Satnavs and the Sunk Cost Fallacy

One of the best pieces of economic advice I know is ”Don't throw good money after bad”. Or in other words, when you consider whether you should continue to invest in a project, don't let the sum you've already invested figure into your decision. To do so is known as the ”sunk cost fallacy”, and leads to ”escalation of commitment”. A good way to avoid this is to decide beforehand what your exit conditions will be, and then stick to them. A bit like saying “I'm going to play the slot machines until six o'clock or until I've lost $50, whatever comes first”.

Google Maps offers a beautiful real-life illustration of this. It's particularly instructive because unlike the case with your financial investments, this software can tell you with a high degree of accuracy what the outcome will be if you escalate your commitment – and if you don't.

I've never had a dedicated satnav in my car. My route planning has always been “look at the map, find the shortest path, see if any major highways are near that shortest path”. But recently I've begun using Google Maps. I tell it the endpoints of my trip and the means of transportation I'm using. It computes a couple of itineraries and tells me how long each is likely to take. “Path A is the fastest and path B will take 17 minutes longer”. I then start driving along path A. I invest in it.

Now, I don't check Google Maps continuously. My smartphone is not in a dashboard holder and I have no lighter cable for powering it. I check the status of my investment maybe once every two hours of driving, or when I get the feeling I may have taken a wrong turn. And sometimes in unfamiliar territory I find that I have indeed veered off path A. Should I retread my tracks and get back on that path? I ask Google Maps again.

The software has no memory of recommending path A to me. It has no sense of commitment whatsoever. It just knows where I am now and where I want to go. So in many cases it tells me “Path C is the fastest. You could take the later stages of Paths A or B instead, but that would take 10 and 13 minutes longer, respectively.” Escalated investment in my original project would clearly just be stupid. I chose path A at the outset because I trusted Google Maps, so why shouldn't I trust it now that I've left that path? On the strength of new information, I invest in path C instead.

More like this

As an archaeologist I often need to plot coordinates on maps and plans. At every scale, really: from individual finds on the plan of an excavation trench to the distribution of something across Europe. Just dots of varying shapes and colours on various background maps. Most often, it's GPS data…
Another episode in the continuing saga, "Janet is a tremendous Luddite." Back when I was "between Ph.D.s" one of the things I did so I could pay rent was work as an SAT-prep tutor. The company I worked for didn't do classroom presentations to a group of students, but rather sent us out on "house…
I read about this in Bio-IT World and had to go check it out. It's called the Genome Projector and it has to be the coolest genome browser I've ever seen. They have 320 bacterial genomes to play with. Naturally, I chose our friend E. coli. The little red pins in the picture below mark the…
This is a classic problem. You are in a car heading straight towards a wall. Should you try to stop or should you try to turn to avoid the wall? Bonus question: what if the wall is not really wide so you don't have to turn 90 degrees? Assumption: Let me assume that I can use the normal model of…

The sunk-cost fallacy is obvious in this kind of case of course. But often it's far from easy to recognize. Consider "Law is really, really boring, isn't it. But I better stick with it; I only have a year left to graduate."

Or "My spouse feels like a stranger to me. But I wouldn't want a divorce; it'd feel like throwing away all those good years we had."

Or, for that matter, "Dull book. Um, only about a hundred pages left, finally."

Hey there Janne!

Indeed, it's obvious, that's what makes it such a good illustration.

But in my opinion none of the examples you offer are situations where it's difficult to make a rational choice. There are situations where you simply don't know if it's time to fold or not, rationally speaking.

I, too, have never had a dedicated satnav system in my car (these things weren't available to the general public when I learned to drive). So I use Google Maps or Apple Maps in a similar way. Except that sometimes I know I will be making the drive several hours later, during rush hour, and I know that some of the possible routes will be congested, and the map software doesn't always know when I will be making the drive. For instance, there are two main ways I can drive to Nashua: via Lowell or via Manchester. During evening commuting hours the Lowell to Nashua segment will always be slow, while the route via Manchester will be free-flowing unless there is an accident or snowfall along the route. So I take the Manchester route, even if the software claims the route via Lowell is faster. Once I'm underway, I will stick with the route unless I have reason to believe I've missed a turnoff, at which point I pull off and consult the map, and take whatever route it suggests from there.

