Wednesday, February 17, 2010

Habits of highly efficient bus routes

For many years, the MTA released only the bare minimum amount of information necessary to satisfy its legal requirements. Under Jay Walder, that is beginning to change, and more and more information is being made available on the agency's website.

One of the most jealously guarded secrets was ridership broken down by route. Apparently, the MTA didn't want people to know who was being subsidized more than others, for fear of jealous riders attacking each other. That too is changing. Although they are not putting all the data up as soon as they have it, blogger Larry Littlefield noticed that they did put a lot of it up in the documents supporting the most recent service cut proposal. For example, this PDF file gives ridership data and per-rider cost estimates for every New York City Transit bus line.

As Larry says, this is a document that everyone should read. For your improved consultation, I have extracted the data tables and converted them into a Google spreadsheet that you can download and play with. Since they give average fare data for weekdays and weekends, I've estimated the farebox recovery numbers for each line.

Interestingly, although there are no lines where fares cover the total costs (including the costs of the buses and garages), there are 22 that run an operating profit on weekdays, and ten that run an operating profit seven days a week. The average fare paid for weekday local bus service is $1.14, and on weekends it's $1.29.
RouteNotesWeekday direct operating cost per riderWeekday farebox recovery ratioWeekend direct operating cost per riderWeekend farebox recovery ratio
M8686th St. crosstown$ 0.66173 %$ 0.75152 % 172 %
Bx12Fordham Rd. Select$ 0.84136 %$ 0.93123 % 139 %
M7979th St. crosstown$ 0.86133 %$ 0.96119 % 134 %
Bx19145th St./149th St./Southern Blvd.$ 0.91125 %$ 0.96119 % 134 %
B74Coney Island projects to subway$ 0.81141 %$ 0.98116 % 132 %
M2323rd St. crosstown$ 0.71161 %$ 1.05109 % 123 %
B3539th St./Church Ave.$ 1.10104 %$ 1.10104 % 117 %
Bx9Kingsbridge Rd./Fordham Rd.$ 1.12102 %$ 1.11103 % 116 %
Bx35181st. St/167th St.$ 1.05109 %$ 1.12102 % 115 %
M1414th St. crosstown$ 1.03111 %$ 1.14100 % 113 %

Some things I noted:
  • All the routes except the B74 have multiple transfer points to the subway.
  • Four Manhattan crosstown routes (the rest of the crosstown routes make an operating profit on weekdays).
  • Four Bronx crosstown routes, three of which cross the Harlem River into Manhattan.
  • The pilot Select Bus route makes an operating profit.
  • Two routes in southern Brooklyn.
  • No Staten Island routes; the most profitable Staten Island bus, the S48/S98, has a 67% weekday farebox recovery ratio, and 70% on weekends.
  • No Queens routes (most Queens local buses are operated by the MTA Bus Company). The most profitable Queens NYC Transit bus, the Q58, has a 98% farebox recovery ratio on weekdays and 96% on weekends.
  • The most profitable NYC Transit express bus, the X27/X28, is one of the shortest. It serves Bay Ridge with a 66% recovery ratio.
I'd be interested to see what other information you all are able to extract from these.


Alon Levy said...

There are weird routes on the spreadsheet, such as an M223. What are they?

Cap'n Transit said...

Those were footnotes that got munged by Google Docs. A 3 footnote indicated "All weekend service discontinued," and a 4 indicated "Route entirely discontinued." I've taken them out; you can see them on the original PDF.

saosebastiao said...

Do they release the same information for their subways? If some bus lines have FRRs this high, I can imagine some of the subway lines reaching into the 200%-300% range.

James Dowden said...

Okay, here's an interesting one:

1) Sort by farebox recovery.
2) Subtract 100% from the farebox recovery to get the surplus.
3) Multiply this by the number of riders.
4) Make a continuous function.
5) Look at where it goes negative -- i.e. the point above which the routes taken as a whole recover 100%.

