Singapore Management University

Institutional Knowledge at Singapore Management University

Research Collection Lee Kong Chian School Of Lee Kong Chian School of Business
Business

4-2021

Urban consolidation center or peer-to-peer platform? The solution

to urban last-mile delivery
Qiyuan DENG
Singapore Management University, [email protected]

Xin FANG
Singapore Management University, [email protected]

Yun Fong LIM

Singapore Management University, [email protected]

Follow this and additional works at: https://ink.library.smu.edu.sg/lkcsb_research

Part of the Operations and Supply Chain Management Commons, Operations Research, Systems
Engineering and Industrial Engineering Commons, and the Transportation Commons

Citation
DENG, Qiyuan; FANG, Xin; and LIM, Yun Fong. Urban consolidation center or peer-to-peer platform? The
solution to urban last-mile delivery. (2021). Production and Operations Management. 30, (4), 997-1013.
Available at: https://ink.library.smu.edu.sg/lkcsb_research/6695

This Journal Article is brought to you for free and open access by the Lee Kong Chian School of Business at
Institutional Knowledge at Singapore Management University. It has been accepted for inclusion in Research
Collection Lee Kong Chian School Of Business by an authorized administrator of Institutional Knowledge at
Singapore Management University. For more information, please email [email protected].
 Urban Consolidation Center or Peer-to-Peer Platform?

The Solution to Urban Last-Mile Delivery

Qiyuan Deng • Xin Fang • Yun Fong Lim∗

Lee Kong Chian School of Business, Singapore Management University, Singapore

[email protected] • [email protected] • [email protected]

Oct 02, 2020

Abstract

The growing population in cities and booming e-commerce activities create huge demand for

urban last-mile delivery, exerting intense pressure on the cities’ well-being. To keep conges-

tion and pollution under control, a consolidator can operate an urban consolidation center

(UCC) to bundle shipments from multiple carriers before the last-mile delivery. Alterna-

tively, the consolidator can operate a peer-to-peer platform for the carriers to share delivery

capacity. We provide guidance for the consolidator to choose between these two business

models by comparative analysis. We capture the interactions between the consolidator and

carriers using a game-theoretical framework. Under each business model, the consolidator

first decides a price of delivery service to maximize her expected profit. Each carrier then

observes his task volume, and decides whether to deliver on his own or use the consolida-

tor’s service to minimize his delivery cost. Under the UCC model, the carriers become more

dependent on the UCC as their variable delivery cost increases. Under the platform model,

the demand and the supply of delivery capacity on the platform are balanced in equilibrium.

Between the two business models, the UCC is more profitable than the platform if and only

if the carriers’ variable delivery cost is sufficiently large. Moreover, the UCC becomes more

dominant as there are more carriers. The UCC is also more efficient than the platform in

mitigating the negative social-environmental impact of urban last-mile delivery if and only

if there are sufficiently many carriers. Otherwise, the platform performs better.

Keywords: last-mile delivery, urban consolidation center, peer-to-peer platform, game theory

History: Received: December 2019; Accepted: September 2020 by Vijay Mookerjee, after 2

revisions.
∗
: Corresponding author

1 Introduction

Last-mile delivery is the last leg of a supply chain that transfers freight or packages from

a distribution center to a receiver. It comprises up to 28% of the total delivery cost of a

supply chain (Lopez, 2017, Wang et al., 2016). Managing last-mile delivery becomes especially

1
challenging if it is performed in an urban area, where congestion increases fuel consumption,

causes delay of delivery, and lowers delivery efficiency (Ranieri et al., 2018). In addition, last-

mile delivery is the most expensive and critical operation for companies engaged in e-commerce

(Lee and Whang, 2001). Due to the continuous growth of urban population and e-commerce

activities, last-mile delivery to a city center exerts intense pressure on the city’s economic, social,

and environmental well-being (Quak and Tavasszy, 2011).

The economic impact of urban last-mile delivery includes the waste of resources due to extra

waiting in traffic congestion and low utilization of uncoordinated vehicles transporting freight

to the city center. The large number of small, individual customer orders in e-commerce further

complicates urban last-mile delivery and incurs significant costs. The social-environmental

impact includes the vicious effect of the increasing traffic incidents and pollution due to transport

vehicles, which degrades the quality of life in the city. For example, based on the Beijing

Municipal Environmental Monitoring Center’s statistics, emissions of transport vehicles are the

main source of PM2.5 that causes hazardous haze in Beijing (http://www.bjmemc.com.cn/).

To build a smart city with congestion and pollution under control, an urban consolidation

center (UCC) is a potential solution to mitigate the repercussion of urban last-mile delivery.

Also known as a city distribution center (van Duin et al., 2008) or an urban distribution center

(Boudoin et al., 2014), a UCC consolidates shipments from multiple carriers and then deliv-

ers them to the city center using the UCC’s own fleet of trucks. A consolidator operating a

UCC usually requires a facility to sort the shipments according to their destinations before they

are delivered. As a result of the consolidation with fewer trucks, higher truck utilization can

be achieved, leading to a lower delivery cost. This shipment consolidation not only economi-

cally benefits stakeholders, including the consolidator, the carriers, and the public authorities

(Ambrosini and Routhier, 2004), but also mitigates the social-environmental impact because of

reduced traffic. Ideally, the resultant cost savings can be shared among the carriers, motivating

them to use the UCC’s service.

Despite the potential benefits, many UCC projects in practice are not successful. The UCCs

of the Port Authority of New York and New Jersey were closed after five years of operations

(Doig, 2001). Dablanc (2011) reports that 150 UCC projects were started in Europe during the

last 25 years, but only five projects survive. Even if they survive, they usually have difficulty to

break even and require significant subsidies from the government. For example, it costs a UCC

in La Rochelle 3.8e to deliver a parcel to a customer, who is charged only 1.7–3e. A UCC

2
in Monaco charges her customers 2.30e/100Kg, and receives 2.59e/100Kg as a subsidy from

the local government (Dablanc, 2005). Many UCC projects failed because the carriers were

reluctant to use their service. This is supported by a survey in the NYC metro, which reveals

that less than 20% of the carriers would like to participate in a UCC project (Holguin-Veras et

al., 2008).

More recently, some peer-to-peer platforms have been established for carriers to share their

delivery capacity. Notable examples include Saloodo! by DHL, Freightos and Convoy in Europe,

Loadsmart in U.S., and Cainiao and Truck Alliance in China. On such a platform, a carrier can

sell his unused capacity to another carrier to fulfill the latter’s delivery needs. It is attractive for

a consolidator to operate a platform because it requires neither a sorting facility nor a fleet of

delivery trucks. The peer-to-peer platform business model typically follows a sharing-economy

approach: The platform takes a revenue share from each transaction of capacity for providing

market access to the carriers and for processing the transaction (Gesing, 2017). In contrast to

the UCCs’ low success rate, the emergence of the capacity sharing platforms motivates us to

investigate whether the latter can be a better alternative for a city to address the challenges of

urban last-mile delivery.

Although bearing the delivery costs, a UCC can achieve a larger economy of scale as each

truck of the UCC may consolidate the tasks of many carriers. In contrast, a capacity sharing

platform does not incur any delivery cost, but each individual carrier on the platform has only

very limited delivery capacity compared to the UCC’s fleet. In this paper, we compare the above

two business models for urban last-mile delivery in terms of the consolidator’s profit and the

efficiency in mitigating the negative social-environmental impact. Specifically, the consolidator

can operate a UCC to bundle shipments from multiple carriers before the last-mile delivery,

or operate a peer-to-peer platform for the carriers to share their delivery capacity. For each

business model, we develop a game-theoretical model to capture the interactions between the

consolidator and the carriers. In each model, knowing that each carrier has a delivery task

with a random volume to fulfill, the consolidator first determines the price of delivery service

to maximize her expected profit. Then, after knowing his task volume, each carrier decides

whether to deliver his task to the city center on his own or use the consolidator’s service such

that his delivery cost is minimized. By identifying subgame perfect Nash Equilibrium with

rational expectations, we have obtained the following insights.

(i) If the consolidator operates a UCC, as the carriers’ variable delivery cost increases, the

3
carriers become more dependent on the UCC for delivering their tasks to the city center. This

result confirms the carriers’ motivation to save their delivery costs by using the UCC’s service

in practice.

(ii) If the consolidator operates a capacity sharing platform, we find that the consolidator

charges a price such that the demand and the supply of the capacity are balanced on the plat-

form. Since the consolidator can always earn a revenue share from each successful transaction,

our result implies that the platform business model can be more financially promising. This

may explain the increasing popularity of the capacity sharing platforms in practice.

(iii) In terms of the consolidator’s expected profit, we find that it is more profitable for the

consolidator to operate the UCC than the platform if and only if the carriers’ variable delivery

cost is sufficiently large. Furthermore, the UCC becomes more dominant when there are more

carriers. In terms of mitigating the negative social-environmental impact of urban last-mile

delivery, our analysis shows that it is more efficient for the consolidator to operate the UCC than

the platform if and only if there are sufficiently many carriers. Therefore, the UCC outperforms

the platform both economically and social-environmentally when the carriers’ variable delivery

cost and the number of carriers are both sufficiently large. Otherwise, the platform performs

better. This comparative analysis provides guidance for a consolidator to choose between the

two business models in practice.

After reviewing the related literature in §2, we formulate the problem between the consol-

idator and the carriers in §3. We analyze the two business models in which the consolidator

operates a UCC and a capacity sharing platform in §3.1 and §3.2 respectively. We compare the

two business models in terms of the consolidator’s expected profit and the efficiency in mitigat-

ing the social-environmental impact in §4. We extend our modeling framework in several ways

in §5, before we conclude in §6. All proofs are provided in Online Supplement A.

2 Related literature

This paper is mainly related to two streams of literature. The first stream consists of papers

on UCCs and the second stream is about peer-to-peer platforms. The majority of studies

on UCCs is conceptual and descriptive. McDermott (1975) shows in a survey conducted in

Columbus, Ohio that operating a UCC could bring substantial benefits to the shippers, carriers,

consumers, society, and government. Based on a program in the European network, Dablanc

(2007) concludes that the provision of urban logistics services emerges slowly despite their

4
growing demand. Allen et al. (2012) review the feasibility studies, trials, and fully operational

schemes of UCCs in 17 countries in the last 40 years.

Some analytical papers on UCCs focus on planning and allocation of delivery jobs among

the carriers. For example, Crainic et al. (2009) consider a two-tier distribution structure and

propose an optimization model to deal with job scheduling, resource management, and route

selection. Handoko et al. (2016) propose an auction mechanism for last-mile delivery to match

a UCC’s truck capacity to the shipments such that the UCC’s profit is maximized. Wang et

al. (2015) study a rolling-horizon auction mechanism with virtual pricing of shipping capacity.

Wang et al. (2018) consider cost uncertainty in last-mile delivery through a UCC and propose

approaches to solve the winner determination problem of an auction. Özener and Ergun (2008)

study a logistics network in which shippers collaborate and bundle their shipment requests to

negotiate better rates with a common carrier. They determine an optimal route covering all

the demands such that the total cost is minimized. To the best of our knowledge, no paper has

formally analyzed the stakeholders’ incentives for a UCC project. Our paper fills the gaps in

the literature by providing a game-theoretical analysis of the carriers’ incentive to participate

in a UCC project.

The ideas of the capacity sharing platform relate our paper to the literature on two-sided

markets (Rochet and Tirole, 2006, Weyl, 2010, Hagiu and Wright, 2015). A typical setting of

a two-sided market involves two types of players. On a platform, independent providers (such

as drivers) offer service to consumers (such as riders). See, for example, Cachon et al. (2017),

Bai et al. (2018), Taylor (2018), Bimpikis et al. (2016), Cohen and Zhang (2017), and Hu and

Zhou (2017). In contrast, a carrier on the platform in our paper is flexible to choose either to

sell his remaining capacity like a service provider or to buy capacity like a consumer.

Several papers in operations management deal with peer-to-peer rental platforms, which are

similar to our capacity sharing platform in spirit. For example, Fraiberger and Sundararajan

(2015) analyze a peer-to-peer rental market where each consumer is either a supplier or a buyer.

Benjaafar et al. (2018) analyze a model where players with different usage levels make decisions

on whether to own a product. Non-owners can access the product through renting from owners

on a needed basis. Jiang and Tian (2016) consider a setting in which consumers who purchased

a product can derive different usage values and generate income by renting out their purchased

product through a third-party sharing platform. Tian and Jiang (2018) further study how this

consumer-to-consumer product sharing affects a distribution channel. Abhishek et al. (2016)

5
consider a setting in which a consumer decides whether to purchase a durable good and whether

to rent it when the rental market is available. In the stream of literature above, if an owner

decides to rent out his product, he cannot use the product during the rental period. In contrast,

a carrier on our capacity sharing platform does not rent out his entire truck. Instead, he uses his

remaining truck capacity to deliver goods for another carrier to earn extra revenue. Benjaafar

et al. (2017) consider a ride sharing platform on which individuals may rent out empty seats

from their cars or find a ride. However, different from ride sharing, the carriers’ random task

volumes play a significant role in matching supply with demand of capacity on our capacity

sharing platform.

The collaboration among the carriers considered in our paper shares some similarity with

the paper by Agarwal and Ergun (2010), which considers the alliance formation among carri-

ers. They study the design of large-scale networks and the allocation of limited capacity on a

transportation network among the carriers in the alliance. Our paper is also related to the lit-

erature of inventory transshipment, which typically considers a wholesaler distributes inventory

to multiple retailers and the inventory can be transshipped among the retailers to fulfill de-

mand. Papers most relevant to our work include Rudi et al. (2001) and Dong and Rudi (2004),

where both the wholesaler’s and the retailers’ profits are considered. However, in this stream of

literature, a player with demand must work with another player with supply to generate profits.

In contrast, the carriers on our platform have the option to deliver by themselves and sell their

remaining capacity to the platform, allowing them to be a seller or a buyer. Our platform model

is also related to the literature of secondary markets, where resellers can buy and sell excess

inventory (see, for example, Lee and Whang (2002), Mendelson and Tunca (2007), Milner and

Kouvelis (2007), Broner et al. (2010), and Chen et al. (2013)). This stream of research focuses

on the impact of secondary markets on supply chains’ or firms’ performance. In contrast, our

paper compares the UCC with the capacity sharing platform. We do not see such a comparison

in this stream of literature.

3 Problem formulation

We consider a consolidator interacts with multiple carriers i = 1, 2, . . . , n. Each carrier i has

a delivery task with volume vi . We assume vi equals vL with a probability λ, or equals vH

(> vL ) with a probability 1 − λ, where λ ∈ [0, 1]. We refer the former as a low type carrier and

the latter as a high type carrier. All the carriers’ delivery tasks must be fulfilled. We assume

6
each carrier is initially equipped with logistics capability that has a limited delivery capacity

sufficient for his own task.

The consolidator first decides the pricing of the delivery service and each carrier then decides

whether to deliver on his own or outsource his task to the consolidator. If carrier i delivers on

his own, then the carrier incurs a fixed cost c > 0 and a variable cost per unit volume m > 0.

The fixed cost c includes the maintenance cost for the trucks, the license and permit fees for

the trucks, and the salary of drivers. The variable cost includes the fuel cost and the loading-

unloading cost (ASEAN, 2014).

Based on the above problem setting, we provide additional details of each business model

and analyze the equilibrium decisions of all the parties in §3.1 and §3.2, where the consolidator

operates a UCC and a capacity sharing platform respectively. We then compare the two business

models in §4 in terms of maximizing the consolidator’s expected profit and reducing the negative

social-environmental impact of urban last-mile delivery.

3.1 Business model 1: An urban consolidation center

3.1.1 Model description

In this section, we consider the consolidator operates a UCC to serve the carriers for their last-

mile delivery to the city center. We assume that the UCC owns a fleet of vehicles with a total

capacity that is sufficiently large to accommodate all the carriers’ tasks. The decision process

is as follows. The UCC first decides the price per unit volume p̄ of her delivery service. After

observing p̄, each carrier i waits until his delivery task volume vi is realized. We assume each

carrier i only knows his own realized task volume and decides independently on how to deliver

his task to the city center. Let d¯i denote the decision of carrier i, who has two possible options

defined as follows. (i) d¯i = 0: Carrier i delivers on his own. (ii) d¯i = 1: Carrier i uses the UCC’s

service. Figure 1 shows the sequence of decisions.

To serve any carrier(s), the UCC incurs a fixed cost C > 0 and a variable cost per unit

volume M > 0. To be consistent with reality, we assume that the UCC receives a subsidy S ≥ 0

per unit volume of shipments from the local government or authority. Note that S = 0 means

the UCC is not subsidized, which is a special case of our general model. As illustrated in Figure

1, the UCC first sets the price per unit volume p̄ of her service to maximize her expected profit.

Given the price p̄ and the realized task volume vi , each carrier i determines his decision d¯i to

minimize his cost.

7
 Decides the price
p̄ of her service

Task volumes
UCC Carriers
are realized

Decide whether
to use the
UCC’s service

Figure 1: The sequence of decisions under the UCC business model

We can solve the problem in Figure 1 backward by first determining the optimal decisions

of the carriers, before we find the optimal decision of the UCC. Given the price p̄, each carrier i

determines his optimal decision d¯∗i to minimize his cost. Define φ̄i d¯i ; p̄ as the cost of carrier i,

which is a function of d¯i given p̄. Each carrier i minimizes his cost φ̄i d¯i ; p̄ by comparing the

following two options: (i) d¯i = 0: Carrier i delivers on his own, which incurs a cost φ̄i (0; p̄) =

c + mvi . (ii) d¯i = 1: Carrier i uses the UCC’s service, which incurs a cost φ̄i (1; p̄) = p̄vi . If the

costs of the two options are identical, we assume that the carrier will use the UCC’s service.

