358 passing mural at North Park C Store
Something old, something new.

Much of the debate surrounding many of the recently-proposed Metro restructures boil down to frequency versus coverage: given Metro’s tightly-constrained operating budget (among other reasons), Metro can’t operate numerous frequent closely-spaced fixed-route services. Rather, Metro must choose between numerous infrequent routes that attempt to provide service within a few minute’s walk of almost everyone’s doorstep, and a smaller number of frequent routes that focus service on ridership centers, serving riders in less-productive areas either with peak-only service or requiring them to walk further.

In addition to policy concerns and value judgements, there are some general empirical observations that can inform these debates. One I wrote about recently is that more-frequent routes are often much more efficient to schedule than infrequent routes. Another, which I plan to address in this post, is whether riders — regardless of what they tend to say in public hearings — prefer frequent, direct routes, with simple schedules and service patterns, or a selection of less-frequent routes that might minimize walking or slightly reduce travel time but run less frequently and only at certain times.

Conveniently, Metro performed a restructure that can serve as a test case for this question more than a decade ago one of its most important and highest-ridership corridors, and I have obtained historical data to show the results of this change. History, data and analysis after the jump.

Evolution of Aurora Corridor Service
Evolution of Aurora Corridor Service // Map by Oran

The corridor in question is primarily Aurora, but due to the historical and current structure of service, it is tied in with service to Wallingford, Green Lake, and Northgate. Oran’s beautiful map shows in three panes the evolution of service on this corridor, including some possibilities for future service changes; note that the map was created before Metro announced its proposed 2012 service changes, so the “Future?” pane does not include any information from that proposal.

First, a description of the service pattern prior to the 1999 restructure:

  • Route 6, derived from the historic streetcar alignment that encircled Green Lake, served Aurora at roughly two-block stop spacing every 30 minutes, deviating off Aurora to serve Stone Way, Woodland Park, and Linden Ave before returning to Aurora at Winona. During the weekday, Route 6 terminated at 145th St; nights and weekends, it extended to Aurora Village TC.
  • Route 16, with a similar history, also operated every 30 minutes, staggered to provide 15-minute headways as far as Stone Way, before heading up Wallingford Ave to join the current alignment of the 16 at 45th St. At night, the 16 was replaced by the 16-Shuttle, which provided a timed transfer to the 6 at Stone Way. Both Routes 6 and 16 served the Seattle Center deviation.
  • Route 359X operated during the weekday, again with 30 minute headways, with limited stops between Denny and 45th St, then five-block stop spacing all the way to AVTC. The bus deviated from Aurora to serve Linden Ave before rejoining Aurora at Winona (the “Linden deviation”).
  • Route 360X operated as a one-way peak service during rush-hour. Its alignment was very similar to Route 359X, except the 360X did not serve the Linden deviation, and skipped all stops south of the ship canal.

This pattern is very heavy on coverage over frequency, with only one brief section of high-frequency all-day service, and providing closely-spaced service in Wallingford, even though Wallingford Ave was not particularly well patronized. The service pattern is also complex, seemingly attempting to provide a route tailored for every desire. The 1999 restructure created the service pattern most of us are familiar with:

  • Route 358X serves essentially the same alignment as route 359X, with the 359X’s wider stop spacing, but made full-time and improved to 20-minute headways during the weekday, subsequently improved to 15-minute headways in 2003.
  • Route 6 was deleted; Route 16’s section on Wallingford Ave was moved to Stone Way, picking up the most productive section of Route 6. The 358X continues to serve the Linden deviation.
  • Route 16 was improved to 20-minute weekday headways; Route 16-Shuttle was abolished in favor of full-time direct service to downtown.
  • Route 358X was renumbered from 359X , that number being retired due to the shooting of Metro driver Mark MacLaughlin on the Aurora Bridge in November 1998.

This restructure collapsed a set of six distinct route patterns — 359X, 360X, 6 (145th St), 6 (AVTC), 16, 16-Shuttle — into two routes that cover Wallingford slightly less thoroughly, and make some trips slightly slower (e.g. riders on the prior 360X), but do so at higher frequencies with simple schedules. The following chart shows what happened.

