Search This Blog

Monday, August 29, 2016

Bollocks

While researching the previous post about the University of Chicago, I came across this testimony to the bankruptcy of the social sciences, from, of all places, the UoC: Behavioral economics helps boost fuel and carbon efficiency of airline captains

The large-scale study, which incorporated data from more than 40,000 unique flights, found significant savings in carbon emissions and monetary costs when airline captains were provided with tailored monthly information on fuel efficiency, along with targets and individualized feedback. The behavioral effects of such interventions are currently estimated as the most cost-effective way to prevent a metric ton of carbon dioxide from entering the atmosphere.

This is what the study (hidden behind a paywall) had to say for itself:

Understanding motivations in the workplace remains of utmost import as economies around the world rely on increases in labor productivity to foster sustainable economic growth. This study makes use of a unique opportunity to “look under the hood” of an organization that critically relies on worker effort and performance. By partnering with Virgin Atlantic Airways on a field experiment that includes over 40,000 unique flights covering an eight-month period, we explore how information and incentives affect captains’ performance. Making use of more than 110,000 captain-level observations, we find that our set of treatments—which include performance information, personal targets, and prosocial incentives—induces captains to improve efficiency in all three key flight areas: pre-flight, in-flight, and post-flight. We estimate that our treatments saved between 266,000-704,000 kg of fuel for the airline over the eight-month experimental period.

Unless Virgin Atlantic Flight Operations is hopelessly negligent, this cannot possibly be true.

At my airline, flight plans specify the most economical available altitude and airspeed. Flight Management Systems are exquisitely tuned to use the best power setting for takeoff and climb, and are equally focused on attaining the best descent profile.

The Flight Operations Manual sets out criteria for Auxiliary Power Unit usage, single engine taxi, speeds and configurations on departure, and energy management on arrival.

All of this is backed up by frequent line checks, quality assurance observations, and continuous data downlinks.

The requirements and guidance are clear; the only variable is pilots' skill in achieving them within the context of the operational environment.

Spoiler alert: performance information, personal targets and prosocial incentives — uhh, oh never mind — have nothing to do with that.

Which is why I'm quite certain this is the last anyone will hear of this seminal! study!

21 comments:

erp said...

Skipper, come on. Everyone knows you guys spill fuel all over the oceans for fun. That is when you're not flying crisscross over the landscape to keep from getting bored flying over the same routes.

Fess up. s/off

Hey Skipper said...

Okay, that was a bit of a rant.

But I am absolutely certain they didn't ask any pilots about their conclusions.

erp said...

Skipper we thank you for your service to our country and the job you do now keeping commerce going and making sure the skies are safe.

Why would academics ask those with their metaphorical boots on ground anything. What do you guys know? Be thankful you don't work in academe.

Hey Skipper said...

This reminds me of the EPA's proposed rule making on increasing airliner fuel efficiency.

As if manufacturers haven't already been doing that to the very limit of economics and technology. I think it as likely that the knuckleheads at the EPA are any more aware of the reality of airliner economics than these "researchers" ("prosocial incentives", really?) are of airline operations.

So let's say the EPA gets its way, and the cost of doing so doubles the price of tickets. As a consequence, tourism to places like Hawaii sinks like a greased safe.

Now what?

erp said...

Teacher, teacher, pick me!

Hawaiian socialism takes a hit.

Clovis e Adri said...


I can't get myself to read their paper right now, Skipper, so I need to ask: their final clain only makes sense if they had something to compare it to, like the behavior right before their data was taken. What they are using as comparative?

Hey Skipper said...

[Clovis:] Skipper, so I need to ask: their final claim only makes sense if they had something to compare it to, like the behavior right before their data was taken. What they are using as comparative.

I have no earthly idea.

There is something, though, that I can't believe eluded me the first time. Here is there claim:

By partnering with Virgin Atlantic Airways on a field experiment that includes over 40,000 unique flights covering an eight-month period, we explore how information and incentives affect captains’ performance. Making use of more than 110,000 captain-level observations, we find that our set of treatments—which include performance information, personal targets, and prosocial incentives—induces captains to improve efficiency in all three key flight areas: pre-flight, in-flight, and post-flight. We estimate that our treatments saved between 266,000-704,000 kg of fuel for the airline over the eight-month experimental period.

