When to double-track?

[oltmannd]

Let's see if I can get this.  Using Michael Sol's figures from way back, on a typical 120-mile track running 35 trains/day, they average 184 moving hours and .88 x 184 = 162 siding hours.  Someone else mentioned $1000/hr for waiting in a siding.  That's $162,000/day for waiting in sidings.  Give them 300 days/year, which comes to about $48.6 million/year for each 120 miles of track - wasted crew time.

If it cost $4 million/mile for a second track, that comes to $480 million for a 120-mile added track.

Keeping it simple, I assume no siding time for double track.  Then the new track pays for itself in ten years in avoided wasted crew time.

At the higher costs, say ten times $4 million/mile, it would take 100 years to break even.

It seems to me that that would be a good investment, considering that the ROW doesn't deteriorate, once built.  I.e., the lifetime of the double track easily is 10-100 years.

This does not take into account the likelihood that the RR will get more business if it can deliver more quickly.  I.e., the new double track line may well run more than 35 trains/day.

 

A simple pay-back period of ten years is most likely well below a RRs ROI hurdle for new investment.  It isn't helped by that much of the savings is "soft", .e.g. you are saving crew hours, but enough that it reduces staffing or hours paid or allows long pools?  Are you really saving car ownership cost or just moving the dwell from the road to the yard?

It is really difficult to get your hands around all the benefits double tracking in hard dollars - which make it hard to say when is the right time to do the double tracking.

[cordon]

Let's see if I can get this.  Using Michael Sol's figures from way back, on a typical 120-mile track running 35 trains/day, they average 184 moving hours and .88 x 184 = 162 siding hours.  Someone else mentioned $1000/hr for waiting in a siding.  That's $162,000/day for waiting in sidings.  Give them 300 days/year, which comes to about $48.6 million/year for each 120 miles of track - wasted crew time.

If it cost $4 million/mile for a second track, that comes to $480 million for a 120-mile added track.

Keeping it simple, I assume no siding time for double track.  Then the new track pays for itself in ten years in avoided wasted crew time.

At the higher costs, say ten times $4 million/mile, it would take 100 years to break even.

It seems to me that that would be a good investment, considering that the ROW doesn't deteriorate, once built.  I.e., the lifetime of the double track easily is 10-100 years.

This does not take into account the likelihood that the RR will get more business if it can deliver more quickly.  I.e., the new double track line may well run more than 35 trains/day.

 

[beaulieu]

"demarketed" often results in raising pricing to the prevailing price, with the understanding that if you retain it, then it will be profitable, if you lose it, then capacity increases.

This is a very interesting time in a number of markets (not just rail) as low demand is beginning to result in lower volumes, which often leads to aggressively lower pricing.  However, with so many commodity costs at historic high levels, the resulting price cuts are not quite as prevelent.  The company I sell for today increased "labor costs standards" for the first time in years.  Raw material prices have been bumping upwards now for about three years.

In good times you can be selective with your pricing and orders...this is not the time to do that, yet costs are up. 

ed

 

This is a result of the Global Economy, demand maybe down within your marketing area for your finished product but still strong globally for your raw feedstock, so you get squeezed in between.  

[MP173]

"demarketed" often results in raising pricing to the prevailing price, with the understanding that if you retain it, then it will be profitable, if you lose it, then capacity increases.

This is a very interesting time in a number of markets (not just rail) as low demand is beginning to result in lower volumes, which often leads to aggressively lower pricing.  However, with so many commodity costs at historic high levels, the resulting price cuts are not quite as prevelent.  The company I sell for today increased "labor costs standards" for the first time in years.  Raw material prices have been bumping upwards now for about three years.

In good times you can be selective with your pricing and orders...this is not the time to do that, yet costs are up. 

ed

[SALfan]

This is the most interesting thread in a long time.  Some of the posts I've had to read two or three times to have a dim understanding of what is being said, but that is a function of my lack of knowledge and not the fault of the poster.  Thanks everyone for all the thought-provoking information.

RWM made a comment about traffic on lines he was formerly responsible for fluctuating violently from one day to the next.  Is that generally still the case?

Does anyone know of instances where less-profitable traffic was "demarketed" to increase capacity for more-profitable traffic?  Is that being done now, due to the capacity problems in the industry?  Just curious how often it happens.

 

[n012944]

n012944:

Thanks for the response.  It sure seems that a dispatcher would have a better handle on the operations of a line, and particularly the trains, crews, terrain, etc.

You live in Valpo?

ed

Yea, the automatic feature sounds good in theory, but thats about it. 

 I just moved to Valpo a couple of months ago.

[MP173]

n012944:

Thanks for the response.  It sure seems that a dispatcher would have a better handle on the operations of a line, and particularly the trains, crews, terrain, etc.

