The Role of Surge Pricing on a Service Platform with Self-Scheduling Capacity

TLDR: It is concluded that all stakeholders can benefit from the use of surge pricing on a platform with self-scheduling capacity, and as labor becomes more expensive, providers and consumers are better off with surge pricing.

Abstract: Recent platforms, like Uber and Lyft, offer service to consumers via “self-scheduling” providers who decide for themselves how often to work. These platforms may charge consumers prices and pay providers wages that both adjust based on prevailing demand conditions. For example, Uber uses a “surge pricing” policy, which pays providers a fixed commission of its dynamic price. We find that the optimal contract substantially increases the platform's profit relative to contracts that have a fixed price or fixed wage (or both) and although surge pricing is not optimal, it generally achieves nearly the optimal profit. Despite its merits for the platform, surge pricing has been criticized in the press and has garnered the attention of regulators due to concerns for the welfare of providers and consumers. However, we find that providers and consumers are generally better off with surge pricing because providers are better utilized and consumers benefit both from lower prices during normal demand and expanded access to service during peak demand. We conclude, in contrast to popular criticism, that all stakeholders can benefit from the use of surge pricing on a platform with self-scheduling capacity.

Figures

Table 7: Profit performance of the four suboptimal contracts relative to the optimal contract in the 2,253 scenarios in which the fixed contract serves both demand states. f, w, p, β, and o to refer to the fixed, dynamic wage, dynamic price, commission and optimal contracts, respectively.

Table 8: Correlations between scenario parameters and the profit ratio of the commission and optimal contracts, Uβ/Uo, in the 2,253 scenarios in which the fixed contract serves both demand states.

Figure 2: An example of demand and capacity rationing with a particular fixed contract.

Table 3: Contract terms (pl, ph, N, β = wl/pl = wh/ph) and profits (Uβ) with three potential versions of the commission contract - each of these versions can be the best solution for the platform.

Figure 3: The ratio of consumer surplus with the commission contract to consumer surplus with the fixed contract as a function of the commission earned by providers with the commission contract in the 2,253 scenarios in which the fixed contract serves both demand states. Red squares indicate the average ratio for scenarios grouped by the commission contract commission in 0.05 intervals.

Table 4: A summary of tested parameter values. All combinations of these values constitute 14,700 scenarios.

Full text

#3.:*67.8=4+*337=1:&3.&#3.:*67.8=4+*337=1:&3.&
!(-41&61=422437!(-41&61=422437
.3&3(*&5*67 $-&6843&(918= *7*&6(-

"-* 41*4+!96,*6.(.3,43&!*6:.(*1&8+462;.8-!*1+"-* 41*4+!96,*6.(.3,43&!*6:.(*1&8+462;.8-!*1+
!(-*)91.3,&5&(.8=!(-*)91.3,&5&(.8=
*6&6)&(-43
#3.:*67.8=4+*337=1:&3.&
&.81.3&3.*17
9'*34'*1
4114;8-.7&3)&)).8.43&1;4607&8-88576*547.846=95*33*)9+3(*%5&5*67
&684+8-*.3&3(*&3).3&3(.&1&3&,*2*38422437&3)8-*&60*8.3,422437
*(422*3)*).8&8.43 *(422*3)*).8&8.43
&(-43&3.*174'*1 "-* 41*4+!96,*6.(.3,43&!*6:.(*1&8+462;.8-!*1+
!(-*)91.3,&5&(.8=
&39+&(896.3,!*6:.(*5*6&8.437&3&,*2*38



-885)<)4.46,2742
"-.75&5*6.75478*)&8!(-41&61=422437-88576*547.846=95*33*)9+3(*%5&5*67
46246*.3+462&8.4351*&7*(438&(86*547.846=54'4<95*33*)9