After identifying each carrier’s optimal decision, we can substitute their optimal responses

into the UCC’s problem to find her optimal price p̄∗ . Let V denote the expected total task

volume of the carriers who use the UCC’s service. The UCC will choose a price p̄ to maximize

her expected profit:

π̄ (p̄) = (p̄ + S − M ) V − C. (1)

3.1.2 Equilibrium outcomes

h   i
v
c λ 1− v L −(1−λ) +(1−λ)(M −S)vH
Let m̄ = H
(1−λ)vH . Theorem 1 summarizes the equilibrium decisions.

Theorem 1. (Equilibrium decisions under the UCC model)

1. If m > m̄, then the UCC charges a price p̄∗ = m + c/vH to attract all the carriers to use

her service in equilibrium.

2. If m ≤ m̄, then the UCC charges a price p̄∗ = m + c/vL to attract the low type carriers to

use her service in equilibrium.

If the variable delivery cost m of each carrier is sufficiently large (m > m̄), the carriers

have a strong incentive to use the UCC’s service to save their delivery cost. In this case, the

8
UCC charges a lower price to attract all the carriers to use her service. On the other hand,

if m is small (m ≤ m̄), the carriers have a weaker incentive to use the UCC’s service. They

will use the service primarily to avoid the fixed delivery cost c. Anticipating that the low type

carriers are more likely to use her service (because they do not want to incur the fixed delivery

cost), the UCC charges a price to target only the low type carriers. Theorem 1 implies that

as m increases, all the carriers will eventually become dependent on the UCC to deliver their

tasks. This result confirms the carriers’ motivation to save their delivery cost through the

UCC’s service in practice. The following proposition determines the equilibrium profit of the

consolidator by operating the UCC.

Proposition 1. (Equilibrium profit under the UCC model)

1. If m > m̄, the UCC’s expected profit in equilibrium is π̄ ∗ (p̄∗ ) = (m + vcH + S − M )(λnvL +

(1 − λ)nvH ) − C.

2. If m ≤ m̄, the UCC’s expected profit in equilibrium is π̄ ∗ (p̄∗ ) = (m+ vcL +S −M )λnvL −C.

In addition to the expected profit, we also analyze the efficiency of the UCC in reducing the

negative social-environmental impact of urban last-mile delivery. As a result of the consolida-

tion, the UCC can mitigate the social-environmental impact by reducing traffic congestion and

pollution in the city center. Define ψ as a social-environmental cost associated with a carrier’s

delivery to the city center. This includes the cost to the society due to congestion and the cost

to the environment due to pollution. Without the UCC, the social-environmental cost incurred

by n carriers is nψ.

The UCC can consolidate the delivery tasks of multiple carriers to reduce the total trans-

portation distance. We adopt an approximation scheme from the travelling salesman problem

to quantify the transportation distance after consolidation. Specifically, we assume that the
√
UCC’s transportation distance of delivering n carriers’ tasks is at the scale of n (Steinerberger,

2015). Let nu denote the expected number of carriers served by the UCC in equilibrium. We

can approximate the expected social-environmental cost with the consolidation by the UCC as
√
nu ψ + V S + (n − nu )ψ, where V S is the total subsidy incurred to support the UCC. Define
¯ ψ = nψ−[√nu ψ+V S +(n−nu )ψ] as the expected social-environmental cost reduction achieved
∆

by the UCC. The following proposition summarizes the equilibrium social-environmental cost

reduction achieved by the UCC.

9
Proposition 2. (Equilibrium social-environmental cost reduction under the UCC

model)

1. If m > m̄, then the expected social-environmental cost reduction achieved by the UCC in
¯ ψ = n ψ − (λvL + (1 − λ)vH )S − √nψ.
 
equilibrium is ∆

2. If m ≤ m̄, then the expected social-environmental cost reduction achieved by the UCC in
  √
¯ ψ = λn ψ − vL S − λnψ.
equilibrium is ∆

3.2 Business model 2: A capacity sharing platform

3.2.1 Model description

Instead of having a physical UCC, we consider that the consolidator operates a platform for the

carriers to share their delivery capacity in this section. On the platform, a carrier delivering by

himself to the city center can sell his remaining truck capacity to another carrier, so that the

latter can outsource his delivery task by paying a fee. If the transaction is successful, then the

platform retains a portion of this fee as her revenue.

Motivated by the fact that each individual carrier usually has very limited delivery capacity

compared to the UCC’s fleet, we assume that a high task volume means a full or nearly-full

truckload for a carrier. Thus, in contrast to the UCC model, a high type carrier has to deliver

by himself to the city center (the other carriers cannot help him) and his remaining capacity

is insufficient to help any other carrier to deliver. Only the low type carriers will participate

(purchase or sell capacity) in the capacity sharing platform. We assume that each carrier

participating in the platform can serve (or can be served by) at most one other carrier on the

platform. We will relax this assumption in §5.2.

Figure 2 shows the sequence of decisions. The platform first decides the price per unit volume

p̂ of the delivery service. After observing the price p̂, each carrier waits until his delivery task

volume is realized. Each low type carrier i then decides independently on how to deliver his

task to the city center. Let dˆi denote his decision, which has two possible options. (i) dˆi = 0:

Delivers on his own and sells his remaining capacity to the platform. (ii) dˆi = 1: Purchases

capacity from the platform to fulfill his task.

If a low type carrier wants to sell his remaining capacity to the platform, whether his capacity

can be successfully sold depends on the demand and the supply of capacity on the platform.

If the demand is no less than the supply, then all the carriers who wish to sell their remaining

10
 Decides the
price p̂ of the
delivery service

Task volumes Low type

Platform
are realized carriers

Decide whether
to purchase
from the
platform

Figure 2: The sequence of decisions under the capacity sharing platform business model

capacity can successfully sell it. However, if the demand is less than the supply, then only a

subset of these carriers can sell their remaining capacity. In this situation, the platform will

randomly distribute the tasks with an equal probability to the carriers willing to sell their

remaining capacity.

Given that all the delivery tasks must be fulfilled, if a low type carrier wants to purchase

capacity from the platform to fulfill his task, we assume the carrier can always obtain his required

capacity vL . The platform can guarantee this by outsourcing the delivery task to an external

party, if necessary. We assume that the platform does not make any profit in this outsourcing

process. If a low type carrier i purchases capacity from the platform (dˆi = 1), then he pays

p̂vL . If there is enough supply on the platform, the platform receives a portion αp̂vL , where

α ∈ (0, 1) represents the platform’s revenue share. The remaining portion (1 − α)p̂vL goes to

the other carrier on the platform who serves carrier i. To ensure that selling remaining capacity

on the platform is profitable, we assume (1 − α)p̂ > m.

For notational convenience, define ns as the expected number of carriers who deliver on their

own and sell their remaining capacity to the platform (that is, the carriers who choose dˆi = 0).

Define np as the expected number of carriers who purchase capacity from the platform to fulfill

their tasks (that is, the carriers who choose dˆi = 1). Therefore, the supply and the demand of

capacity on the platform are proportional to ns and np respectively. To facilitate the transactions

among the carriers, the capacity sharing platform incurs an administrative cost that depends

on the number of successful transactions. Specifically, we assume that the administrative cost

equals min{ns , np } · A, where A > 0 is the administrative cost per transaction. To ensure that

the platform is profitable, we assume A < αp̂vL .

As illustrated in Figure 2, the platform first sets the price per unit volume p̂ of the delivery

11
service to maximize her expected profit. Given the price p̂ and the realized task volume vi ,

each low type carrier i determines his optimal decision dˆ∗i to minimize his expected cost. We

can solve the problem in Figure 2 backward by first identifying the optimal decision of each low

type carrier before we find the optimal decision of the capacity sharing platform.

Given the price p̂, each low type carrier i minimizes his expected cost by comparing the

two options: dˆi = 0 or 1. If carrier i delivers by himself and sells his remaining capacity to

the platform (dˆi = 0), then the expected revenue generated from selling his remaining capacity

depends on the supply (proportional to ns ) and the demand (proportional to np ) of capacity

on the platform. Following Su and Zhang (2008) and Cachon and Swinney (2009), we aim

to identify a subgame perfect Nash Equilibrium with rational expectations. We assume all

the low type carriers form the same rational beliefs ñs and ñp about ns and np , respectively,
 
when they optimize their decisions. Furthermore, for each low type carrier i, ñs = ns dˆ∗i and
 
ñp = np dˆ∗i in equilibrium. Define θ = min{ñp /ñs , 1}.
 
Specifically, each low type carrier i minimizes his expected cost φ̂i dˆi ; p̂ by comparing the

following options. (i) dˆi = 0: Carrier i delivers on his own and sells his remaining capacity

to the platform, which incurs an expected cost φ̂i (0; p̂) = c + mvL − θ [(1 − α)p̂ − m] vL . (ii)

dˆi = 1: Carrier i purchases capacity from the platform to fulfill his task, which incurs a cost

φ̂i (1; p̂) = p̂vL . If the cost of delivering by himself is identical to the cost of purchasing capacity

from the platform, we assume that carrier i will choose either option with an equal probability.

This random tie-breaking rule is to avoid the extreme situation where the carriers with identical

costs choose the same option on the platform.

After we determine the optimal decision dˆ∗i of each low type carrier i, we can substitute it

into the platform’s problem to find her optimal price. The platform will choose a price p̂ to

maximize her expected profit:

π̂ (p̂) = (αp̂vL − A) min {ns , np } . (2)

3.2.2 Equilibrium outcomes

Theorem 2. (Equilibrium decisions under the platform model) In equilibrium, the

c+2mvL
platform charges a price p̂∗ = (2−α)vL to ensure that each low type carrier purchases capacity

from the platform, or delivers on his own and sells his remaining capacity on the platform with

an equal probability.

It is optimal for the platform to charge the market clearance price in Theorem 2 because

her expected profit is the highest when the demand and the supply of capacity are balanced,

12
resulting in the largest number of successful transactions on the platform. Since the consolidator

can always earn a revenue share from each transaction, Theorem 2 shows that the platform

business model is financially promising in equilibrium. This result may explain the increasing

popularity of the capacity sharing platforms in practice. The following proposition determines

the equilibrium profit of the consolidator by operating the capacity sharing platform.

Proposition 3. (Equilibrium profit under the platform model) In equilibrium, the ca-
αλn(c+2mvL )
pacity sharing platform’s expected profit is π̂ ∗ (p̂∗ ) = 2(2−α) − λnA
2 .

Similar to the UCC model, we also analyze the efficiency of the capacity sharing platform in

mitigating the negative social-environmental impact. By sharing capacity among the low type

carriers, the platform can result in fewer trucks used, which reduces congestion and pollution.

Recall that ψ represents a social-environmental cost associated with a carrier’s delivery to

the city center. Since the delivery task of a carrier who purchases capacity from the platform is

fulfilled by another carrier, it leads to a reduction of ψ in the social-environmental cost. Recall

that np represents the expected number of carriers who purchase capacity from the platform

in equilibrium. Theorem 2 reveals that the demand and the supply of delivery capacity on

the platform are balanced in equilibrium. Therefore, the expected social-environmental cost
ˆ ψ = np ψ, which is determined by the
reduction achieved by the platform in equilibrium is ∆

following proposition.

Proposition 4. (Equilibrium social-environmental cost reduction under the platform

model) In equilibrium, the expected social-environmental cost reduction achieved by the capacity
ˆ ψ = λnψ/2.
sharing platform is ∆

4 Comparing the UCC and the capacity sharing platform

We compare the UCC and the capacity sharing platform in terms of their expected profits and

their efficiency in mitigating the negative social-environmental impact.

4.1 Expected profit

Between the UCC and the capacity sharing platform, which business model is more profitable

for the consolidator? As discussed in §1, it is important to make the consolidator financially

sustainable in order to achieve the benefits of consolidation. We determine the consolidator’s

preference by comparing the equilibrium profits π̄ ∗ (p̄∗ ) of the UCC in §3.1 and π̂ ∗ (p̂∗ ) of the

13
 C− λnA αλnc
+ 2(2−α) +(M −S− vc )(λnvL +(1−λ)nvH )
capacity sharing platform in §3.2. Let m̄0 = 2
α
λnvL (1− 2−α
H
)+(1−λ)nvH and m̄00 =
C− λnA
2
αλnc
+ 2(2−α) +(M −S− vc )λnvL
α
λnvL (1− 2−α )
L
. The following theorem identifies conditions under which the UCC

(or the platform) is more profitable for the consolidator.

Theorem 3. (Comparing the UCC’s and the platform’s profits)

1. If m > m̄, then the UCC is more profitable than the platform if and only if m > m̄0 , where

m̄0 decreases in n.

2. If m ≤ m̄, then the UCC is more profitable than the platform if and only if m > m̄00 ,

where m̄00 decreases in n.

Theorem 3 shows that the UCC is more profitable than the platform if the carriers’ variable

delivery cost m is large. This is because when their delivery cost is large, the carriers have

a strong incentive to outsource their delivery tasks. This will benefit the consolidator if she

operates a UCC because there will be many carriers using her service. On the other hand, if

the consolidator operates a platform, there will not be many successful transactions. This is

because the carriers are reluctant to deliver on their own and sell their remaining capacity on

the platform, and thus the supply of capacity is low. Furthermore, Theorem 3 shows that the

thresholds m̄0 and m̄00 decrease with the number of carriers n. As n increases, the UCC can

achieve a larger economy of scale and thus become more likely to outperform the platform.

We now consider a fixed market entry cost for the consolidator. Let Ē and Ê denote the

fixed market entry costs if the consolidator operates a UCC and a platform respectively. Let

m̄0E = m̄0 + λnvL (1− Ē−Ê

α
)+(1−λ)nvH and m̄00E = m̄00 + λnvLĒ−
(1−
Ê
α
) . The following corollary identifies
2−α 2−α

conditions under which the UCC (or the platform) is more profitable for the consolidator if there

is a fixed market entry cost for each business model.

Corollary 1. (Comparing the UCC’s and the platform’s profits under fixed market

entry costs)

1. If m > m̄, then the UCC is more profitable than the platform if and only if m > m̄0E ,

where m̄0E decreases in n.

2. If m ≤ m̄, then the UCC is more profitable than the platform if and only if m > m̄00E ,

where m̄00E decreases in n.

In practice, the market entry cost Ē of the UCC may be larger than the market entry cost Ê

of the platform. In that case, we have m̄0E > m̄0 and m̄00E > m̄00 .

14
4.2 Social-environmental impact

Between the UCC and the capacity sharing platform, which business model is more efficient to

mitigate the negative social-environmental impact of urban last-mile delivery? As a result of the

consolidation, both the UCC and the platform can mitigate the negative social-environmental

impact (in terms of reduced congestion and pollution). Under the UCC model, although ad-

ditional trucks are required, each UCC’s truck can potentially consolidate multiple tasks. In

contrast, under the platform model, although no additional trucks are required, each carrier can

at most serve one other carrier’s task. It is unclear that which business model is more effective

in reducing the negative social-environmental impact.

¯ ψ un-
In this section, we compare the equilibrium social-environmental cost reductions ∆

der the UCC model in §3.1 and ∆

ˆ ψ under the capacity sharing platform model in §3.2. Let
ψ2 ψ2
n̄ =  2 and n̄0 = ψ
. The following theorem identifies condi-
2
( 1− λ
)ψ−(λvL +(1−λ)vH )S λ( 2)−vL S
2
tions under which the UCC (or the platform) is more efficient to reduce the negative social-

environmental impact.

Theorem 4. (Comparing the UCC’s and the platform’s social-environmental cost

reductions)

1. If m > m̄, then the UCC is more efficient in reducing the social-environmental cost than

the platform if and only if n > n̄.

2. If m ≤ m̄, then the UCC is more efficient in reducing the social-environmental cost than

the platform if and only if n > n̄0 .

Theorem 4 shows that if the number of carriers n is large, then the UCC is more efficient

in reducing the negative social-environmental impact than the platform. This is because if n is

large, the UCC’s trucks (each can serve multiple tasks) can achieve a larger economy of scale in

shipment consolidation. This significantly reduces the traffic congestion and pollution caused

by the last-mile delivery. On the other hand, if n is small, the UCC may not be efficient because

of the additional UCC trucks. In contrast, the platform, which matches a carrier’s task with

another carrier without employing any additional trucks, becomes more efficient.

After comparing the expected profits and the expected social-environmental cost reductions

of the UCC and the capacity sharing platform, we illustrate the predominant business model

with respect to these two performance measures for different parameter regions in Figure 3. In

general, if the carriers’ variable delivery cost m and the number of carriers n are both large,

15
then the consolidator should operate the UCC which is more profitable and also more efficient

in reducing the negative social-environmental impact. On the other hand, if both m and n

are small, then the consolidator should choose the capacity sharing platform, which performs

better both economically and environmentally. In contrast, in the shaded areas of Figure 3, the

UCC or the capacity sharing platform only outperforms the other in one aspect. Therefore, the

consolidator may prefer either business model depending on her objective. This comparative

analysis provides guidance for consolidators to make better choices between the two business

models in practice.

m̄
UCC more m̄0
profitable and m̄00
n̄
more efficient
n̄0
m

Platform more
profitable and
more efficient

n
UCC more profitable; Platform more efficient
Platform more profitable; UCC more efficient

Figure 3: The preferred business model for the consolidator

5 Extensions

5.1 A hybrid model

In our model, we assume that the consolidator can only choose one business model, that is,

operating either the UCC or the capacity sharing platform. In this section, we consider a hybrid

business model that combines the UCC and the capacity sharing platform. In the hybrid model,

the consolidator can simultaneously operate a UCC that fulfills the carriers’ delivery tasks, and

a platform that matches supply and demand for capacity among the carriers. This hybrid model

is inspired by Amazon, which sells products to its consumers and also allows peer-to-peer selling

on its platform.

