Comparing Route Segment Fares

[niemi24s]

Ryan recently posted a link to a study which compared fares between various stations along a route - the Crescent, as I recall. Thought it was interesting so did something similar for the train I use the most, the Empire Builder:



The graphics are a bit crude, but what's depicted are fares for five different levels of dissection for the route: halves; thirds; quarters; fifths; and twelfths. The stations closest to those dividing points were used without regard for anything else. The cost per mile for the segments in each tier were fairly consistent until the lower tier with 12 segments. In the lower tier, the highest cost per mile was more than twice that of the lowest cost per mile. But discounting the lower tier, there seems nothing unusual about any of the other fares.

Those who might want to do something similar for their favorite route are advised to first use AmSnag to find a block of dates having several consecutive days in which the fares to be studied are constant as well as the same in both directions. After selecting a date in the middle of that block, use Arrow to conduct the fare searches. If a fare seems higher than it should be, check the previous or following date - you may have stumbled across a situation where the sleeper fare has been jacked up (for whatever reason sleeper fares get jacked up).

 

[willem]

What fare are you illustrating? Adult with roomette?

Edit. The heading was too obvious, so I missed it. Duh.

 

[niemi24s]

Don't feel bad, happens to me more often than I'd care to admit.

Of course, there are many other ways of dividing up the route. One way would be to select a segment size - say 1/5 of the route or 441 miles - and "slide" that segment along the 30+ increments determined by the stations. Then compare the fares (or cost per mile) for each increment. Gawd, talk about excitement! :blink: <yawn>

 

[neroden]

I am absolutely certain that this is not the optimal pricing scheme for filling the train. The discounts for travelling further are too large, or put another way, the prices on the short segments should not *all* be that expensive.

 

[niemi24s]

I am absolutely certain that this is not the optimal pricing scheme for filling the train. The discounts for travelling further are too large, or put another way, the prices on the short segments should not *all* be that expensive.

Even considering these fares are for as Roomette? Be interesting to do the same sort of thing for a different route for comparison, but I'm not that bored!

 

[neroden]

The prices just aren't matching up to any plausible demand profile.

You'd expect the "peak segment" to have particularly high prices for short hops (to save it for long-distance travellers). You'd also expect the "weakest segment" to have lower prices for short hops (to fill in the empty roomettes: for instance, if there's a lot of Chicago-Willison traffic, you'd expect cheaper rates for Williston-Spokane).

There is simply no such pattern visible in the data here. What is visible is *huge* discounts for riding straight through rather than doing a stopover. I don't think this is financially sound.

 

[niemi24s]

I don't think this is financially sound.

You'd best contact Amtrak then, because it's a bit worse for the Silver Meteor. The average costs per mile for the segmented route, the entire route and their ratios are (based on 1 adult, low bucket sleeper):

• Empire Builder: $0.74 per mile segmented (average segment = 184 miles), $0.20 per mile not segmented (whole) for a ratio of 3.7

• Silver Meteor: $1.20 per mile segmented (average segment = 154 miles), $0.29 per mile not segmented (whole) for a ratio of 4.1

FWIW, low bucket Roomette on the Meteor for the whole route has increased from $239 to $251 since last checked on 29 Jan 2016. High bucket remains at $584, but there's now a total of 6 buckets for roomette fares. [shows how quickly I can get bored, huh?]

 

[neroden]

I'm doubtful that Amtrak has done serious detail-level travel demand modelling on most routes in recent years. I suspect their analytics division is short-staffed.

A tidbit I learned at the NARP/ESPA meeting: Apparently they only do a breakdown of riders/revenue by segment for one train route each year. Since Amtrak has more than 45 routes and is only 45 years old, this would mean that some of these routes have *never* had a breakdown done. At any rate, on most routes, the most recent breakdown would be way, way out of date.

Anyway, given this, I think the extreme level of the pricing bias against short trips -- all short trips, not just "peak segment" short trips -- and in favor of long trips -- is just habit on Amtrak's part. I doubt there's any rational basis for it.