"-* 41*4+!96,*6.(.3,43&!*6:.(*1&8+462;.8-!*1+!(-*)91.3,&5&(.8="-* 41*4+!96,*6.(.3,43&!*6:.(*1&8+462;.8-!*1+!(-*)91.3,&5&(.8=
'786&(8'786&(8
*(*3851&8+46271.0*#'*6&3)=+84++*67*6:.(*84(43792*67:.&A7*1+7(-*)91.3,B564:.)*67;-4
)*(.)*+468-*27*1:*7-4;4+8*384;460"-*7*51&8+46272&=(-&6,*(43792*6756.(*7&3)5&=
564:.)*67;&,*78-&8'48-&)/978'&7*)4356*:&.1.3,)*2&3)(43).8.43746*<&251*#'*697*7&
A796,*56.(.3,B541.(=;-.(-5&=7564:.)*67&C<*)(422.77.434+.87)=3&2.(56.(*$.8-&78=1.>*)
24)*18-&8=.*1)7&3&1=8.(&1&3)392*6.(&16*79187;*789)=7*:*6&156.(.3,7(-*2*78-&8(491)'*
.251*2*38*)43&7*6:.(*51&8+462.3(19).3,796,*56.(.3,$*C3)8-&88-*458.2&1(4386&(8
79'78&38.&11=.3(6*&7*78-*51&8+462@7564C86*1&8.:*84(4386&(878-&8-&:*&C<*)56.(*46C<*);&,*46
'48-&3)&18-49,-796,*56.(.3,.7348458.2&1.8,*3*6&11=&(-.*:*73*&61=8-*458.2&1564C8*75.8*.87
2*6.87+468-*51&8+462796,*56.(.3,-&7'**3(6.8.(.>*)'*(&97*4+(43(*637+468-*;*1+&6*4+564:.)*67
&3)(43792*67349624)*1&71&'46'*(42*7246**<5*37.:*564:.)*67&3)(43792*67&6*'*88*64++
;.8-796,*56.(.3,'*(&97*564:.)*67&6*'*88*698.1.>*)&3)(43792*67'*3*C8'48-+64214;*656.(*7
)96.3,3462&1)*2&3)&3)*<5&3)*)&((*77847*6:.(*)96.3,5*&0)*2&3)$*(43(19)*.3(4386&7884
54591&6(6.8.(.728-&8&1178&0*-41)*67(&3'*3*C8+6428-*97*4+796,*56.(.3,43&51&8+462;.8-7*1+
7(-*)91.3,(&5&(.8=
*=;46)7*=;46)7
7*1+7(-*)91.3,(&5&(.8=5**6845**62&60*87(4386&(8)*7.,3)=3&2.(56.(.3,7*6:.(*45*6&8.4376.)*
7-&6.3,
.7(.51.3*7.7(.51.3*7
97.3*77?.3&3(*&3).3&3(.&1&3&,*2*38?&60*8.3,
"-.7/4963&1&68.(1*.7&:&.1&'1*&8!(-41&61=422437-88576*547.846=95*33*)9+3(*%5&5*67

The Role of Surge Pricing on a Service Platform with
Self-Scheduling Capacity
G´erard P. Cachon, Kaitlin M. Daniels, Ruben Lobel
∗
December 2, 2015; revised
June 14, 2016
Abstract
Recent platforms, like Uber and Lyft, offer service to consumers via “self-scheduling” providers
who decide for themselves how often to work. These platforms may charge consumers prices
and pay providers wages that both adjust based on prevailing demand conditions. For example,
Uber uses a “surge pricing” policy, which pays providers a fixed commission of its dynamic
price. We find that the optimal contract substantially increases the platform’s profit relative
to contracts that have a fixed price or fixed wage (or both) and although surge pricing is not
optimal, it generally achieves nearly the optimal profit. Despite its merits for the platform,
surge pricing has been criticized in the press and has garnered the attention of regulators due
to concerns for the welfare of providers and consumers. However, we find that providers and
consumers are generally better off with surge pricing because providers are better utilized and
consumers benefit both from lower prices during normal demand and expanded access to service
during peak demand. We conclude, in contrast to popular criticism, that all stakeholders can
benefit from the use of surge pricing on a platform with self-scheduling capacity.
Keywords: self-scheduling capacity, peer-to-peer markets, contract design, dynamic pricing,
service operations, ride-sharing
1 Introduction
The rise of the “sharing economy” has transformed the way firms can deliver service to consumers.
The firm no longer must centrally schedule its capacity by assigning workers to shifts. Instead,
∗
Operations, Information, and Decisions Department, Wharton School, University of Pennsylvania, email:
{cachon, kaitd, rlobel}@wharton.upenn.edu. Thanks is extended to seminar participants at Harvard University,
Northwestern University, University of California Los Angeles, University of Pittsburg, University of North Carolina
Chapel Hill, and Washington University.
1