Specifically, we consider a setting in which the consolidator operates both the UCC and

the platform. Through the UCC, the consolidator charges the carriers for her delivery service.

Through the platform, the consolidator receives a revenue share α ∈ (0, 1) from each successful

16
transaction of capacity. The consolidator first chooses the prices p̄ and p̂ per unit volume of

delivery service for the UCC and the platform, respectively, to maximize her expected profit.

After observing the prices p̄ and p̂, each carrier i waits until his delivery task volume vi is

realized. Depending on vi , each carrier i has different options to fulfill his task. A low type

carrier i has three possible options: (i) He delivers on his own and sells his remaining capacity to

the platform. (ii) He uses the UCC’s service. (iii) He purchases capacity from the platform. A

high type carrier i has two possible options: (i) He delivers on his own. (ii) He uses the UCC’s

service. Each carrier independently decides how to fulfill his task to minimize his expected

cost. To ensure that selling capacity on the platform is profitable and the options do not always

dominate each other, we assume m < (1 − α)p̂ < (1/vL − 1/vH ) c. The following theorem

summarizes the consolidator’s and the carriers’ equilibrium decisions.

Theorem 5. (Equilibrium decisions under the hybrid model)

1. If m ≤ min{m̄0 , m̄00 }, then the consolidator charges any p̄∗ > (c + 2mvL )/((2 − α)vL ) and

p̂∗ = (c + 2mvL )/((2 − α)vL ) in equilibrium. Under these prices, each high type carrier delivers

on his own, and each low type carrier delivers on his own (and sells his remaining capacity on

the platform) or purchases capacity from the platform with an equal probability.

2. If m̄00 < m ≤ m̄, then the consolidator charges p̄∗ = m + c/vL and any p̂∗ ≥ m + c/vL in

equilibrium. Under these prices, each high type carrier delivers on his own, and each low type

carrier uses the UCC’s service.

3. If m > max{m̄, m̄0 }, then the consolidator charges p̄∗ = m + c/vH and any p̂∗ ≥ m + c/vL in

equilibrium. Under these prices, all the carriers use the UCC’s service.

There are three parameter regions in which the consolidator generates profit from different

sources. If the carriers’ variable delivery cost m is small, then the consolidator will only generate

profit from the platform. This is because the affordable delivery cost m makes it difficult to

attract the carriers to use the UCC’s service. However, as m becomes moderate or large, more

carriers are willing to outsource their delivery tasks. Specifically, if m is moderate, then only

the carriers with a high task volume will deliver on their own. If m is large, then no carriers will

make their own delivery. Both cases eliminate the supply of capacity on the platform. Thus,

the consolidator will optimize her prices to induce the carriers to engage the UCC (rather than

the platform), such that her expected profit is maximized. In both cases, the consolidator only

generates profit from the UCC.

In this hybrid model, although the consolidator operates both the UCC and the platform

17
concurrently, the consolidator charges a price to incentivize the carriers to participate only in

the more profitable business model rather than both. This is consistent with Theorem 3 of our

base model, which compares the expected profits of the UCC and the capacity sharing platform.

5.2 A generalized platform model

Motivated by the fact that each individual carrier’s delivery capacity is usually very limited,

we assume that each carrier on the platform can serve (or can be served by) at most one other

carrier in the platform model. In this section, we generalize the platform model such that each

carrier can serve (or can be served by) multiple other carriers. Specifically, we assume that each

carrier has a finite delivery capacity v̄. If a carrier delivers on his own, he can sell his remaining

capacity v̄ − vL on the platform, which can be purchased by other carriers. If a carrier purchases

capacity from the platform, his delivery task can be fulfilled by multiple other carriers. The

number of successful transactions on the platform will be proportional to min{ns (v̄ −vL ), np vL }.

The rest of the generalized platform model is identical to that in §3.2. The following theorem

summarizes the platform’s and the carriers’ equilibrium decisions.

Theorem 6. (Equilibrium decisions under the generalized platform model)

c+2mvL
1. If v̄ > 2vL , the platform’s price is p̂∗ = (2−α)vL in equilibrium.

2. If v̄ ≤ 2vL , the platform’s price is p̂∗ = c+mv̄

(1−α)v̄+αvL in equilibrium.

For both cases, each low type carrier purchases capacity from the platform, or delivers on his

own and sells his remaining capacity on the platform with an equal probability.

If the carriers’ delivery capacity v̄ > 2vL , then a carrier is able to fulfill at least one other

carrier’s delivery task. If v̄ ≤ 2vL , the delivery task of a carrier, who purchases capacity from

the platform, will be served by multiple other carriers. Although the platform charges a different

price in equilibrium for the different cases, the carriers’ equilibrium decisions and the insights

remain the same as that of the original platform model in §3.2.

5.3 Heterogeneous fixed delivery costs

We assume that the carriers have the same fixed delivery cost c if they deliver on their own,

although they may have different total delivery costs given their heterogeneous task volumes.

In this section, we generalize the UCC and the platform models in §3.1 and §3.2, respectively,

to incorporate heterogeneous fixed delivery costs. Specifically, we assume that a fraction η of

the carriers has a low fixed delivery cost cL , and the remaining fraction (1 − η) of the carriers

18
has a high fixed delivery cost cH (> cL ). Note that the carriers’ variable delivery costs are also

heterogeneous given the heterogeneity of their delivery task volumes. A detailed analysis of

this generalized setting is presented in Online Supplement B. Here, we focus on the impact of

heterogeneous fixed delivery costs on the profits of the UCC and the platform.

UCC more
UCC more profitable
profitable

m
m

Platform more
Platform more profitable
profitable

n n
1
(a) η ≤ η̄ (b) η̄ < η ≤ 2

UCC more UCC more

profitable profitable
Platform more Platform more
profitable profitable
m

UCC more UCC more

profitable profitable
Platform more Platform more
profitable profitable
n n
(c) 1
2 < η ≤ η̄ 0
(d) η > η̄ 0

Figure 4: Comparing the UCC’s and platform’s profits under heterogeneous fixed delivery costs

Figures 4a and 4b show that if the carriers are more likely to have the high fixed delivery

cost cH (that is, η ≤ 1/2), then the UCC tends to be more profitable as m and n become larger.

This is similar to the observation from Figure 3 in the original setting in §4.

However, Figures 4c and 4d show that if the carriers are more likely to have the low fixed

delivery cost cL (that is, η > 1/2), the results become different from that of the original setting.

Specifically, the platform may be more profitable even if m is large, whereas the UCC may be

more profitable if m is small. This is because in the original setting in §4, it is difficult for

the platform to be more profitable (than the UCC) if m is large because of the low supply of

capacity. However, in the generalized setting, more carriers deliver on their own and sell their

remaining capacity on the platform because of their low fixed delivery cost. This creates more

successful transactions on the platform, making it more profitable. In contrast, if m is small,

19
many carriers deliver on their own given that both their fixed and variable delivery costs are

low. This leads to excessive supply of capacity on the platform, which may make the UCC more

profitable than the platform.

5.4 A two-period setting

As mentioned in §1, some UCC projects are not successful in practice because carriers are re-

luctant to use a UCC’s service. The carriers are concerned with the potential loss of their own

logistics capability in the long run if they rely on the UCC. It is possible that after a carrier

starts using the UCC’s service, he reduces or eliminates his own logistics capability to cut down

expenses (Snapp, 2012, Vivaldini et al., 2012, Choe et al., 2017). For example, the logistics

department of GOME, a Chinese retailer for electrical appliances, reduces its investment in

delivery trucks and drivers after engaging a consolidation service (National Express, 2010). In

this case, if the UCC charges a high price in the future, it will be difficult for the carriers to

reestablish their logistics capability and perform their own delivery. The substantial reestab-

lishment cost, which may include the costs to purchase trucks, recruit drivers, obtain licenses,

and gain knowledge about local clients (Browne et al., 2005), makes the carriers reluctant to

rely on the UCC’s service.

With this observation, we consider a two-period setting to understand how this long-term

factor affects our results. Specifically, we develop a two-period model for each of the UCC and

the capacity sharing platform. In each period, the consolidator first determines the price of

delivery service to maximize her expected profit. Then, after knowing his task volume, each

carrier decides whether to deliver on his own or use the consolidator’s service to deliver his task

to the city center such that his expected cost is minimized.

In period 1, if a carrier decides to use the consolidator’s service, then he can also choose

to eliminate or keep his logistics capability for the future. It will incur a holding cost h > 0

to the carrier for keeping his logistics capability. In period 2, if a carrier decides to deliver

on his own, then he needs to reestablish his logistics capability if it is eliminated in period 1.

It will incur a setup cost f > 0 to the carrier for reestablishing his logistics capability. We

assume that each carrier’s delivery capacity has no value after period 2. Let δ ∈ (0, 1) denote

a discount factor across the two periods. To rule out uninteresting cases, such as the carriers

never keep their logistics capability, we assume h ≤ δ(1 − λ)c(vH − vL )/vL , f > c(vH − vL )/vL ,

and λ ≤ vH /(2vH − vL ). The rest of the models remains the same as the original models in §3.1

20
and §3.2.

5.4.1 The UCC model

h   i  
v v
c λ 1− v L −(1−λ) +(1−λ)(M −S)vH λ2 (c+f )− λ v L +1−λ c+(M −S)(1−λ)(vH +λvL )
Recall that m̄ = H
(1−λ)vH . Define m̂1 = H
(1−λ)(vH +λvL ) ,
    
v v
λf −(1−λ)c+(M −S)(1−λ)vH λ 1− v L −(1−λ) 1−λ vH c+(M −S)(1−λ)2 vH
m̂2 = (1−λ)vH , and m̂3 = H L
(1−λ)2 vH
. The fol-

lowing theorem characterizes the equilibrium decisions of the UCC and the carriers under the

two-period UCC model.

Theorem 7. (Equilibrium decisions under the two-period UCC model)

1. If m > max{m̂1 , m̂2 , m̂3 }, then the equilibrium decisions in each period are as follows.

Period 1: The UCC’s price is p̄∗1 = m + c/vL . Under this price, each carrier i uses the

UCC’s service and eliminates his logistics capability if vi1 = vL , and delivers on his

own otherwise.

Period 2: The UCC’s price is p̄∗2 = m + c/vH . Under this price, all the carriers use the

UCC’s service.

2. If m ≤ m̄, then the equilibrium decisions in each period are as follows.

Period 1: The UCC’s price is p̄∗1 = m + (c − h)/vL . Under this price, each carrier i uses

the UCC’s service and keeps his logistics capability if vi1 = vL , and delivers on his

own otherwise.

Period 2: The UCC’s price is p̄∗2 = m + c/vL . Under this price, each carrier i uses the

UCC’s service if vi1 = vL , and delivers on his own otherwise.

Note that the threshold max{m̂1 , m̂2 , m̂3 } > m̄, which means the two intervals of m in

Theorem 7 do not overlap, and there is no equilibrium for m̄ < m ≤ max{m̂1 , m̂2 , m̂3 }. As the

threshold max{m̂1 , m̂2 , m̂3 } depends on the carriers’ reestablishment cost f , we can characterize

the carriers’ equilibrium decisions using their variable delivery cost m and their reestablishment

cost f . Figures 5a and 5b illustrate the equilibrium decisions of the carriers in periods 1

and 2 respectively. If the carriers’ variable delivery cost is small (m ≤ m̄), then the carriers’

equilibrium decisions are consistent with that of the one-period model in Theorem 1. That is,

for both periods, only the low type carriers use the UCC’s service and the high type carriers

deliver on their own. In addition, the carriers who use the UCC’s service in period 1 keep their

own logistics capability. Since delivering on their own is not costly given the small variable

21
delivery cost m, keeping logistics capability ensures that the carriers can easily switch back to

delivery on their own in period 2 if the UCC’s service becomes expensive.

Carriers with vL use UCC and m̂1

eliminate logistics capability; m̂2
All carriers use UCC m̂3
carriers with vH deliver
m̄
on their own

m
m

No equilibrium No equilibrium

Carriers with vL use UCC but

Carriers with vL use UCC;
keep logistics capability;
carriers with vH deliver on their own carriers with vH deliver on their own

f f

(a) Period 1 (b) Period 2

Figure 5: The carriers’ equilibrium decisions under the two-period UCC model

On the other hand, if the carriers’ variable delivery cost is large (m > m̄), there may or

may not be an equilibrium outcome, which is different from the one-period model in §3.1. If the

reestablishment cost f is sufficiently small (corresponding to the top-left corner of Figure 5a),

then the carriers, who use the UCC’s service, eliminate their logistics capability. This is because

the carriers anticipate that they are likely to use the UCC’s service in period 2 given the large

m. Having said that, given the small reestablishment cost f , it is easier for these carriers to

switch to delivery on their own if the UCC charges a high price in period 2. Anticipating this,

the UCC will charge a suitable price in period 2 to retain these carriers for using her service,

which results in an equilibrium.

As the reestablishment cost f increases, the equilibrium becomes less achievable (correspond-

ing to the top-right corner of Figure 5a). The expensive reestablishment cost makes it difficult

for these carriers (who eliminate their logistics capability in period 1) to switch to delivery on

their own in period 2. The UCC anticipates that these carriers are likely to continue using her

service in period 2, and therefore she will increase her price in period 2. Knowing that this will

happen, these carriers would rather not use the UCC’s service in period 1 to avoid the expensive

UCC’s service or capacity reestablishment in period 2. This leads to no equilibrium.

The above finding echoes our observation in practice. When a carrier finds that the UCC’s

service is cheaper than delivering on his own (under a large m), he may start using the UCC’s

service and eliminate his logistics capability to cut down expenses (to avoid the holding cost

h). However, if the UCC charges a high price in the future, it will be difficult for the carrier

22
to reestablish his logistics capability and switch back to delivery on his own (given a large f ).

This deters the carriers from using the UCC’s service in the first place. From this perspective,

our result partially explains why the carriers are reluctant to use a UCC’s service in practice,

causing many UCC projects not successful.

5.4.2 The capacity sharing platform model

The following theorem summarizes the equilibrium decisions of the capacity sharing platform

and the carriers under the two-period platform model.

Theorem 8. (Equilibrium decisions under the two-period platform model)

Period 1: The platform’s price is p̂∗1 = [c + 2mvL − h]/[(2 − α)vL ] in equilibrium. Under this

price, each low type carrier purchases capacity from the platform (but keeps his logistics

capability), or delivers on his own (and sells his remaining capacity on the platform) with

an equal probability.

Period 2: The platform’s price is p̂∗2 = (c + 2mvL )/[(2 − α)vL ] in equilibrium. Under this price,

each low type carrier purchases capacity from the platform, or delivers on his own (and

sells his remaining capacity on the platform) with an equal probability.

In equilibrium, the carriers always keep their logistics capability in period 1 even if they

purchase capacity from the platform. This means that each carrier always has his logistics

capability on hand for both periods, which allows them to flexibly choose a delivery option

(without paying a reestablishment cost). Since the carriers can deliver on their own and sell

remaining capacity on the platform, the platform will have sufficient capacity available to facil-

itate successful transactions. This makes the platform financially promising as she can always

earn a revenue share from each successful transaction. Furthermore, unlike the two-period UCC

model, the two-period platform model can always achieve an equilibrium. Our result provides

explanations for the increasing popularity of the capacity sharing platforms in practice.

The relative performance of the UCC and the platform in terms of the profitability and

the social-environmental cost reduction under the two-period setting is similar to that of the

one-period setting. All the insights in §4 continue to hold, so we omit the details here.

23
5.5 Correlated demands

In the two-period setting, the carriers can keep or eliminate their logistics capability if they use

the consolidator’s service in the first period. This decision depends on the carriers’ expectation

of their delivery task volumes for the next period. In practice, the carriers’ demands across the

periods can be correlated. In this section, we analyze the UCC and the platform models under

the two-period setting with correlated task volumes between the two periods for each carrier.

Specifically, we assume that if the carrier’s task volume is low in period 1, then in period 2 his

task volume will be low with probability ξL ∈ [0, 1], or high with probability (1 − ξL ). If the car-

rier’s task volume is high in period 1, then in period 2 his task volume will be high with probabil-

ity ξH ∈ [0, 1], or low with probability (1−ξH ). To rule out uninteresting cases where the carriers
n    o
never keep their logistics capability, we assume h ≤ min δ(1 − ξL )c vvHL − 1 , δξH c vvHL − 1 .

The rest of the models is identical to that of §5.4.

5.5.1 The UCC model

 
v
λ[λξL +(1−λ)(1−ξH )](c+f )− [λξL +(1−λ)(1−ξH )] vH +λ(1−ξ)+(1−λ)ξH c
Define m̂4 = L
(λ(1−ξL )+(1−λ)ξH )vH +[λξL +(1−λ)(1−ξH )](1−λ)vL + M − S,
     h   i
v v v
[λξL +(1−λ)(1−ξH )] 1− vH −(λ(1−ξL )+(1−λ)ξH ) 1−λ v L c (c−h) λ 1− v L −(1−λ)
m̂5 = L
(1−λ)(λ(1−ξL )+(1−λ)ξH )vH
H
+M − S, m̂6 = H
(1−λ)vH +
   
vH
[λξL +(1−λ)(1−ξH )] 1− v −(λ(1−ξL )+(1−λ)ξH ) c
M − S, and m̂7 = L
(λ(1−ξL )+(1−λ)ξH )vH + M − S . The following theorem

summarizes the carriers’ equilibrium decisions under the UCC model.

