Cost Drivers of Electric Bus Contracts: Analysis of 33 Indian Cities

Abstract

Electric buses (e-buses) are rapidly gaining global interest for their energy efficiency and emission reduction benefits. Even though e-bus technology has leapt forward in recent years, the business models are still evolving. Insights from recent e-bus procurements can be used to understand the key factors driving their costs and to improve the efficiency of future procurements. However, there is almost no quantitative literature on e-bus contracting. This paper analyzes data from 36 e-bus contracts across 33 cities in India to identify the key variables that influence the outputs of procurement. The number of bids attracted by the tender and the cost quoted per kilometer are identified as the key outputs of a procurement exercise. Pairwise correlation analysis shows statistically significant associations of the two outputs with the following inputs: number of buses tendered, bus length, contract duration, payment for minimum assured kilometers and any additional kilometers of operation, specifying daily operational hours per bus, performance guarantee to be furnished by the operator, penalty for delay in bus delivery and time allowed for bid submission. The role of effective service planning to ensure accurate representation of operational variables in the contract is also highlighted. These variables are passed through the data envelopment analysis to derive the efficiency of the contracts for the number of bids and the cost quoted per kilometer. The relative efficiency levels between cities and the scale efficiency derived for each contract provide useful insights to improve the cost efficiency of future procurements.

Keywords

public transportation electric trolley buses contracting procurement productivity

Electric buses (e-buses) offer significant benefits such as better air quality, lower greenhouse gas (GHG) emissions and higher energy efficiency compared with buses powered by internal combustion engines (ICE) using diesel or compressed natural gas (CNG). For these reasons, cities around the world are increasingly prioritizing e-buses in their fleet procurement. About half a million electric buses are already in circulation, most of which are in China ( 1 ). Over the next decade, e-buses are likely to become the predominant choice of procurement by cities ( 2 ). The rapid decline in the price of batteries is resulting in the life-cycle cost of e-buses being close to or even lower than that of ICE buses, depending on the operating context ( 1 ). However, the cost of fleet-wide electrification depends on procurement, contracting and business model decisions in addition to the costs associated with technology. Factors like bankability of the contract, payment and penalty terms, and financial and technology risk-sharing between different stakeholders can significantly affect the overall cost of electrification ( 2 ). Improving cost efficiency by minimizing the additional costs imposed by these factors is crucial for cities to achieve their ambitious electrification goals and meet their GHG mitigation targets. Analyzing recent e-bus procurements and the key factors driving their costs can provide crucial inputs on avenues for efficiency improvement for future rounds of procurement.

India offers a unique setting for the study as many Indian cities have issued e-bus contracts since 2019. Requests for Proposals (RfPs) (commonly known as tenders) to procure about 3,500 e-buses were issued in this period, out of which more than 1,500 buses are already operational while the rest are at various stages of contracting and deployment. China, Latin America and Europe are the other regions with large-scale e-bus deployments in recent years ( 1 ). China’s large-scale electrification across many cities is a consequence of continued policy and fiscal support by national and regional governments over many years ( 3 ), which is difficult to emulate in other regions with constrained resources for public transport. In Latin America and Europe, bus electrification efforts are generally concentrated in one or two cities in a country (4, 5). Further, public access to data on e-bus contracts in these regions is limited as they are normally shared only between the contracting authorities and the bidders.

Analyzing the data on recent e-bus procurement across many Indian cities provides a unique opportunity to understand the key cost drivers of e-bus procurement ( 6 ). The dataset is made publicly available by the International Association of Public Transport (UITP) India. Specifically, 36 recent e-bus contracts in India are analyzed to establish the correlation between key technical and financial specifications of contracts and their outcomes (e.g., cost per kilometer). First the analysis finds the covariates with statistically significant associations with the contract outcomes and subsequently uses them to benchmark the efficiency levels of these contracts using data envelopment analysis (DEA). Such benchmarking helps cities to maximize their efficiency by varying their specifications to meet the most efficient contract. Even though e-bus contracts reflect local operational and financial conditions, they broadly comprise similar elements across regions—eligibility criteria, fleet specifications, terms of payments and penalties and other obligations of the contracting authority and operator. Thus, the findings from the analysis of Indian contracts and the approach for benchmarking e-bus contracts will also help cities in other regions design efficient contracts that minimize the cost of procurement.

