How Google’s Estimated Driving Costs Misses the Train

I am a huge fan of Google maps, using them pretty much every day as a core part of my navigation through life. They’ve done a lot of great things (like offering built-in transit directions and walking maps). I’m a fan.


Google's "estimated driving costs" tool

But, but… Google has added a new tool — “estimated driving costs” — that’s broken, conceptually and factually. I think it may even be actively un-helpful. I want to riff a bit how it’s broken, and why that matters, and what it says about the American debate on transportation and cities.

First, a caveat: I am sure that this tool was provided with the best of intentions. I’d bet money that the developers’ thought process went something like this: more metrics must equal more informed decisions, which is an obvious good, right? I’d also bet that at least part of the intention was to show transit users that for many trips, they’re saving money, thus bolstering the case for taking transit. I’d further bet that the budget for creating this tool was far from vast, and the developers did the best they could with a new idea and limited resources.

But the conceptual map underlying the “estimated driving costs” tool and the calculation chosen to provide its result are both deeply flawed, meaning that the tool doesn’t actually offer good metrics for decision-making, and in some ways undercuts the case for transit by not actually comparing apples and apples when setting cars and transit against each other.

This a blog post, not a research paper, and I don’t have time to fully explore and argue every aspect of the challenge. I hope this can start a discussion and others will debate the issues at greater length. Here, though, are obvious flaws I can see without digging:

A) Flaws in how the number is calculated:

1. “The cost of driving is based on the distance driving between your start and end addresses, multiplied by the standard cost per mile that tax regulations allow businesses to deduct.” Because the mileage deduction is calculated for “average” American drivers across the entire rural – exurban – suburban – urban transect, and vehicle miles traveled (VMT) rises as density drops, the IRS deduction is almost certainly not an accurate prediction of the cost of urban car ownership per mile (because if you drive less, but pay the same in car payments, maintenance and insurance, each mile you drive costs you more). The direct costs of car ownership are likely considerably higher per mile for urbanites who drive infrequently.

2a. I strongly suspect the IRS mileage numbers don’t actually include all the direct costs of ownership (for instance, I doubt the cost of owning and paying a mortgage on (or paying a rent premium for) a garage or driveway is ever factored in)… but others can probably address this better than I can.

2b. On an even larger scale, of course, the driving costs engine doesn’t even allow for the possibility of not owning a car at all. Simply owning a car with the attendant costs of purchasing, maintaining, storing and insuring that car, but also the costs of financing all of the above and paying the appropriate taxes, is expensive, even before you ever turn the ignition. Because car-free life means you are saving all those costs, at this point, the driving-vs.-transit cost comparisons for individual trips break down completely for many urban dwellers. (This is also why, when the time spent earning the money to pay for owning a car is factored in, driving is actually seen to be considerably less of a time-saver than it appears.)

3. Google driving costs engine “doesn’t consider” tolls or parking fees. This is sort of crazy, as in many cities the cost of parking alone runs higher than the entire driving costs Google engine cites for some of the short trips I searched out of curiosity.

4. The driving costs engine obviously doesn’t count at all the indirect costs to drivers of driving (like the costs to your health of auto-dependence and its increased risk or obesity and heart disease, or the increased likelihood of death in a car accident [riding transit is dramatically safer than driving], and so on).

5. On an even geekier note, the Google comparison fails to account for the fact that driving and transit are differentkinds of trips. People who drive go one place, park, go another, park, and so on, scattering their trips across the landscape based on a variety of factors, of which distance is only one concern (and depending on the person, perhaps a not very important concern, as evidenced by people who will drive two miles to save 10 cents a gallon on the price of gas). People who walk/bike and take transit (the modes are inexorably linked) tend to string their trips, meaning they plan routes that will allow them to accomplish multiple tasks along the route (for instance, grabbing coffee on the way to the transit stop, meeting a friend for lunch nearby the office, picking up the dry cleaning or doing the shopping on the way home) instead of making a number of discrete trips to complete those tasks. The result is that people who use transit take fewer trips overall, and travel shorter distances. This produces my favorite odd urbanism effect, that of“transit leverage.” Transit leverage, simply put, tells us that a person traveling a mile by public transportation does not therefore simply travel a mile less by car: in the U.S., he or she actually drives between 2.9 and 9.0 miles less (depending on the characteristics of the neighborhood and transit system). So, to be a fair comparison, the driving costs engine should in theory figure out how to account for the trips that the transit user isn’t making because she or he didn’t drive in the first place.**

B) Flaws in how the driving costs tool is framed, and in the conceptual map underlying the tool:

1) As far as I can tell, it only shows up when asking for transit directions, not when asking driving directions. Therefore, the obvious implication is that when taking transit, we should consider whether or not we’re saving money; but when driving, we should just worry about the fastest way to get there.

2) In a time when exurban conservative politicians are fiercely attacking public transportation (while lauding the car and promising $2/gallon gas), we need to be more rigorous in looking at the true costs of car ownership both to us as individuals and families and to our neighborhoods, cities and nation. To provide a tool that severely underestimates the cost of using cars by limiting what it counts is to offer a certain amount of momentum to ideological and counter-factual arguments about transit’s costs to society. If we’re going to put a number to these things, let’s make it a true-cost number.

3) There is no mention at all made of externalities. Because, though the costs of driving are inarguably higher for many-to-most people in cities than the costs of taking transit, the costs of cars to society and the planet are huge. From climate change to oil spills, geopolitical instability to highway welfare, sprawling exurbs to broken health care, cars exact a terrible toll on all of us. Choosing not to drive once is a small, perhaps positive choice. Choosing to drive rarely or never is probably the single best lifestyle change a person who cares about the world can make. (And though the societal costs of driving are the subject of great debate, and range from large to inconceivable based on what we choose to count, a cost calculator that in the 21st century completely excludes the externalities of an action is not making the contribution it might.)

I could go on, but you get the point.

What could Google do to fix this? Well, given the incredible access to data and in-house genius that Google has, it could choose to completely remake this tool, creating a completely new measurement that aims at a full-and-complete cost accounting of these individual trips. It could look to find a way to service trip-stringing and compare it to trip-scattering, showing transit users how many trips they’re avoiding by using transit and walking. It could remake the whole service, moving partially towards meeting Walkscore‘s excellent work offering information for people who live in neighborhoods and not at exit ramps: this could mean, for instance, offering an engine that would attempt to calculate the VMTs a person would likely accumulate living in different places with a menu of different destinations and trip choices; or a tool that allowed individual users to aggregate trip data and calculate their actual transportation expenditures over time, whether driving or riding/walking. Again, I could go on and on, and Google could probably find a way to come up a practical version with every idea I reeled off. All it would take is the will to make Google maps as excellent a service as it could be. The result would be an extremely useful set of tools, and perhaps some momentum towards a country that makes smarter choices about where to live and how to get around.


**This, by the way, is also why many arguments against the utility of transit in lowering climate emissions are so deeply flawed. For, even in cases where it’s true that a mile ride on a train produces no fewer (or even, conceivably, more) direct emissions than a mile of driving, they are not actually comparable on a mile-by-mile basis, and the overall life-system that transit in a walkable neighborhood produces is without doubt a much lower emissions way of life than an auto-dependent one. The more things you include in your transportation emissions calculations, the better and better walkable urbanism served by transit looks.