[sustran] Re: Thinking Outside the Bus

Paul Barter paulbarter at reinventingtransport.org
Mon Jun 4 21:42:13 JST 2012

Thanks Karl for the thoughtful comments and Guangzhou examples. Gives me a
chance to clarify a little.

I don't think we will agree but it is worthwhile getting to a better
understanding of where the disagreements might be. In that spirit, here

I'll make a few comments amongst your text below but here are two key

-  Connective networks are not the answer to every problem or situation. I
am claiming the idea often deserves more consideration. I am not saying
that every city must make its network more connective no matter what.

-  The point is to get to a "turn-up-and-go" level of service on more lines
to make public transport more attractive overall. If your service
frequencies are currently poor and you have a complex network with lots of
overlapping services, then reorganising towards a more "connective network"
(with fewer route kms) can usually help.

- A key empirical issue here is the question of how much high frequency
service matters. Those who are more sympathetic to connective networks tend
to see the evidence as demonstrating that short headways matter a lot to
the attractiveness of public transport.

On 4 June 2012 17:31, Karl Fjellstrom <kfjellstrom at gmail.com> wrote:

> To me there seems to be a contradiction between objectives of maximizing
> transit ridership and saving passenger time, and many of the people writing
> books and reports about transit, especially the ones who laud 'transfers'
> as if they are something actually good.

PB: You are correct of course that connections are only a means to an end.
I haven't heard of anyone saying they are a good thing except in order to
get something else which is very good:  namely, high frequencies/short
headways which give people the freedom to "turn up and go" at stops (and
which fortunately also reduces the pain of waiting for those connections).

But the key point is that sometimes reforms that increase the number of
transfers does increase ridership by helping to increase frequencies.
Minimising transfers will not maximise ridership if it means low levels of
service on each line.

> Some people making this argument presumably come from a fixed-rail network
> background, and/or are promoting fixed rail systems, which perhaps helps
> explain why they try to impose this fixed rail network thinking on buses.

PB: Interesting. Yes, there may be some correlation. Obviously, if you are
a die-hard rail advocate you will certainly want to reorganise buses around
the rail spines.

But that doesn't mean everyone who is sympathetic to connective networks is
die-hard pro-metro! I am not. And ironically, Jarrett Walker is often
accused in the US of being pro-bus and anti-rail.

> We saw this argument presented a lot in Guangzhou. People saying Guangzhou
> doesn't have enough transfers, and has too many overlapping bus routes, so
> we need to build transfer hubs and cut the bus routes so we can have more
> transfers. To me it is a lot like arguing that Guangzhou's food is too
> delicious, so we need to cut it back, make it less tasty so it's closer to
> the average. Anyway, these consultant proposals, which culminated in a
> transport master plan in 2006 funded by a World Bank loan recommending
> dozens of transfer hubs and cutting the bus routes accordingly, are usually
> thankfully and rightly dismissed by the city. And then the Guangzhou BRT
> opened in early 2010. The Guangzhou BRT is based on an opposite premise.
> It's a direct-service model, the idea being to minimize transfers and
> maximize ridership and passenger time savings.

PB: I don't know GZ well so I am guessing here, but:

-  If most routes ALREADY have attractive headways throughout the day then
there would be no headway-based argument to simplify the network.  Already
the case across most of Guangzhou? If so, there would be no point creating
more connections or a simpler network for their own sake.

-  On the other hand, could it be that, even though all key corridors are
served wonderfully by overlapping routes, many routes in outer areas may
have low headways (eg more than 15 or 20 minutes)? If that were the case,
there might be some merit to some shift in the direction of a connective
network. (Not necessarily any extreme change -- it is a spectrum of

-  In addition, in dense cities like Guangzhou there is another common
argument for reorganisation of bus lines (mentioned by Eric Bruun the other
day): bus congestion on the busiest corridors with the overlapping routes.
 I guess Guangzhou's amazing open BRT has now shown a new answer to this
problem.  But until the GZ BRT, consultants probably assumed that a shift
to closed BRT or to rail would be necessary to cope with a corridor like
that (which would force more connections). Maybe the consultants you
mentioned were thinking along those lines. An honest mistake based on prior
experience but now in need of updating in light of the GZ experience?

> I took a quick look at the first link you provide below.
> http://www.humantransit.org/2009/04/why-transferring-is-good-for-you-and-good-for-your-city.html.
> The travel time argument is key, but the longer travel times are clearly
> not the only disadvantage of transfers.

PB: Just to be clear, despite his attention-seeking headline, he is not
really arguing transfers are good in themselves. But he is saying that if
the network simplification can achieve high enough frequencies for the same
input of resources, then you can actually get shorter total travel times,
despite the need for the connections.