Literature Review

Contracting has received relatively little attention within public transport research compared with areas like travel demand, operations planning, policy and regulations ( 7 ). Sheng and Meng ( 8 ) present a detailed review of the literature on public bus contracting. The paper highlights that most of the existing research focuses on the approaches to the overall bus contracting, such as negotiation versus competitive tendering-based contracts and incentive design to drive efficiency in contracting. Wong and Hensher ( 9 ) present a detailed analysis of the evolution of competition and ownership of land-based public transport since the late 1980s and present perspectives on their likely evolution in the future. Pedro and Macário ( 10 ) present a more detailed review of recent bus contracts and their transition to bus rapid transit (BRT) systems. The study presents a detailed qualitative analysis of the contract design elements adopted in various countries, such as types of contracts, size and length of contracts, financing, incentives, penalties, risk allocation and monitoring. They highlight the need to contextualize contract design alignments based on local regulatory and institutional regimes. Currie and Fournier ( 11 ) compare good practice in public transport contracting with examples like London (UK) and Melbourne (Australia) by explaining how the contracting models in these cities evolved over successive rounds of procurement. The study outlines key principles for effective contracting like competitive tendering, contract design elements, trust-building, incentives and penalties, risk allocation, entry barriers and regulatory management.

Empirical research on bus contracting and methods to drive its cost efficiency is limited. Hensher et al. ( 12 ) study alternative performance-based bus contracts to identify the risks associated with contracting. They use revealed and stated preference surveys to establish the risk perception of various stakeholders and propose various incentives to mitigate the risks identified. Aarhaug et al. ( 13 ) analyze 232 urban bus contracts awarded in Norway between 1995 and 2015 to identify the key cost drivers of bus contracts. They use regression analysis to find the impact of key variables such vehicle-kilometer and experience on the number of bids received and the cost per kilometer quoted in the winning bids. Among previous studies, the objectives and approach of this study are the closest to the current study.

E-buses are a recent phenomenon and their contracting needs to incorporate additional technological challenges, such as the limited range offered by battery-powered buses compared with ICE buses and the additional hours of charging time needed for these buses while the ICE buses can be refueled within a few minutes. Additional uncertainties such as performance also need to be considered by the operators. However, in gross cost contracts (GCC), the predominant model of contracting globally, the technological and performance risks are borne by the operator and not the transport authority. The existing literature focuses on ICE buses and needs to be adapted to incorporate the specific features and issues that authorities face while contracting out e-buses. Li et al. ( 14 ) present a global case study of e-bus contracting and financing mechanisms covering 22 cities globally. The study presents a qualitative assessment of the financial incentives and legal arrangements for e-bus contracting in these cities. They establish that efficient e-bus contracting is driven by allocation of responsibility for implementation, reduced cost of financing and legal arrangements for balanced risk allocation. The study does not cover a quantitative analysis of the cost implication of these variables.

In summary, the literature on bus contracting predominantly uses qualitative tools to analyze contract design elements. Even the limited quantitative studies do not cover e-buses and their cost drivers. This paper adds to the existing literature on bus contracting by providing empirical insights from a quantitative analysis of recent e-bus tenders in India.

Approach and Methodology

A two-pronged approach is adopted to analyze the e-bus tenders (RfPs) floated between April 2019 and March 2021 in India. The number of bids and their cost quoted per kilometer are selected as the key outputs of the tender, as recommended by Aarhaug et al. ( 13 ). First, correlation analysis is carried out to identify the key input variables that have statistically significant association with the output variables. Subsequently, DEA is adopted to benchmark the efficiency of the contracts. DEA is a non-parametric frontier-based benchmarking technique that puts the most efficient services, the ones with the maximum number of bids or unit cost, on the frontier and evaluates the rest of the services relatively ( 15 ). Although DEA does not propose any mechanism to attain efficiency, it quantifies the changes needed for the inefficient unit to become efficient according to the outputs being sought ( 16 ). It also determines the weightages of each input and output variable within the overall efficiency as an output ( 17 ).