You live in Valpo?

ed

[MichaelSol]

One of the mistakes people make with simulators is they will use the simulator to count how often a particular siding is used or a particular turnout is thrown, and they pick out the low outliers and decide those particular sidings are not needed.  That ignores network effects, and often when they rerun the scenario without that siding the traffic congeals. 

When the Frisco took over the BN, they brought the "Lou Menk" attitude with them in spades and were in full fury to "rationalize" the system by eliminating "little used" or "redundant" yards and sidings. In our neck of the woods, that was sidings here and there, a yard at Wishram, and some other things. One of the BN dispatchers, a man of long experience, estimated that the "Frisco Toads" eliminated about one-third of the fluidity of the lines in the Seattle region, and gradually, their mistakes had to be undone. Whether or not that was based on simulations, I do not know, but suspect as much ...

 

[n012944]

How much real "dispatching " is done with today's modern systems?  In other words does the system essentially set up the meets with the ability for there to be override?  I understand with maintainence, inspections, etc there will always be a need for dispatcher input, but what are the systems of today ( and tomorrow) capable of.

One of the biggest deterents to the single line CTC/siding situation is the grade crossings.  I often hear the NS having to decide whether or not to cut the train.  That time loss can really make certain sidings ineffective for trains of sizes that exceed the effective length of the siding (from switch to grade crossing).

No doubt if the railroads could go back 100 years and redo their lines, it would be different.

ed

 

On our dispatching systems, there is an automatic feature, which will allow the computer to "run" the railroad.  It will make meets and passes according to a trains priority.  I do not know of any dispatcher that uses it in our office, as it does not do that great of a job.  It will make some questionable meets, and does some other wierd things.  It does not factor in such things as road crossing in the sidings, or physical characteristics of the road. 

[Anonymous]

railroads are constantly short on crews & power..only when "trains held" statistics start piling up do they realize "maybe we could run more trains with more crews/power" a crew sitting on a train that is stopped is a crew that could be sitting at home getting rested & ready to move another train...it's a classic case of losing a dollar to save a dime.there's nothing sophistacted about this very basic fact...you can't run trains without crews...the carriers will one day have to abandon the long held belief that crews have to be worked as close to the hog law as possible...if nothing else it is a safety issue that is gaining more recognition by the public.

The analysis which RR's apply to all facets of their operations, expenses, revenues and projected revenues is so much more sophisticated today than it was even ten years ago. They are more 'on top' of their business than most posters here realize.

Since I am 17 years retired I do not know how crews are compensated; perhaps Dweezil or someone else can tell me if crew costs are extra for waiting on meets if they do not exceed the 12 hour law.

[MP173]

Ok, let me shift gears here slightly.  We have the simulation software and hardware to run all of these scenarios. 

How much real "dispatching " is done with today's modern systems?  In other words does the system essentially set up the meets with the ability for there to be override?  I understand with maintainence, inspections, etc there will always be a need for dispatcher input, but what are the systems of today ( and tomorrow) capable of.

One of the biggest deterents to the single line CTC/siding situation is the grade crossings.  I often hear the NS having to decide whether or not to cut the train.  That time loss can really make certain sidings ineffective for trains of sizes that exceed the effective length of the siding (from switch to grade crossing).

No doubt if the railroads could go back 100 years and redo their lines, it would be different.

ed

[erikem]

The software will set you back mid-five numbers plus annual support contract.

In other words about the same order of magnitude as many engineering packages (e.g. Ansys) and annual support is typically 10% of the purchase price. Also similar to many engineering packages, if you need the capability, the price is typically worth it. The type of simulation is esoteric enough that I doubt that there would be sufficient interest in the Free Software community to write something similar - though it would make for a great 'engine' for a dispatcher game.

I imagine that the costs of collecting and inputting the data would easily equal or exceed the purchase price for the software.

Appreciate the info.

- Erik 

[Railway Man]

Lost irony ... it does take a heck of a laptop but if you have that laptop, it will run it.

The software will set you back mid-five numbers plus annual support contract.

RWM 

[erikem]

That said, we now have something better than the formula -- the computer simulation -- which solves the formula simultaneously and allows for variables that are not present in the forumula. 

Is the computer simulation package a proprietary program (sounds a bit like Espee's TOPS)? If not, it sounds like a fun package for us computer oriented types to play with. What might be even more 'fun' is to modify it to simulate an electrification (i.e. effects of line voltage, substation spacing, etc.).

I'm assuming that the package can take weather into account, with allowances for reduced factor of adhesion for wet weather, longer times to pump up the trainline in extremely cold weather, increased number of CWR pull-aparts in cold weather, CWR sun kinks in hot weather, etc.

This reminds me of a story I heard about BART (not vouching for the truth of the story) - train operations were simulated on a computer and system capacity was projected from those simulations. When it came time to run trains on the system, the capacity was substantially less than what was simulated. When the simulation program was analyzed, there was a line of code that set the headway to 120 sec if the headway ever dropped below that figure. 