> The analysis is wrong, for several reason, and I'm surprised that you are
> promoting this material. When looking at frequencies and hence waiting
> times, it assumes no overlap between the direct-service routes. The reality
> is with direct service routes that you end up with a lot of route overlap
> at key, high demand points. This provides many passengers - especially at
> the high demand areas where they are most useful - with multiple route
> options, at high frequencies. Secondly, the analysis assumes a 5 minute
> transfer cost, which is far too low. Even in the best transfer situation
> you should probably assume a 10 minute delay. And that is in the best
> situation, e.g. where you just need to cross a platform. In other transfer
> situations you may e.g. need to alight, cross a road, and walk to another
> bus stop or platform, which could easily already exceed the 5 minutes
> transfer time that the analysis lists. Plus you may need to pay again, and
> you are uncertain about the waiting time. Plus perhaps the next bus is
> full, or there is no seat on the next bus, etc.

PB: Yes he glosses over lots of these issues in order to make his key point
via an oversimplified example. There is no denying that making a connection
can be painful and we should only increase connections in a network if the
payoff is worthwhile. It is one step in a wider argument that such reforms
can often offer more attractive public transport, despite the problems with
connections and the difficulties with making them easy enough.

We should only reform towards a more connective network arrangement in
cases where this really delivers better service not worse. An empirical
question for specific cases.

For example, if you already have very good frequencies without reforming
your network (as in GZ?), then making it more connective may very well be
pain without gain. I don't blame you for being sceptical in such a

Plus, in order to access some transfer facility, vehicles typically have to
> do some additional manoeuvring, which adds to trip time and hence fleet
> requirements and system costs.

PB: Agreed. And in hubs-and-spokes type networks the interchanges can also
become bottlenecks for buses. Some of Singapore's interchanges have reached
this point I think. You wouldn't want to over do it.

But don't forget other kinds of connective networks, such as the simple
grid, for which these transfer-point problems are less of an issue. But a
grid raises other issues like how to get the bus stops close enough to the
intersections without screwing them up. Singapore's bus stops are 150m or
more from intersections: hence no grid of bus routes here. Lots of
trade-offs, no free lunches ... Didn't mean to imply that network planning
is easy.

> Plus there's the cost of building and operating the transfer
> facilities. It's why you almost always see when looking at fare levels that
> what you misleadingly call 'connective' networks have higher fare levels
> than the 'direct-service' networks.

PB: Fair point. Shouldn't ignore such costs if comparing the options. They
should be counted when asking if the changes are worthwhile on balance or
not worthwhile.

It's typically disingenuous of people advocating transfers to gloss over
> these issues of the actual physical transfer requirements and time and
> other costs of transfers.

PB: Maybe some do gloss over them in their zeal. That's a pity. But in my
experience, people advocating this kind of reorganisation are sincerely
aiming for the benefits that flow from short-headway service. They
genuinely want public transport to improve to attract more users. They are
generally transit advocates. They are generally acutely aware that
transfers are still a pain and that they need to be made as painless as
possible. But if not, then yes, they are not being honest about the tricky
trade-off to be faced here.

The key point is to see that there is a trade-off between frequency and
connections. And it runs both ways. If a city can't support frequent
service on a 'direct network' that aims to minimise connections, then it
will either have abysmal frequencies or require heroic levels of subsidy.
In such a city, more transfers may be a price worth paying to get the
frequencies up to a level that makes public transport more attractive for
more people.

> And misusing the word 'integration' as a way of describing proposals to
> cut bus routes and connect them with other routes at hubs is one of the
> reasons the term 'integration' now has so little actual meaning. Similarly,
> calling these cut-up bus networks imposing high transfer costs 'connective'
> is just another piece of doublespeak.

PB: I agree that integration has too many different meanings now, which
causes confusion.

Maybe you have seen some inappropriate proposals for bus reorganisation
that are giving the idea of connective networks a bad name.  If such a
proposal imposes high transfer costs without large benefits in terms of
headways and strenuous efforts to make the transfers less painful, then,
yes, it would probably be a bad idea.

I don't see why 'connective network' is double-speak. If anything, doesn't
it honestly acknowledge that the approach involves more connections in the

Maybe the problem with the term is that it doesn't make the hoped-for
benefit obvious enough! So perhaps the proponents should talk about
"high-frequency connective networks" to highlight that the point is to get
better frequencies. (Remember, no point doing it if you already have high

Hope this helps.

Working to make urban transport and parking enrich our lives more and harm
us all less.
paulbarter at reinventingtransport.org
http://www.reinventingtransport.org  http://www.reinventingparking.org

More information about the Sustran-discuss mailing list