workers may act as independent service providers who determine their own work schedules, and
the firm’s role becomes that of a platform which connects providers to consumers. (See Katz
and Krueger (2016) for data on the growth of alternative work arrangements in the United States.)
Although the platform has far less control over how many providers work at any one time, providers
gain the freedom to “self-schedule” the hours they work, presumably allowing them to better
integrate their work with the other activities in their lives (Hall and Krueger (2015)). To make
these new relationships viable, customers must be charged a reasonable fee and be adequately
served.
Examples of relatively new platforms that feature self-scheduling capacity include Uber and
Lyft for local transportation, and Postmates for local delivery. A potential provider for one of
these platforms must first make the long-term decision of whether to join the platform or not. This
decision has implications for several months or years, and providers join only if they expect to earn
more with the platform than with their next best alternative. If a person joins a platform as a
provider, then they must make short-term decisions about when and how often to work. These
decisions are made on a daily or hourly basis, so the participation decision is relevant over a much
shorter time interval than the joining decision. The participation decision is based in part on the
wage providers receive per service. It is also based on providers’ expectations of how likely they are
to get work, which is a function of the overall level of demand and the number of providers working
at that time on the platform. For example, an Uber driver may know that demand is higher on
rainy days but may also know that other drivers are more likely to drive as a consequence. What
matters to the provider is the amount of demand relative to the amount of offered capacity at a
particular time.
In this paper we focus on the contractual forms a platform could select to make a viable market
with self-scheduling capacity. We study a stylized model with the following features: (i) there exists
a large pool of potential providers, (ii) providers join the platform only if their rational expectation
of their earnings from participation on the platform exceeds the opportunity cost of their next best
activity, (iii) the platform sets a price for consumers, a wage paid to providers for work completed
and regulates the maximum number of providers who join the platform, (iv) the platform cannot
directly determine when providers work and, instead, the providers who joined the platform self-
schedule their offered capacity, (v) demand varies in predictable ways (e.g., more consumers seek
transportation on a rainy evening), (vi) if the offered capacity exceeds demand, providers share the
available demand equally, but if the offered capacity is less than demand, then demand is randomly
rationed (i.e. all consumers are equally likely to receive the scarce service), and (vii) the platform’s
2

price and wage can depend on the current level of demand.
There are two key features of the model that make this environment distinctive. First, providers
self-schedule their offered capacity. Consequently, even if the number of providers who have joined
the platform is sufficient to satisfy demand, it is possible that demand rationing can occur because
too few providers may choose to work. However, it is also possible that capacity rationing can
occur, because too many providers may choose to work leaving some underutilized. Both forms of
rationing represent costly inefficiencies for the platform. Second, the platform can offer demand-
contingent prices and wages. Demand-contingent prices are often called dynamic prices. Uber and
Lyft employ versions of dynamic prices and wages called surge pricing and prime time respectively.
There is a large literature on dynamic prices, while the literature on dynamic wages is far less
extensive, and there is no work on the interaction between dynamic prices and dynamic wages.
The platform’s primary goal with the design of its contract is to maximize its profit. Doing so
requires a contract that assures providers that join sufficient expected profit. However, the contract
must not give providers too much of an incentive to participate, which could lead to an excess of
providers, nor too little incentive, which could entice too little participation from providers to
satisfy demand.
Although maximizing profit is a clear objective for the platform, it is not the platform’s only
concern. A number of controversies have emerged with this new business model. Some people
believe providers are not adequately compensated because they are not given benefits and rights
associated with being employees (Isaac and Singer (2015), Scheiber (2015)). Others worry that
customers are unfairly discriminated against as a result of dynamic pricing (Kosoff (2015), Stoller
(2014)). Consequently, with a view towards potential litigation and regulation, a platform should
be concerned with both provider and consumer welfare. In particular, it is important to understand
the degree to which there is a tension between maximizing the platform’s profit and the surplus
earned by the other relevant stakeholders, the providers and consumers.
We focus on five possible operating models, or contracts, for the platform. With the simplest
possible contract, called the fixed contract, the platform offers providers a fixed wage and charges
consumers a fixed price. Next, we consider contracts in which the the platform either chooses
dynamic prices (with a fixed wage), or dynamic wages (with a fixed price). We refer to the former
as the dynamic price contract and the later as the dynamic wage contract. A commission contract,
which resembles surge pricing used in practice, allows the platform to dynamically adjust both
prices and wages in response to demand, but imposes the contraint of a fixed commission, i.e., a
fixed ratio between the two. It has been argued that this constraint may substantially lower the
3