As for your reply to Michael about laptops, bear in mind that some laptops are now shipping with 3GB of RAM, which is probably enough to simulate a pretty long chunk of track. Processor performance should be within a factor of two of any existing single processor system. Though I doubt that Excel would be the best tool for this kind of modeling (could be done, but it would be painful).

[Railway Man]

I am loving this discussion, but having difficulty grasping it. 

For one thing, as stated by RWM, there are so many variables attached to this that it is very difficult to make a definative determination, unless you have the "inside information".  The traffic mix (coal, intermodal, single car, grain, etc) and the pricing involved with each type of traffic, plus the performance requirements of each, makes this a computation that my slide rule "cannot compute".

As in any business, the key component is going to be the additional revenue per unit that the expansion brings.  Looking at the Transcon's expansion, which is based on low rated intermodal, it is obvious (sort of) why BNSF held off as long as they did. 

Now, I turn my attention to a couple of real life examples.

First, one of my hometown lines is the ex NKP between Ft Wayne and Chicago.  Pretty decent track on level Indiana cornfields, with 6,000 - 8,000 foot sidings every 8 miles or so.  Traffic is as follows 16 intermodal/auto racks daily, 11 boxcar trains daily, plus a round trip local, and an average of 2 loaded and 2 empty coal trains.  Figure 30-35 trains daily.  The intermodals run pretty hot and hold a tight schedule, particularly out of Chicago.  Inbound to Chicago is a different item as they must hold for terminal space.  Also, of those 30-35 trains, about 10 peal off the mainline at Gary/Hammond and must receive clearance prior to moving to those lines (IHB and EJE).  I have noticed that about 35 per day is the tipping point.  Anything more than that and trains fill the sidings, recrews occur and tension builds.

RWM...how does your simulation account for the uneven spacing of availability of "landing slots" in a terminal such as Chicago?  Or the offline movements such as to the Harbor or J?  Is it based on historic movements of these trains? 

Second point.  I grew up in Southern Illinois near Olney which had the B&O St Louis - Cincinnati line.  Decent track today, not great, probably 40mph, signaled with old B&O CPL (honorary semaphore signals in my book) and hand thrown sidings, about 15 miles apart.  CSX, has really wanted to get rid of this line but never did.  They talked of it 20 years ago, but somehow the line has hung on, even after the Conrail merger.  With it's 8-12 trains a day, it has to be a money maker these days...or does it?

New coal moves has meant tie and rail investment.  East of Mitchell they were even doing some serious work last spring to get the line up to speed to St. Louis - Cincy traffic again. 

Yet, CSX over the years diverted considerable traffic off the line to the Indy line, possibly to take advantage of the Avon yard.  So, along with the obvious issues of when to expand, how does a carrier decide to cut a line? 

If the above didnt make sense, I have had a couple of phone interruptions plus a couple of homework questions...I guess that is called multitasking.

ed

Intermodal isn't low-rated anymore.  Rates have been advanced quite a bit and they're sticking.  It's the coal business that's getting pushed off the track in some cases. 

Simulations are only as good as the scenario you feed them.  You can simulate any condition you wish and as many as you wish until your budget runs out, but how do you know you have chosen the conditions that will obtain in the future?  Well, you don't!  That's where experience and intuition come into play.  Generally in a simulation I'm looking to see how marginal the railroad appears for a given traffic level, traffic type, and fixed plant, and for recurring choke points under different scenarios.  Sometimes when you run a scenario for, say, 7-20 days (sped up), you see things that are quite surprising.

One of the mistakes people make with simulators is they will use the simulator to count how often a particular siding is used or a particular turnout is thrown, and they pick out the low outliers and decide those particular sidings are not needed.  That ignores network effects, and often when they rerun the scenario without that siding the traffic congeals. 

RWM

[Railway Man]

     Is there a great jump in efficiency, in areas where directional running is used? (Two parallel tracks, that work like one way streets).

Yes, if you can start running trains that previously would have been too long to fit in the sidings, or if you were otherwise going to have to invest in CTC or longer sidings.  Directional running is no different in effect from double-track (current-of-traffic, as opposed to 2 main tracks which are bidirectional) except that the two tracks are usually not on the same right of way.  Early examples include the Joint Line of Santa Fe and Rio Grande from South Denver to Bragdon (Pueblo), Colo., and the Paired Track of Southern Pacific and Western Pacific from Weso (Winnemucca) to Alazon (Wells), Nevada, both of which date from the USRA during WWI.  A notable example of directional running on the same right of way is the dark-double track of Rio Grande and Colorado & Southern between Southern Junction (Pueblo) and Walsenburg, Colo., about 50 miles.