Theorem 9. (Equilibrium decisions under the two-period UCC model with corre-

lated demands)

1. If m > max{m̂2 , m̂4 , m̂5 }, the carriers’ equilibrium decisions in each period are as follows.

Period 1: Each carrier i uses the UCC’s service and eliminates his logistics capability if

vi1 = vL , and delivers on his own otherwise.

Period 2: Each carrier i uses the UCC’s service.

2. If m̂6 < m ≤ m̂7 , the carriers’ equilibrium decisions in each period are as follows.

Period 1: Each carrier i uses the UCC’s service but keeps his logistics capability.

Period 2: Each carrier i uses the UCC’s service if vi2 = vL , and delivers on his own

otherwise.

3. If m ≤ min{m̂6 , m̂7 }, the carriers’ equilibrium decisions in each period are as follows.

24
 Period 1: Each carrier i uses the UCC’s service but keep his logistics capability if vi1 = vL ,

and delivers on his own otherwise.

Period 2: Each carrier i uses the UCC’s service if vi2 = vL , and delivers on his own

otherwise.

Carriers with vL use UCC and

eliminate logistics capability;
carriers with vH deliver All carriers use UCC
on their own
No equilibrium No equilibrium
m

m
All carriers use UCC but keep
logistics capability
Carriers with vL use UCC;
Carriers with vL use UCC but carriers with vH deliver on their own
keep logistics capability;
carriers with vH deliver on their own
f f

(a) Period 1 (b) Period 2

Figure 6: The carriers’ equilibrium decisions under the two-period UCC model with correlated
demands

Figure 6 shows the carriers’ equilibrium decisions under the two-period UCC model with

correlated demands. Compared to the equilibrium results with independent demands in Figure

5, we observe one additional case in equilibrium: All the carriers use the UCC’s service but keep

their logistics capability in period 1. If the demands are correlated, the carriers become less

conservative in using the UCC’s service. This is because the carriers can make a more informed

decision about their logistics capability, and thus they are less concerned about the increase of

the UCC’s price in period 2. This implies that the UCC’s service may become more appealing

when the demands are correlated.

5.5.2 The capacity sharing platform model

The following theorem summarizes the equilibrium decisions of the low type carriers under the

platform model.

Theorem 10. (Equilibrium decisions under the two-period platform model with

correlated demands) If f > A/α − (c + mvL ), then the carriers’ equilibrium decisions in each

period are as follows.

Period 1: Each low type carrier purchases capacity from the platform (but keeps his logistics

capability), or delivers on his own (and sells his remaining capacity on the platform) with

an equal probability.

25
 Period 2: Each low type carrier purchases capacity from the platform, or delivers on his own

(and sells his remaining capacity on the platform) with an equal probability.

Although the carriers make the same decisions as in the platform model with independent

demands (see Theorem 8), we find that the correlated demands make the platform model less

viable. Specifically, there is no equilibrium when the reestablishment cost f is too small. The

reason is as follows. When the demands are correlated, if the carriers anticipate that they are

likely to have a high task volume (and have to deliver on their own) in period 2, they would keep

their logistics capability in period 1. However, a small reestablishment cost f incentivizes the

carriers to eliminate their logistics capability, leading to deviation. On the other hand, if the

carriers anticipate a low task volume in period 2, they would eliminate their logistics capability

in period 1 given a small f . The platform then will charge a high price in period 2, deterring

the carriers from eliminating their logistics capability in period 1, leading to deviation.

6 Conclusion

The consistent growth of urban population and e-commerce activities significantly increases

the demand for last-mile delivery to urban areas. This causes more traffic congestion in a city

center, creating negative impact on the well-being of the city economically, socially, and envi-

ronmentally. We study how a consolidator can make urban last-mile delivery more economically

and social-environmentally sustainable.

Specifically, the consolidator can choose to operate a UCC or a capacity sharing platform

to consolidate delivery tasks from multiple carriers. Under the UCC business model, the con-

solidator requires a sorting facility and a fleet of trucks to deliver the tasks of the carriers. The

consolidator bears the delivery costs, but charges the carriers a service fee for the last-mile deliv-

ery. Under the capacity sharing platform business model, the consolidator operates a platform

for the carriers to share their delivery capacity. The consolidator does not need a facility or

trucks, but there is an administrative cost incurred to the consolidator for matching the supply

and demand of the capacity on the platform. She receives a revenue share from each successful

transaction of capacity on the platform.

For each business model, we develop a game-theoretical model capturing the interactions

between the consolidator and the multiple carriers. Under each model, the consolidator first

determines the price per unit volume of the delivery service to maximize her expected profit.

26
After knowing his task volume, each carrier then minimizes the cost to fulfill his delivery task

by (i) delivering on his own, or (ii) using the consolidator’s service.

In practice, by using the UCC, the carriers can potentially save their delivery costs because

of the consolidation. Our game-theoretical model demonstrates this potential benefit through

its equilibrium results. We observe that as the carriers’ variable delivery cost increases, they

become more dependent on the UCC to fulfill their delivery tasks and eventually all the carriers

will use the UCC’s service in equilibrium.

Under the capacity sharing platform model, the consolidator will charge a market clearance

price such that the supply matches the demand on the platform in equilibrium. This facilitates

sufficient successful transactions and ensures that the platform can always earn the revenue

share from each transaction in equilibrium. This makes the platform business model financially

promising. Our equilibrium result may explain the increasing popularity of the capacity sharing

platforms in practice.

To provide guidance for the consolidator to choose between the two business models, we first

compare their expected profits. In general, the UCC is more profitable than the platform if the

carriers’ variable delivery cost is large. In this situation, the carriers have a strong incentive to

outsource their delivery tasks. This will benefit the consolidator if she operates a UCC because

there will be many carriers using her service. On the other hand, this situation will lead to

high demand but low supply of capacity on the platform as the carriers tend to outsource their

delivery tasks. The low supply of capacity makes the platform less profitable than the UCC.

We also determine which business model is more efficient in mitigating the negative social-

environmental impact of urban last-mile delivery. Although additional trucks are required by

the UCC model, each truck of the UCC can potentially consolidate multiple carriers’ tasks.

In contrast, no additional trucks are required by the platform model, but each carrier on the

platform has limited capacity. We find that if the total number of carriers is sufficiently large,

then the UCC is more efficient than the platform in mitigating the social-environmental impact.

In this case, the UCC’s trucks (each can serve multiple tasks) can achieve a larger economy of

scale in shipment consolidation. This significantly reduces the traffic congestion and pollution

of urban last-mile delivery.

Based on the above findings, we identify the predominant business model for the consolidator

with respect to both the profitability and the efficiency in reducing negative social-environmental

impact. In general, the UCC outperforms the platform in both perspectives as long as the

27
carriers’ variable delivery cost and the number of carriers are sufficiently large. Otherwise, the

platform performs better, even though the UCC and the platform may outperform each other

in only one perspective. Following our comparative analysis, the consolidator can always choose

a business model that suits her essential need (profitability or efficiency in reducing the social-

environmental cost). For example, a nonprofit organization in practice may prefer the business

model that is more efficient in reducing the negative social-environmental impact, whereas a

private company may prefer the business model that yields a larger profit.

We study several extensions of our modeling framework. The first extension considers a

hybrid model in which the consolidator concurrently operates a UCC and a platform. We have

also generalized the platform model such that each carrier can serve or can be served by multiple

other carriers. The third extension assumes heterogeneous fixed delivery costs of the carriers.

Finally, we analyze both the UCC and the platform models in a two-period setting to capture

their long-term impact. We also consider demand correlation across the two periods.

Acknowledgment

The authors thank the senior editor and the two anonymous referees for their valuable com-

ments that have substantially improved the paper. Portions of this paper have been presented

in Arizona State University, 2018; Cai Niao (Alibaba Group), 2018; Nanyang Technological

University, 2018; Northeastern University (China), 2018; Zhejiang University, 2018; Chinese

University of Hong Kong (Shenzhen), 2020; and Southwest Jiaotong University, 2020. The au-

thors thank the audiences for many insightful comments and stimulating questions. The second

author is grateful for the support from the Singapore Management University (SMU) under

the Lee Kong Chian Fellowship. The third author thanks the support from SMU under the

Maritime and Port Authority (MPA) Research Fellowship and the Key Program of National

Natural Science Foundation of China (No. 71931009). Both second and third authors thank the

support from SMU under the Retail Centre of Excellence (RCoE) research grant.

References
Abhishek V, Guajardo JA, Zhang Z (2016) Business models in the sharing economy: Manufac-
turing durable goods in the presence of peer-to-peer rental markets. Working paper, Carnegie
Mellon University.

Agarwal R, Ergun Ö (2010) Network design and allocation mechanisms for carrier alliances in
liner shipping. Oper. Res. 58(6):1726–1742.

28
Allen J, Browne M, Woodburn A, Leonardi J (2012) The role of urban consolidation centres in
sustainable freight transport. Transport Rev. 32(4):473–490.

Ambrosini C, Routhier J (2004) Objectives, methods and results of surveys carried out in the
field of urban freight transport: An international comparison. Transport Rev. 24(1):57–77.

ASEAN (2014) Logistics Planning and Analysis.

Bai J, So KC, Tang CS, Chen X, Wang H (2018) Coordinating supply and demand on an
on-demand service platform with impatient customers. Manufacturing Service Oper. Man-
agement, forthcoming.

Benjaafar S, Bernhard H, Courcoubetis C (2017) Drivers, riders and service providers: The
impact of the sharing economy on mobility. Proceedings of the 12th workshop on the economics
of networks, systems and computation, 27 June 2017, pp.1-6.

Benjaafar S, Kong G, Li X, Courcoubetis C (2018) Peer-to-peer product sharing: Implications

for ownership, usage and social welfare in the sharing economy. Management Sci. 65(2):477–
493.

Bimpikis K, Candogan O, Saban D (2016) Spatial pricing in ride-sharing networks. Working

paper, Stanford University.

Boudoin D, Morel C, Gardat M (2014) Supply chains and urban logistics platforms, in Gonzalez-
Feliu J, Semet F, Routhier J-L (eds), Sustainable Urban Logistics: Concepts, Methods and
Information Systems, EcoProduction, Springer Berlin Heidelberg, pp. 1–20.

Broner F, Martin A, Ventura J (2010) Sovereign risk and secondary markets. Amer. Econom.
Rev. 100(4):1523–1555.

Browne M, Sweet M, Woodburn A, Allen J (2005) Urban Freight Consolidation Centres Final
Report. Transport Studies Group, University of Westminster, London.

Cachon GP, Swinney R (2009) Purchasing, pricing, and quick response in the presence of strate-
gic consumers. Management Sci. 55(3):497–511.

Cachon GP, Daniels KM, Lobel R (2017) The role of surge pricing on a service platform with
self-scheduling capacity. Manufacturing Service Oper. Management 19(3):368–384.

Chen J, Esteban S, Shum M (2013) When do secondary markets harm firms? Amer. Econom.
Rev. 103(7):2911–2934.

Choe T, Scott A, Mauricio G, Jon W (2017) The future of freight: How new technology and
new thinking can transform how goods are moved. Deloitte Insights.

Cohen M, Zhang R (2017) Coopetition and profit sharing for ride-sharing platforms. Working
paper, New York University.

Crainic R, Ricciardi N, Storchi G (2009) Models for evaluating and planning city logistics
systems. Transportation Sci. 43(4):432–454.

Dablanc L (2005) French strategic approach to urban consolidation. Bestufs II First Workshop
13th–14th January 2005, London.

Dablanc L (2007) Goods transport in large European cities: Difficult to organize, difficult to
modernize. Transportation Res. Part A 41(3):280–285.

Dablanc L (2011) City logistic best practices: A Handbook for Authorities, SUGAR, Bologna,
Italy.

Doig JW (2001) Empire on the Hudson: Entrepreneurial Vision and Political Power at the Port
of New York Authority. Columbia University Press, New York.

29
Dong L, Rudi N (2004) Who benefits from transshipment? Exogenous vs. endogenous wholesale
price. Management Sci. 50:645–657.

Fraiberger S, Sundararajan A (2015) Peer-to-peer rental markets in the sharing economy. Work-
ing paper, New York University.

Gesing B (2017) Transport capacity sharing, in Sharing Economy Logistics: Rethinking Logistics
with Access over Ownership, DHL Trend Research, pp. 21–22.

Hagiu A, Wright J (2015) Marketplace or reseller? Management Sci. 61(1):184–203.

Handoko D, Lau HC, Cheng SF (2016) Achieving economic and environmental sustainability
in urban consolidation center with bi-criteria auction. IEEE Trans. Automation Science and
Engineering 13(4):1471–1479.

Holguin-Veras J, Silas M, Polimeni J (2008) An investigation into the attitudinal factors deter-
mining participation in cooperative multi-carrier delivery initiatives, in Taniguchi E, Thomson
R (eds), Innovations in City Logistics IV, Nova Science Publishers, pp. 55–68.

Hu M, Zhou Y (2017) Price, wage and fixed commission in on-demand matching. Working
paper, University of Toronto.

Jiang B, Tian L (2016) Collaborative consumption: Strategic and economic implications of

product sharing. Management Sci. 64(3):1171–1188.

Lee H, Whang S (2001) Winning the last mile of e-commerce. MIT Sloan Management Rev.
42(4):54–62.

Lee H, Whang S (2002) The impact of the secondary market on the supply chain. Management
Sci. 48(6):719–731.

Lopez E (2017) Why is the last mile so inefficient? Supply Chain Dive.

McDermott DR (1975) An alternative framework for urban goods distribution: Consolidation.

Transportation J. 15(1):29–39.

Mendelson H, Tunc T (2007) Strategic spot trading in supply chains. Management Sci.
53(5):742–759.

Milner J, Kouvelis P(2007) Inventory, speculation, and sourcing strategies in the presence of
online exchanges. Manufacturing Service Oper. Management 9(3):312–331.

National Express (2010) National Express International Ltd case. http://www.ne56.com/

experiential/ne56109111369BB3F5.html.

Özener OÖ, Ergun Ö (2008) Allocating costs in a collaborative transportation procurement

network. Transportation Sci. 42(2):146–165.

Quak H, Tavasszy L (2011) Customized solutions for sustainable city logistics: The viability
of urban freight consolidation centres, in van Nunen JAEE, Huijbregts P, Rietveld P (eds),
Transitions Towards Sustainable Mobility, Springer Berlin Heidelberg.

Ranieri L, Digiesi S, Silvestri B, Roccotelli M (2018) A review of last mile logistics innovations
in an externalities cost reduction vision. Sustainability 10(3):782.

Rochet JC, Tirole J (2006) Two-sided markets: A progress report. The RAND J. Econom.
37(3):645–667.

Rudi N, Kapur S, Pyke D (2001) A two-location inventory model with transshipment and local
decision making. Management Sci. 47:1668–1680.

30
Snapp S (2012) The overestimation of the benefits of 3PL versus private fleets. Brightwork
research.

Steinerberger S (2015) New bounds for the traveling salesman constant. Advances Appl. Prob-
ability 47(1).

Su X, Zhang F (2008) Strategic customer behavior, commitment, and supply chain performance.
Management Sci. 54(10):1759–1773.

Taylor T (2018) On-demand service platforms. Manufacturing Service Oper. Management,

20(4):704–720.

Tian L, Jiang B (2018) Effects of consumer-to-consumer product sharing on distribution channel.

Production Oper. Management 27(2):350–367.

van Duin R, Quak H, Munuzuri J (2008) Revival of the cost benefit analysis for evaluating the
city distribution center concept? In Taniguchi E, Thompson RG (eds), Innovations in City
Logistics, Nova Publishers, New York, pp. 97–114.

Vivaldini M, Pires S, de Souza F (2012) Improving logistics services through the technology
used in fleet management. J. Information Systems and Technology Management: JISTEM
9(3): 541–562.

Wang C, Lim YF, Lau HC (2018) An auction mechanism for last-mile delivery with cost uncer-
tainty. Working paper, Singapore Management University.

Wang C, Lau HC, Lim YF (2015) A rolling horizon auction mechanism and virtual pricing of
shipping capacity for urban consolidation centers, in Corman F, VoßS, Negenborn RR (eds),
Computational Logistics, Springer International Publishing Switzerland, pp. 422-436.

Wang Y, Zhang D, Liu Q, Shen F, Lee LH (2016) Towards enhancing the last-mile delivery: An
effective crowd-tasking model with scalable solutions. Transportation Res. Part E 93:279–293.

Weyl EG (2010) A price theory of multi-sided platforms. Amer. Econom. Rev. 100(4):1642–
1672.

