November 6, 2017

Why regulating ride-sourcing fares is risky

Update: Thanks to Harry the Rideshare Guy for pointing me to this paper (also here) by Uber’s Jonathan Hall and Dan Knoepfle, and NYU’s John Horton. The paper addresses the same question covered in this post, albeit in much more depth, and with actual data. The paper finds that the supply of driver labor faced by Uber is highly elastic, to the point that drivers’ hourly earnings would be virtually unchanged within 8 weeks of an Uber-initiated fare increase.

However, this result may be due in part to the fact that a significant number of drivers drive on multiple platforms, but the study examined only Uber’s labor supply. So the increase in Uber’s labor supply is due in part to drivers spending more time on the Uber app than on competing apps (such as Lyft). As such, a government-mandated fare increase for all companies would probably generate a smaller labor supply response for the industry as a whole, than that observed for Uber unilaterally changing its fares. The smaller labor supply response means that driver earnings could still increase, though as explained in the post below, this increase is still likely to be much smaller than the increase in fares.

Tangentially:  The paper also makes an offhand observation that “With the move to up-front pricing, Uber could potetially charge more for utilization-reducing trips, such as those to areas where a return trip with a paying passenger is unlikely.” This is, of course, a big part of the logic behind New York taxi drivers not wanting to pick up black passengers — the driver worries that he’ll have a long deadhead before his next paying fare if he takes a passenger to a “black neighborhood.” While pricing utilization-reducing trips might be economically efficient, it runs a serious risk of creating discriminatory effects in pricing for certain demographic groups.

---

Rumor has it that Seattle City Council may start regulating fares of transportation network companies (TNCs, such as Uber and Lyft),  as is already done for taxis. I assume this is driven by a desire to help drivers earn a better wage, but such a policy would be an expensive way to achieve this goal, and could have significant unintended consequences. Imagine government trying to help fast food workers by dictating the price of a hamburger!

Is this necessary?

First of all, does the TNC market need further intervention to ensure that drivers earn a minimum wage? A while back I shared a back-of-the envelope analysis that suggested that UberX drivers in Seattle could be earning in the neighborhood of $15 per hour, although this result was very sensitive to assumptions about how drivers’ time is split between transporting passengers, deadheading to passenger pickups, and waiting to be matched with a ride.

Uber has recently stated their median driver in Seattle makes between $19 and $21 per hour before expenses. They also cited an NBER working paper by Uber’s Jonathan Hall and Princeton’s Alan Krueger, which estimated driver expenses of $2.94 – $6.46 per hour spent with the app on, depending on vehicle type and whether the driver is a full-time or part-time driver. Drivers with expenses at the high end of the range could very well be earning less than Seattle’s $15/hour minimum wage. Moreover, the fact that these are median hourly earnings implies that half of Uber drivers are earning less than this amount – though we don’t know how much less. Finally, Hall and Krueger’s expense estimates seem fairly low, considering that Consumer Reports estimates a total cost of ownership of $0.47 per mile for the most cost-effective vehicle (a Toyota Prius), or $7.52 / hour for someone driving 16 miles in each working hour.

So, if we accept the premise that TNC drivers ought to be entitled to the same minimum wage protections as other workers, then some government intervention in the market might be justified, given the information we have available.

Is Regulating Fares the Right Approach?

From what I’ve heard, Council is considering reducing the regulated taxi fare by 10%, and requiring both taxis and TNCs to charge these rates. This would represent an increase of approximately 70% from standard UberX fares today.

Current UberX, Current Taxi, and Proposed UberX fare structures in Seattle. Total fares are based on a 5-mile trip.

	UberX Current	Taxi Current	UberX Proposed
Booking Fee	$1.95		$1.95
Base Fare	$1.35	$2.60	$2.34
Rate per mile	$1.35	$2.70	$2.43
Rate per minute	$0.24	$0.50	$0.45
Total Fare	$13.65	$23.60	$23.19
Total Fare Per Mile	$2.73	$4.72	$4.64

If TNC fares were to be increased by government fiat, we can expect several things to happen.

• Driver earnings per mile would increase.
• This increased profitability of working as a TNC driver would attract additional drivers into the business.
• Higher prices would mean that travelers would take fewer trips by TNCs.

These latter two changes mean that drivers would spend more time sitting idle and less time actually transporting passengers, putting downward pressure on earnings per hour.

To understand how these effects net out, I built a simple model of supply and demand for TNC trips.

The demand model is fairly straightforward:

D = D₀(F/F₀)^ε_D

Where D is the amount of TNC travel demanded when the per-mile fare is F, and D₀ and F₀ are the current demand and fare levels, respectively, and ε_D is the elasticity of demand.

Similarly, the supply model is:

S = S₀(P/P₀)^ε_S

Where S is the quantity of TNC travel available from drivers in the market, and P is the hourly profit from working as a TNC driver, and ε_S is the elasticity of TNC capacity supplied.