Directional running gives you a lot more ton-miles per crew start.  There's also a savings in fuel and time spent decelerating to a meet, hogging into the siding, and accelerating out the other end after a meet.  In the long run there's a very large savings in not having to build sidings that fit trains, and you might not even have to install bidirectional (CTC) signaling but just make do with directional ABS.  The first drawback is if the directional lines do not have co-located crew terminals there's a van and crew time cost to take a crew from the tie-up terminal to the other terminal heading back the other way.  The second drawback is when you want to take one line out of service for maintenance or you have a wreck; the other line cannot accommodate bi-directional traffic and you might have to fleet trains, clogging yards at each end badly and melting down the subdivision in a matter of hours.  The third drawback is that wrong-way trains, such as locals or unit trains running to or from a customer that's off one track only, can be a big killer of directional running fluidity.

There are some single-track lines that run trains that fit into the sidings in one direction, and overlong trains in the other direction, which will move more ton-miles per crew start in at least one direction but results in a crew surplus in the long-train direction.  If the line has a terminal at one end that trains frequently have trouble getting into but not the other end -- the end with the held-out terminal is the dogcatch direction and the dogcatch crews use up the surplus crews that would have been deadheaded the other way.

RWM 

[Murphy Siding]

     Is there a great jump in efficiency, in areas where directional running is used? (Two parallel tracks, that work like one way streets).

[MLG4812]

It was a tongue in cheek comment.  Clearly, the cost of having to maintain a second mainline far exceeds the incremental add on cost of crews burning up hours waiting on meets.

In fact, it's only because the RR's have managed to shed many miles of double track (and it's related cost to maintain)  in this post-passenger train era, that crew costs are next in line as a factor to be dealt with.

Putting in more miles of dual main as a strategy to control payroll costs, would be folly

And....enter remote control..Seriously though, it would seem that the only cost effective and sensible reason to add a second mainline nowadays would be to facilitate "boom" traffic. The intermodal "boom" of the past couple of decades. The Powder River coal "boom" of the past several years. (Maybe there will be a "corn boom" for the extra ethanol in the near future.) Many double mains of yesteryear can handle the same level of traffic through CTC today. If you study most of the dual mains in this country...take away the passenger mix, don't count the duals over mountainous grades...we're talking just plain traffic...the only dual lines I've found that are actually needed are those carrying large amounts of intermodal or coal (along with the other traffic.) Mr. Dweezil, it would be folly indeed to add otherwise.

MLG4'8.5"

  

[Anonymous]

One would think that with all the whining you hear from the RRs about cutting crew costs, they'd see how much they are pouring down the drain paying crews to wait for hours on end for meets, in single track land

You would be surprised at how long trains sit at a CTC control point in CTC/two main track territory waiting for more "important" trains.

Train to Dispatcher:  "UP (number) West looking for a signal at CP A300."

Dispatcher to Train:  "I'm going to hold you at CP A300 to get the bird (Z-train) around you.

Train to Dispatcher:  "OK, where's he at?"

Dispatcher to Train:  "Going thru Boone now."

Boone is MP 202, CP A300 is MP 300.  

Jeff  

 

It was a tongue in cheek comment.  Clearly, the cost of having to maintain a second mainline far exceeds the incremental add on cost of crews burning up hours waiting on meets.

In fact, it's only because the RR's have managed to shed many miles of double track (and it's related cost to maintain)  in this post-passenger train era, that crew costs are next in line as a factor to be dealt with.

Putting in more miles of dual main as a strategy to control payroll costs, would be folly

[MP173]

I am loving this discussion, but having difficulty grasping it. 

For one thing, as stated by RWM, there are so many variables attached to this that it is very difficult to make a definative determination, unless you have the "inside information".  The traffic mix (coal, intermodal, single car, grain, etc) and the pricing involved with each type of traffic, plus the performance requirements of each, makes this a computation that my slide rule "cannot compute".

As in any business, the key component is going to be the additional revenue per unit that the expansion brings.  Looking at the Transcon's expansion, which is based on low rated intermodal, it is obvious (sort of) why BNSF held off as long as they did. 

Now, I turn my attention to a couple of real life examples.

First, one of my hometown lines is the ex NKP between Ft Wayne and Chicago.  Pretty decent track on level Indiana cornfields, with 6,000 - 8,000 foot sidings every 8 miles or so.  Traffic is as follows 16 intermodal/auto racks daily, 11 boxcar trains daily, plus a round trip local, and an average of 2 loaded and 2 empty coal trains.  Figure 30-35 trains daily.  The intermodals run pretty hot and hold a tight schedule, particularly out of Chicago.  Inbound to Chicago is a different item as they must hold for terminal space.  Also, of those 30-35 trains, about 10 peal off the mainline at Gary/Hammond and must receive clearance prior to moving to those lines (IHB and EJE).  I have noticed that about 35 per day is the tipping point.  Anything more than that and trains fill the sidings, recrews occur and tension builds.