[UPDATE: Chart scale corrected.]

Aurora Corridor Historial Weekday Ridership
Aurora Corridor Historial Weekday Ridership

Combined ridership dropped from 1998 to 1999, recovering above its previous levels in 2000; some ridership shifted categories permanently. Between 2000 to 2008, ridership increased 34%. 2008 was a systemwide peak for Metro ridership; since then, the Great Recession has sapped ridership. Between 2008 and 2011, ridership declined 7%.

Of course, these numbers need a frame of reference to be meaningful. An ideal comparison dataset would be the average weekday ridership for all routes in North Seattle. I don’t have that information to hand, but I was able to obtain the following natty chart from Metro’s excellent systemwide statistics page. Reading from the chart, Metro’s systemwide boardings increased roughly 17% from 2000 to 2008, then declined roughly 8% from 2008 to 2010, suggesting that ridership improvements on this corridor — mostly on the 358 — have outpaced the growth of the system overall.

Systemwide Historical Annual Ridership
Systemwide Historical Annual Ridership

The reader may well ask at what cost these additional riders came; after all, it’s possible (to some extent) to stimulate ridership on any service no matter how well or badly designed, simply by throwing money at the service to increase frequency. Fortunately, I have productivity data for the Aurora corridor, in the form rides per platform hour, a metric that captures the full quantity of service hours (including deadhead time) per rider.

As the following chart shows, performance generally increases after an initial steep drop in 1999, although again the 358 leads the way while the 16 improves only modestly. Two notable data points are 2003 and 2004: new service subsidy was added to the 358 in 2003, boosting midday headways from 20 to 15 minutes, which resulted in both more riders and better productivity, suggesting it was a smart investment.

Aurora Corridor Historical Weekday Productivity
Aurora Corridor Historical Weekday Productivity (2011 data projected based on first half year)

I’m going to venture a few conclusions based on the data I’ve presented above:

  • Riders flock to routes that are frequent, direct and straightforward. Contrary to much of the feedback at public hearings, riders value these attributes more than service to their doorstep. These features are, not by coincidence, many of the features that make rail systems much more attractive than bus systems in the public mind.
  • Very high frequencies may not matter much for developing ridership. The 358 began with 20-minute frequencies; more service subsidy for 15-minute headways came four years later. At least in this case, 15- or sub-15-minute headways were not a prerequisite for success.
  • Wide stop spacing doesn’t hurt ridership. Stop reduction programs invariably draw vocal opposition from existing riders, but switching from the two-block stop spacing of the 6 to the five-block stop spacing of the 358 has simply led to a smaller number of very-well-used stops. For all that the 358 can crawl sometimes, it’s painful to imagine what it would be like with more than twice as many stops.
  • It takes several years for major route changes to “bed in”. Performance and ridership initially declined before exceeding prior levels.

As most readers probably know, the next steps in the evolution of the Aurora corridor will take place in the next few years, with the proposed Fall 2012 service change and the opening of RapidRide E in 2013. RapidRide, put forth as Bus Rapid Transit, is regularly derided in the comments of this blog as little more than what all local service should look like. It’s hard to disagree with that, but only if one overlooks the financial chasm Metro has fallen into since 2008 and the protracted effort required to point a large public-sector institution on a new and different course.

Personally, I’m very optimistic about the potential for RapidRide E to drive ridership on this corridor. A very similar route in South King, RapidRide A (previously Route 174) has posted good gains in ridership, and anecdotally has become a much more pleasant route to ride and drive due to the presence of uniformed fare inspectors. Some off-board payment, better quality, well lit stops, the presence of fare inspectors, deleting a few little-used stops, and adding slightly more service in the evenings and Sundays will make Route 358 even more popular and productive, for very little money.