31
 Online Supplement

A Proofs
Lemma 1. (Optimal decision of carrier i) Carrier i uses the UCC’s service (d¯∗i = 1) if p̄ ≤ m+c/vi ,
or delivers on his own (d¯∗i = 0) otherwise.
Proof of Lemma 1. By solving φ̄i (1; p̄) ≤ φ̄i (0; p̄) for vi , we obtain p̄ ≤ m + c/vi . Thus, d¯∗i = 1 if
p̄ ≤ m + c/vi , and d¯∗i = 0 otherwise.

h   i
v
c λ 1− v L −(1−λ) +(1−λ)(M −S)vH
H
Lemma 2. (Optimal decision of the UCC) If m > (1−λ)vH , then the UCC’s
optimal price is p̄∗ = m + c/vH . Otherwise, p̄∗ = m + c/vL .
Proof of Lemma 2. According to Lemma 1, we derive V (which is in the UCC’s expected profit function
in Equation (1)) for the two types of carriers as below.
Low type (vi = vL ): Each carrier i of this type uses the UCC’s service (d¯∗i = 1) if p̄ ≤ m + c/vL .
The expected number of carriers of this type is λn, and if those carriers use the UCC’s service, then the
expected task volumes served by the UCC are λnvL .
High type (vi = vH ): Each carrier i of this type uses the UCC’s service (d¯∗i = 1) if p̄ ≤ m + c/vH .
The expected number of carriers of this type is (1 − λ)n, and if those carriers use the UCC’s service,
then the expected task volumes served by the UCC are (1 − λ)nvH .
Note that m + c/vL > m + c/vH . So, the optimal choice of the UCC is among the following two:
1. Choose a price p̄ ∈ (m + c/vH , m + c/vL ] to attract Type 1 carriers, then V equals to the total
expected task volumes of those carriers, that is V = λnvL . Substituting V into Equation (1), the UCC’s
expected profit is
π̄(p̄) = (p̄ + S − M )(λnvL ) − C, (3)
∗
which increases in p̄, so it is optimal for the
 UCC to
 choose p̄ = m+c/v Lto maximize profit. Substituting
p̄∗ into Equation (3), we obtain that π̄ m + c
vL = m+ c
vL + S − M λnvL − C.
2. Choose a price p̄ ∈ (0, m + c/vH ] to attract both Type 1 and Type 2 carriers, then V equals to
the total expected task volumes of all the carriers, that is V = λnvL + (1 − λ)nvH . Substituting V into
Equation (1), the UCC’s expected profit is
π̄(p̄) = (p̄ + S − M )(λnvL + (1 − λ)nvH ) − C, (4)
∗
which increases in p̄, so it is optimal for the UCC to choose
  p̄ = m+c/v H to
 maximize profit. Substituting
p̄∗ into Equation (4), we obtain that π̄ m + c
vH = m+ c
+ S − M (λnvL + (1 − λ)nvH ) − C.
vH
 
By comparing the profits of the UCC under choices 1 and 2, we can obtain that π̄ m + vcL ≥
v
c[λ(1− v L )−(1−λ)]+(1−λ)(M −S)vH
 
π̄ m + vcH if m ≤ H
(1−λ)vH . Thus, the results in Lemma 2 follow.

Proof of Theorem 1. According to lemmas 1 and 2, the results in Theorem 1 follows.

Lemma 3. (Optimal decision of low type carrier i) Low type carrier i purchases capacity from the
c+(1+θ)mvL
platform to fulfil his task (dˆ∗i = 1) if p̂ < [1+θ(1−α)]v L
, or delivers on his own and sells his remaining
ˆ∗ c+(1+θ)mvL
capacity to platform (d = 0) if p̂ >
i [1+θ(1−α)]vL, or chooses either option with probability 0.5 if p̂ =
c+(1+θ)mvL
[1+θ(1−α)]vL .

c+(1+θ)mvL
Proof of Lemma 3. By solving φ̂i (1; p̂) < φ̂i (0; p̂) for vi , we obtain that p̂ < [1+θ(1−α)]v L
. Thus, dˆ∗i = 1
c+(1+θ)mvL c+(1+θ)mv
if p̂ < [1+θ(1−α)]vL , dˆ∗i = 0 if p̂ > [1+θ(1−α)]vLL , and dˆ∗i = 1 or 0 with an equal probability if p̂ =
c+(1+θ)mvL
[1+θ(1−α)]vL .

Lemma 4. (Optimal decision of the platform) The optimal price of the capacity sharing platform
c+2mvL
is p̂∗ = (2−α)v L
.
Proof of Lemma 4. According to Lemma 3, the optimal choice of the platform is choosing a price p̂ =
c+(1+θ)mvL
[1+θ(1−α)]vL to incentivize the low type carriers to purchase capacity from the platform or deliver on
their own and sell remaining capacity to the platform with an equal probability. Otherwise, there will
be either no supply or no demand on the platform. We can obtain that ns = λn/2 , np = λn/2, and
c+(1+θ)mvL c+2mvL
θ = 1. Substituting them into p̂ = [1+θ(1−α)]v L
and Equation (2), we can obtain p̂∗ = (2−α)v L
and
αλn(c+2mvL )
π̂ ∗ = 2(2−α) − λnA
2 .

32
Proof of Theorem 2. According to lemmas 3 and 4, the results in Theorem 2 follows.

Proof of Theorem 3. According to propositions 1 and 3, by solving π̄ ∗ (p̄∗ ) > π̂ ∗ (p̂∗ ) for m in the cases
m > m̄ and m ≤ m̄, we obtain the results in Theorem 3.

Proof of Corollary 1. Given the market entry cost Ē, the UCC’s profit in equilibrium becomes π̄ ∗ (p̄∗ ) =
(m + vcH + S − M )(λnvL + (1 − λ)nvH ) − C − Ē if m > m̄, and π̄ ∗ (p̄∗ ) = (m + vcL + S − M )λnvL − C − Ē
if m ≤ m̄. Given the market entry cost Ê, the platform’s profit in equilibrium becomes π̂ ∗ (p̂∗ ) =
αλn(c+2mvL ) ∗ ∗ ∗ ∗
2(2−α) − λnA
2 − Ê. By solving π̄ (p̄ ) > π̂ (p̂ ) for m in the cases m > m̄ and m ≤ m̄, we obtain
the results in Corollary 1.

Proof of Theorem 4. According to propositions 2 and 4, by solving ∆ ¯ψ > ∆ ˆ ψ for n in the cases m > m̄
and m ≤ m̄, we obtain the results in Theorem 4.
n o
ñ
Proof of Theorem 5. Define θ = min ñps , 1 , where ñp and ñs are the rational beliefs about the number
of carriers who purchase capacity from the platform and who sell capacity on the platform, respectively.
Similar to the proof of Lemma 1, we can derive the optimal decision of each carrier as follows.
1. Each high type carrier uses the UCC’s service if p̄ ≤ m + vcH , or delivers on his own if p̄ > m + vcH .
2. Each low type carrier uses the UCC’s service if p̄ ≤ m + vcL − θ[(1 − α)p̂ − m] and p̂ ≥ p̄,
c+(1+θ)mvL
or purchases capacity from the platform if p̄ > p̂ and p̂ < [1+θ(1−α)]vL , or delivers on his own if p̄ >
c c+(1+θ)mvL
m+ vL − θ[(1 − α)p̂ − m] and p̂ > [1+θ(1−α)]vL . Note that the carrier is indifferent between purchasing
c+(1+θ)mvL
capacity from the platform and delivering on his own if p̂ = [1+θ(1−α)]v L
.
1
According to the assumption (1−α)p̂−m < ( vL − vH )c, we can obtain that m+ vcL −θ[(1−α)p̂−m] >
1

m+ vcH . Similar to the proof of Lemma 2, we can obtain that the optimal pricing choice of the consolidator
is among the following:
c+2mvL c+2mvL
1. Choose p̄∗ > (2−α)v L
and p̂∗ = (2−α)v L
. Under these prices, each high type carrier delivers on
his own, and each low type carrier is indifferent between delivering on his own (and selling his remaining
capacity to the platform) and purchasing capacity on the platform. The consolidator’s expected profit
is αλn(c+2mv
2(2−α)
L)
− λnA.
2. Choose p̄∗ = m+ vcL and p̂∗ ≥ m+ vcL . Under these prices, each high type carrier delivers on his own,
 
and each low type carrier uses the UCC’s service. The consolidator’s profit is m + vcL + S − M λnvL −
C.
3. Choose p̄∗ = m + vcH and p̂∗ ≥ m + vcL . Under these prices, each carrier uses the UCC’s service.
 
The consolidator’s profit is m + vcH + S − M (λnvL + (1 − λ)nvH ) − C.
It is optimal for the consolidator to choose the price that leads to a largest profit. Comparing the
consolidator’s profit under the above three choices, we can obtain the results in Theorem 5.

Proof of Theorem 6. The proof is similar to the proof of Theorem 2 and thus omitted.

φ̄i2 d¯i2 ; d¯i1 , p̄2 as the cost of carrier i in period 2. Each

Under the two-period UCC model, we define
carrier i minimizes his cost φ̄i2 d¯i2 ; d¯i1 , p̄2 by comparing the following
two options: (i) d¯i2 = −1:
Carrier

i delivers on his own in period 2, which incurs a cost φ̄i2 −1; d¯i1 , p̄2 = c + mvi2 − d
¯i1 − 1 f . (ii)
d¯i2 = 0: Carrier i uses the UCC’s service in period 2, which incurs a cost φ̄i2 0; d¯i1 , p̄2 = p̄2 vi2 . The
following lemma shows the optimal decision of each carrier i in period 2.

Lemma 5. (Optimal decision of carrier i in period 2 under the UCC model)

1. If carrier i delivers on his own or uses the UCC’s service and keeps his logistics capability in period 1
(d¯i1 = −1 or 1), then in period 2, carrier i uses the UCC’s service and eliminates his logistics capability
(d¯∗i2 = 0) if p̄2 ≤ m + c/vi2 , or delivers on his own (d¯∗i2 = −1) otherwise.
2. If carrier i uses the UCC’s service and eliminates his logistics capability in period 1 (d¯i1 = 0), then
in period 2, carrier i uses the UCC’s service (d¯∗i2 = 0) if p̄2 ≤ m + (c + f )/vi2 , or delivers on his own
(d¯∗i2 = −1) otherwise.

Proof of Lemma 5. By solving φ̄i2 (0; d¯i1 , p̄2 ) ≤ φ̄i2 (−1; d¯i1 , p̄2 ) for vi2 , we obtain that
1. p̄2 ≤ m + vci2 if d¯i1 = −1 or 1. Thus, d¯∗i2 = 0 if p̄2 ≤ m + vci2 , and d¯∗i2 = −1 otherwise.
2. p̄2 ≤ m + c+f ¯ ¯∗ c+f ¯∗
vi2 if di1 = 0. Thus, di2 = 0 if p̄2 ≤ m + vi2 , and di2 = −1 otherwise.

33
Let V2 denote the expected total task volume of the carriers who use the UCC’s service in period 2. Given
the carriers’ optimal responses in Lemma 5, the UCC chooses the price p̄2 to maximize her expected
profit in period 2:
π̄2 (p̄2 ) = (p̄2 + S − M ) V2 − C. (5)
Define ne as the number of carriers who use the UCC’s service and eliminate their logistics capability in
period 1 (that is, the carriers with d¯i1 = 0). Note that ne is known in period 2. The following lemma
shows the UCC’s optimal pricing decision in period 2.

Lemma 6. (Optimal decision of the UCC in period 2 under the UCC model)
1. If ne > 0, the optimal price of the UCC’s service in period 2 is

 m + (c + f )/vL , if m ≤ min{m1 , m2 , m3 };

∗ m + (c + f )/vH , if m1 < m ≤ min{m4 , m5 };
p̄2 = if max{m2 , m4 } < m ≤ m6 ;
 m + c/vL ,

m + c/vH , if m > max{m3 , m5 , m6 }.
2. If ne = 0, the optimal price of the UCC’s service in period 2 is

m + c/vL , if m ≤ m̄;
p̄∗2 =
m + c/vH , if m > m̄.

h   i  
v v
(c+f ) λ 1− v L −(1−λ) λne f − (1−λ)ne vH +λ(n−ne ) c
Proof of Lemma 6. Define m1 = M −S+ H
(1−λ)vH , m2 = M −S+ L
,
    (1−λ)ne vH +λ(n−n
 e )vL
v vL v
λ(c+f )ne − λ v L +(1−λ) cn λ v +(1−λ) (c+f )ne − λn+(1−λ)ne vH c
m3 = M − S + (1−λ)nv
H
H +λ(n−n e )vL
, m4 = M −S+ H L
, m5 =
      λ(n−ne )vL 
v v vL vH
λ vL +(1−λ) (c+f )ne − λ v L +(1−λ) cn λn 1− v +(1−λ)ne v −(1−λ)n c
M −S+ H H
λ(n−ne )vL +(1−λ)(n−ne )vH , m6 = M − S + H L
(1−λ)(n−ne )vH .
To derive V2 in the UCC’s expected profit function in Equation (5), we need to distinguish the
following four types of carriers:
Type 1 (d¯i1 = −1 or 1, and vi2 = vL ): Each carrier i of this type uses the UCC’s service and
eliminates his logistics capability in period 2 (d¯∗i2 = 0) if p̄2 ≤ m + vcL . The expected number of carriers
of this type is λ(n − ne ), and if those carriers use the UCC’s service in period 2, then the expected task
volumes served by the UCC are λ(n − ne )vL .
Type 2 (d¯i1 = −1 or 1, and vi2 = vH ): Each carrier i of this type uses the UCC’s service and
eliminates his logistics capability in period 2 (d¯∗i2 = 0) if p̄2 ≤ m + vcH . The expected number of carriers
of this type is (1 − λ)(n − ne ), and if those carriers use the UCC’s service in period 2, then the expected
task volumes served by the UCC in period 2 are (1 − λ)(n − ne )vH .
Type 3 (d¯i1 = 0, and vi2 = vL ): Each carrier i of this type uses the UCC’s service in period 2
(d¯∗i2 = 0) if p̄2 ≤ m + c+f vL . The expected number of carriers of this type is λne , and if those carriers use
the UCC’s service in period 2, then the expected task volumes served by the UCC in period 2 are λne vL .
Type 4 (d¯i1 = 0, and vi2 = vH ): Each carrier i of this type uses the UCC’s service in period 2
(d¯∗i2 = 0) if p̄2 ≤ m + c+fvH . The expected number of carriers of this type is (1 − λ)ne , and if those carriers
use the UCC’s service in period 2, then the expected task volumes served by the UCC in period 2 are
(1 − λ)ne vH .
Note that the expected number of Type 3 and Type 4 carriers will be 0 if ne = 0, which will
be identical to the proof of Lemma 2. Thus, we omit the analysis for the case that ne = 0. In the
following, we analyze the UCC’s optimal decision in the case that ne > 0. According to the assumption
f > c(vHv−vL
L)
, we can derive m + c+f c+f c c
vL > m + vH > m + vL > m + vH , so the optimal choice of the UCC
is among the following four:  i
1. Choose a price p̄2 ∈ m + c+f vH , m + c+f
vL to attract type 3 carriers only, then V2 = λne vL .
Substituting it into Equation (5), the UCC’s expected profit is
π̄2 (p̄2 ) = (p̄2 + S − M )λne vL − C, (6)
which increases in p̄2 , so it is optimal for the UCC to choose p̄∗2 = m+ c+f to maximize profit. Substituting
   vL 
p̄∗2 = m + c+f
vL into Equation (6), we obtain that π̄ 2 m + c+f
vL = m + c+f
vL + S − M λne vL − C.
 i
2. Choose a price p̄2 ∈ m + vcL , m + c+f vH to attract type 3 and type 4 carriers, then V2 = λne vL +
(1 − λ)ne vH . Substituting it into Equation (5), the UCC’s expected profit is
π̄2 (p̄2 ) = (p̄2 + S − M )(λne vL + (1 − λ)ne vH ) − C, (7)
which increases in p̄2 , so it is optimal for the UCC to choose p̄∗2 = m+ c+f vH to maximize profit. Substituting
   
p̄∗2 = m + c+f
vH into Equation (7), we obtain that π̄ 2 m + c+f
vH = m + c+f
vH + S − M (λne vL + (1 −
λ)ne vH ) − C.

34
  i
c c
3. Choose a price p̄2 ∈ m+ vH , m + vL to attract type 3, type 4, and type 1 carriers, then
V2 = λne vL + (1 − λ)ne vH + λ(n − ne )vL . Substituting it into Equation (5), the UCC’s expected profit is
π̄2 (p̄2 ) = (p̄2 + S − M )(λne vL + (1 − λ)ne vH + λ(n − ne )vL ) − C, (8)
which increases in p̄2 , so it is optimal for the UCC to choose p̄∗2 = m+ vcL to maximize profit. Substituting
   
p̄∗2 = m+ vcL into Equation (8), we obtain that π̄2 m + vcL = m + vcL + S − M (λnvL +(1−λ)ne vH )−
C.  i
4. Choose a price p̄2 ∈ 0, m + vcH to attract all types of carriers, then V2 = λne vL + (1 − λ)ne vH +
λ(n − ne )vL + (1 − λ)(n − ne )vH = λnvL + (1 − λ)nvH . Substituting it into Equation (5), the UCC’s
expected profit is
π̄2 (p̄2 ) = (p̄2 + S − M )(λnvL + (1 − λ)nvH ) − C, (9)
which increases in p̄2 , so it is optimal for the UCC to choose p̄∗2 = m+ vcH to maximize profit. Substituting
   
p̄∗2 = m+ vcH into Equation (9), we obtain that π̄2 m + vcH = m + vcH + S − M (λnvL +(1−λ)nvH )−
C.  
By comparing the profits of the UCC under choices 1, 2, 3, and 4, we can obtain that π̄2 m + c+f vL
 
is the maximum if m ≤ min{m1 , m2 , m3 }; π̄2 m + c+f vH is the maximum if m1 < m ≤ min{m4 , m5 };
   
π̄2 m + vcL is the maximum if max{m2 , m4 } < m ≤ m6 ; and π̄2 m + vcH is the maximum if m >
max{m3 , m5 , m6 }. Therefore, the corresponding prices p̄∗2 under those choices are optimal for the UCC,
and the results in Lemma 6 follow.