S and D are related by:

D = S·R

Where R is the fraction of time drivers spend transporting passengers. And the profit of working as a TNC driver can be calculated as:

P = F·V·R·η – E·V·(R+H)

Where η is the fraction of fares paid by passengers that are passed on to the drivers, V is the average driving speed, E is the per-mile expense of owning and operating a TNC vehicle, and H is the fraction of time that drivers spend deadheading to passenger pickups. We assume that deadheading time is a constant fraction k of time spent transporting passengers:

H = R·k

To solve this system of equations, I started out following the same general approach that I used in my previous post. I assume that TNC drivers currently spend R₀ = 60% of their time transporting passengers, H₀ = 20% deadheading to pickups, and 20% waiting idle for their next trip request. Let’s also assume an average trip length of 5 miles and driving speed of V = 20 miles / hour. Under the current UberX pricing structure, this translates into driver revenue of $21.06 per hour, after Uber retains booking fees ($1.95 / trip, 2.4 trips/hour) and a 25% commission. Overall, this corresponds to 64% of passenger revenue being passed on to drivers. This is consistent with the $19-21 median hourly earnings reported by Uber for Seattle drivers. Deducting $7.52 per hour for expenses, this leaves $13.54 per hour in net earnings under current conditions.

Next, I assume that the elasticity of demand for TNC services is -0.5, based on this working paper that used Uber request data. This is a short-run elasticity for UberX only, rather than a long-run elasticity for all TNC services, but is the best (well, only) source that I could find.

I could find even less about the supply elasticity of TNC drivers. This analysis of Uber’s surge pricing by Hall, Kendrick, and Nosko shows an increase in drivers of about 60% in response to a time-averaged surge rate of 1.46, suggesting an elasticity of 1.25. However, this pertains to the number of UberX drivers in a certain part of a city at a certain time, not to the overall number of TNC drivers working in the city as a whole. Since I’m not sure of a reasonable value for long-run supply elasticity, I’ll explore how sensitive the results are to different assumed values of the supply elasticity of TNC drivers.

Results

With a demand elasticity of -0.5, we would expect to see a 23% reduction in the quantity of TNC travel demanded in response to the 70% increase in price.

On the driver side, the equilibrium solutions are summarized in the table below, for different assumed values of supply elasticity. In contrast to the baseline case in which drivers are busy (transporting passengers or deadheading to a pickup) approximately 80% of the time, drivers under the new fare structure would spend roughly half their time idle, only transporting passengers about 35-40% of the time. And while the total amount of TNC travel demanded would decrease, the number of drivers offering rides would increase, by about 10-30%, depending on the elasticity of driver supply. Although per-mile fares would increase 70%, hourly earnings to drivers (net of fees and expenses) would go up by far less than this amount, under the range of elasticity values explored here. In fact, if the supply of TNC drivers proved to be highly elastic, then we might see the number of drivers increase by one-third, and the hourly profits per driver increase by less than $3 per hour.

εS	Time Spent Transporting Passengers	Time Spent Deadheading to Pickups	Hourly Profit	% Increase in Available Supply	% Increase in Hourly Profits per Driver
0.2	42.6%	14.2%	$20.04	8%	48%
0.4	40.2%	13.4%	$18.95	14%	40%
0.6	38.6%	12.9%	$18.17	19%	34%
0.8	37.3%	12.4%	$17.59	23%	30%
1	36.4%	12.1%	$17.13	27%	27%
1.2	35.6%	11.9%	$16.77	29%	24%
1.4	35.0%	11.7%	$16.47	32%	22%

Other Considerations

It is not clear to me whether this proposal would force fares to be exactly the prescribed level, or whether it would simply put a floor on prices. If the former, then it would also preclude the TNCs from using dynamic pricing (e.g. Uber’s surge, or Lyft’s prime time) to match supply and demand. Dynamic pricing encourages some travelers to shift their travel times or modes, and encourages more drivers to be available in peak hours. Eliminating this ability would mean longer waits for a ride at peak times.

It is also not clear how this would affect ride-sharing services such as Uber Pool and Lyft Line. If the regulated fares applied to each traveler (or each separate travel party), then there would be no price incentive for travelers to share rides through a ride-matching service like Pool or Line. This would be bad from the perspective of reducing traffic and the emissions intensity of travel.

Is There a Better Approach?

It seems to me that if driver earnings are deemed to be too low and that drivers deserve the same minimum wage protections as other workers in Seattle, then policymakers might be better off addressing driver earnings more directly. Mandating a minimum wage for drivers would encourage TNCs to use their drivers and vehicles more efficiently, rather than encouraging drivers to sit idle in the system waiting for a “big score” of a trip. In addition to addressing the wage problem more directly, it would align the incentives of the TNCs with other priorities such as reducing traffic and reliance on personally owned cars. However, I suspect that the city is unable to mandate higher wages directly because of TNC drivers’ status as independent contractors. If this is the case, I’m not sure there’s really a good solution available to local government – though I would love to hear some more creative ideas!

Limitations of Analysis
The analysis here is a simplified representation of the TNC market. Results are sensitive to the elasticities of supply and demand. The analysis does not consider the role of surge pricing in determining overall driver earnings. The policy change in question might induce responses beyond those considered here, such as changes in the commission rate or service fee charged by TNCs. Further research and analysis would be helpful in predicting more precisely the overall effects of such a change in fares.