03 September, 2015

Uber (Part 4b) - The limits of cost savings from electric vehicles

In Uber (Part 4a), we discussed the expert consensus that the on-demand ride industry is headed towards marginal cost pricing, with or without autonomy.

Autonomy could help Uber in the long-term, by enabling it to capture the revenue currently allocated to the drivers.  In next week's Part 4c, we'll look at the overall vehicle cost curves and discuss how the run-rate economics of Uber will likely impact individually owned cars in the medium- and long-term.

Today, I want to consider only marginal cost pricing in the short term, from the driver's perspective.  Since drivers can still make money in a marginal pricing environment if their costs are cheaper than the marginal cost supplier, they should be using low-cost cars, probably electric powered.  If/when drivers are replaced by Uber-owned vehicles (driven or autonomous), these should also be electric.

The fuel cost savings of electric vehicles are real, however the environmental benefits are much less than I used to believe.  Electric vehicles will also have savings from reduced maintenance costs, and a potentially longer life-cycle would also impact mileage-based vehicle depreciation.

At the end of today's post we'll examine the economics of a recent Uber ride I took: outside of surge pricing, it does not appear to pay to be an Uber driver in Pittsburgh unless Uber is still offering extra driver incentives.

Switching to a cheap electric vehicle instead of an E-class Mercedes would help my driver, but doesn't seem like enough to keep me driving (if I were him).  On the other hand, if my trip were representative of the average Uber passenger, and I made it ~3-4 times each month, Uber's $50b valuation begins to look aggressively achievable.

Conventional wisdom says that electric vehicles are good for the environment and save money on gasoline

In the summer of 2014 I read this article on GigaOm, covering Tesla's announcement that its superchargers were then offsetting 4.2 million pounds of CO2 and pushing 1 GwH of electricty annually.

GigaOm claimed this was half the power of the Hoover Dam - enough to power 700,000 homes.  I immediately grimaced: that claim was so clearly ridiculous it should never have been published (I'm the Greg in the comments section, read from bottom up).

Katie Fehrenbacher has a distinctive reputation for quality reporting on clean tech, yet stumbled in an egregiously innumerate way while doing little more than republishing Tesla's press release.

GigaOm ran out of money and shut down less than a year later, so perhaps corners were already being cut and that wasn't her journalism at its finest.
Side note of interest: Even the article's correction is now misleading: GigaOm lined out the headline claim but at some point rewrote the body of the article to appear to have been "less wrong", including removing the original claim of 700,000 homes.  See the original version archived here on the wayback machine about six hours after its publication.  
This ability to "rewrite history" is important to be aware of in the internet age.  For example, in 2013, a study demonstrated that 49% of the URLs cited in United States Supreme Court briefs were "dead links."  Wouldn't it be interesting to know how many others have been subtly altered?  Last week, Wikipedia banned 350+ editors caught in a "for-profit" editing ring.  To the victors, history, indeed.

Alas, in my own haste to demonstrate the author's obvious errors from the comfort of my couch on a Saturday morning, I made an error of my own: my conclusion that use of carbon pricing from reducing CO2 emissions could significantly subsidize electric cars was off by a factor of ten.

To understand my current beliefs on the cost savings from electric vehicles, let's first see what some experts recently had to say about the benefits of switching away from gasoline.

Electricity as a fuel is much cheaper than gas.  In some cases it's worse for the environment, particularly distant neighbors

A few weeks ago, a group of economists wrote a nice post in VoxEU titled "Analysing environmental benefits from driving electric vehicles."  One of the authors, Nick Muller, is currently a visiting professor at Carnegie Mellon.  Perhaps I'll look him up.

The VoxEU piece summarized a lot of current work in this area, including a 2011 paper by my (now deceased) mentor Lester Lave.

The VOX authors' more recent working paper ($5, the appendix is ungated here) is titled "Environmental Benefits from Driving Electric Vehicles?".   This paper adds interesting nuance around several ideas readers may not be familiar with:
  • Electricity is cheaper than gasoline on a per-mile basis, and the gap is most pronounced on low-mileage vehicles
  • Emission benefits from replacement of gasoline with electricity are dependent on the type of power generation used
  • These emission differences are net positive for some locations, and surprisingly net negative for others