DEA further has two main models of benchmarking, from Charnes et al. ( 18 ) (CCR model) and Banker et al. ( 19 ) (BCC model). The CCR model is based on constant returns to scale, which assumes that an increase in the inputs by a factor will have a proportional increase in the output by the same factor, and vice versa ( 18 ). The efficiencies calculated for each decision making unit (DMU) using this CCR model are known as overall technical efficiencies (OTE). The CCR model does not consider the scale-size of a DMU to be relevant in assessing technical efficiency. Therefore, to know whether inefficiency in any DMU is the result of inefficient production operation or unfavourable conditions displayed by the size of DMU, the BCC input model is also applied. The BCC model is based on variable returns to scale, which assumes that an increase in all the inputs by a factor may not change the output by the same factor, and vice versa ( 19 ). The efficiencies calculated using the BCC model are known as pure technical efficiencies (PTE). PTE measures how efficiently inputs are converted into output(s) regardless of the size of the DMU. The ratio of these two efficiencies (OTE/PTE) is known as scale efficiency, which measures the impact of scale/size on the efficiency of a DMU ( 20 ). The OTE of a DMU can never exceed its PTE. All the three efficiencies (OTE, PTE and scale efficiency) are bounded by zero and one. The outputs from scale efficiency will provide insights on increasing or diminishing returns to scale of the DMU, which can be used to understand if a firm needs to increase or decrease its size to maximize its efficiency ( 21 ).

Formulation of DEA

Equations 1 to 4 provide the formulation adopted for DEA. The objective function of the DEA formulation is to maximize efficiency h of target DMU j₀ where a total of n DMUs operate with m inputs and s outputs; y_rj is the amount of rth output from entity j, and x_ij is the amount of ith input from the same entity j. The decision variables u = (u₁, u₂,…, u_r, …, us) and v = (v₁, v₂, …, v_r, …, v_m) are weights given to the s outputs and m inputs, respectively. The objective function in Equation 1 is iterated n times to calculate the relative efficiencies of one entity at a time. Equation 2 ensures that the efficiency of any entity is not greater than one. Equations 3 and 4 ensure positive weights by assigning an infinitesimally small positive value to weightages.

max h_{j_{0}} (u, v) = \frac{\sum_{r = 1}^{s} u_{r} y_{r j_{0}}}{\sum_{i = 1}^{m} v_{i} x_{i j_{0}}}

(1)

subject to:

\frac{\sum_{r = 1}^{s} u_{r} y_{rj}}{\sum_{i = 1}^{m} v_{i} x_{ij}} \leq 1, j = 1, 2, \dots n,

(2)

\frac{u_{r j_{0}}}{\sum_{i = 1}^{m} v_{i} x_{ij}} \geq ε, r = 1, 2, \dots, s

(3)

\frac{v_{i j_{0}}}{\sum_{i = 1}^{m} v_{i} x_{ij}} \geq ε, i = 1, 2, \dots, m

(4)

Data From Indian E-Bus Contracts

Overview of Recent E-Bus Contracts in India

Recent e-bus procurement in India is driven by the Faster Adoption and Manufacturing of Electric and Hybrid Vehicles (FAME) scheme of the Government of India (GoI). The scheme incentivizes e-bus procurement through a subsidy of INR 4.5 million (approx. USD 60,000) for 9 m buses and INR 5.5 million (approx. USD 73,300) for 12 m buses. To comply, the bus agency receiving the subsidy needs to adopt a GCC-based procurement model using the Model Concession Agreement (MCA) framework issued by GoI. GCC-based procurement involves private ownership and operation of e-buses and their charging infrastructure in lieu of per-kilometer based payments to be made periodically by the contracting authority. The MCA includes the key procurement specifications to make the GCC contract bankable for the issuing authority and the operator providing services. It has adequate flexibility for cities to modify various clauses to suit local needs (MCA, NITI 2019). The procurement process involved the cities issuing a tender (also known as RfP), accompanied by their MCA. Interested service providers submit their bids, out of which the least cost (L1) bidder receives the contract. During the roll-out of FAME II, cities adopted the overall GCC structure suggested by the FAME II MCA but modified many of the specifications to suit their local context. This resulted in significant variation in costs realized by different cities procuring similar buses ( 6 ).

A total of 36 MCAs across 33 cities for a combined procurement of 2,875 e-buses are analyzed. Three of the MCAs are for intercity services while the rest correspond to urban buses. Delhi had three separate tenders while Mumbai had two separate tenders. The fleet type tendered included 2,065 9 m long buses (also known as midibuses) and 810 12 m long standard size buses. The number of bids received across tenders varied from one to four while the cost quoted per kilometer varied between INR 52.2 (USD 0.70 approx.) and INR 86 (USD 1.15 approx.) per kilometer. Such significant variation in number of bids and costs despite adopting a standardized procurement model and a similar MCA indicates scope for improvements in the specifications of the tenders to reduce costs. Table 1 presents the summary of the output and input variables used for correlation and DEA analysis. Their descriptions are explained below.