To identify a subgame perfect Nash Equilibrium with rational expectations, we assume all the carriers
and the UCC form the same rational belief ñe about
ne when they optimize their decisions in period

1, and in equilibrium, ñe = ne p̄∗1 ; d¯∗i1 , i = 1, 2, . . . , n . Given p̄1 , each carrier i minimizes Φ̄i d¯i1 ; p̄1
by choosing one of the following options: (i) d¯i1 = −1: Carrier i delivers on his own, which incurs
an expected total discounted cost Φ̄i (−1; p̄1 ) = c + mvi1 + δ φ̄∗i2 (−1). (ii) d¯i1 = 0: Carrier i uses the
UCC’s service and eliminates his logistics capability, which incurs an expected total discounted cost
Φ̄i (0; p̄1 ) = p̄1 vi1 + δ φ̄∗i2 (0). (iii) d¯i1 = 1: Carrier i uses the UCC’s service and keeps his logistics
capability, which incurs an expected total discounted cost Φ̄i (1; p̄1 ) = p̄1 vi1 + h + δ φ̄∗i2 (1). The following
lemma shows the optimal decision of carrier i in period 1.
Lemma 7. (Optimal decision of carrier i in period 1 under the UCC model)
1. If ñe > 0, the optimal decision of carrier i is determined as follows.
(a) If m ≤ min {m1 , m̃2 , m̃3 }, then

¯∗ 1, if p̄1 ≤ m + (c − h)/vi1 ;
di1 = −1, otherwise.
(b) If m1 < m ≤ min {m̃4 , m̃5 }, then

 1,
 if p̄1 ≤ m + (c − h)/vi1 and h ≤ δ(c + f )(λvL /vH + 1 − λ) − δc;
0, if p̄1 ≤ m + (1 + δ)c/vi1 − δ(c + f )(λvL /vH + 1 − λ)/vi1
d¯i1 =
∗
 and h > δ(c + f )(λvL /vH + 1 − λ) − δc;
−1, otherwise.

(c) If max {m̃2 , m̃4 } < m ≤ m̃6 , then

 1,
 if p̄1 ≤ m + (c − h)/vi1 and h ≤ δc(λ + (1 − λ)vH /vL ) − δc;
¯∗ 0, if p̄1 ≤ m + (1 + δ)c/vi1 − δc(λ + (1 − λ)vH /vL )/vi1
di1 =
 and h > δc(λ + (1 − λ)vH /vL ) − δc;
−1, otherwise.

(d) If m > max {m̃3 , m̃5 , m̃6 }, then

0, if p̄1 ≤ m + c/vi1 ;
d¯∗i1 = −1, otherwise.
2. If ñe = 0, the optimal decision of carrier i is determined as follows.
(a) If m ≤ m̄, then

 1,
 if p̄1 ≤ m + (c − h)/vi1 and h ≤ δc(λ + (1 − λ)vH /vL ) − δc;
0, if p̄1 ≤ m + (1 + δ)c/vi1 − δc(λ + (1 − λ)vH /vL )/vi1
d¯i1 =
∗
 and h > δc(λ + (1 − λ)vH /vL ) − δc;
−1, otherwise.

(b) If m > m̄, then

¯∗ 0, if p̄1 ≤ m + c/vi1 ;
di1 = −1, otherwise.

35
  
v
λñe f − (1−λ)ñe vH +λ(n−ñe ) c
Proof of Lemma 7. Define m̃2 = M − S + L
(1−λ)ñe vH +λ(n−ñ e )vL
,
     
v v v
λ(c+f )ñe − λ v L +(1−λ) cn λ vL +(1−λ) (c+f )ñe − λn+(1−λ)ñe vH c
m̃3 = M − S + H
, m̃4 = M − S+ H L
,
 (1−λ)nvH +λ(n−ñ
 e )vL
   λ(n−ñ
 e )v
L 
vL vL vL vH
λ v +(1−λ) (c+f )ñe − λ v +(1−λ) cn λn 1− v +(1−λ)ñe v −(1−λ)n c
m̃5 = M − S + H H
λ(n−ñe )vL +(1−λ)(n−ñe )vH , m̃6 = M − S + H
(1−λ)(n−ñe )vH
L
.
We first determine a carrier’s optimal decision when ñe > 0. Note that ñe is rational and hence is
equal, in equilibrium, to the corresponding actual value ne . Thus, according to case 1(a) of Lemma 6, if
m ≤ min {m1 , m̃2 , m̃3 }, then p̄∗2 = m + c+f ∗
vL . Given p̄2 , each carrier i minimizes his total discounted cost
Φ̄i (d¯i1 ; p̄1 ) by comparing the following 3 options:
1. d¯i1 = −1: In this case, according to Lemmas 5 and 6, carrier i will deliver on his own in period
2. This incurs an expected cost Φ̄i (−1; p̄1 ) = c + mvi1 + δ(λ(c + mvL ) + (1 − λ)(c + mvH )).
2. d¯i1 = 0: In this case, according to Lemmas 5 and 6, carrier i will use the UCC’s service in period 2
if vi2 = vL and deliver on his own otherwise. This incurs an expected cost Φ̄i (0; p̄1 ) = p̄1 vi1 + δ(λp̄∗2 vL +
(1 − λ)(c + mvH + f )) = p̄1 vi1 + δ(λ(c + mvL ) + (1 − λ)(c + mvH ) + f ).
3. d¯i1 = 1: In this case, according to Lemmas 5 and 6, carrier i will deliver on his own in period 2.
This incurs an expected cost Φ̄i (1; p̄1 ) = p̄1 vi1 + h + δ(λ(c + mvL ) + (1 − λ)(c + mvH )).
By comparing the above three options, we obtain that d¯∗i1 = 1 if p̄1 ≤ m + c−h ¯∗
vi1 , and di1 = −1
otherwise. This proves case 1(a) of Lemma 7. Next we determine the carrier’s optimal decision in
case 1(b) of Lemma 7. Similarly, according to case 1 of Lemma 6, if m1 < m ≤ min {m̃4 , m̃5 }, then
p̄∗2 = m + c+f ∗ ¯
vH . Given p̄2 , each carrier i minimizes his total discounted cost Φ̄i (di1 ; p̄1 ) by comparing the
following three options:
1. d¯i1 = −1: In this case, according to Lemmas 5 and 6, carrier i will deliver on his own in period
2. This incurs an expected cost Φ̄i (−1; p̄1 ) = c + mvi1 + δ(λ(c + mvL ) + (1 − λ)(c + mvH )).
2. d¯i1 = 0: In this case, according to Lemmas 5 and 6, carrier i will use the UCC’s  service in
 period
∗ ∗ c+f
2. This incurs an expected cost Φ̄i (0; p̄1 ) = p̄1 vi1 + δ(λp̄2 vL + (1 − λ)p̄2 vH ) = p̄1 vi1 + δ m + vH (λvL +
(1 − λ)vH ).
3. d¯i1 = 1: In this case, according to Lemmas 5 and 6, carrier i will deliver on his own in period 2.
This incurs an expected cost Φ̄i (1; p̄1 ) = p̄1 vi1 + h + δ(λ(c + mvL ) + (1 − λ)(c + mvH )).
By comparing the above three options, we obtain that d¯∗i1 = 1 if p̄1 ≤ m + c−h vi1 and h ≤ δ(c +
 
vL
  δ(c+f ) λ +1−λ  
f ) λ vvHL + 1 − λ −δc, d¯∗i1 = 0 if p̄1 ≤ m+ (1+δ)c vL
vH
vi1 − vi1 and h > δ(c+f ) λ vH + 1 − λ −δc,
¯∗
and di1 = −1 otherwise. This proves case 1(b) of Lemma 7. The proofs of cases 1(c) and 2(a) are similar
to the proof of case 1(b), and the proofs of cases 1(d) and 2(b) are similar to the proof of case 1(a), and
thus omitted.

Let V1 denote the expected total task volume of the carriers who use the UCC’s service in period 1.
Assuming all the carriers respond optimally according to Lemma 7, the UCC optimizes her price p̄1 to
maximize her expected total discounted profit over the two periods:
Π̄ (p̄1 ) = (p̄1 + S − M ) V1 − C + δπ̄2 (p̄∗2 (p̄1 )) . (10)

Proof of Theorem 7. The UCC’s expected profit Π̄(p̄1 ) in Equation (10) depends on V1 . The different
cases in Lemma 7 corresponding to different decisions of each carrier will lead to different values of V1 .
In the following, we analyze each case of Lemma 7 to derive V1 and obtain the UCC’s expected total
discounted profit and then determine the equilibrium price. To derive V1 , we need to distinguish the
following two types of carriers:
Type 1 (vi1 = vL ): The expected number of carriers of this type is λn.
Type 2 (vi1 = vH ): The expected number of carriers of this type is (1 − λ)n.
We first analyze the cases that ñe > 0 of Lemma 7, that is cases 1(a), 1(b), 1(c), and 1(d). Note that ñe is
rational and hence equal to to the corresponding actual value in equilibrium, and thus ñe = ne (p̄∗1 , p̄∗2 ) > 0,
m̃2 = m2 , m̃3 = m3 , m̃4 = m4 , m̃5 = m5 , and m̃6 = m6 .
In case 1(a) (ñe > 0 and m ≤ min {m1 , m̃2 , m̃3 }), d¯∗i1 = 1 if p̄1 ≤ m + c−h ¯∗
vi1 , or di1 = −1 otherwise.
Thus, type 1 carriers use the UCC’s service and keep their logistics capability if p̄1 ≤ m + c−h vL , and
c−h
type 2 carriers use the UCC’s service and keep their logistics capability if p̄1 ≤ m + vH . In this case,
no carrier will use the UCC’s service and eliminate logistics capability, which means ne = 0, and thus
cannot happen in equilibrium.
In cases 1(b) (ñe > 0 and mn1 < m ≤ min
 {m̃4 , m̃5 })and 1(c) (ñe > 0 and max  {m̃2 ,om̃4 } < m ≤ m̃6 ),
since we assume that h ≤ min δ(c + f ) λ vL
+ 1 − λ − δc, δc λ + (1 − λ)
vH
vH
− δc , thus d¯∗ = 1 if
vL i1
p̄1 ≤ m + c−h
vi1 , or d¯∗i1 = −1 otherwise. Similar to the above case 1(a), these cases will never happen in
equilibrium.

36
 In case 1(d) (ñe > 0 and m ≥ max {m̃3 , m̃5 , m̃6 }), d¯∗i1 = 0 if p̄1 ≤ m + vci1 , or d¯∗i1 = −1 otherwise.
Thus, type 1 carriers use the UCC’s service and eliminate their logistics capability if p̄1 ≤ m + vcL , and
type 2 carriers use the UCC’s service and eliminate their logistics capability if p̄1 ≤ m + vcH . In case
1(d), we have obtained that p̄∗2 = m + vcH according to Lemma 6. The optimal choice of the UCC in
period 1 is among the following
 two: i
1. Choose a price p̄1 ∈ m + vcH , m + vcL to attract type 1 carriers only, then V1 = λnvL . Since
type 1 carriers use the UCC’s service and eliminate their logistics capability, thus ne = λn. Substituting
them into Equation (10), the UCC’s expeted total discounted profit is
 
Π̄(p̄1 ) = (p̄1 + S − M )λnvL − C + δπ̄2 m + vcH
h  i (11)
= (p̄1 + S − M )λnvL − C + δ m + vcH + S − M (λnvL + (1 − λ)nvH ) − C ,
which increases in p̄1 , so it is optimal for the UCC to choose p̄∗1 = m + vcL to maximize profit. This could
be in equilirbium only if ne > 0 (which is satisfied) and m > max{m3 , m5 , m6 }. Substituting ne = λn
into m3 , m5 , and m6 , we can rewrite the latter conition as m > max{m̂1 , m̂2 , m̂3 }. This leads to the
results in case 1 of Theorem 7. i
2. Choose a price p̄1 ∈ 0, m + vcH to attract both types of carriers, then V1 = λnvL + (1 −
λ)nvH . Since all the carriers use the UCC’s service and eliminate their logistics capability, thus ne = n.
Substituting them into Equation (10), the UCC’s expeted total discounted profit is
 
Π̄(p̄1 ) = (p̄1 + S − M )(λnvL + (1 − λ)nvH ) − C + δπ̄2 m + vcH
= S − M )(λnvL + (1 − λ)nvH ) − C
(p̄1 +h i (12)
c
+δ m+ vH + S − M (λnvL + (1 − λ)nvH ) − C ,
which increases in p̄1 , so it is optimal for the UCC to choose p̄∗1 = m + vcH to maximize profit. This could
be in equilirbium only if ne > 0 (which is satisfied) and m > max{m3 , m5 , m6 }. Substituting ne = n into
m3 , m5 , and m6 , we find that the latter condition can never be satisfied as m5 and m6 go to infinity.
Thus, this choice is not in equilibrium.
Next we analyze the cases that ñe = 0 of Lemma 7, that is cases 2(a) and 2(b). Similarly, since ñe is
rational and hence equal to the corresponding actual value in equilibrium, and n thus ne(p̄∗1 , p̄∗2 ) = ñe =
 0.
In case 2(a) (ñe = 0 and m ≤ m̄, since we assume that h ≤ min δ(c + f ) λ vvHL + 1 − λ −
  o
δc, δc λ + (1 − λ) vvHL − δc , and thus we can obtain d¯∗i1 = 1 if p̄1 ≤ m + c−h ¯∗
vi1 , or di1 = −1 other-
wise. Therefore, type 1 carriers use the UCC’s service and keep their logistics capability if p̄1 ≤ m + c−h
vL ,
and type 2 carriers use the UCC’s service and keep their logistics capability if p̄1 ≤ m + c−h vH . In case
2(a), we have obtained that p̄∗2 = m + vcL according to Lemma 6. Thus, the optimal choice of the UCC
in period 1 is among the following
 two: i
1. Choose a price p̄1 ∈ m + vH , m + c−h
c−h
vL to attract type 1 carriers only, then V1 = λnvL . Since
no carrier will use the UCC’s service and eliminate his logistics capability, thus ne = 0. Substituting
them into Equation (10), the UCC’s expeted total discounted profit is
 
Π̄(p̄1 ) = (p̄1 + S − M )λnvL − C + δπ̄2 m + vcL
h  i (13)
= (p̄1 + S − M )λnvL − C + δ m + vcL + S − M λnvL − C ,
which increases in p̄1 , so it is optimal for the UCC to choose p̄∗1 = m + c−h vL to maximize profit. This
could be in equilirbium only if ne = 0 (which is satisfied) and m ≤ m̄. Substituting p̄∗1 = m + c−h vL into
 
c−h
Equation (13), we can obtain that Π̄ m + vL = (1 + δ)(m + S − M )λnvL + ((1 + δ)c − h)λn − (1 + δ)C.
 i
2. Choose a price p̄1 ∈ 0, m + c−h vH to attract both types of carriers, then V1 = λnvL + (1 − λ)nvH .
Since no carrier will use the UCC’s service and eliminate logistics capability, thus ne = 0. Substituting
them into Equation (10), the UCC’s expeted total discounted profit is
 
Π̄(p̄1 ) = (p̄1 + S − M )(λnvL + (1 − λ)nvH ) − C + δπ̄2 m + vcL
h  i (14)
= (p̄1 + S − M )(λnvL + (1 − λ)nvH ) − C + δ m + vcL + S − M λnvL − C ,
which increases in p̄1 , so it is optimal for the UCC to choose p̄∗1 = m + c−h
vH to maximize profit. This
could be in equilirbium only if ne = 0 (which is satisfied) and m ≤ m̄. Substituting p̄∗1 = m + c−h
vH into
 
c−h
Equation (14), we can obtain that Π̄ m + vH = (m + S − M )((1 + δ)λnvL + (1 − λ)nvH ) + δλcn +
 
(c − h)n λ vvHL + 1 − λ − (1 + δ)C.

37
      
By comparing Π̄ m + c−h vL and Π̄ m + c−h
vH with respect to m, we obtain that Π̄ m + c−h
vL ≥
   
vL
  λ 1− v −(1−λ) (c−h)+(M −S)(1−λ)vH
Π̄ m + c−h
vH if m ≤ H
(1−λ)vH . Since we assume that λ ≤ vH /(2vH − vL ),
   
vL
λ 1− v −(1−λ) (c−h)+(M −S)(1−λ)vH    
then we can obtain H
(1−λ)vH ≥ m̄. Therefore, Π̄ m + c−h
vL ≥ Π̄ m + c−h vH
always holds if m ≤ m̄, and thus in equilibrium, the optimal price is p̄∗1 = m + c−h vL . This leads to the
results in case 2 of Theorem 7.
In case 2(b) (ñe = 0 and m > m̄), d¯∗i1 = 0 if p̄1 ≤ m + vci1 , or d¯∗i1 = −1 otherwise. Thus, type 1
carriers use the UCC’s service and eliminate their logistics capability if p̄1 ≤ m + vcL , and type 2 carriers
use the UCC’s service and eliminate their logistics capability if p̄1 ≤ m + vcH . In this case, ne = 0 will
never happen which indicates that this case is not in equilibrium.
 
Under the two-period capacity sharing platform model, we define φ̂i2 dˆi2 ; dˆi1 , p̂2 as the expected cost
of period 2 low type carrier i. If carrier i delivers by himself and sells his remaining capacity to the
platform, then the expected revenue generated from selling his remaining capacity depends on the supply
(proportional to the number of carriers who sell remaining capacity on the platform, which is denoted
by ns,2 ) and the demand (proportional to the number of carriers who purchase capacity on the platform,
which is denoted by np,2 ) of capacity on the platform in period 2. To identify a subgame perfect Nash
Equilibrium with rational expectations, we assume all the carriers form the same rational beliefs ñs,2
and ñp,2 about
  ns,2 and np,2 , respectively,
  when they optimize their decisions in period 2. Furthermore,
ˆ∗ ˆ∗
ñs,2 = ns,2 di2 and ñp,2 = np,2 di2 in equilibrium. Define θt = min {ñp,t /ñs,t , 1}, for t = 1, 2. Each
 
low type carrier i minimizes φ̂i2 dˆi2 ; dˆi1 , p̂2 by comparing the following options. (i) dˆi2 = −1: Carrier
i delivers
 on his  own and sells his remaining
 capacity to the platform, which incurs an expected cost
φ̂i2 −1; di1 , p̂2 = c + mvL − di1 − 1 f − θ2 [(1 − α)p̂2 − m] vL . (ii) dˆi2 = 0: Carrier i purchases
ˆ ˆ
 
capacity from the platform, incurring a cost φ̂i2 0; dˆi1 , p̂2 = p̂2 vL . Note that same as in the one-period
platform model, for both periods 1 and 2, if the cost of delivering by himself is identical to the cost of
purchasing capacity from the platform, we assume that carrier i will choose either option with an equal
probability.

