumption? Probably not posted anywhere.
        What was your basis for showing a radically different slope than all the SIP’s used in inflating ridership year to year for your estimate?
        Wishful thinking that 10% is sustainable for years to come?

      2. I heard somewhere that Link actually has pretty low ORCA usage.

        As to the% 10% growth, there is a reason the line is dotted. More of a visualization than an actual projection, I have no training in these things at all, I’m just going on past performance.

        Over the past 12 months (ending in August) Link grew at an average rate of 10.7%. The 12 months before that it was 10.3%. In other words Link’s growth rate is actually INCREASING.

        No, that obviously cannot continue forever, but considering the factors involved in the initial slow start (discussed here) I don’t think it impossible for the 10% to be sustained over the next couple of years.

        I think it more likely to slowly decrease to about 6 or 7% by the time University Link opens, but that is 100% completely and totally just a guess.

  7. Excellent post, thank you. It proves what I’ve been saying re: any ridership projections: the further out you’re speaking about, the more iffy they are. That’s why staff giving reports, policy makers, and other interested parties shouldn’t use ridership projections as their opening salvo/primary reason for a certain, for instance, alignment.

    For example, in the Link headed to Lynnwood, many were writing and Sound Transit staff leading their remarks by saying that such-and-such a stop would result in 200 less riders. I’d expect that the true results will be, as with the chart in this story, show a ridership estimate that’s off in the thousands at many junctures. Translated: small ridership differences should be taken with a grain of salt. More important are the tangible “knowns” in deciding a station location: how many properties would be taken, environmental impacts, capital costs, costs to ensure access to that location. Those should be the primary focus of policy discussions, not pie-in-the-sky ridership estimates.

    Another interesting projection that would probably make this point is a ridership projection by station for C-Link.

Comments are closed.