Table 1.

Summary of E-Bus Tender Input and Output Variables Analyzed

Variable	Description	Units	No. of tenders	Mean	Minimum	Maximum	Standard deviation
nb	Number of bids received	na	34	2.2	1.0	4.0	0.9
cpk	Cost quoted per km of operation by the least-cost bidder	INR per km	34	65.2	52.2	86.0	7.9
bt	Number of buses tendered	na	34	80.4	25.0	300.0	68.1
bl	Type of bus (9 m/12 m long)	meters	34	9.6	9.0	12.0	1.2
rsc	Vehicle range per single charge	km	34	199.4	120.0	300.0	41.0
tbs	Time available for bid submission	Days	34	26.5	6.0	42.0	11.4
cd	Assured contract duration	Years	34	10.0	7.0	12.0	0.7
cey	Contract extension years	Years	34	0.5	0.0	2.0	0.9
ak	Assured km of payment per month	km	34	6041.0	4200.0	18000.0	2191.7
pea	Payment for electricity: authority/operator	Yes/No	34	0.3	0.0	1.0	0.5
dp	Duration of payment secured in an escrow account dedicated to the contract	Days	34	2.2	0.0	3.0	1.0
aepad	Adjusted equity paid in case of contract termination caused by authority’s default	%	34	126.5	0.0	150.0	54.0
stm	Statutory tax obligations clarified?	Yes/No	34	0.8	0.0	1.0	0.4
idp	Interest on delayed payments	% per annum	33	1.5	0.0	3.0	1.2
pga	Performance bank guarantee amount	INR per bus	34	123411.8	3854.0	875000.0	204064.2
pbd	Performance bank guarantee duration	Days	34	137.9	0.0	365.0	78.0
bsd	Bid security deposit	INR per bus	34	0.9	0.1	9.0	1.6
pak	Payment for additional km	%	34	0.6	0.0	0.8	0.2
puk	Payment for under-utilized km	%	34	0.5	0.0	1.0	0.3
pf	Payment frequency (every 15 days/30 days)	Days	34	16.3	0.0	30.0	5.7
cpi	Consumer Price Index-Industrial Workers (CPI-IW) index (in %) within annual payment revision formula	%	34	16.8	0.0	25.0	9.4
wpi	Wholesale Price Index (WPI) (in %) in annual payment revision formula	%	34	16.0	0.0	42.0	14.4
ete	Electricity Tariff (ET) index (in %) in annual payment revision formula	%	34	18.3	0.0	45.0	13.0
pa	Penalty for authority not meeting payment obligations (as a % of total payment)	%	34	0.0	0.0	0.1	0.0
pc	Penalty capping mentioned?	Yes/No	34	0.7	0.0	1.0	0.5
pnsla	Penalty for non-adherence to SLAs	in INR	24	168.8	11.8	875.0	232.4
pld	Penalty for delay in delivery of buses	in INR	34	2494.2	0.0	10000.0	2522.9
ds	Depots specified?	Yes/No	34	0.6	0.0	1.0	0.5
rs	No. of depots	–	34	0.4	0.0	1.0	0.5
pn	Routes specified?	Yes/No	34	0.9	0.0	1.0	0.2
ohs	Permit needed?	Yes/No	34	0.8	0.0	1.0	0.4
nod	Operational hours specified?	Yes/No	34	1.3	0.0	9.0	1.9
toc	Time for opportunity charging	Minutes	34	45.4	0.0	120.0	32.5
eooem	Eligibility: Operator+OEM?	Yes/No	34	0.7	0.0	1.0	0.4
efep	Eligibility: Financial entity part or whole of consortium?	Yes/No	34	0.9	0.0	1.0	0.3
cngm	CNG buses manufactured by bidder?	Yes/No	32	127.5	0.0	500.0	187.2
ebm	Electric buses manufactured by bidder?	Yes/No	33	28.8	0.0	50.0	15.3
cngo	CNG buses operated by bidder?	Yes/No	34	73.8	25.0	500.0	81.8
eeo	Electric buses operated by bidder?	Yes/No	34	9.2	0.0	32.0	11.2

Note: na = not applicable; CNG = compressed natural gas powered; OEM = Original Equipment Manufacturer; SLA = Service Level Agreements.