 

[niemi24s]

For both of those previous studies (EB & SM), care was taken to insure fares (whether for the entire route or any of its segments of various length) were gleaned in the middle of a time period having constant low bucket Roomette fares in both directions. I suspect (but am not certain) if this is not done, there's some chance that as many as three different Roomette buckets could find themselves being included in a carelessly conducted study. On the other hand, a very brief look at SWC fares for high and low bucket dates showed that fares remained constant for a few short segments while the fare for the entire route just about doubled. Maybe bucket changes only trickle down a limited amount?

Beats me - I just find these things without bothering to explain them. They may well be unexplainable.

 

[jis]

Anyway, given this, I think the extreme level of the pricing bias against short trips -- all short trips, not just "peak segment" short trips -- and in favor of long trips -- is just habit on Amtrak's part. I doubt there's any rational basis for it.

It is not just Amtrak. This is standard practice in many parts of the world which run way many more LD trains in a day than Amtrak runs in a week. But usually that is conjunction with a second set of fares more suited for use on local/regional trains.

 

[neroden]

Well, it's totally different if you're explicitly trying to drive the short-haul passengers to *another train*; so I'd expect high prices for short hops in New Jersey where Amtrak would prefer people to take NJ Transit. When there isn't another train, it's a different situation; maximizing ridership and revenue calls for lowering the price on the "weakest link" on the route so that the train will fill up.

I don't see this in the pricing data anywhere, and I'm sure each of these trains *has* a "weakest link", though I'm not sure where it is. As an example of one of the few trains where I'm quite sure about the demand pattern, Reno-Grand Junction ought to be extra cheap compared to (say) Chicago-Omaha. But is it? Likewise, Birmingham-New Orleans should be quite cheap compared to anything on the north end of the Crescent.

 

[niemi24s]

One important thing I forgot to add about both of those previous studies was the direction of the trains: Empire Builder Westbound; Silver Meteor Southbound. While the direction didn't make any difference in the fares for the entire route, it usually does make a difference in the fares for the segments. For example, here's a brief portion of a high bucket Roomette fare search for the Empire Builder:

• Total route = $929 for either direction

• Eastern half, WTN to CHI = $283, CHI to WTN = $350

• Western half, WTN to SEA or SEA to WTN = $683 for either direction and about double the cost per mile.

Will try to find a way graphically display all that information for easier comprehension. Hmmm. Wonder if a fare search using the middle bucket Roomette upcharge would reveal anything different or more revealing.

 

[niemi24s]

After analyzing Empire Builder fares for 1 adult in a Roomette in both directions it was found that:

• Low bucket fares were the same in either direction for each of the 26 segments

• High bucket fares differed with direction in 21 of the 26 segments, sometimes by as much as 2X for the same segment

This is getting complicated! But mildly interesting!!

 

[PaulM]

One thing missing in this and similar discussions is what's already booked. My favorite case is Mount Pleasant to Denver. A couple of times I've booked a roomette and latter discovered I had to board in Galesburg, 2 stops earlier. The cost to add the Galesburg to MTP segment in a roomette was very high, as your analysis would predict.

Now once a room was booked MTP - DEN, GBB - MTP became virtually worthless to Amtrak since it is unlikely any one would be filling the segment. On the other hand, for a room that had not been booked, a high short segment price makes sense to keep the space open for CHI-DEN or even CHI-EMY.

Naturally, I booked coach GBB to MTP. My approach at GBB was "Sir, would you like us to schlepp our luggage, including folding bikes, from coach to sleeper through the train or via the platform?" In both cases the conductor had us board the sleeper. Clearly a case where the revenue managers shot themselves in the foot.

 

[niemi24s]

One thing missing in this and similar discussions is what's already booked.

And I see no reason to include it because what's already booked is (to us mortals outside of Amtrak's inner circle) both unknown and unknowable - up until it's sold out.

