Cost Estimates
As Reich notes in his study of ridesharing in New York City, “a comprehensive accounting of all vehicle-related expenses is essential” to understanding driver earnings. This remains true in California, and is precisely why we also do not rely on the IRS rate, but instead conduct our own study on the cost of driving. Our internal analysis, summarized below, estimated the average cost of a marginal mile to be $0.258.
It is important to note that our results are not without context. Similar studies of cost have found values similar to ours. The Rideshare Guy found his marginal expenses in a Prius to be $0.195 per mile. Zoepf et al. calculated a cost of $0.30, using more conservative assumptions about MPG, and allocating fixed insurance costs to miles driven with network companies.
We can also use the AAA’s costs pamphlet to get quick contextualizing numbers. We adjust their fuel estimate to account for California’s higher cost, add in maintenance, and split out per-mile depreciation to arrive at $0.215 and $0.259 for Hybrids and Medium Sedans, respectively.
Cost Study
The population of drivers and couriers active across these platforms in California is disproportionately composed of part-time workers; only 17.5% of all engaged time in a quarter is attributable to Uber drivers who averaged at least 30 hours a week [1]. For couriers, we would expect this figure to be even lower. As such, we infer that most drivers did not purchase their vehicles for the purpose of driving with a network company. With this backdrop, we consider the marginal cost of a vehicle mile to be most relevant. This figure is composed of fuel costs, the cost of variable depreciation, and the cost of maintenance and repairs.
We use internal data to understand our mix of vehicle types, which skews much more towards hybrids and high-MPG vehicles than either national or state averages. Data on the cost of ownership for each vehicle type is taken from publicly available external sources, as well as vendors that partner with Uber.
Fuel Costs
We combine internal VIN data with DataONE’s VIN decoder to get each vehicle’s combined fuel economy. We then multiply by regional gas price data from the Energy Information Administration (EIA) to account for the relatively high cost of gas in California. The average cost of fuel per mile logged on our platform was $0.112 for the period studied.
Depreciation
To understand the amount a vehicle on the platform depreciates with each mile, we used publicly available data on depreciation amounts for specific make/model/years for various mileage levels. We queried this data at broad intervals to get the information needed to allow us to sketch out a cost curve for each vehicle (see example plot). At each point, we can use the slope given by this data to understand the depreciation associated with an incremental mile. The final component of this analysis is an assumption about total mileage to-date for these vehicles. As we do not observe this, we make a conservative assumption, and use average usage data from the National Highway Traffic Safety Administration. This is conservative, as driver-partners likely drive more than the average person. Using a smaller figure for miles to-date puts the vehicle on a steeper part of the curve, inflating likely depreciation costs. We find an average depreciation per-mile of $0.049.
Maintenance and Repairs
We used public estimates of the cost of maintenance and repairs per mile by make/model/year. We join that to our internal data to estimate the average cost of maintenance and repairs, which we find to be $0.097 per mile.
Ultimately, we chose to move forward with $0.30, a figure higher than our own internal estimate, to be conservative and have committed to pegging our estimate to inflation to ensure that it remains an accurate measure over time.
__________________________
- Reich cites an external study that finds that “10 percent of transportation platform drivers account for about 57 percent of driver earnings” to imply that a large share of work on the platform is done by drivers that work close to full time. We do not observe the same pattern in supply for this cohort using our internal data.