To round out this piece, I’m including a ridership chart based on the usual format:

Ridership Patterns on King County Metro 358
Ridership Patterns on King County Metro 358

I’ve marked the stops tentatively slated for deletion with an asterisk, and stops where the exact location is still undetermined with a question mark. Here’s what I see in the chart:

  • Excellent performance as far as Shoreline. South of 145th St, almost all stops are very well used, loads are high, many stops exhibit the churn that is a hallmark of popular urban service, especially where there is connecting service at 105th/Northgate and 85th. This is enabled by the “Main Street” character of Aurora, where commercial properties abut the road, and moderately-dense residential property extends back from that.
  • Good performance through Shoreline. North of 145th St, loads are good and stops are well used, although the pattern changes to a mostly-suburban pattern of loading up inbound and unloading outbound, except at the terminal stop.
  • Bidirectional demand in all time periods throughout the route. This bus isn’t just used as a work shuttle — riders use it thoughout the day and evening. Interestingly, bidirectional night demand continues into Shoreline, something usually hitherto only seen in truly urban routes.
  • Stops deleted for RapidRide won’t be missed.  Most of the stops slated for elimination are quite close to other stops and underperform significantly compared to their peers; of the remaining stops, only Galer street is relatively lightly-used, and it arguably has significant geographic value, due to the presence of a pedestrian bridge and staircases that extend the walkshed into Queen Anne and Westlake.

Two sections of the route are worth discussing in detail:

  • The Linden Deviation. The precise fate of the Linden deviation is unclear. Northbound, adjacent to Green Lake, there are no sidewalks, and only one crosswalk between Woodland and Winona, making it unacceptably difficult to access any stop located on the east side of Aurora. Therefore, I suspect the choice Metro will face is between keeping the deviation in both directions, or creating a reversed couplet where southbound buses stay on Aurora and northbound buses serve Linden.
  • South of Mercer. In 2016, the AVW replacement project will rebuild the street grid between Harrison and Denny, allowing the stops over Mercer to be moved south to the reconnected Harrison, a much nicer pedestrian environment altogether. When that happens, the current southbound stops on John and Wall will be replaced by one stop near Denny.

The Aurora corridor is just one in Metro’s large system, but trading coverage (closely-spaced stops and close parallel routes) and schedule complexity for frequency and simplicity has been a success here, a process I expect to continue with the introduction of RapidRide E; this result should inform future debates about ridership-oriented restructures such as Metro’s initial proposal for Fall 2012.

61 Replies to “Aurora: A Case Study in Frequency versus Coverage”

  1. Excellent work! I do have to take issue with drawing any conclusions from the A line, though.

    Doubling frequency, and getting only 50% increase in ridership (at least four months in — it may have continued to grow since then) would show as a 25% drop in passengers per platform hour. Does Metro have a more up-to-date ridership count on the Line A?

    Second, I’m sure Metro has figured this out by now, but the way to not sell stop consolidations is to put up signs just at the stops being removed. The way to sell them is to put up signs at the remaining stops as well (and to put out word on this or other blogs that transit enthusiasts will see).

    I’ve only seen a handful of targetted stops get saved by That’s-my-stop! pressure. (One of them almost led to a much more important corner-and-transfer stop pairing getting removed instead. So, I guess I’m guilty of helping save a closely-spaced stop since I understood the value of transfer stops.)

    It’s useful to keep in mind that the 1999 358 improvements didn’t include transit lanes, queue jumps, bus bulbs, signal priority, etc. Part of selling the 2012 plan may be to advertise what improvements are being considered, and enabled by route consolidation. It won’t change the minds of some of the YOMFSers (Yes, on my front step), but it will hopefully generate an excited support lobby, beyond just us number-crunching geeks.

    1. Yes, that would be a great companion comparison to make, plus very little has been written about RapidRide since it started. That seems to be where the hours are headed these days, so it would be nice to know how it’s going from time to time.
      A small correction, a premise of RapidRide is that the buses are faster than before, so you can get more frequent service without doubling the number of buses on the route. I’m not sure RR-A has achieved the speed goals it set for itself, and I’m not sure how much layover time was removed from the schedule to keep the bus count down. All that would be useful to see if riders per hour is trending towards a ‘good deal’ for Metro.