Output variables: The key output variables for these tenders are:

Number of bids received (nb): Number of qualified bids received by the city.

Cost quoted per kilometer of operation by the L1 bidder (cpk): Least cost quote finalized by the city, the key bidding parameter.

Cost quoted per kilometer is key to financial sustainability and is an obvious output of any bus procurement tender. Additionally, the number of bids was analyzed as it indicates increased competition which typically leads to better cost efficiency in tendering ( 13 ).

Input variables: Data from 48 different input variables that cover all the key aspects of bus procurement were used, including: eligibility criteria for bidders, technical specifications of buses and chargers, operational requirements, contractual obligations, payment terms and penalties. The city level MCAs matched with the MCA structure issued by GoI with modifications to some of the variables to meet local requirements. Such variations are the key to explaining the difference in outputs across cities. Conversely, some variables had the same values across all tenders and are screened out for further analysis as they would not result in any variance in outputs. A total of 39 variables were selected for the 36 tenders after removing those with constant values across cities. The variables selected for analysis and their significance for the current analysis along with their category and a brief description of their relevance for analysis are listed below:

Fleet specifications

Number of buses tendered (bt).

Length of bus (9 m/12 m long) (bl): Two length categories were procured under the FAME II scheme, which has an impact on the vehicle model availability and cost.

Vehicle range per single charge (rsc): Kilometers of range the vehicle needs to deliver per full charge.

Bid process management and payments

4. Time available for bid submission (tbs): Time between the date of issuance of the tender and last date for bid submission.

Obligations on authority

5. Assured contract duration (cd): Minimum number of years for which the contract will be awarded.

6. Contract extension years (cey): Number of years the contract can be extended beyond the assured contract duration.

7. Assured kilometers of payment per month (ak): The minimum kilometers of operations and its corresponding payment to be assured to the operator each month. Penalties for underperformance are calculated separately.

8. Payment for electricity (pea): Whether the payment for the electricity consumed during operations is paid by the contracting authority or the operator.

9. Duration of payment secured (dp): The contracting authority shall open and maintain a separate escrow account dedicated to the contract. Payments to the operator are secured in this account in advance and shall be released as per contractual payment timelines. However, the duration for which such payments are secured varies between cities, thereby impacting the operators confidence in receiving timely payments.

10. Percentage of adjusted equity paid in case of contract termination because of authority’s default (aepad): In case the contract is terminated as a result of the contracting authority defaulting on payments, it is still liable to pay a share of the debt incurred and adjusted equity to the operator. Debt incurred is fixed at 90% by all cities and therefore does not cause any variance in outputs. However, the adjusted equity component varied between cities and therefore is analyzed here.

Obligations on operator

11. Statutory tax obligations of operator clarified (Yes/No?) (stm): If operators are liable to any local taxes, the MCA needs to state them for better clarity in costing. However, some states have not clarified this, leading to uncertainty among bidders.

12. Interest on delayed payments (idp): In case authority delays payments beyond scheduled frequency the additional interest paid by the authority is defined here.

13. Performance bank guarantee amount (pga): Amount of advance performance guarantee amount to be furnished by operator at the time of agreement. This shall be forfeited in case of non-performance and termination of the contract.

14. Performance bank guarantee duration (pbd): Duration for which performance guarantee amount shall be committed.

15. Bid security deposit (bsd): Amount to be deposited by bidders participating in the tender, generally used to ensure participation only from serious bidders.

Payment terms

16. Payment for additional kilometers (pak): If the operator performs for additional kilometers beyond their assured kilometers of payment, authority pays them at an equivalent or lower per-kilometer rate indicated as a percentage of per-kilometer payment for assured kilometers. The lower per-kilometer rate is derived assuming that the fixed costs of the operator like staff and infrastructure costs are already covered by the assured kilometers. The payment for additional kilometers covers the variable costs like energy costs and incentives for operators for operating extra kilometers.

17. Payment for underutilized kilometers (puk): Even if the operator does not perform the assured kilometers of service in a day, it is paid for the actual operated kilometers as a percentage of per-kilometer payment for assured kilometers, to cover for its fixed costs.