 

[niemi24s]

Another thing that complicates the analysis of fares for the Empire Builder is that the train splits/combines in Spokane. This somehow has the effect of causing occasional drastic differences in fares between two intermediate stations away from Spokane. As an example, the 1 adult Roomette fare from Malta to Libby MT is $393 on #7, but is only $180 on #27 on the same date of 29 Mar 2016. And 29 March is in the center of a 5 day span of constant low bucket fares in both directions. However when going from Libby to Malta on that same date, the fare is the same ($393) for both train #8 and #28.

This same sort of incongruity probably occurs on other "split" trains like the LSL & TE/SL.

 

[neroden]

Oh yeah -- on the LSL you usually want to book the "Boston section" if your trip is entirely west of Albany.

 

[neroden]

Better revenue management may actually depend on replacing ARROW. I mean, a modern system could probably be programmed to figure out that Mt. Pleasant-Denver is booked and discount the next Chicago-Mt. Pleasant ticket... no way that can be done with ARROW.

 

[niemi24s]

Oh yeah -- on the LSL you usually want to book the "Boston section" if your trip is entirely west of Albany.

I don't think booking presents any problem provided the booker(?) recognizes there are two different train numbers and books the cheapest of the two. But the different train numbers between the same two points with their different fares in the same or different directions throws a monkey wrench into a study like I've been trying to do. Simplest thing to do is look at a route that's not split - like the one already done for the Silver Meteor (no graphics yet for that one).

Plans for a look at mid-bucket fares on the EB were foiled by the lack of a suitably long period with the same fares in both directions.

Been planning on a trip to Seattle using the Zephyr, so maybe I'll do that one next - if I decide to do another one. Lots of eye strain, however.

 

[PaulM]

One thing missing in this and similar discussions is what's already booked.

And I see no reason to include it because what's already booked is (to us mortals outside of Amtrak's inner circle) both unknown and unknowable - up until it's sold out.

A discussion of whether the high short distance prices you observed make sense naturally followed. And I maintain that the answer depends on what's been booked.

 

[niemi24s]

One thing missing in this and similar discussions is what's already booked.

And I see no reason to include it because what's already booked is (to us mortals outside of Amtrak's inner circle) both unknown and unknowable - up until it's sold out.

A discussion of whether the high short distance prices you observed make sense naturally followed. And I maintain that the answer depends on what's been booked.

I agree that it depends on what's been booked. But not only do we not know what has already been booked, we cannot know what has already been booked until they're all booked - and then Sold Out sppears. Only then can we know what's been booked - all of them.

 

[niemi24s]

Here's the chart of the segment fares for the Silver Meteor I'd mentioned previously:



As usual and as expected, the short segments at the bottom have the highest cost per mile of the segments on this chart. They average $1.29 per mile with the RMT to FAY segment being the highest at $2.12 per mile. But for travel on a high bucket date such as 16 Apr 2016 it's a pricey $4.55 per mile! A ticket for that 90 mile trip costs $410 and because its about midnight, you don't even get a bite to eat!

The shortest possible segments are not shown on this chart and can be expected to have costs exceeding $10 per mile. Economy of scale in reverse. And that's perfectly reasonable as the booking of a roomette for a really short (20 or 30 mile) segment blocks the sale of that same roomette for any other segment connecting points on opposite sides of the shorter segment.

 

[niemi24s]

And finally, here's one for the California Zephyr:



Nothing unusual was noted in any of the cost per mile figures for the 24 segments. Here's a comparison of sorts for all three routes:

• Calif. Zephyr: $0.61 per mile segmented (average segment = 244 miles), $0.19 per mile not segmented (whole) for a ration of 3.2

• Empire Builder: $0.74 per mile segmented (average segment = 184 miles), $0.20 per mile not segmented (whole) for a ratio of 3.7

• Silver Meteor: $1.20 per mile segmented (average segment = 154 miles), $0.29 per mile not segmented (whole) for a ratio of 4.1

Cost per mile (in blue) for each segment added to each of the three graphics.

 

[neroden]

I don't think Amtrak's revenue yield management has progressed out of the 1970s. I suppose I should not be surprised. It probably requires a better underlying system than ARROW to do a better job.