RWM...how does your simulation account for the uneven spacing of availability of "landing slots" in a terminal such as Chicago?  Or the offline movements such as to the Harbor or J?  Is it based on historic movements of these trains? 

Second point.  I grew up in Southern Illinois near Olney which had the B&O St Louis - Cincinnati line.  Decent track today, not great, probably 40mph, signaled with old B&O CPL (honorary semaphore signals in my book) and hand thrown sidings, about 15 miles apart.  CSX, has really wanted to get rid of this line but never did.  They talked of it 20 years ago, but somehow the line has hung on, even after the Conrail merger.  With it's 8-12 trains a day, it has to be a money maker these days...or does it?

New coal moves has meant tie and rail investment.  East of Mitchell they were even doing some serious work last spring to get the line up to speed to St. Louis - Cincy traffic again. 

Yet, CSX over the years diverted considerable traffic off the line to the Indy line, possibly to take advantage of the Avon yard.  So, along with the obvious issues of when to expand, how does a carrier decide to cut a line? 

If the above didnt make sense, I have had a couple of phone interruptions plus a couple of homework questions...I guess that is called multitasking.

ed

[jeffhergert]

One would think that with all the whining you hear from the RRs about cutting crew costs, they'd see how much they are pouring down the drain paying crews to wait for hours on end for meets, in single track land

You would be surprised at how long trains sit at a CTC control point in CTC/two main track territory waiting for more "important" trains.

Train to Dispatcher:  "UP (number) West looking for a signal at CP A300."

Dispatcher to Train:  "I'm going to hold you at CP A300 to get the bird (Z-train) around you.

Train to Dispatcher:  "OK, where's he at?"

Dispatcher to Train:  "Going thru Boone now."

Boone is MP 202, CP A300 is MP 300.  

Jeff  

[J. Edgar]

i dont know about cost percentages and traffic volumes but i know a case in point locally....the Pere Marquette spent good late 1930's money "preparing" to double track its main from Plymouth MI to Porter IN...this included many state roads bridged with double track abutments that still stand carrying CSX trains on single track....appearantly it was decided that CTC with DSPR controled remote switches and signals would be sufficent.....whether this desicion was biased on the pending/looming merger with C&O i havent seen.....yet

[MichaelSol]

....  what is the primary cost factor when the profitability falls?  Or is it simply an accumulation of several factors?

As you noted, with increasing traffic density, movement slows. Railroads are networks and that is a network reality whether its railroad cars or information packets.

In the example cited above, you can see that the transit time increases by 50% from the ten train per day optimum to the capacity of 35 trains. From the standpoint of train crew, most of that time increase falls into the upper margin of the pay scale -- overtime and even dead on hours. More importantly, however, assuming each example is a steady state, the railroad needs 50% more employees per carload. So, three costs have leveraged -- the pay of the individual employee, the cost of additional employees, and the railroad's cost of changing dead crews.

As you noted, the same thing happens with equipment. Locomotive utilization declines by 50% between a 10 train matrix and a 35 train matrix. Further, each locomotive is using a lot of energy starting and stopping trains for all those meets in the 35 train matrix. Fuel use per carload is substantially higher on the 35 train matrix as a result. Again, assuming a steady state, the railroads needs 50% more locomotive hp to move each ton of freight because such a large proportion of hp is sitting on a siding somewhere.

The same is true for the car fleet. The car cycle time increases by 25% in this example (because not all of the cycle time is spent on a train). Either the railroad or the shipper has to pony up for a substantially larger car fleet to carry the same tonnage.

Too, there is a maintenance penalty. The straight mileage prorate that used to be a good rule of thumb for estimating costs by the amount of tons being carried is no longer valid. Two things: at 613 meets there are a lot of moving parts being used on the physical plant that are not being moved in the ten train matrix (50 meets). More significantly, higher tonnages resulting from the use of 100 ton+ cars and the installation of higher weight rail has caused the cost of maintenance of the high utilization track to be higher than, say, the expected proportional cost of the ten train track. That's not my baliwick, but I may have a formula for that around here somewhere; but, it is an additional cost consideration that raises the variable cost. 

The interesting thing about the models is that they show the optimum cost efficiency is always between 8 and 11 trains per day. Doesn't matter what you do, because the variable cost increases that occur above that range (and the proportion of fixed costs below that range) occur because of the laws of physics and there's nothing the Marketing Department can do about those. So profitability will always maximize at that traffic level.

To be clear, and I'm not addressing this to you Ed because I know you know the difference, but to others who may not: "profitability" and "profit" are different concepts; one's a percentage and one's a number.

 

[Railway Man]

Today, Microsoft Excel and a standard laptop can handle the most sophisticated of such models and provide instantaneous graphic representations of the results across a broad spectrum of alternatives. Using "Solver" can manipulate a wide array of the variables to determine an optimum of any given characteristic.