18. Payment frequency (pf), that is, whether the payments are made once in 15 days or 30 days. Frequency of payment is crucial to estimate the cash flow requirements of operators.

19. Annual payment revision formula: The per-kilometer payments to the operators are revised annually to cover for increase in key cost items, like staff, consumable items and energy, using economic indicators. The MCA gives the formula for payment indexation as a combination of the following three macro-economic variables:

i. CPI: Consumer Price Index-Industrial Workers (CPI-IW) (in %) to incorporate increasing staff costs.

ii. Wholesale Price Index (WPI) (in %) to incorporate increasing consumable costs like spare parts, tires, and so forth.

iii. Electricity Tariff (ET) index (in %) to cover for increasing electricity tariff within annual payment revision

Penalties on payments

20. Penalty on authority (pa) for not meeting payment timelines, as percentage of total payment.

21. Penalty capping (pc) mentioned (Yes/No)?: Penalties are imposed on operators for non-adherence to SLAs like punctuality, fleet utilization, and so forth. Caps on penalties to a maximum threshold of the total payment will help operators estimate the minimum payment they can expect each month. However, many cities have SLA-wise penalty without mentioning the cap on total payment. This variable helps measure the association between penalty capping and the two output variables.

22. Penalty for non-adherence to SLAs (pnsla) gives the actual penalty specified for underperformance on SLAs.

23. Penalty for delay in delivery of buses (pld) imposed on operators for every day of delay beyond the date of commissioning the project.

Operational readiness variables

The following variables establish the level of preparedness of the authority issuing the tender. They also help operators with accurate estimation of their costs. Given the criticality of depots for efficient operations, it is analyzed in two variables.

24. Depots of operation specified (ds): Yes/No

25. Number of depots (nod) allocated. This variable goes further into the depot planning to check the association of the actual number of depots with the output variables.

26. Routes specified (rs): Yes/No.

27. Separate permit needed for the operator (pn): Yes/No. In some cities operators need to apply for a separate permit to provide public transport services, which leads to additional financial and time costs in implementing the contract.

28. Operational hours specified (ohs): Yes/No. Clear guidance on the number of hours of operation will help with accurate planning of staff, battery sizing and battery charging.

29. Time for opportunity charging (toc). Opportunity charging allows for operators to downsize their batteries and save on capital expenses. However, adequate time availability for opportunity charging is crucial to ensure the daily range requirements are met.

Eligibility criteria for the bidders

These variables are included to test the association of various eligibility criteria of bidders with the output variables.

30. Is Original Equipment Manufacturer (OEM) membership in consortium mandatory for eligibility? (eooem): Yes/No.

31. Eligibility: Can a financial entity be part or whole of the bidding consortium? (efep): Yes/No

32. Eligibility based on CNG buses manufactured by bidder (cngm): Yes/No.

33. Eligibility based on electric buses manufactured by bidder (ebm): Yes/No.

34. Eligibility based on CNG buses operated by bidder (cngo).

35. Eligibility based on electric buses operated by bidder (eeo).

Analysis and Results

Correlation Between Input and Output Variables

Correlation analysis is used to assess the level of association of each of the 37 input variables with the number of bids and cost quoted per kilometer. Even though the data collected from all the 36 tenders are used for this analysis, the exact number of tenders analyzed for each variable varies between 32 and 34 after excluding the tenders with missing values. The coefficient of correlation and its p-value for each variable are summarized in Table 2.

Table 2.