Frequently asked questions

Q1. What are the contributions in "The role of surge pricing on a service platform with self-scheduling capacity" ?

With a stylized model that yields analytical and numerical results, the authors study several pricing schemes that could be implemented on a service platform, including surge pricing. Despite its merits for the platform, surge pricing has been criticized because of concerns for the welfare of providers and consumers. The authors conclude, in contrast to popular criticism, that all stakeholders can benefit from the use of surge pricing on a platform with selfscheduling capacity. 

Q2. How many evenly spaced observations yield a minimum profit ratio?

Evaluation of 3,600 evenly spaced observations throughout the feasible parameter space yields a minimum profit ratio close to the lower bound, Uβ/Uo = 0.646. 

Q3. What is the key advantage of central-scheduling?

The key advantage of central-scheduling is that it allows the platform to eliminate the inefficiency of capacity rationing: the platform would never choose to have more providers working than necessary as that lowers the providers’ earnings, making recruiting them more costly. 

Q4. How does the platform mitigate capacity rationing in the low demand state?

Once a dynamic wage is allowed, the platform can mitigate capacity rationing in the low demand state by lowering the wage in that state. 

Q5. What are some examples of new platforms that feature self-scheduling capacity?

Examples of relatively new platforms that feature self-scheduling capacity include Uber and Lyft for local transportation, and Postmates for local delivery. 

Q6. What is the average performance of the fixed, dynamic wage and dynamic price contracts?

On average, the fixed, dynamic wage and dynamic price contracts perform poorly relative to the optimal contract, earning only on average 75.5%, 76.2% and 79.1% of the optimal profit respectively. 

Q7. What is the optimal contract for a given platform?

The optimal contract allows the platform complete flexibility: both wages and prices may vary according to the demand state without the constraint of a fixed ratio between the two. 

Q8. What is the reason why consumers are skeptical about dynamic pricing?

The authors find that consumers indeed have a reason to be skeptical about dynamic pricing: relative to the fixed contract, adding dynamic pricing (with a fixed wage) reduces consumer surplus. 

Q9. What is the way to determine the optimal commission contract?

As expected, the commission contract does best when there is less variability in demand (i.e., δ is high): the commission contract is identical to the optimal contract if there is no variability in demand. 

Q10. What is the optimal profit of a platform with central-scheduling?

the platform’s optimal profit with central-scheduling, Uc, is only 35.7% of the optimal profit with self-scheduling providers, Uo. 

Q11. What is the penalty associated with central-scheduling?

As intuition suggests, the penalty associated with central-scheduling is smaller when the providers experience less variability in their participation cost (σ/µ is small) and the average participation cost is small relative to consumer willingness to pay (G(a) is large). 

Q12. What is the main benefit of self-scheduling?

self-scheduling provides a benefit to the providers: they can work when it is most desirable for them to do so, rather than being forced to work at times the platform dictates. 

Q13. How does the commission contract perform in the worst case?

there are a few scenarios in which the commission contract performs poorly - in the worst scenario the commission contract earns only 63.7% of the optimal profit. 

Q14. What is the difference between the two versions of the dynamic wage contract?

Relative to the fixed contract, the dynamic wage contract allows the platform to address the issue of capacity rationing due to excessive provider participation. 

Q15. What is the median ratio of the profit with the membership fee contract to the optimal profit?

In their preferred sample of 2,253 scenarios, the median ratio of the platform’s profit with the membership fee contract to the optimal profit, Um/Uo, is only 0.858 and the lowest ratio is 0.565.