      1. I think one take away from Bruce’s graph is that any change takes a couple years to build up to it’s potential. I think we might be judging RR-A a bit early.

      2. a premise of RapidRide is that the buses are faster than before

        RR-B would like to have a word with you. Hopefully D & E will have more transit-centric street improvements to avoid the same outcome.

        I’d really like to see a good analysis on “what went wrong” with B’s travel times, so we can avoid those problems with C/D/E/F.

      3. Well RR-A certainly doesn’t crawl nearly as much as the 174 did. The stop consolidation seems to have helped a lot, as have the bulbs, signal priority, and BAT lanes.

        The environment on the RR-A is much improved over the 174 as well. The couple of times I’ve ridden it there simply weren’t nearly as many disruptive passengers as the 174 used to attract. Hopefully a similar effect will be seen when the 358 becomes RR-E.

        Clean transit people feel safe on attracts ridership just as much as service frequency, speed, and reliability do.

  2. Bus routes that are geared toward frequency over closer stop spacing are biased against the elderly and disabled and favor the young and able-bodied. RapidRide B is a example of this (a line clearly designed for Microsoft workers), where mostly empty buses run frequently throughout the day. What a waste. I’ve now ridden the B line a number of times, and BRT is wrong for that area. The alignment is also wrong, in my opinion. The ridership numbers don’t support the service, except for a few hours a day during peak times. I’ve also noticed that headway is completely out of whack, and very soon I’m going to be doing my own little investigation to see if Metro’s claim of 10 minute headway is true. If they aren’t ensuring proper headway is being adhered to, then one of the main advantages of BRT is gone.

    1. The couple of times I’ve ridden the B-line (around 6:00 on a weekday evening), judging by the number of buses I saw going by the opposite direction, if anything, their headway seemed better than what was advertised. It sort of makes sense, in a way. If you promise 10 minute headways, you can’t schedule 6 buses per hour because they’ll inevitably get bunched leading to an effective headway of more like 15 minutes. So, you under the covers schedule 8 buses per hour, figuring that, with bunching, the advertised 10 minute headway is what you get in practice.

      Is anybody in the know aware of whether Metro is doing this, or whether it’s just my imagination?

      1. Considering how much confidence some people on this blog have in Metro, it would not surprise me if they are not doing this.

        Besides, how do you determine the conversion ratio between buses per hour and advertised frequency?

      2. “Besides, how do you determine the conversion ratio between buses per hour and advertised frequency?”

        I ran a computer simulation on this once. The model I used is that a bus is scheduled to arrive every h minutes, and each minute starting from the scheduled time, the bus has probability p of actually arriving. Thus, every scheduled bus will, for sure, arrive eventually, but not necessarily at the scheduled time.

        I experimented with various combinations of h and p, having the computer simulate a random arrival at the bus stop and printing out how long the person would actually wait. The result was with p=0.2 (each bus averaging about 5 minutes late), 6 buses per hour led to a de-facto 15 minute headway (average wait time 7.5 minutes) and 4 buses per hour lead to a de-facto 20 minute headway (average wait time 10 minutes).

        Another interesting conclusion was that once headway dips below 10-15 minutes, improving reliability (raising the p value) becomes substantially more effective in reducing wait time than throwing more buses at the route to reduce headway (lowering the h value). I don’t remember the exact numbers, but the number of buses required to achieve a real headway of 5 minutes with p=0.2 was downright insane – a whole lot more than 12.

        While I realize my model made a number of simplifying assumptions, the conclusions drawn from it do fit with my gut feeling on the issue and I think it’s reasonably accurate. At some point, it might be interesting to use real data from OneBusAway to estimate the “p” value for some of the heavily traveled routes at various times of day.

      3. Your opinion about it making more sense to perform capital improvements below ten minute headways (rather than just throw buses at the problem, except as a temporary solution) is one that’s fairly widely shared among transit professionals, I think; although it probably derives from experience and handed-down wisdom rather than simulation.

        Of course, the first step is getting to a route network that’s sufficiently consolidated that you have a manageably small number of well-used frequent-service corridors that you can then put capital money into.