Just 45 extra routes are included by this, namely:

B1*, B3, B6, B8*, B11, B17, B25*, B36, B41, B44, B46, B49*, B54, B64, B68, B70*, B77*, B82, B83
Bx1/2, Bx5, Bx7, Bx11, Bx13, Bx15, Bx17*, Bx21, Bx22, Bx27*, Bx31*, Bx36, Bx41
M15, M31*, M42, M57, M60, M72, M100, M101, M102, M103
Q55*, Q56*, Q58

So the next question I asked is by how much the $1.14 average fare would have to rise to expand this set to encompass the entire list. The answer is a 20¢ rise to $1.34, or for those who like percentages, an 18% rise. Incidentally, this would move 33 of the above 45 routes (all those except the asterisked ones) over 100%, which does "feel" right, considering how exceptionally well-used some of those routes are.

Now of course it isn't as simple as the trivial extrapolation that raising the base fare to $2.65 would mean that no bus cuts would have to be made, but it does show that a bus network with 100% farebox recovery isn't absurd.

Cap'n Transit said...

Saosebastiao, I haven't seen the same information for the subways, but they do have higher operating costs. Interesting analysis, James!

Please also see my subsequent post on the MTA Bus Company data.

Alon Levy said...

The numbers in the table aren't the actual farebox recovery numbers. They're at most farebox revenue divided by allocated operating costs. If you do all operating costs, NYC Transit buses have a 40% recovery ratio, not as claimed in the table.

Jarrett at said...

Cap'n. I don't know your network well enough to evaluate these routes, but in most of the networks where I've looked at such rankings, top performing routes are either (a) commuter express services running only on the peak or (b) all-day high-frequency services in very dense corridors with two-way demand.

saosebastiao said...

James D,

Interesting observation. How politically possible would it be to have a fare increase like you say? Have they tried it before? I'm not familiar with NYC politics, so I don't know how sensitive they are to it. For me, I wouldn't mind paying 25 cents more to keep riding the bus, even if I wasn't riding those in danger of cancellation.

In general, rail has higher operating costs, but they also have much higher capacity, which if used right, lowers the per person operating costs dramatically.

What operating costs are they missing?

James Dowden said...


Well, it's fairly crude, and one would have to expect that a fare hike would be met by some drop in ridership canceling part of it out. But it is worth noting that New York's fares are roughly half of Tokyo's and a third of London's. The political aspect is what has held New York's fares so low -- and it's ultimately what sent the IRT bust all those years back.

The real failure is one to couple the fares and contract negotiations -- the MTA is forced into a bilateral with the TWU over pay, then into another bilateral with the riding public over fares, but the riding public and the TWU only debate through the intermediary of the MTA, with a near inevitability of the MTA getting squeezed. There needs to be a move to a trilateral approach by which the TWU and riding public agree on a price-point option.

We do ultimately somehow need to face the counterintuitive reality that losing some trivial amount of money per rider on the B44 adds up to far more than losing a colossal amount of money per rider on the S60: and that does imply some sort of fare hike. Cutting low-ridership routes isn't generally going to be very effective, as it will just increase the problem next year as more high-ridership routes drop through 100% (okay, Alon, it's a fake 100%, but it's at least the metric they're using internally).

But I wouldn't advocate a simple jack-up-the-base-fare fare-hike. There are ways in which we can minimize the number of discretionary riders we scare off:

1) The MTA is a bus and railroad company, not an instrument of social and educational policy. The MTA should sell ordinary pay-per-ride metrocards in bulk to schools (and for that matter, anyone else who wants to buy a box of a few hundred metrocards) at the regular rate, and it should be a matter for the schools how much they choose to sell them on for. And if they think they can do it cheaper with their own yellow buses, good luck to them.

2) Unlimited metrocards are priced based on erroneous assumptions. There are roughly 22 working days in a 30-day period. The 30-day metrocard assumes just 46 rides for the rider to "win": two more than just going to work and home again -- perhaps 1950s man is going to take Junior to Coney Island once a month for a treat whilst Mom cleans the apartment. Okay, that's a slightly snide stereotype, but the point is that the average rider in fact doesn't take around 50 rides with his 30-day metrocard: the true figure is 71.51. This is ultimately a failure to distinguish between two market segments: 1950s man should be on pay-per-ride, whilst unlimiteds (and especially the 30-day one) should be priced to reflect the rather heavier usage their users make of them.