Taking roughly the central value, that is 500,000kg saved over 40,000 flights, or 12.5kg per flight; roughly 27 lbs.

Airliner fuel quantity measurement systems have resolution to the nearest 100 lbs. 27 lbs of fuel amounts to about 7 mins of auxiliary power unit operation on my airplane; much less on a wide body.

The result they are claiming is less than measurement error. And APU usage -- which one could measure -- is driven by environmental and timing considerations. Plus, it isn't as if the airlines aren't hawking APUs.

erp said...

Skipper, I read the paragraph above earlier and as it made no sense to me, I inferred/assumed, it was some kind inside baseball jargon. Also, I've been pondering and pondering and can't imagine how "prosocial incentives" can incentivize pilots to improve post flight efficiency.

Clovis e Adri said...

Skipper,

---
The result they are claiming is less than measurement error.
---

Oh my. Now you just promoted this article to my "must read" list.

The authors are from the University of Chicago and the London School of Economics. If you are right, they made a mistake my sophomore students don't get to do with impunity.

Clovis e Adri said...

Skipper,

I went to look the paper in more detail. It is even worse than you thought.

All you need to know how bunk is their work is displayed on pgs 8 and 24.

To make it short: they assign points to the captains if they get to load fuel into the Heavy Machine within a 200 Kg error of the ideal calculation (pg 8).

Then on page 24 they finally explain how they converted such measure to actual fuel usage during flight: "VAA projects an average fuel savings of 250 kg per flight as a result of proper execution of Fuel Load."

Which means that, for every captain making an 'error' of more than 200 Kg on Fuel Load, they give them a 250kg Fuel 'penalty'. But if a Captain gets to Fuel the Machine, say, 199 in error, he will be counted as doing right, and no penalty.

The error introduced by this 'rounding' procedure alone far exceeds the 12.5 kg per flight of economy they claim. They have absolutely no way to defend their claim when their own setup is built with far greater imprecision.

And I am only pointing out one mistake (Fuel Load phase), which is the worst I've seen, though I see problems with how they consider the other phases too.

I kind of feel better when I read such kind of work. There is always someone making worse of their time than you, even if they get to keep positions in the University of Chicago famous economics department.

erp said...

Clovis, I hope your last paragraph above is missing the s/off.

The object in all things, not only academic "research" is not to further knowledge, but to further the narrative.

Hey Skipper said...

Clovis:

Man, you got me going full geek here. I should never go full geek.

Here is are the fuel calculations for a Barcelona - Paris I had last week:

FLIGHTPLAN RELEASE # 1 08/25/16 00:30:15
.
MEM FEDEX IFR DISPATCH RELEASE (DOMESTIC OR FLAG) BCFM
.
FLT DTE DEP DEST ALTN EQUIP ACFT NR CRZ WND/C RTE OUT/OFF
FDX5173 25 LFPG LEBL B757-200 N901FD 80 M014 RR 0222/0242
RB211-535E4 3
TIME FUEL CORR
DEST 01:23 10862 ..../......
ADJ 00:00 0 ..../......
MAF 00:04 500 ..../......
RES 00:08 957 ..../......
HOLD 00:30 3511 ..../......
BALST XX:XX 0 ..../......
MEL XX:XX 0 ..../......
MFWF XX:XX 0 ..../......
CONT 00:02 236 ..../......
-RQR- 02:08 16066 ..../......
EXTRA 00:28 3354 ..../......
-TKOF- 02:36 19420 ..../......
TAXI XX:XX 780 ..../......
TOTAL 02:36 20200 ..../......

FOD 01:13 8558 ..../......
EMERG 00:42 5000 ..../......

It doesn't really matter what all those things mean, but there are a couple things worth noting. (Fuel numbers are in pounds.)

— TOTAL is the target for the aircraft fueler. He loads the difference between fuel on board (FOB) and the target.

— TOTAL is the sum of the numbers above it, rounded up to the nearest hundred.

— All the numbers (save MAF) are calculated to the nearest pound. How can that be? All modern airliners continually send data back to Operations. Part of that stream is airframe/engine fuel consumption. The numbers on this flight plan are specific to this specific plane, and the engines on the wings. (As engines get older, their specific fuel consumption changes; some airframes are straighter than others, and have less drag.)