    1. The 358 is a much more “complete” line than the currently-active section of Central Link, so that really helps its numbers.

    2. It doesn’t make sense to refer to one car on a Link train as a “vehicle.” One Link Train has one driver, so it is one vehicle even if it is two cars hooked together. The key is passengers/driver, since labor is the bulk of the cost of running transit.

      1. Tell that to ST. Their metrics are mostly based on per vehicle statistics (vehicle cost per hour, riders per vehicle per ??), and no, driver salary alone is not a significant cost of operating the trains. Look at vehicle cost per hour and driver salaries.

      2. Look up driver benefits, Mike. The current situation in *all* of American business is that benefits (for jobs with benefits) cost *more* than wages, so more-than-double your number for salaries if you want to know the cost of having the driver.

        This is because this crazy country doesn’t have single-payer health care (Canada-style) or a national health service (UK-style), and instead expects health insurance to be paid by employers. Something Obama did not even try to fix.

        But anyway, you have to know that fact in order to make any meaningful economic analyses about the cost of hiring an employee. Wages are less than half the cost of an employee — often less than a third. So, again, Mike, you don’t know what you’re talking about economically, but hopefully now you know more than you did before.

      3. I think the reason ST uses cost per vehicle is to make the trains look really good because most people think that it’s cost for the whole train and don’t realize they have to double or triple the cost per hour. Note however that the operating cost of running two car trains is almost identical to that of a three car train (~$40/hr less) but you do save on maintenance and depreciation costs.

  3. One other thing that might have been related to the 1999 dip: Wasn’t that the year someone shot the driver of a 6 and the bus drove off the Aurora bridge? I can see how some regular riders might have selected other routes for a little while after that.

    1. That was November ’98. So yes, that’s a possibility, although it’s hard to quantify that effect, and I wasn’t around then, so I’m not in a position to evaluate whether people did respond that way.

      1. I was around then, but I don’t remember anyone avoiding transit for that reason. Most of the people who would avoid transit for fear of lawlessness aboard were already avoiding it.

      2. I was around then. My sense was that folks thought it was a freak event and while they may have been a bit more observant about watching other passengers as the bus got closer to the bridge it really didn’t change their overall behavior. I don’t have any hard data though.

      3. I had co-workers who either stopped taking the bus or who took other routes after that incident. The question is were enough people behaving that way to impact the ridership.

    2. Everyone was making bank on dot-coms in 98-99 and driving their new cars to work instead of taking transit.

    3. One other thing that might have been related to the 1999 dip:

      The 800# gorilla is always gas prices. Plus, when ever you make a major change it takes 12-18 months for people to adjust. Look at what happened with the DT Redmond P&R/Transit Center. The old surface lot was chock full but after the closure and opening of the parking garage it was 1/2 empty. A year later it was back to being nearly full. Link’s another example; ridership takes about 18 months following additions or changes in bus routes to realize the potential. The other big driver is population density. How many residential units have been added along this corridor? The effect of that and gas prices (which can be inferred from overall system ridership) need to be factored into a baseline just like inflation has to be factored into pricing. Oh, and yes, bus fares should also be factored in. Really the relative fare vs. the price of gas. But again the comparison to overall system ridership should adjust for that.

  4. The heavy ons/offs just south of Shoreline and lower boardings in Shoreline might have something to do with the zone boundary. I’d be curious to see what would happen if the whole route were to be designated “local” service.

    How is e-purse handled on the 358 right now? Does it default to one or two zones?

  5. Anyone want to take any bets on whether the store with the RapidRide bus mural will oppose removing parking from the future BAT lane?

  6. Why is Aurora and 115th listed on the southbound chart twice? Is that an error in the chart, or does the bus stop before and after the intersection?

  7. This is a good argument for why we should have built LINK straight North-South from Sea-Tac to Northgate, right along the old Interurban route, right up highway 99.

    If we started in 1993, when the planning was being done for light rail, and skipped all the monorail nonsense, we would now have a fast, core, transit system, built at a cost of $35 million per mile and would be working on various crossways extensions (also at or near grade) from that main line.