Lemma 8. (Optimal decision of the period 2 low type carrier i under the platform model)
1. If dˆi1 = −1 or 1, then in period 2 carrier i purchases capacity from the platform (dˆ∗i2 = 0) if
c+(1+θ2 )mvL
p̂2 < [1+θ 2 (1−α)]vL
, or delivers on his own and sells remaining capacity on the platform (dˆ∗i2 = −1) if
c+(1+θ2 )mvL c+(1+θ2 )mvL
p̂2 > [1+θ2 (1−α)]vL , or chooses either option with probability 0.5 if p̂2 = [1+θ2 (1−α)]vL .
2. If dˆi1 = 0, then in period 2, carrier i purchases capacity from the platform (dˆ∗i2 = 0) if p̂2 <
c+f +(1+θ2 )mvL ˆ∗
[1+θ2 (1−α)]vL , or delivers on his own and sells remaining capacity on the platform (di2 = −1) if p̂2 >
c+f +(1+θ2 )mvL c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL , or chooses either option with probability 0.5 if p̂2 = [1+θ2 (1−α)]vL .

Proof of Lemma 8. By solving φ̂i2 (0; dˆi1 , p̂2 ) < φ̂i2 (−1; dˆi1 , p̂2 ) for vi2 , we obtain that
c+(1+θ2 )mvL c+(1+θ2 )mvL
1. p̂2 < [1+θ 2 (1−α)]vL
if dˆi1 = −1 or 1. Thus, dˆ∗i2 = 0 if p̂2 < [1+θ 2 (1−α)]vL
, and dˆ∗i2 = −1 if p̂2 >
c+(1+θ2 )mvL
[1+θ2 (1−α)]vL .
2. p̂2 < c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL if dˆi1 = 0. Thus, dˆ∗i2 = 0 if p̂2 < c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL , and dˆ∗i2 = −1 if p̂2 >
c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL .

Anticipating the carriers’ optimal decision dˆ∗i2 , the capacity sharing platform chooses p̂2 to maximize her
expected profit in period 2:
π̂2 (p̂2 ) = (αp̂2 vL − A) min {ns,2 , np,2 } . (15)
Lemma 9. (Optimal decision of the platform in period 2) Define ne as the number of carriers
who purchase capacity on the platform and elimate their logistics capability in period 1.
1. If ne > n/2, the optimal price of the platform in period 2 is as follows.

38
 
(2n−ne )(c+f )+2nmvL

 (2n−αne )vL , if (2 − α)ne [(2n − ne )(c + f ) + 2nmvL ]

e )(2ne −n)A
−(2n − αne )(n − ne )(c + 2mvL ) > (2−α)(2n−αn




 α and
(2 − α)ne [(2n − ne )(c + f ) + 2nmvL ]







 −2(2n − αne )(n − ne )(c + f + 2mvL )
> (2−α)(3ne −2n)(2n−αn e )A

;


α


 c+f +2mvL
 − , if (2 − α)ne [(2n − ne )(c + f ) + 2nmvL ]
p̂∗2 = (2−α)vL

 −2(2n − αne )(n − ne )(c + f + 2mvL )
≤ (2−α)(3ne −2n)(2n−αn

e )A
and


α


A




 c + 2f + 2mvL > α ;
c+2mvL
(2−α)vL , if (2 − α)ne [(2n − ne )(c + f ) + 2nmvL ]




(2−α)(2n−αne )(2ne −n)A

−(2n − αne )(n − ne )(c + 2mvL ) ≤



 α and
c + 2f + 2mvL ≤ A

α.


2. If ne ≤ n/2, the optimal price of the UCC’s service in period 2 is as follows.

 (2n−n )(c+f )+2nmv


e
(2n−αne )vL
L
, if 2n2 mvL + (c + f )[2n(n − ne ) − ne (n − αne )] ≤

 (n−αne )(2n−αne )A


 α and
(2 − α)n [(2n − n )(c + f ) + 2nmvL ]

e e



e )(2ne −n)A
−(2n − − + 2mvL ) > (2−α)(2n−αn



 αn e )(n n e )(c α ;
(n−n )(c+f )+nmv
 2
e L
− , if 2n mv + (c + f )[2n(n − n ) − n (n − αn )] >

L e e e


 (n−αne )vL
(n−αne )(2n−αne )A

∗
p̂2 = α and

 2(2 − α)ne [(n − ne )(c + f ) + nmvL ]
−(n − αne )(n − ne )(c + 2mvL ) > (2−α)(n−αnαe )(3ne −n)A ;




 c+2mv
(2−α)vL ,
L
if (2 − α)ne [(2n − ne )(c + f ) + 2nmvL ]




e )(2ne −n)A
−(2n − αne )(n − ne )(c + 2mvL ) ≤ (2−α)(2n−αn

and



 α


 2(2 − α)n e [(n − n e )(c + f ) + nmv L ]
−(n − αne )(n − ne )(c + 2mvL ) ≤ (2−α)(n−αnαe )(3ne −n)A .



Proof of Lemma 9. To derive ns,2 and np,2 in the platform’s expected profit function in Equation (15),
we need to distinguish the following two types of carriers:
Type 1 (dˆi1 = −1 or 1, and vi2 = vL ): Each carrier i of this type purchases capacity from the
c+(1+θ2 )mvL
platform in period 2 if p̂2 < [1+θ 2 (1−α)]vL
, or delivers on his own and sell remaining capacity on the
c+(1+θ2 )mvL c+(1+θ2 )mvL
platform if p̂2 > [1+θ 2 (1−α)]vL
, or chooses either option with same probability if p̂2 = [1+θ2 (1−α)]vL
. The
expected number of carriers of this type is λ(λn − ne ) + λ(1 − λ)n.
Type 2 (dˆi1 = 0, and vi2 = vL ): Each carrier i of this type purchases capacity from the platform
in period 2 if p̂2 < c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL , or delivers on his own and sell capacity on the platform if p̂2 >
c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL , or chooses either option with same probability if p̂2 = c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL . The expected
number of carriers of this type is λne .
To maximize her profit, the optimal choice of the platform is among the following three:
1. Choose a price p̂2 = c+f +(1+θ2 )mvL
[1+θ2 (1−α)]vL to incentivize type 1 carriers to deliver on their own and sell
their remaining capacity, and type 2 carriers to purchase or sell capacity with same probability. Then
we can obtain ns,2 = λ(λn − ne ) + λ(1 − λ)n + λn2 e and np,2 = λn2 e . Substituting them into θ2 , we obtain
ne
θ2 = 2n−n e
. Substituting them to p̂2 and Equation (15), we obtain the platform’s expected profit:
   
π̂2 (2n−n(2n−αn
e )(c+f )+2nmvL α[(2n−ne )(c+f )+2nmvL ]
min λ(λn − ne ) + λ(1 − λ)n + λn2 e , λn2 e

e )vL
= 2n−αne − A
αλne [(2n−ne )(c+f )+2nmvL ]
= 2(2n−αne ) − λn2e A .
c+f +(1+θ2 )mvL
2. Choose a price p̂2 = [1+θ2 (1−α)]vL − to incentivize type 1 carriers
to deliver on their own and sell
their remaining capacity, and type 2 carriers to purchase capacity from the platform. Then we can obtain
ne
ns,2 = λ(λn − ne ) + λ(1 − λ)n and np,2 = λne . Substituting them into θ2 , we obtain θ2 = min{ n−n e
, 1}.
Substituting them to p̂2 , we obtain that
( (n−n )(c+f )+nmv
e
(n−αne )vL
L
− , if ne ≤ n/2;
p̂2 = c+f +2mvL
(2−α)vL − , if ne > n/2.
Substituting p̂2 , ns,2 , and np,2 into Equation (15), we obtain the platform’s expected profit:
(1) If ne ≤ n/2, then

39
    
(n−ne )(c+f )+nmvL α[(n−ne )(c+f )+nmvL ]
π̂2 (n−αne )vL − = n−αne − A −  min{λ(λn − ne ) + λ(1 − λ)n, λne }
αλne [(n−ne )(c+f )+nmvL ]
= n−αne − λne A − .
(2) If ne > n/2, then
   
c+f +2mvL α[(n−ne )(c+f )+nmvL ]
π̂2 (2−α)vL − = n−αne − A −  min{λ(λn − ne ) + λ(1 − λ)n, λne }
αλ(n−ne )(c+f +2mvL )
= 2−α − λ(n − ne )A − .
c+(1+θ2 )mvL
3. Choose a price p̂2 = [1+θ2 (1−α)]vL to incentivize type 1 carriers to sell or purchase capacity
with
same probability, and type 2 carriers to purchase capacity from the platform. Then we can obtain that
ns,2 = λ(λn−ne )+λ(1−λ)n
2 , and np,2 = λ(λn−ne )+λ(1−λ)n
2 + λne . Substituting them into θ2 , we obtain
θ2 = 1. Substituting them to p̂2 and Equation (15), we obtain the platform’s expected profit:
    n o
c+2mvL α(c+2mvL ) λ(λn−ne )+λ(1−λ)n λ(λn−ne )+λ(1−λ)n
π̂2 (2−α)v L
= 2−α − A min 2 , 2 + λn e

= αλ(c+2mv L )(n−ne )
2(2−α) − λ(n−n
2
e )A
.
The capacity sharing platform’s optimal price will be the one yields the largest profit, and thus by
comparing the profits of the platform under choices 1, 2 and 3, we can obtain the results in Lemma 9.

In period 1, if a low type carrier delivers on his own and sells his remaining capacity on the platform,
then the expected cost of this carrier depends on ns,1 and np,1 , which are the numbers of carriers who sell
capacity and who purchase capacity respectively. To identify a subgame perfect Nash Equilibrium with
rational expectations, we assume all the low type carriers in period 1 form the same rational beliefs ñs,1
and ñp,1 about ns,1 and np,1 , respectively, when optimizing their decisions in period
  1. In addition, we 
ˆ
assume they form the same rational belief ñe about ne . Furthermore, ñs,1 = ns,1 di1 , ñp,1 = np,1 dˆ∗i1 ,
∗
 
and ñe = ne in equilibrium. Define Φ̂i dˆi1 ; p̂1 as the expected total discounted cost of each low type
 
carrier i in period 1. Given p̂1 , each carrier i minimizes Φ̂i dˆi1 ; p̂1 by choosing one of the following
options:
(i) dˆi1 = −1: Carrier i delivers on his own and sells his remaining capacity to the platform, which incurs
an expected total discounted cost Φ̂i (−1; p̂1 ) = c + mvL − θ1 [(1 − α)p̂1 − m] vL + δ φ̂∗i2 (−1).
(ii) dˆi1 = 0: Carrier i purchases capacity from the platform and eliminates his logistics capability, which
incurs an expected total discounted cost Φ̂i (0; p̂1 ) = p̂1 vL + δ φ̂∗i2 (0).
(iii) dˆi1 = 1: Carrier i purchases capacity from the platform and keeps his logistics capability, which
incurs an expected total discounted cost Φ̂i (1; p̂1 ) = p̂1 vL + h + δ φ̂∗i2 (1).
Lemma 10. (Optimal decision of the period 1 low type carrier i under the platform model)
Each carrier i purchases capacity from the platform and keep his logistics capability (dˆ∗i1 = 1) if p̂1 <
c+2mvL −h ˆ∗
(2−α)vL , or delivers on his own and sells remaining capacity on the platform (di1 = −1) if p̂1 >
c+2mvL −h c+2mvL −h
(2−α)vL , or chooses either option with probability 0.5 if p̂1 = (2−α)vL .

Proof of Lemma 10. We first determine the carrier’s optimal decision in the case that ñe > n/2, (2 −
e )(2ñe −n)A
α)ñe [(2n − ñe )(c + f ) + 2nmvL ] − (2n − αñe )(n − ñe )(c + 2mvL ) > (2−α)(2n−αñ α and (2 −
(2−α)(3ñe −2n)(2n−αñe )A
α)ñe [(2n − ñe )(c + f ) + 2nmvL ] − 2(2n − αñe )(n − ñe )(c + f + 2mvL ) > α ; or the
case that ñe ≤ n/2, 2n2 mvL + (c + f )[2n(n − ñe ) − ñe (n − αñe )] ≤ (n−αñe )(2n−αñ
α
e )A
and (2 − α)ñ e [(2n −
(2−α)(2n−αñe )(2ñe −n)A
ñe )(c + f ) + 2nmvL ] − (2n − αñe )(n − ñe )(c + 2mvL ) > α . In these cases, according
 
(2n−ñe )(c+f )+2nmvL
to Lemma 9, p̂ = ∗
2 . Each carrier i minimizes his total discounted cost Φ̂i dˆi1 ; p̂1
(2n−αñe )vL
by comparing the following three options:
1. dˆi1 = −1: In this case, according to lemmas 8 and 9, carrier i will deliver on his own and sell
capacity in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (−1; p̂1 ) = c + mvL − θ1 [(1 − α)p1 − m]vL + δ(λ(c + mvL − θ2 [(1 − α)p̂∗2 − m]vL ) + (1 −
λ)(c + mvH )).
2. dˆi1 = 0: In this case, according to lemmas 8 and 9, carrier i will purchase capacity from the
platform or deliver on his own and sell capacity in period 2 with the same probability (if he has low
task volume in period 2, i.e., vi2 = vL ). This incurs an expected total discounted cost Φ̂i (0; p̂1 ) =
p1 vL + δ(λ(p̂∗2 vL /2 + (c + mvL − θ2 [(1 − α)p̂∗2 − m]vL + f )/2) + (1 − λ)(c + mvH + f )).
3. dˆi1 = 1: In this case, according to lemmas 8 and 9, carrier i will deliver on his own and sell
capacity in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (1; p̂1 ) = p1 vL + h + δ(λ(c + mvL − θ2 [(1 − α)p̂∗2 − m]vL ) + (1 − λ)(c + mvH )).

40
By comparing the above three options, we obtain that
(
−1, if p1 > (c + 2mvL − h)/((2 − α)vL );
dˆi1 =
1, if p1 < (c + 2mvL − h)/((2 − α)vL ).
Next we determine the carrier’s optimal decision in the case that ñe > n/2, (2 − α)ñe [(2n − ñe )(c +
f ) + 2nmvL ] − 2(2n − αñe )(n − ñe )(c + f + 2mvL ) ≤ (2−α)(3ñe −2n)(2n−αñ α
e )A
, and c + 2f + 2mvL > A α.
∗ c+f +2mvL
In this case, according to Lemma 9, p̂2 = (2−α)vL − . Each carrier i minimizes his total discounted
 
cost Φ̂i dˆi1 ; p̂1 by comparing the following three options:
1. dˆi1 = −1: In this case, according to lemmas 8 and 9, carrier i will deliver on his own and sell
capacity in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (−1; p̂1 ) = c + mvL − θ1 [(1 − α)p1 − m]vL + δ(λ(c + mvL − θ2 [(1 − α)p̂∗2 − m]vL ) + (1 −
λ)(c + mvH )).
2. dˆi1 = 0: In this case, according to lemmas 8 and 9, carrier i will purchase capacity from the
platform in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (0; p̂1 ) = p1 vL + δ(λp̂∗2 vL + (1 − λ)(c + mvH + f )).
3. dˆi1 = 1: In this case, according to lemmas 8 and 9, carrier i will deliver on his own and sell
capacity in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (1; p1 ) = p1 vL + h + δ(λ(c + mvL − θ2 [(1 − α)p̂∗2 − m]vL ) + (1 − λ)(c + mvH )). By
comparing the above three options, we obtain the same results as in the above case.
Next, we determine the carrier’s optimal decision in the case that ñe ≤ n/2, 2n2 mvL + (c + f )[2n(n −
ñe ) − ñe (n − αñe )] > (n−αñe )(2n−αñ
α
e )A
, and 2(2 − α)ñe [(n − ñe )(c + f ) + nmvL ] − (n − αñe )(n − ñe )(c +
(2−α)(n−αñe )(3ñe −n)A
2mvL ) > α . In this case, according to Lemma 9, p̂∗2 = (n−ñ(n−αñ e )(c+f )+nmvL
e )vL
− . Each
 