Overreliance on satnav systems can be dangerous. There have been many stories of, e.g., GPS receivers suggesting routes along US Forestry Service roads through Oregon's coast range in winter (these roads are generally not plowed). And people following these directions until they get stuck in the snow.

By Eric Lund (not verified) on 07 Mar 2016 #permalink

Martin, I do think they are at least in part this fallacy. Both cases are really "This is not for me, but I've spent too much time on it to quit."

To take a real example, I'll likely leave science soon. And the most common reaction I get is "Mottainai!" - "What a waste!" The idea is that I've spent so many years working and improving my skills, and switching careers would be throwing it all away. But it doesn't really matter how much time I spent learning the job if I'm no longer interested in doing it.

Nonsense on what they say, Janne - making a major career change in mid-career is now very common, and very liberating.

By John Massey (not verified) on 08 Mar 2016 #permalink

My use of Satnav tends to go the other way, if I have missed a turn I would rather use a new route than have to turn back.
For me the weakness of Satnav or Google Maps route selection is that it ignores risk. I don't want to take a route that looks as if it is 2 minutes faster but includes significant risk of being delayed by 1/2 hour. Google Maps provides the best current solution in that if you select an arrival or departure time/day it will show the best and worst case times.

I agree, there's room for improvement.

Is your moniker related to programming, progressive politics, progressive rock?

Martin: both progressive rock and politics, together with a tendency to consider change to more likely than not to make things better.

Martin, if I knew what I'll be doing I would tell you :) I'm currently interviewing for one rather special job that feels like an almost perfect fit. I'd be very happy if I got it; chances are I won't get it though, so then I'll have to start thinking more widely about what to do next.

Hope you get the job! Or failing that, I hope you have the means to spend the necessary time thinking before you jump careers.

Sorry, should have included a trigger warning.

By Birgerjohansson (not verified) on 08 Mar 2016 #permalink

Birger@14: Some context would help. In my line of work, the initials FGM usually stand for "fluxgate magnetometer", but it makes no sense to me that refusing one would be fatal. I assume that in that Guardian story it stands for some other piece of jargon, which being outside my field, I would have no idea what that might be. And if it is actually a British colloquial expression, that doesn't mean I would understand it: the US and the UK are famously said (with good reason) to be two countries divided by a common language.

Headlines don't always give adequate context. In some cases, they're even misleading.

By Eric Lund (not verified) on 08 Mar 2016 #permalink

" But so far, no research has been able to place a lower bound on the value."

I haven't (yet) read the paper, but whatever its merits, bad journalism here. I myself am author of a paper which has "a lower limit on the cosmological constant" in its title, almost 20 years ago. (Note that we use different units---those commonly used in astronomy---for the cosmological constant, but it is easy enough to convert one to the other.)

By Phillip Helbig (not verified) on 09 Mar 2016 #permalink

It helps to read the paper. Maybe the journalist should have. The paper discusses not a measurement of the value of the cosmological constant, but rather presents an argument as to why it is likely to be greater than some value, which is something completely different.

By Phillip Helbig (not verified) on 09 Mar 2016 #permalink

Philip@19: As a general rule (at least with US news sites), the editor, not the writer, writes the headline. I haven't read either the press release (what appears on tends to be press releases) or the paper it reports, so I don't know how accurately the article reports the contents of the paper, or how accurately the headline reports the contents of the article. It also happens frequently that discoveries are forgotten, then found again 10-20 years later (this is something of a running joke in my field). So without taking the time to look at it further, I couldn't say whether the fault lies with the editor, writer, or paper authors.

By Eric Lund (not verified) on 09 Mar 2016 #permalink

If software people ignored sunk cost rule nothing would ever get shipped. It is almost impossible to finish any significant piece of code without realizing that it could have been done much better. This leads to a choice between pushing on to the bitter end or going back and doing it right. This cost of getting something working is what software people call "technical debt". Getting something out into the world means owing the code a rewrite, this time correctly. Worse, technical debt collectors make those guys who break your kneecaps look like sweetie pies.