    1. Some problems with this:
      * getting through downtown would require a tunnel
      * crossing the ship canal would require a new bridge or a tunnel
      * low demand for most of the route between downtown and Tukwilla
      * entirely misses two of the highest nodes of transit demand in the state (UW, Northgate) as well as some of the highest population density North of the Ship Canal.
      * Much of the alignment would need to be elevated, especially north of the Ship Canal due to conflicts between the trains, cross-traffic, and turn movements.

      While it might have been cheaper than what we’re currently paying for link there is no way it would have been $35 million/mile. Nobody has built light rail in North America for that little.

      1. The quoted survey is 10 years old. Also these:
        * getting through downtown would require a tunnel
        * crossing the ship canal would require a new bridge or a tunnel
        * Much of the alignment would need to be elevated, especially north of the Ship Canal due to conflicts between the trains, cross-traffic, and turn movements.
        are going to cost you much more than $35 million a mile.

      2. Also there is no indication if the system costs at adjusted for inflation or not. A $35 million/mile line completed in 1987 would cost much more if built today.

        Recent surface-only lines seem to be running over $55 million/mile.

        Furthermore many cities have been able to take advantage of old rail ROW, highway ROW, or very wide streets for their light rail lines. Not really an option in most of Seattle.

      3. It would have been cheaper if the Interurban had never been *removed*, of course. Because all the “conflicts” would have been resolved in favor of the trains long ago, as in the few interurbans which were eventually upgraded to grade-separated rail (South Shore Line comes to mind). But that’s another matter, water under the bridge, decisions made long ago.

      4. * entirely misses two of the highest nodes of transit demand in the state (UW, Northgate) as well as some of the highest population density North of the Ship Canal.

        Doesn’t miss U-Dub if East Link had been incorporated into the new shiny ferro-cemment 520 crossing. But Seattle politics of social justice and ST’s growing pains mean we’re building a hundred year project and got it way wrong.

      5. Bernie,

        Still waiting on your response to the I-90 article arguments. :)

        Given the existing Link routing, East Link via 520 would require a transfer to get to Bellevue from downtown, or from the north, or maybe both (if it was a connection rather than a branch).

        And an Interurban routing (which, at this point, is about as plausible as rerouting I-5) has so many other problems, it’s almost not worth listing them all. That would make about as much sense as a routing via the I-5 express lanes. You build trains where people actually want to go, not just where it’s convenient.

    1. I still think something like this could happen someday, though my preference would be for a second downtown tunnel to serve Belltown, Lower Queen Anne, Queen Anne, and Fremont, surfacing at 99 just north of 46th. That serves the dense neighborhoods with a grade-separated line, and you could even split the route to add a surface route to Ballard that curves north/northeast along 15th/Holman to Northgate. The south end of the line could split to serve West Seattle and an airport express route.

      The new tunnel would have to be cleverly sited to cross the Viaduct replacement tunnel, though.

      More directly on topic, the next round of bus improvements look like another step in the right direction. I really liked the way this post was presented–it seems persuasive.

  8. Good stuff. However, the ridership chart should show total corridor ridership in the area of 15,000 rather than 25,000. Something is out of whack.

    1. If route 16 ridership is subtracted that leaves about 18,000 on the 358. Is that chart actual boardings on Aurora or is it simply the sum of route 358 + 16 ridership? Only part of the current 16 serves the Aurora corridor, unlike the old 6 local.

    2. Thanks for bringing this to my attention, I should have cross-checked my data better. The data I was given was semiannual, and I combined the daily ridership incorrectly, summing rather than averaging, effectively multiplying everything by two. The rides/platform hr numbers were correct, and the conclusion is unchanged, of course — just the scale was incorrect.