carrier i minimizes his total discounted cost Φ̂i dˆi1 ; p̂1 by comparing the following three options:
1. dˆi1 = −1: In this case, according to lemmas 8 and 9, carrier i will deliver on his own and sell
capacity in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (−1; p̂1 ) = c + mvL − θ1 [(1 − α)p1 − m]vL + δ(λ(c + mvL − θ2 [(1 − α)p̂∗2 − m]vL ) + (1 −
λ)(c + mvH )).
2. dˆi1 = 0: In this case, according to lemmas 8 and 9, carrier i will purchase capacity from the
platform in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (0; p̂1 ) = p1 vL + δ(λp̂∗2 vL + (1 − λ)(c + mvH + f )).
3. dˆi1 = 1: In this case, according to lemmas 8 and 9, carrier i will deliver on his own and sell
capacity in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (1; p1 ) = p1 vL + h + δ(λ(c + mvL − θ2 [(1 − α)p̂∗2 − m]vL ) + (1 − λ)(c + mvH )). By
comparing the above three options, we obtain the same results as in the above two cases.
Finally, we determine the carrier’s optimal decision in the case that ñe > n/2, (2 − α)ñe [(2n − ñe )(c +
e )(2ñe −n)A
f )+2nmvL ]−(2n−αñe )(n− ñe )(c+2mvL ) ≤ (2−α)(2n−αñ α , and c+2f +2mvL ≤ A α ; or the case
(2−α)(2n−αñe )(2ñe −n)A
that ñe ≤ n/2, (2−α)ñe [(2n−ñe )(c+f )+2nmvL ]−(2n−αñe )(n−ñe )(c+2mvL ) ≤ α ,
(2−α)(n−αñe )(3ñe −n)A
and 2(2 − α)ñe [(n − ñe )(c + f ) + nmvL ] − (n − αñe )(n − ñe )(c + 2mvL ) ≤ α . In these

cases, according to Lemma 9, p̂∗ = c+2mvL . Each carrier i minimizes his total discounted cost Φ̂i dˆi1 ; p̂1
2 (2−α)vL
by comparing the following three options:
1. dˆi1 = −1: In this case, according to lemmas 8 and 9, carrier i will purchase capacity from
the platform or deliver on his own and sell capacity in period 2 with same probability (if he has low
task volume in period 2, i.e., vi2 = vL ). This incurs an expected total discounted cost Φ̂i (−1; p1 ) =
c + mvL − θ1 [(1 − α)p1 − m]vL + δ(λ(p̂∗2 vL /2 + (c + mvL − θ2 [(1 − α)p̂∗2 − m]vL )/2) + (1 − λ)(c + mvH )).
2. dˆi1 = 0: In this case, according to lemmas 8 and 9, carrier i will purchase capacity from the
platform in period 2 (if he has low task volume in period 2, i.e., vi2 = vL ). This incurs an expected total
discounted cost Φ̂i (0; p1 ) = p1 vL + δ(λp̂∗2 vL + (1 − λ)(c + mvH + f )).
3. dˆi1 = 1: In this case, according to lemmas 8 and 9, carrier i will purchase capacity from the
platform or deliver on his own and sell capacity in period 2 with same probability (if he has low task
volume in period 2, i.e., vi2 = vL ). This incurs an expected total discounted cost Φ̂i (1; p1 ) = p1 vL + h +
δ(λ(p̂∗2 vL /2 + (c + mvL − θ2 [(1 − α)p̂∗2 − m]vL )/2) + (1 − λ)(c + mvH )).
By comparing the above three options, we obtain the same results as in the above three cases. Therefore,
we have the same result for all the cases, and thus Lemma 10 follows.

Proof of Theorem 8. Anticipating the carriers’ responses dˆ∗i1 , the platform optimize her price p̂1 to maxi-
mize her expected total discounted profit Π̂ (p̂1 ) = (αp̂1 vL − A) min {ns,1 , np,1 } + δπ̂2 (p̂∗2 (p̂1 )). According
to Lemma 10, each low type carrier in period 1 either delivers on his own and sells his remaining capacity
on the platform, or purchases capacity on the platform and keeps his own logistics capability. Thus,

41
ñe = ne = 0 in equilibrium. Substituting it into lemmas 8, 9 and 10, we can identify the following three
candidates for equilibrium.
n o
(1) If m ≤ min A−α(c+f
αvL
) (2−α)A−αc
, 2αvL , then in period 1, each carrier purchases capacity from the
c+2mvL −h
platform and keeps his logistics capability if p̂1 < (2−α)vL , or delivers on his own and sell remaining
c+2mvL −h c+2mvL −h
capacity if p̂1 > (2−α)vL , or choose either option with 0.5 probability if p̂1 = (2−α)vL . In period
c+f
2, p̂∗2
= m+ vL and each carrier delivers on his own and sells remaining capacity on the platform.
L −h
It’s optimal for the platform to choose p̂∗1 = c+2mv
(2−α)v to maximize her expected total profit. Since
nL o
∗ A−α(c+f ) (2−α)A−αc
A < αp̂1 vL and A < αp̂ ∗2 vL , the condition m ≤ min αvL , 2αvL cannot hold, and thus this
case is not in equilibrium.
(2) If A−α(c+f
αvL
)
< m ≤ (2−α)A−αc
2αvL , then in period 1, each carrier purchases capacity from the
L −h
platform and keeps his logistics capability if p̂1 < c+2mv
(2−α)vL , or delivers on his own and sell remaining
c+2mvL −h c+2mvL −h
capacity if p̂1 > (2−α)vL , or choose either option with 0.5 probability if p̂1 = (2−α)vL . In period 2,
c+f
p̂∗2= m+ −  and each carrier delivers on his own and sells remaining capacity on the platform.
vL
L −h
It’s optimal for the platform to choose p̂∗1 = c+2mv
(2−α)vL to maximize her expected total profit. Since
(2−α)A−αc
A < αp̂∗1 vL , the condition m ≤ 2αvL cannot hold, and thus this case is not in equilibrium.
(2−α)A−αc
(3) If m > 2αvL ,
then in period 1, each carrier purchases capacity from the platform and
L −h
keeps his logistics capability if p̂1 < c+2mv
(2−α)vL , or delivers on his own and sell remaining capacity if
L −h c+2mvL −h
p̂1 > c+2mv ∗ c+2mvL
(2−α)vL , or choose either option with 0.5 probability if p̂1 = (2−α)vL . In period 2, p̂2 = (2−α)vL
and each carrier delivers on his own and sells remaining capacity on the platform or purchase capacity
L −h
from the platform with the same probability . It’s optimal for the platform to choose p̂∗1 = c+2mv
(2−α)vL to
(2−α)A−αc
maximize her expected total profit. Since A < αp̂∗1 vL and A < αp̂ ∗2 vL , the condition m > 2αvL
always holds, and thus this case leads to the result in this theorem.

42
B Heterogeneous fixed delivery costs
Suppose that there are two groups of carriers: a fraction η of them have low fixed delivery cost cL and
a fraction (1 − η) of them have high fixed delivery cost cH (> cL ). Everything else remains the same as
the base model. The following two theorems describe the equilibrium outcomes under the UCC model
and under the platform model respectively.

Theorem 11. (Equilibrium decisions under the UCC


model with heterogenous carriers’
v
λ(1−η)cH −cL λ v L +1−λ
H −cL
delivery cost) Define m̄1 = M − S + ληv
H
, m̄2 = M − S + (1−η)c ,
h    i L +(1−λ)vH  ηvH
vL vH vH
cL λ 1− v +(1−λ) (1−η) v −1 (1−η) λcH −(1−λ)cL v −λcL
m̄3 = M − S + H
η(1−λ)v
L
, m̄4 = M − S + ληvL +(1−λ)(1−η)v
L
,
h i H   H
v v v
λ (1−η)cH v L −cL +(1−λ)(1−η)(cH −cL vH ) λcH 1− v L −(1−λ)cH
m̄5 = M − S + H
ληv
L
, m̄6 = M − S + H
(1−λ)vH ,
h   iL  
vL vL vL
cL λ 1− v −(1−λ) cH λ v +(1−λ)(1−η) −cL (λ v +1−λ)
m̄02 = M − S + H
, m̄03 = M − S + H H
, m̄04 = M −S+
h i (1−λ)vH   η(1−λ)vH
vL vL
cH (2λ−1)(1−η)−λ v λ cL −cH v −(1−λ)(1−η)cH
H −cL
0
ληvL +(1−λ)(1−η)vH , m̄5 = M − S +
H H
(1−λ)(1−η)vH , and m̄06 = M − S + (1−η)c ηvL .
cL vL
1. If cH ≤ vH , and
(a) m > max{m̄1 , m̄2 , m̄3 }, then the UCC’s optimal price is p̄∗ = m + cL /vH . Under this price,
each carrier uses the UCC’s service.
(b) max{m̄4 , m̄5 } < m ≤ m̄3 , then the UCC’s optimal price is p̄∗ = m + cL /vL . Under this
price, each low type carrier uses the UCC’s service, and each high type carrier uses the
UCC’s service if he has high fixed cost cH and delivers on his own otherwise.
(c) m̄6 < m ≤ min{m̄2 , m̄5 }, then the UCC’s optimal price is p̄∗ = m + cH /vH . Under this
price, each carrier uses the UCC’s service if he has high fixed cost cH and delivers on his
own otherwise.
(d) m ≤ min{m̄1 , m̄4 , m̄6 }, then the UCC’s optimal price is p̄∗ = m + cH /vL . Under this price,
each low type carrier uses the UCC’s service if he has high fixed cost cH , and delivers on his
own otherwise. Each high type carrier delivers on his own.
cL vL
2. If cH > vH , and
(a) m > max{m̄1 , m̄02 , m̄03 }, then the UCC’s optimal price is p̄∗ = m + cL /vH . Under this price,
each carrier uses the UCC’s service.
(b) max{m̄04 , m̄05 } < m ≤ m̄03 , then the UCC’s optimal price is p̄∗ = m + cH /vH . Under this
price, each low type carrier uses the UCC’s service, and each high type carrier uses the UCC’s
service if he has high fixed cost cH and delivers on his own otherwise.
(c) m̄06 < m ≤ min{m̄02 , m̄05 }, then the UCC’s optimal price is p̄∗ = m + cL /vL . Under this price,
each low type carrier uses the UCC’s service and each high type carrier delivers on his own.
(d) m ≤ min{m̄01 , m̄04 , m̄06 }, then the UCC’s optimal price is p̄∗ = m + cH /vL . Under this price,
each low type carrier uses the UCC’s service if he has high fixed cost cH , and delivers on his
own otherwise. Each high type carrier delivers on his own.

Theorem 12. (Equilibrium decisions under the platform model with heterogenous carriers’
delivery cost) Define m̄7 = [2−α(1−η)][1−α(1−η)]A−2αηcH2αv
(2−α(1−η))+αcH (1+η)(1−α(1−η))
L
,
[1−α(1−η)](2−α)(2−3η)A+αη[1−α(1−η)]cL −2α(2−α)(1−η)ηcH
m̄8 = 2αvL ((2−α)(1−η)−η[1−α(1−η)]) ,
(2η−1)(2−α)(2−α(1−η))A+α(2−α)(1−η)(1+η)cH −αη(2−α(1−η))cL
m̄9 = 2αvL (η(2−α(1−η))−(2−α)(1−η)) , and
(2−α)(2−α(1−η))(1−3η)A+2αη(2−α(1−η))cH −α(2−α)(1+η)cH
m̄10 = 2αvL (2−α−2η[2−α(1−η)]) .

1. If η ≤≡ η̄, and

(a) m̄10 < m ≤ m̄9 , then the platform’s optimal price is p̂∗ = (1+η)cH +2mvL
[2−α(1−η)]vL . Under this price,
each low type carrier delivers on his own if he has low fixed cost cL . Otherwise, he delivers
on his own or purchases capacity from the platform with same probability.
(b) m ≤ m̄10 , then the platform’s optimal price is p̂∗ = c(2−α)v
H +2mvL
L
− . Under this price, each low
type carrier purchases capacity from the platform if he has high fixed cost cH and delivers on
his own otherwise.
2. If η̄ < η ≤ 12 , and

43
 (a) m > m̄10 , then the platform’s optimal price is p̂∗ = c(2−α)v
H +2mvL
L
− . Under this price, each low
type carrier purchases capacity from the platform if he has high fixed cost cH and delivers on
his own otherwise.
(b) m ≤ min{m̄9 , m̄10 }, then the platform’s optimal price is p̂∗ = (1+η)c H +2mvL
[2−α(1−η)]vL . Under this
price, each low type carrier delivers on his own if he has low fixed cost cL . Otherwise, he
delivers on his own or purchases capacity from the platform with same probability.
√
3− 9−4α
3. If 1
2 <η ≤1− 2α ≡ η̄ 0 , and

(a) m > max{m̄8 , m̄9 }, then the platform’s optimal price is p̂∗ = c(2−α)v
L +2mvL
L
. Under this price, each
low type carrier purchases capacity from the platform if he has high fixed cost cH . Otherwise,
he delivers on his own or purchases capacity from the platform with same probability.
ηcH +mvL
(b) m̄7 < m ≤ m̄8 , then the platform’s optimal price is p̂∗ = [1−α(1−η)]v L
− . Under this price,
each low type carrier purchases capacity from the platform if he has high fixed cost cH and
delivers on his own otherwise.
(c) m ≤ min{m̄7 , m̄9 }, then the platform’s optimal price is p̂∗ = (1+η)cH +2mvL
[2−α(1−η)]vL . Under this price,
each low type carrier delivers on his own if he has low fixed cost cL . Otherwise, he delivers
on his own or purchases capacity from the platform with same probability.
4. If η > η̄ 0 , and
ηcH +mvL
(a) m > max{m̄7 , m̄8 }, then the platform’s optimal price is p̂∗ = [1−α(1−η)]v L
− . Under this
price, each low type carrier purchases capacity from the platform if he has high fixed cost cH
and delivers on his own otherwise.
(b) m̄9 < m ≤ m̄8 , then the platform’s optimal price is p̂∗ = c(2−α)v
L +2mvL
L
. Under this price, each
low type carrier purchases capacity from the platform if he has high fixed cost cH . Otherwise,
he delivers on his own or purchases capacity from the platform with same probability.
(c) m ≤ min{m̄7 , m̄9 }, then the platform’s optimal price is p̂∗ = (1+η)cH +2mvL
[2−α(1−η)]vL . Under this price,
each low type carrier delivers on his own if he has low fixed cost cL . Otherwise, he delivers
on his own or purchases capacity from the platform with same probability.

The following two propositions summarize the equilibrium profits of the consolidator under the UCC
model and under the platform model respectively.

Proposition 5. (Equilibrium profit of the UCC under heterogenous carriers’ delivery cost)
cL vL
1. If cH ≤ vH , and
(a) m > max{m̄1 , m̄2 , m̄3 }, then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m+ vcH
L
+S −M )(λnvL +
(1 − λ)nvH ) − C.
(b) max{m̄4 , m̄5 } < m ≤ m̄3 , then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m+ vcLL +S−M )(λnvL +
(1 − λ)(1 − η)nvH ) − C.
cH
(c) m̄6 < m ≤ min{m̄2 , m̄5 }, then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m + vH + S − M )(1 −
η)(λnvL + (1 − λ)nvH ) − C.
cH
(d) m ≤ min{m̄1 , m̄4 , m̄6 }, then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m + vL + S − M )(1 −
η)λnvL − C.
cL vL
2. If cH > vH , and
(a) m > max{m̄1 , m̄02 , m̄03 }, then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m+ vcH
L
+S −M )(λnvL +
(1 − λ)nvH ) − C.
(b) max{m̄04 , m̄05 } < m ≤ m̄03 , then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m+ vcH
H
+S−M )(λnvL +
(1 − λ)(1 − η)nvH ) − C.
(c) m̄06 < m ≤ min{m̄02 , m̄05 }, then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m+ vcLL +S −M )λnvL −
C.
cH
(d) m ≤ min{m̄01 , m̄04 , m̄06 }, then the UCC’s expected profit is π̄ ∗ (p̄∗ ) = (m + vL + S − M )(1 −
η)λnvL − C.

Proposition 6. (Equilibrium profit of the platform under heterogenous carriers’ delivery

cost)
1. If η ≤ η̄, and

44
 (a) m̄10 < m ≤ m̄9 , then
 the platform’s expected
 profit is
∗ ∗ λ(1−η)n α((1+η)cH +2mvL )
π̂ (p̂ ) = 2 2−α(1−η) −A .
 
α(cH +2mvL )
(b) m ≤ m̄10 , then the platform’s expected profit is π̂ ∗ (p̂∗ ) = ληn 2−α − A − .

2. If η̄ < η ≤ 12 , and
 
α(cH +2mvL )
(a) m > m̄10 , then the platform’s expected profit is π̂ ∗ (p̂∗ ) = ληn 2−α − A − .
(b) m ≤ min{m̄9 , m̄10}, then the platform’sexpected profit is
π̂ ∗ (p̂∗ ) = λ(1−η)n
2
α((1+η)cH +2mvL )
2−α(1−η) −A .

3. If 1
2 < η ≤ η̄ 0 , and
 
ληn α(cL +2mvL )
(a) m > max{m̄8 , m̄9 }, then the platform’s expected profit is π̂ ∗ (p̂∗ ) = 2 2−α −A .
(b) m̄7 < m ≤ m̄8 , thenthe platform’s expected
 profit is
∗ ∗ α(ηcH +mvL )
π̂ (p̂ ) = λ(1 − η)n 1−α(1−η) − A − .
(c) m ≤ min{m̄7 , m̄9 },
 then the platform’s expected
 profit is
∗ ∗ λ(1−η)n α((1+η)cH +2mvL )
π̂ (p̂ ) = 2 2−α(1−η) −A .

4. If η > η̄ 0 , and
(a) m > max{m̄7 , m̄8 }, then the platform’s expected profit is
π̂ ∗ (p̂∗ ) = λ(1 − η)n α(ηcH +mvL )
1−α(1−η) − A − .
 
ληn α(cL +2mvL )
(b) m̄9 < m ≤ m̄8 , then the platform’s expected profit is π̂ ∗ (p̂∗ ) = 2 2−α −A .
(c) m ≤ min{m̄7 , m̄9 },  then the platform’s expected
 profit is
π̂ ∗ (p̂∗ ) = λ(1−η)n
2
α((1+η)cH +2mvL )
2−α(1−η) − A .

45