If you're curious about the details, you should read their paper and its appendix; my corrected back-of-envelope for the "value" of CO2 savings illustrates the nature of the much more detailed analysis performed by Nick Mueller et al:
  • Basic facts on emissions
    •  A gallon of non-ethanol gas causes about 19.6 pounds of CO2 to be formed because carbon in exhaust reacts with atmosphere (eia.gov)
    • At 31mpg, that's about 1.58 pounds of CO2 per mile generated by driving a car with gasoline
    • Electric plant CO2 emissions are about 2.1 pounds/kWh (coal) and 1.2 (natural gas).  Nuclear/hydro/wind/solar are negligible
  • Calculate CO2 emissions/mile
    • An electric car uses about .31 kWh/mile, so would cause .65 pounds/mile of CO2 with coal, and .37 with natural gas
    • "Coal fired" electric cars therefore save ~.93 pounds/mile of CO2; "gas-fired" electric cars save ~1.21 pounds/mile.  "Clean-fired" electric cars could save the full 1.58 pounds/mile
  • Price CO2 emissions/mile
    • The "social cost" of carbon mentioned in the paper is $41/ton, this is $.0186 per pound (metric ton is 2,204 pounds)
    • Actual cost of trading permits (where used) seem to vary, ranging from maybe $10-50
    • At "social cost" a coal-fired electric car saves $0.0173/mile in avoided CO2; a gas-fired electric car saves $0.0225/mile.  A "clean-fired" electric car could save $.0294/mile
  • Summarize the lifetime cost savings of electric vehicles
    • Fuel savings average $.0737/mile, or $11,055 over a 150k mile lifetime accruing to owner
      • The paper calibrates electric vehicles as using $.0389/mile in electricity; gas vehicles as using $.1126/mile in gasoline (at $3.49/gallon)
      • This implies a $.0737/mile direct fuel savings of electric vehicles; or $11,055 over a 150k vehicle lifetime
    • CO2 savings depends on electricity source, and (currently) requires government intervention to price
      • A "coal-fired" electric car would save an additional $2,595 in lifetime CO2 emissions; a "clean-fired" electric car could save $4,410

In their paper, Nick and the others reach three interesting and somewhat non-intuitive conclusions:
  1. Net of the social cost of all types of emissions (NOx, particulates, etc), the benefits of switching to electric-fueled vehicles aggregate at a state level from as much as +$3,000 to -$4,500 (depending on the type of electricity replaced).  The US-wide average is -$750.
  2. Negative benefits often accrued downstream of the car's use.  Even when the value of switching is net negative for society, it's often net positive when viewed as a local decision
  3. Federal tax credits of $7,500/vehicle far exceed any conceivable environmental benefit, though local subsidies may be more justifiable on these grounds

Anecdote as case study: Electric vehicles may be well suited to on-demand rides

The situation: Uber is a low-cost ride

Recently I Uber'd to my car dealership, a 2 mile drive (uphill, hot day) if you avoid a heavily trafficked main road.  Driver used a Mercedes E-class sedan to earn $6.21.

When I asked if he selected his car for Uber Black the driver said yes - Uber was encouraging drivers to purchase nice cars because "an UberBlack launch is just around the corner."

Driver drove 12 minutes (~3 miles) to pick me up; our trip seemed to take about ten minutes and we drove exactly 2 miles.  UberX pricing in Pittsburgh appears to be $1.50 + $0.20/min + $1.20/mile (implying it took us 11.5 minutes).  (Remember that Pittsburgh Uber lowered rates at the beginning of the summer by 15% to stimulate demand.)

From a passenger perspective, Uber (sans surge) is much better than a cab.  Pittsburgh cab rates are $5 + $.35/min + $2.45/mile; if available that cab would have cost me $13.93 - more than double.

The complication: Uber driver doesn't make enough money under "normal" conditions

What did my driver earn net of his costs?  The IRS allows a rate of $.575/mile for automobiles, set annually through analysis by a private contractor.

Drivers can deduct more if they can prove their costs are higher.  KBB suggests the operating costs for a Ford Focus are about $.52/mile; for an E-class Mercedes they are $.89/mile.

Using these naive numbers, the five miles my driver drove for the entire fare cost him 5*$.89= $4.45, leaving him earnings of $1.76 for about 25 minutes of work.

My driver is unlikely to stay in UberX very long unless one of two things are true:
  • Uber is guaranteeing driver a minimum wage
    • Problematic for Uber, unless it's subsidized by operating cash flow/surge pricing instead of financed cash flow/debt
  • "Re-positioning" empty-leg costs of this ride were unusually high
    • Note: This consideration is largely what kills the economics of plane sharing
Most people would bet on the second possibility: that the "ride density" is usually higher, so drivers don't have to drive as far without passengers.

While this could be true, in practice how this nets out is complicated and situation-dependent:  
  • One condition that reduces empty travel is higher demand - beyond a certain point, though, this means surge pricing
  • The other condition is higher supply - having a denser network of unused drivers.  This decreases customer wait times/empty driving distances but increases driver (empty, non-earning) wait between rides.  
Uber publishes ride data in several cities; as an exercise to the reader you could take a look at this IRL and let me know what you find in the comments section.  Preferably including a link to the raw data & analysis you did.