  9. I thought route 359 was retired after the Metro bus tragedy where the bus operator was shot in November of 1998.

  10. I don’t really agree with “Very high frequencies may not matter much for developing ridership”. looking at the graphs, especially productivity, you could argue they are essentially flat, and only take off in 2005.
    going from 20 to 15 to 10 minute headway makes a huge difference.
    Different corridor, but from my personal experience, commuting to Bellevue from NE Seattle, when the 271 was at 20 minutes, I’d ride the 243 every day instead, when it went to 15, taking 65/271 combo became a viable alternative, I haven’t taken the 243 at all since 271 is at 10 minute headway.
    Meaning, if it was for me, you could eliminate my one seat peak route, in favor of adding more frequent service to all day routes.
    Simplicity, and direct routes are important too, but high frequency, i.e. service whenever I need it, is the one factor that will get people to use the bus instead of their car.

    1. I’m unconvinced. The headways were improved in 2003, which doesn’t really explain 2005’s increase. 2005-2008 was also where the system as a whole posted its greatest gains. Some of the apparent flatness in the combined graph is due to the 16 losing some riders in the early ’00s, and is one of the disadvantages of choosing to represent this data as a stacked bar chart.

      1. The 800 pound gorilla is gas prices. Gas goes up, ridership goes up. Gas goes down, ridership goes down. Of course it’s a relative difference. Increase fares and it decreases ridership but that’s negligible to gas prices since a 25 cent increase in gas price adds roughly double the cost of commuting vs the cost of a 25 cent cost in fares. That’s because gas price effects the per mile commute where as fare increases only affect the per commute cost. You see a huge shift when gas prices get a dollar a gallon bump or go over a “threshold” like $5 per gallon. Ironically, higher gas prices help the big petroleum companies.

      2. I think it’s the difference between “gradually improving ridership” from existing demographics and “developing ridership” from whole new sources. See my much longer comment below.

  11. It is also extremely important to consider that a route’s reliability and tendency to get bunched has a big effect on wait time, in addition to the paper frequency.

    For example, the 48 during the day, on paper runs every 15 minutes. Link after 10 PM is also, on paper every 15 minutes.

    However, in practice, if I show up at a #48 stop using OneBusAway to try to time the arrival, I still get an average wait time of at least 10 minutes and if I showed up without consulting OneBusAway, I suspect it might be longer. In other words, 15-minutes headways on an unreliable route is equivalent to 20+ minute headways on a reliable route.

    With Link on the other hand, I can hop off a plane at 10:00 in the evening, walk over to the train stop, and actually expect an average wait time of 7.5 minutes, with a maximum of 15 minutes. Big difference!

    1. Reliability is ridiculously important in ridership on any sort of passenger transportation service. There’ve been some fun correlations run on various specific routes in various places when something suddenly changed the reliability (for better or for worse). Massive ridership changes.

      I sometimes suspect that if Amtrak started running on time consistently they’d discover they had about half as many passenger cars as they needed.

  12. Very high frequencies may not matter much for developing ridership.

    It’s not wholly inaccurate, Bruce, but it’s a bit overstated. Really, the lesson here is that any increase in frequency, combined with improved route legibility, will yield an increase.

    20 minutes frequency on a single route is a palpable improvement over 30, and the response you see in your data was palpable. 15 minutes is another palpable improvement, and as you see, productivity continued to rise incrementally after 2003’s headway shortening (the ridership response was enough to compensate for extra vehicles).

    But the 358 has remained a deeply unreliable route, so post-2003 15 minute headways are rarely perceived as such. It has also remained an “unpleasant” route, driving away most elective riders.

    The proof: all that other traffic on Aurora. The 358 still captures a very low percentage of trips in its corridor and exhibits an unusually high (even for Metro) high percentage of “captive” riders, who don’t respond as dramatically to frequency changes as elective ones.

    At least in this case, 15- or sub-15-minute headways were not a prerequisite for success.

    One thing that makes this case special: the 358 is a very low-transfer route, except at its two endpoints. While 46th, 85th, and 105th (major intersections all) are busy stops, I’m always shocked how small a percentage of the 358’s load transfers to the 44, 48, or 75. (It’s never more than 1 or 2 people, even from a packed-to-the-gills bus.)