That will handle the economic models but the train simulation software is expensive and it takes one heck of a laptop to run it.

Some of the economic models are pretty cool and some of them make you wonder what the economists are smoking.

RWM

[MichaelSol]

Actually there is nothing wrong per se with the formula.  ...

That said, we now have something better than the formula -- the computer simulation -- which solves the formula simultaneously and allows for variables that are not present in the forumula.  The steps are as follows:

Exactly, and that is what I meant earlier about what Healy would have been able to identify insofar as what was causing the diseconomies of scale in terms of what he was seeing from the financial results had he had the computing power available.

The usefulness of the basic equation is to demonstrate the frictional costs of traffic. From five trains a day with 13 meets, multiplying the trains to 35 per day doesn't just increase the number of meets proportionately, rather there is a geometric character to the 613 meets that must occur on the section of track with that many trains. And, within the thread context of noticing trains on sidings, an interested observer can look at the equation and instantly get a sense of just how much time has to be spent not doing anything for the railroad except sitting there, and the causes for that.

And also see and understand why the variable costs of operation go up, not down, with increasing utilization of capacity past the economic optimum.

The computer programs were developed around these equations in the mid to late 1960s. The programming with Fortran or Cobol was tedious and time-consuming, running it through the compiler was boring, as was the inevitable "wait" to get to use the IBM 360 that was something of a standard at the time.

Today, Microsoft Excel and a standard laptop can handle the most sophisticated of such models and provide instantaneous graphic representations of the results across a broad spectrum of alternatives. Using "Solver" can manipulate a wide array of the variables to determine an optimum of any given characteristic.

A knowledgeable Economist like Kent Healy would have had a field day because these programs explain so much regarding the underlying causes of broad economic trends that he saw within the industry.

 

[Railway Man]

Ed:

Everything in railroading is case-specific.  The optimum profitability of a line depends on what the trains are hauling and what possible traffic the line might have instead of the existing traffic, as well as the totally idiosyncratic geographic characteristics of the line.  A coal line is very different from a line with mixed manifest, intermodal, and bulk, from a line that's dominated by LTL and TL intermodal. 

We first talk in terms of absolute capacity of a CTC single-track line, which is a range from 25-70 trains per day depending upon grades, siding length, siding spacing, train length, and hp/ton.  Generally speaking a line biased toward premium intermodal might have an absolute capacity in the vicinity of 70 trains per day if and only if the track has the engineering bones to carry that off.  A single-track line that was never super-railroaded in the early 1900s (e.g., not the UP-SP Overland Route, Santa Fe Transcon, PRR main, NYC main), with a lot of curvature and hills, with mixed bulk, unit, intermodal, and manifest, has an absolute capacity in the vicinity of 35 trains per day, as do super-railroaded lines that could not improve grades beyond 1.8-2.0 percent.  A single-track line that is purely bulk with 2.0 percent or steeper grades has an absolute capacity in the vicinity of 25 trains per day.

"Effective capacity" is in the range of 70% of absolute capacity -- by effective I mean the amount of traffic beyond which train delays become significant.  That's a consensus number from us chief dispatchers, network planners, schedulers, maintenance managers, and superintendents.  Beyond 70% it is almost impossible to get MOW the time it needs.  However, a line can accommodate absolute capacity some days so long as you're not trying to do that every day. 

"Ideal capacity" -- the number you are seeking, I think -- is capped by effective and absolute capacity and requires you know a great deal about the future traffic on the line, how much that traffic will pay, and whether if you accept that traffic on that particular line if it will enhance or diminish the value of the traffic you will now or in the future enjoy on all your other lines plus your competitors (including truck and water) and connections.  The formula, for example, might tell you that your most profitable traffic is 10 trains per day of foreign cans at 5.0 cents/ton-mile but if you tried to discourage your 2.0 cents/ton-mile coal in order to make room for more cans, you might provide enough base coal traffic to incentive a coal mining company to open a big new mine on your competitor and take all the coal you don't want plus the coal you do want that is now using another of your lines.

You cannot treat any line in isolation; the marketing decisions you make there ripple to all of your other lines. 

If there were a formula to deliver meaningful outputs that consolidated finance, operating, traffic, marketing, sales, engineering, and mechanical into one function that spat out an answer, then railroads wouldn't need me and a lot of people like me, and the railroads of course would be perfectly happy to dispose of us since they think we charge too much and deliver too little.  There is no button to push that says "calculate business I ought to have, how much I should charge for it, and how much railroad I should build to put it on" just as there is no button on a 747-400 that says "Push here to autoland plane at nearest airport."  The formulas are useful but you have to know what to do with them.  

If I had such a formula, I would treat it as proprietary and make a killing.