3) Much of the cost of running a transit line comes from how many vehicles you have in operation at peak times. Buses and trains that make a run or two in rush hour, then sit in a depot and make no money all day cost vast amounts of money to run (this is partly why the express buses are such a financial disaster, although that is exacerbated by each seat being occupied only once en route). Of course there is a lot less elasticity of demand at peak times (i.e. your ridership isn't just going to go away if you put up the fare), particularly in the morning peak. So it is doubly more efficient to put up the base fare by a smaller amount, but charge, say, 50¢ or $1 more between, say, 5:30am and 9:30am. I wouldn't advocate taking this approach as far as London has (they have some seriously nasty peak fares), but this is where I see there being a relatively effective fare hike.

And now I'm off to look at the MTA Bus Company data... :-)

Cap'n Transit said...

Thanks, Jarrett! These are all (b), if anything; the express buses don't come anywhere near paying for themselves. They're essentially all circumferential routes, complementing and filling gaps in the radial subway network.

The one that doesn't seem to fit under your category (b) is the B74, which goes from several large public housing projects to the Stillwell Avenue terminal, where riders can transfer to the subways. As far as I can tell, though, the demand isn't bi-directional, it's from a residential area to a transfer point.

Christopher Parker said...

Interesting that so much of cost of bus operation are overhead. Those costs presumably will remain constant if you cut service.

Also I'm struck by how the average fare is higher on the weekends. This contradicts the higher rush-hour demand -- and the much higher cost incurred by peak only service.

I'd say it's time for a fare increase (better than cutting service) -- but only for rush hour and passes. Retain cheap off-peak fares to build that ridership.

Alon Levy said...

What operating costs are they missing?

My guess is that they're missing anything that's not directly attributed to operating the line, for example management overhead.

But it is worth noting that New York's fares are roughly half of Tokyo's...

No, they're not. For short trips, Tokyo is cheaper. For medium-length trips, it's about the same, or Tokyo may be a little more expensive.

And independently of trip length, New York City Transit costs about the same as the Berlin U-Bahn and 20% more than the Paris Métro. For almost all trips it costs 40% more than the Madrid Metro, breaking even only on the longest trips.

London's more expensive, but it's an outlier.

James Dowden said...

No, Alon, you're missing the point: yes, the basis of Tokyo's single fares is radically different, but overall, if you want a 30-day unlimited, it'll cost you about $185 for something that's less permissive than a 30-day Metrocard; the equivalent of the Fun Pass (i.e. the one you can use on the buses as well as the subways, before you find the locked-down Tokyo Metro only version) meanwhile costs about $15. The roughly double statement works when you actually think about frequent riders rather than looking for some hypothetical thing a railfan could do to prove a point.

And the Paris Metro is a poor comparison because it serves such a small area: again, you're into a system that charges lower fares for shorter distances, something that just isn't going to happen in New York, thanks to the "one city one fare" policy. Far Rockaway is as far from Midtown as Cergy (in zone 5, way out at the end of RER A) is from central Paris. Once you're paying for a comparable area, Paris is about 65% more expensive than New York.

London is extreme, but you should remember that until the Labour Party came up with an insane bureaucratic part-privatization scheme (known as PPP), the Underground turned a small operating profit. The fares therefore to some extent reflect commercial reality in a developed country, rather than the sort of fares that rely on socialized transit, poverty of the workforce, or both.

Alon Levy said...

The Métro doesn't use farezones; the RER does. If you want to include the RER in your calculations, then you should be including LIRR and Metro-North tickets. Taking the LIRR or Metro-North just one stop outside city limits is more expensive than taking the RER to zone 6.

Far Rockaway is an aberrant low-ridership branch. You should use Jamaica Center or Brooklyn College or Flushing as the outbound terminus for comparison, not Far Rockaway.

George K said...