    If you’re only taking a single vehicle, the difference between 15 and 20 minutes is not the end of the world, even for spontaneous (starting from outside) travel. Waiting 20 minutes in lousy weather sucks, but at least you’re only doing it once.

    But add in a necessary transfers, and the difference between 15 and 20 (or between 10 and 15) is huge, because you might be stuck doing that 20-minutes-in-the-rain-and-wind twice through no fault of your own.

    I have no doubt that a 20-vs-15-minute network is a much bigger selling-point (or dealbreaker) for the average rider than on this isolated example.

    1. If you look at this as scheduled service, meaning you plan your trip rather than just show up and hope a bus will arrive, then I think it’s a quite different perspective. Going from 20 minutes to 15 minutes you (potentially) save five minutes or 10 minutes per commute for only a 1/3rd increase in cost. Going from 15 minutes to 10 minutes you get the same potential time saving but the cost increases by 50%. It’s the law of diminishing returns. Double frequency from 2 minutes to 1 minute and you double the cost and get virtually zero benefit. Of course the length of the trip factors in. A short shuttle needs high frequency; if the escalator only turns on once every two minutes most people will take the stairs ;-)

      1. That isn’t quite true, actually. See Bruce’s article on the Queen Anne restructure for an explanation, but basically, if you want to maintain clock-face schedules (and on routes less frequent than 7-8 minutes, you pretty much have to do that), then doubling frequency does not mean doubling costs.

        You’re definitely right that, past a certain point, decreasing headways isn’t too useful from a frequency perspective. (It can be useful on grade-separated lines from a capacity perspective, but that’s a totally different argument.) But 15 minutes is definitely not that point.

    2. When I wrote that sentence, I knew it draw a comment from you, d.p.; I have been waiting…

      I agree with you that the response of demand with respect to frequency is different in different places, and Aurora having an atypically transit-dependent demographic profile makes the difference between 20- and 15-minute headways matter less here than elsewhere.

      I guess I differ with you on the point at which spontaneous ridership that utilizes transfers becomes likely. I believe the research shows (and I don’t have a link handy, but I’ve read it some official-looking paper somewhere so it must be true) that people stop referring to schedules at headways below ten minutes, but they continue to refer to schedules at 15-minute headways.

      Partly because of that, and my own experience, I don’t think many choice riders will do spontaneous transfers at 15-minute headways, and I don’t see money on the horizon (either new money or from cutting coverage) to run a bunch of routes at 10-minute headways all day; or even to make the capital improvements necessary to get full off-board payment, which is what’s needed to get routes like the 358 with constant churn to run at the close-to-clockwork precision that really should come with that quality of service.

      Not quite sure where I’m going with this, but I think we basically agree, just where we draw the line for spontaneous ridership and transfers is different.

      1. I agree with every word of that, actually.

        I don’t think 15 minutes is ideal for a spontaneous transfer, but it’s certainly better than 20.

        That’s why I worry (even after a very productive-feeling trip to one of Metro’s 2012-revision Open Houses tonight) about what Metro will achieve if their radical restructuring gets hobbled by 20- and 30-minute core routes.

        Even with OneBusAway, there are limits to a passenger’s ability to “schedule” a transfer. Sometimes there is only one possible route, and your first bus drops you when it drops you. Those next 9 versus 14 versus 19 versus 29 minutes are do-or-die for the new network.

      2. and I don’t see money on the horizon (either new money or from cutting coverage) to run a bunch of routes at 10-minute headways all day

        Why not?

        RapidRide C could have 10-minute service just by reusing the 54/55 service hours. RapidRide D could have 10-minute service just by reusing the 15/18 service hours. Deleting the 43 could get us 10-minute service on the 8 (at least to 23rd E) and on Madison, assuming some restructures in line with the proposed TMP. We already have 10-minute service to the U-District and Rainier Valley, and and the 80 super-express restructure would give us 10-minute service to Northgate as well.

        Maybe the message we need to send to Metro is “10 minutes or bust”. And if that means 20-minute service on some corridors that have 15-minute service now, or deleting a downtown connection on a route that has one now, then maybe that’s just the price we have to pay.

Comments are closed.