In the meantime, while I probably can't answer any specifics about any current-day or recent-history line without disclosing proprietary information, or making commentary about business practices of my clients, if you can posit any "test cases" complete with specifics I can at least tell you about the decision limits and data inputs.

RWM

[MP173]

This is a fascinating thread. 

Michael, what is the primary cost factor when the profitability falls?  Or is it simply an accumulation of several factors?

Most companies (industries) find that leveraging their operations result in higher profits.  This is of course simple economics of factoring in fixed and variable costs.  It appears the study sited finds that leverage doesnt apply to the railroads bottom line...or am I missing something.

Railway Man...your current application of the problem is most interesting.  Generally speaking, what does the typical simulation show as far as ideal operations of single track, CTC operations?  In other words, what is the rule of thumb for operating efficiency for a 200 mile terminal to terminal operation without any unusual operating conditions and adequate HP/tonnage, etc?  I know...this is a loaded question as each division is different and there are no typical operations, but this is pretty neat stuff.

One final note.  It seems that asset utilization is critical in profitablity of a company.  On one hand with high usage railroad, the fixed plant assets are highly utilized.  On the other hand, would the increased train meets reduce the asset utilization of rolling stock...thus resulting in the lower profitablity?

Just curious.

 

ed

[Railway Man]

1961  Communication between dispatcher and train crew:  Writing on tissue paper handed up at random locations along the route.

       Communication between train crews:  None.

Chances that a 1961 formula on train movement will produce a picture close to today's real world:  Iffy.

Actually there is nothing wrong per se with the formula.  Some of the other early looks at this included Leonor Fresnel Loree, 1924, "Railroad Freight Transportation," Clement Clarence Williams, "The Design of Railway Location," 1917, and more recently, J. Phillps, "A Method for the Calculation of the Capacity of a Single Track Railroad System, International Heavy Haul Conference 1978."   My favorite is a 1925 master's thesis from a Indian National Railways management trainee working in the U.S. on the Illlinois Central.

That said, we now have something better than the formula -- the computer simulation -- which solves the formula simultaneously and allows for variables that are not present in the forumula.  The steps are as follows:

1.  Input an stringline diagram of the line under study including the exact location and value for every grade, curve, turnout, siding, wayside signal, and permanent speed limit change.

2.  Input the train characteristics: length, hp/ton, A.C. or D.C. power or mixed, drawbar limitations if any, maximum speed allowed for that train type, and priority.  You can use the trains from any historical day for which you have records, or any future day you might want to create.  You can have locals hold the main to switch industries if you want.  If you have grade crossings through sidings the computer will hold trains off them until the opposing train arrives.

3.  Run the model.  The computer will auto-dispatch the trains and give them accurate acceleration and deceleration speeds.  You can watch the model as it runs and see if you like or dislike the dispatching decisions it makes.  You can insert a weather factor, a maintenance of way window or windows, or any other real-life event you care to use.

4.  Analyze the data.  The computer will tell you elapsed time, speed, fuel burn, and hours of delay for each train, and total and average for all trains.  The computer plots speed curve and horsepower ouput (both in dynamic and power) for each train.

5.  Modify the input.  Put in a siding where you think you might want it.  Usually the possible locations for the siding are limited by natural features or other expensive obstacles such as bridges, tunnels, road crossings that would require grade-separations, curves, etc.  Run the model again and compare the improvement, if any.  Try some alternate locations and see if it's any better.  Try double-tracking the whole railroad and see if it's any better.  Or any other improvement you care to try including things like changing unbalance on curves.

6.  Compare the model results for a historic day to the actual results for that day and calibrate.  Often the historic day does much worse because of some event -- a communication failure, a train tripped a detector, etc.  You can input those into the model and see what changes.

7.  After you have the model calibrated and you've located and arranged the physical plant improvements the model suggests are the best option, price the improvements.  Calculate net present value using the finance department's target discount rate and amortization period, and if the value is there, go talk to the board of directors.

RWM 

 

[oltmannd]

If there is an industry welded to its conventional wisdoms, it is undoubtedly the rail industry, and it is no different when it comes to "capacity". Most manufacturing industries know at exactly what level of operation best maximizes the economic efficiency of their machines. And, it's usually not "wide open." In the industry I am currently familiar with, the mills run their best at between 86% and 92% of "capacity". Above or below that and the economic efficiency -- profitability -- of the production falls off for a variety of reasons. The rail industry, based on no actual studies that I am aware of, continues to believe that operating at capacity is economically efficient.

 In 1961, Kent Healy produced his interesting study The Effects of Scale in the Railroad Industry (Cambridge: Yale University Press, 1961). During his study, he found the puzzling result that smaller railroads were generally more profitable than larger railroads, and through his analysis -- he was a Transportation Economics professor at Yale -- discovered that in the rail industry there were economies of scale but that, at a certain point, railroads suffered diseconomies of scale. The problem with railroads was that they were generally staffed by people who had 19th Century views of problems: and bigger was always better to those gentlemen, including the notion that consolidation of lines to increase capacity utilization generated more profit.