Correlations Between Input and Output Variables

Input variable	Number of bids (nb)		Cost per km (cpk)		Number of tenders analyzed
Input variable	Correlation with nb	p-Value	Correlation with cpk	p-Value	N
bt	0.32	0.07	−0.03	0.86	34
bl	−0.02	0.92	0.44	0.01	34
rsc	−0.27	0.13	−0.50	0.00	34
tbs	0.53	0.00	−0.40	0.02	34
cd	0.31	0.08	−0.32	0.06	34
cey	−0.48	0.00	0.03	0.86	34
ak	−0.29	0.10	−0.28	0.12	34
pea	−0.12	0.50	−0.04	0.83	34
dp	0.27	0.14	−0.41	0.02	34
aepad	−0.10	0.61	−0.17	0.36	34
stm	0.00	0.98	−0.07	0.71	34
idp	0.06	0.74	0.17	0.35	33
pga	−0.15	0.43	0.36	0.04	34
pbd	−0.22	0.23	0.25	0.17	34
bsd	0.28	0.11	−0.03	0.85	34
pak	0.36	0.04	−0.42	0.01	34
puk	0.36	0.04	−0.22	0.23	34
pf	0.21	0.24	0.05	0.80	34
cpi	0.45	0.01	−0.36	0.04	34
wpi	−0.45	0.01	0.04	0.81	34
ete	−0.25	0.17	0.09	0.61	34
pa	−0.13	0.49	−0.16	0.38	34
pc	0.20	0.27	−0.28	0.11	34
pnsla	−0.32	0.14	0.50	0.01	24
pld	0.36	0.04	−0.21	0.25	34
ds	0.29	0.10	0.24	0.18	34
nod	−0.03	0.89	0.19	0.29	34
rs	−0.54	0.00	0.45	0.01	34
pn	0.32	0.07	−0.55	0.00	34
ohs	0.34	0.05	−0.42	0.01	34
toc	0.64	0.00	−0.47	0.00	34
eooem	−0.03	0.87	−0.19	0.31	34
efep	−0.28	0.12	−0.40	0.02	34
cngm	−0.28	0.13	−0.14	0.48	32
ebm	−0.19	0.32	0.06	0.73	33
cngo	−0.23	0.20	−0.06	0.73	34
eeo	−0.54	0.00	0.26	0.14	34

In the case of the number of bids (nb), 15 of the 37 input variables show statistically significant correlation at a significance level of 0.1 or below and 19 variables at a significance level of 0.15 or below. Among those with significant correlation, the variables associated with the returns to the operator like the number of buses tendered (bt), contract duration (cd), payment for additional kilometers beyond assured kilometers (pak) and annual payment revision indexed to CPI show a positive correlation. The variables covering obligations on the operator, such as range per single charge (rsc) and eligibility criteria (eeo) predominantly exhibit a negative correlation. Interestingly, input variables indicating the level of preparedness of the authority and operator, like the specification of the depots (ds) and time available for bid submission (tbs) also have a significant correlation with the number of bids attracted. Some counter-intuitive results have also emerged. Assured kilometers of payment has a negative correlation with the number of bids even though the variable is associated with increasing return to the operator. This could be because of the limited availability of e-bus models to meet higher mileage needs. Variables like bid security deposit and requirement of permit show the opposite to expected sign. However, the financial and time cost associated with these variables is limited.

Correlation with cost per kilometer (cpk) is statistically significant for 14 input variables at a significance level of 0.1 or below. A few other key variables—assured kilometers of payment (ak) and penalty capping (pc)—are also correlated with unit cost at a significance level of 0.15 or below. The variables associated with cost inputs like bus length (more length more cost), performance guarantee amount and penalty for underperformance are positively correlated with unit cost. Other revenue contributors, like contract duration, assured excess-kilometer payments and annual indexation, exhibit a negative correlation with unit cost. Variables like routes specified and permit needed have counter-intuitive associations with unit cost, but they play a relatively minor role compared with the depot and operational hours in determining the cost of operations.

DEA Analysis for Efficiency Comparison

Outputs from the correlation analysis are used to create DEA models to compare the relative efficiency of the e-bus tenders across cities. The output-oriented DEA approach is adopted for benchmarking to establish the potential for maximizing each of the outputs. For this purpose, the cost per kilometer (cpk) is transformed to kilometer per cost (kpc), measuring the kilometers of service for a payment of 100 INR to meet the output maximization criteria of DEA.

Separate DEA models are developed for the number of bids (nb) and kilometers per cost (kpc) because of the limited overlap in the set of input variables that are correlated with them. All the input variables with statistically significant correlation with output variables at 0.15 or below significance level are used for the DEA analysis. Thus, 19 and 16 input variables are selected for DEA of the number of bids (nb) and kilometers per cost (kpc), respectively. Out of these variables, pnsla is removed from DEA model for kpc given the limited dataset of just 24 DMUs. The remaining input variables for both nb and kpc DEA models have data for 34 DMUs each. MaxDEA, a Microsoft access-based open-source tool is used for the DEA analysis. A high ratio of DMUs to input variables is recommended to achieve better efficiency frontiers. It is recommended that the number of DMUs should be two to three times the total number of input and output variables ( 22 – 24 ). Further, strong correlation between input and output variables is preferred while correlation amongst input variables and output variables does not have any impact on the DMU’s efficiency. Therefore, using all the variables with statistically significant correlations places many DMUs on the efficiency frontier, while we know from the variance in nb and kpc that there is a significant difference in efficiency between DMUs. Therefore, the following method is used to identify six or fewer input variables to obtain a healthy ratio of DMUs to input variables.