The problem that I see with the cost recovery ratio is that you are dividing the average fare paid ($1.14 weekdays, $1.29 weekends) by the cost per passenger. The problem with this is that that the average fare paid is different on different routes.
For example, the crosstown routes, such as the M23 have a lot of cnnections to subway lines, therefore, this means that a high percentage of the passengers are probably transferring from the subway. I don't know if they are still making a profit, but the cost recovery ratio is probably lower than what is stated in the spreadsheet.
On the other hand, the Q58 is a long route, with fewer connections to the subway. This route probably has cost recovery ratio that is probably slightly over 100%, since the average fare paid on this route is probably higher than the system average.

Cap'n Transit said...

Thanks for your comment, George. Do you think that the use of an unlimited monthly Metrocard counts as a "fare paid"?

George K said...

I would probably say that it would count as a passenger boarding the bus. I would say that it would count the same as a transfer (no fare paid).
I don't think that they would say that ''x'' number of bus rides were taken on the local bus with the Unlimited MetroCard that cost $''y'', and figure out the cost per ride, and do that for the whole system, though I could be wrong.

Cap'n Transit said...

In January, more than half of all non-student trips were paid for with unlimited Metrocards. If they counted all those trips as one passenger, no fare paid they'd get grossly unreliable results.

If I were the one processing the data, I would figure out the cost of every unlimited ride as a share of the monthly price, and use that to calculate the average fare (which is $1.14 on the NYCT buses, not $2.25). I would do the same thing for transfers: divide each transfer by the number of trips it pays for, and use that as the fare paid. I'm pretty sure that's what the MTA is doing.

It's true that that does cause a lot of variation in the fares paid on individual routes that is not reflected in the averages, but I don't think that the MTA would discontinue a route because it attracts lots of transfers, or keep it because it has a relatively low number of senior citizens riding. Their strategy is just to get people riding the system, and I think that's a good one.

George K said...

Transfers aren't really a problem, since the MTA would assume that a person taking a bus to the subway in one direction would reverse the commute in the other direction. So in the end, it balances out. For example, a customer taking the B31 bus to Kings Highway in the morning for the subway would probably take the subway to the B31 in the afternoon. In the morning, the fare is counted on the B31, and in the afternoon, the fare is counted on the subway.
All people who board are definitely counted, since there are buttons to press for farebeats/shortdrops (people with only a partial fare paid on the MetroCard) who board the bus, so in the end, it is almost always cost efficiency that counts when deciding whether or not to discontinue a route.
What I said in the post (in the second paragraph) was that I wasn't sure if the MTA had ways of dividing fares paid over the number of trips taken (like on Unlimited MetroCards and transfers). Now that I think about it, the MTA, which should have pretty sophisticated technology, considering the size of the system they run, would probably have a way of allocating fares to certain routes.

George K said...

I was just looking at this sheet again and realized that the weekday fare paid and weekend fare paid were the same. The M14 is listed as a 100% recovery ratio when the average fare paid on the weekend was $1.29. According to the formula that was used to approximate the fare recovery ratio, it runs at a 113% farebox recovery ratio.
Here are the weekend farebox recovery ratios for the routes specified, using the formula you used:
M86: 172%
Bx12: 139%
M79: 134%
Bx19: 134%
B74: 132%
M23: 123%
B35: 117%
Bx9: 116%
Bx35: 115%
M14: 113%

Also, @ James D:
The only problem I have with your reasoning is that if low-ridership routes are eliminated, many of their their former riders will switch to nearby routes (the average percentage depends on how convenient the other route is), which will cause the surrounding routes to become more cost-efficient.
Some routes just can't be touched, regardless of if they are at the bottom. They are either "coverage" services, and removing them would put neighborhood residents beyond a reasonable distance of a bus route, or just politically can't be touched. Either way, they can't be eliminated. Just look at routes like the S54/S57. They run at about 25% efficiency (like I said, it is probably more since they are long enough that somebody wouldn't have to transfer to another route, resulting in a higher fare paid, but regardless, they are very inefficient.), yet because they serve areas not served by other bus routes, like Seaview Hospital, they are saved.

Cap'n Transit said...

Thanks for catching that, George! I've posted revised numbers.