It didn't.

However, Healy's landmark study never took hold in the rail industry, notwithstanding the statistical evidence that his conclusions were absolutely true, because the conventional wisdom was simply stronger than the actual evidence. It was neither the first, nor the last time that conclusive proof of something would be dismissed by rooms full of cigar-smoking "insiders" proclaiming "that's BS. Everyone knows that such and such is what we need to ....". And they proceeded to rip out perfectly good mainlines to increase utilization of others -- and shot themselves squarely in their collective foot.

At the time that Healy did his study, a useful and well-known formula existed in the industry for calculating line capacity. This goes to the posts above talking about train delays and crew costs -- frictional costs of systems. The formula was developed by Ernest Poole while he was Director of Transportation Research at the Southern Pacific, and the "Poole Formula" is the generally accepted template for calculating the theoretical capacity of a given single track line configuration. It is described by Poole in his book Costs -- a Tool for Railroad Management, published in 1962, and his bibliography plainly shows the influence that modern methods of "cost analysis" had on his refinements, although he had been developing the formula since at least 1943, originally as a tool to analyze the effectiveness of CTC, to improve its utilization, and to predict and manage dispatching delays.

The nice thing about the Poole Formula is that it is readily accessible to econometric modeling -- assigning cost inputs, rates, cost of capital investment, interest rates, crew cost and size, cost of equipment and turnaround time, fuel, etc -- everything that goes into the revenue and expense of a railroad, to the ultimate determination of profitability.

Had Healy had that formula plainly in front of him, and the computing power to utilize it, he would have seen more clearly the economic principles underlying his findings regarding organizational size and the "diseconomies" he saw in the developed statistical record.

The source of most of the diseconomies of scale are shown by the Poole Formula -- which was one of the earliest and most accurate measures of "network" and capacity costs -- to result from simple laws of physics operating in network systems.

To make a long history short, the Poole Formula, adopted to an econometric model, shows why Healy got the results that he did in his analysis. The railroad "machine" works at its most effective economic efficiency at about 20-25% of capacity, 8-10 trains per day on a typically configured mainline. That reflected the typical mainline utilization of smaller railroad companies which did not have the traffic to "load up" their mainlines to achieve the conventional view of efficiency which was to operate as close to capacity as possible. Using the Poole Formula, a manager can clearly see why, at a given tariff rate, he might earn a profit of 12% on 7 trains per day, but suffer a negative 4.4% return on revenues at 20 trains per day, even though line capacity is 35 trains [an example I just took off of a specific model configuration]. System "friction" in the form of train delays and crew costs at 20 trains per day in that specific example adds 18% to the cost of hauling the freight -- a huge increase when considering that a 5% increase in the price of fuel throws everyone into a tailspin ... At capacity, 35 trains per day, the cost of hauling freight is 52% higher per ton than at 7 trains per day.

The effect was to increase the variable costs of operation with each additional ton of freight over the optimum -- with decreasing profitabilty even as revenues could be doubled and tripled by reaching line capacity.

Healy's review and discovery of the diseconomies of scale in the rail industry has a firm foundation in the laws of physics that govern networked systems, and the Poole Formula clearly shows why that is so.

By those measures -- mainlines today operating in general far above the economic optimum for efficient operation -- the rail industry is hugely undercapitalized at the current time, but obtaining the capital for expansion is handicapped by the historically high debt to equity ratios which were one of the fallouts from the Staggers Act.

 

Fascinating post!

The "everyone knows" school of mgt is still alive and well but there are more people attempting some science and availability of data and analysis tools are slowing chipping away at this.

[MichaelSol]

Seems to me that there is another point worthy of consideration.  Assuming a direct correlation between total transit time and operating cost, and further assuming that operating cost is on the order of 50% of revenue for the 5 train volume, total profit for the 25 train volume would be on the order of three times that of the 5 train volumn.

Personally, I'll sacrifice the per train margin for the total profit gain.  

Too simplistic. This is just guessing without any econometric justification.

Depends on the revenues. At a given revenue per carload, 5 trains per day under a given set of conditions can yield a 16% net profit, but at 25 trains per day can lose 2.6% of the revenues. The costs leverage with increasing congestion; the number of meets incurred with increasing numbers of trains is not an arithmetic progression. Revenue doesn't leverage.

You can, under specific circumstances of revenue per carload, ultimately gain more profit ($), even though suffering lower profitability (%), but then the risk factor increases -- which is why analysts look to profitability as a better measure of corporate strength than mere profit, even though under the right circumstances, you might prefer the raw profit. But there is greater risk ...