The dispatcher feeds all the relevant information into the computer (payload, winds, temperatures, fuel penalties for certain maintenance items, alternate airport (if any), filed departure, route, arrival, etc) and it uses the performance parameters to calculate the specific fuel amounts.

I flew this leg two nights in a row. There was a 5,000 pound payload difference which caused a 299 lb difference in DEST — fuel burn from takeoff to touchdown. Winds and temperatures were near as darnnit to the same, but it was a different plane; I'm not sure how much that contributed to the burn difference.

Makes sense — a heavier airplane requires the engines to work harder. It costs fuel to carry fuel. That overhead varies proportionally with the length of the flight (hour and 13 mins in this example). For a five hour flight, a 5000 pound increase in aircraft weight would require (rough guess) an additional 1200 pounds of fuel.

So, clearly carrying fuel above required wastes fuel. And, just as clearly, the company spends a lot of money to a fare-thee-well.

Hey Skipper said...

Where do Captains enter into this? They look at the fuel load against their assessment of conditions and decide if it is sufficient. The answer is almost always yes, and the Capt signs off on the fuel load.

Occasionally, though, the Capt disagrees. When I first got here, some fuel loads, while legal, put us in the position of having to declare "Minimum Fuel" in the event of a missed approach. That call means that anything more than a minor delay to the next approach will result in calling "Emergency Fuel".

Yes, legal, but not really a good plan B. In those cases, the Capts would call the dispatcher and coordinate to add additional fuel (typically around 1000 pounds). It's been months since I've seen that; this flight plan has enough for an additional approach, with allowance for a busy airport.

FedEx is exactly like every other major airline in this regard. (Although it wasn't always, 10 years ago, when I got here, many Capts just insisted on adding arbitrary amounts of fuel, which was a culture thing. Then fuel got really expensive, and the company got religion about the whole subject. One way they did that was showing historical fuel burns for each specific segment.)

Apologies for going full geek — I'll try and wrap this up. In almost all cases, the Capts approve the calculated fuel load. Fuelers use fuel quantity indicators located right next to the fuel filler valve in the wing to load the fuel — they do that by calculating the amount of fuel to be loaded, then telling the machine how much to pump. Our fuel loads are almost always within +/- 100 lbs.

There is no such thing as a Capt fueling error.

Now, if Virgin has an ops culture that throws on fuel above the calculated amount, then that is a problem. I rate the likelihood of that significantly below needing to buy steel umbrellas on account of flying flocks of pigs.

Hey Skipper said...

Sometimes my wife hints that I have a slight tendency to build people clocks when all they are asking is the time.

I have no idea why. Women are so inscrutable.

erp said...

... and women are the inscrutable ones???? I have one of your kind here too. Your wife has my sympathy. A simple answer to a simple question is simply impossible.

Hey Skipper said...

erp:

Yes, quite. But ... two things. Clovis might be a receptive audience (or we might never hear from him again, in which case my guess is very, very bad.)

And, I'm being the anti-Harry, by backing up the assertion that this study is a steaming pile of ordure with actual, verifiable, facts.

erp said...

... No worries. Clovis, like the rest of us is hooked on GG.

Clovis e Adri said...

Bring it on whenever you can, Skipper. Where else could I learn all this geeky stuff about flight logistics?

I could be really happy flying planes. I also could be happy working on any complex logistics like those you guys have. I can only envy you, for you get to have fun and money at the same profession. I've got only the first one in my own.

At least I am under less risk of cancer due to cosmic ray incidence, if that's any comfort :-)

Hey Skipper said...

Clovis:

Do you think this thing got any peer review before getting published?

Clovis e Adri said...

Skipper,

It did not:

"NBER Working Papers have not undergone the review accorded official NBER publications; in particular, they have not been submitted for approval by the Board of directors. They are intended to make results of NBER research available to other economists in preliminary form to encourage discussion and suggestions for revision before publication."

But I think it is a bad job even for a working paper. The whole procedure is flawed and the authors do not even hint at their limitations. Though it is clear why: they wouldn't have a paper at all.

Hey Skipper said...

Clovis -- thanks.