The resolution: Uber drivers should select vehicles based on lowest operating costs

Of course, the KBB numbers aren't all variable costs: about half is depreciation; 25% is fuel, maintenance, and repair; and 25% is fixed cost amortization of insurance and financing.

And a big chunk of depreciation is driving it off the lot; after this, depreciation for a low-use car is mostly age-related but depreciation for a high-use car is based on mileage (and there are tax effects I'm ignoring).

A more realistic look at variable costs might be $.35/mile, "flipping" the above economics: costs of $1.75 and profits of $4.46 for 25 minutes of work.  He'll put this towards his fixed costs and living expenses.  Still not quite living the dream.
  • $.10/mile for gas (28 mpg at $2.75/gal)
  • $.10/mile for maintenance and repairs ($1,500 for 15,000 miles)
  • $.15/mile for long-term depreciation ($2,200 for 15,000 miles)
What could my driver do differently/better?

Basically, have a car with lower variable operating costs.  This means higher mileage, lower maintenance, and less depreciation.  

The driver will want to find the right trade-off between acquisition/financing cost and low operating costs; a slightly complicated question but the basic rules are something like:

First, never buy a new car and put it in a fleet unless the owner has a fetish for depreciation tax deductions.  When possible, probably a 2-3 year old, low mileage car.  Drive it for two or three years, and sell it again before the maintenance requirements get too taxing (or the old car impacts your driver ratings).

Second, find a car with the highest fuel efficiency and lowest maintenance costs possible.  KBB suggests a Honda Civic or Insight (at $.42/mile all-in) or a Chevy Spark (at $.37/mile all-in, some of which is due to electric rebate)

Looking closer, KBB suggests the Spark has variable costs of around $.20/mile: less than 60% of the E-class Mercedes.  That's an extra $.75 in the pocket of our driver for this ride, about a 17% profitability improvement when compared to the Mercedes
  • $.027/mile for fuel ($408 for 15,000 miles)
  • $.067/mile for maintenance and repairs ($1,000 for 15,000 miles)
  • $.10/mile for long-term depreciation ($1,500 for 15,000 miles)
The Spark looks like a better fit than the Honda Civic because electric vehicles save big on both gas and maintenance (at the expense of higher anticipated depreciation, offset by state tax incentive for the electric car): most maintenance in cars is related to engine or drive train or brakes.  In electric cars the first two are virtually eliminated, and brake maintenance is greatly reduced due to regenerative braking.

Bottom line: Uber drivers should choose their vehicles wisely, and it's conceivable that Uber is earning enough money at these prices to justify a significant portion of their current valuation

At current rates conducting intra-city trips in Pittsburgh, it seems challenging for an Uber driver to be paid at/above minimum wage while recouping enough to amortize vehicle fixed costs outside of surge pricing times.  Drivers wanting to drive for Uber should choose their car to minimize operating expenses relative to the purchase/financing cost.

This isn't a problem or a judgement, just an observation.  Perhaps Uber is subsidizing the drivers during non-surge hours and they are earning more than run-rate economics suggest.  Perhaps the drivers are just making poor decisions in the short run.

From the platform's perspective, assuming no driver subsidies, Uber earned 20% of $6.21, or $1.24.  In section 4 of the valuation primer, we discussed the value of a user.  If I took that trip on Uber 3-4x each month, Uber's annual platform revenue from me would be $45-60.

If I were an "average" Uber user, my value at a 3x revenue model could conceivably be $135-180.  Meaning the userbase of 277 million riders at $50b in ride revenue that we hypothesized in the valuation primer would value Uber at up to $49.8b.

Wow, that's a bit surprising (seriously, didn't plan that).  I always thought Uber would need to make money during surge pricing despite using marginal cost contractor drivers.

I have two cars, though, and the only reason I'd Uber back-and-forth with any degree of frequency would be to save the parking fee in the most (Pittsburgh) expensive garages downtown - about $16-18/day.

Unfortunately for Uber, that's about the same as a round-trip Uber would cost and I wouldn't have the flexibility of my car.  Plus I live less than ten minutes from the city core; farther out, even cut-rate Uber prices are prohibitive.

For comparison, though, the global urban population in 2014 is 3.9 billion and expected to increase to 6.4 billion by 2050.  While it's unlikely that many markets have a ride value equivalent to the US, it's possible that "turns" of frequent use by non-car owners in the densest cities could make up for this somewhat.  That's where access to ride-level data would come in handy for analysis.

In next week's post we'll dig deeper into run-rate economics, describing the circumstances under which more frequent Uber use makes most sense as a car replacement.

Thanks for reading,

No comments:

Post a Comment

Focused, topical critiques are welcome though my responses may be limited. Off-topic comments, trolling, spam, or anything not pulling its weight will be deleted.