DEA analysis provides weightage of the input variables for each DMU along with its relative efficiency. While only input variables with significant pairwise correlation with output variables were selected for the analysis, some of these variables have zero weightage, implying that they have no role in determining the efficiency score of a DMU. Therefore, the input variables which were showing zero weightage for a significant number of DMUs were removed. The input variables which have a weightage of zero for 80% or more DMUs (i.e., at least 28 out of the 34 DMUs) are considered as insignificant and removed from the DEA analysis to identify the top five to six input variables exhibiting both pairwise correlation and weightage in determining the efficiency of the DMU. Using this approach, the six and five input variables for nb and kpc are selected for final DEA analysis, respectively. Table 3 shows the most significant input variables for nb and kpc, and Table 4 presents the corresponding data used for the DEA model.

Table 3.

Input and Output Variables Considered for Sensitivity Analysis

Data envelopment analysis model	Output variable	Input variables	Unit
Model 1	Number of bids received (nb)	Number of buses procured (bt)	na
		Contract duration (cd)	Years
		Assured km per month (ak)	km
		Time for bid submission (tbs)	Days
		Operational hours specified (ohs)	Yes/No
		Penalty for delay in delivery of buses (pld)	INR
Model 2	Km per unit cost quoted (kpc)	Bus length (bl)	m
		Assured km per month (ak)	km
		Rate for additional km (rae)	%
		Performance guarantee amount (pga)	INR
		Operational hours specified (ohs)	Yes/No

Table 4.

Data Summary of Output Variables and Input Variables With Significant Correlation

No.	Decision making unit	cpk	kpc	nb	bt	bl	cd	ak	tbs	rae	pga	ohs	bsd	pld
1	Delhi	68.58	1.46	4	300	12	10	6100	25	0.75	500000	0	9.00	2500
2	Mumbai (9 m)	74.00	1.35	2	160	9	10	4200	14	0.7	50000	1	0.50	2000
3	Mumbai (12 m)	83.00	1.20	3	140	12	10	4750	14	0.7	50000	1	0.50	2000
4	Navi Mumbai (9 m)	52.20	1.92	3	30	9	12	5700	36	0.75	50000	1	0.50	50
5	Navi Mumbai (12 m)	69.90	1.43	3	70	12	12	6600	36	0.75	50000	1	0.50	50
6	Kolkata (12 m)	86.00	1.16	1	50	12	10	5000	21	0.3	500000	0	0.50	500
7	Pune	63.95	1.56	2	150	12	10	6844	30	0.5	524667	1	2.62	10000
8	Ahmedabad	54.90	1.82	2	300	9	10	5850	9	0.5	3854	1	2.00	4
9	Surat	55.26	1.81	3	150	9	10	5850	28	0.5	7759	1	1.50	8
10	Jaipur	66.50	1.50	2	100	9	10	6000	10	0	875000	1	3.50	875
11	Bhopal	64.80	1.54	2	100	9	10	6100	28	0.75	23717	1	0.50	1000
12	Indore	63.90	1.56	2	100	9	10	6100	30	0.75	23387	1	0.50	1000
13	Bhubaneshwar	60.22	1.66	2	50	9	10	6600	25	0.65	47694	1	0.25	1000
14	Nagpur	66.30	1.51	1	40	9	10	5700	42	0.75	22222	0	0.30	22
15	Patna	79.83	1.25	1	25	9	7	6000	6	0	500000	0	0.10	500
16	Rajkot	53.91	1.85	2	50	9	10	5850	25	0.5	22707	1	0.65	23
17	Lucknow	62.55	1.60	3	100	9	10	5250	38	0.75	11822	1	0.25	5000
18	Jabalpur	67.23	1.49	1	50	9	10	6100	19	0.75	49212	1	0.25	1000
19	Ujjain	68.40	1.46	1	50	9	10	6100	19	0.75	50069	1	0.25	1000
20	Gwalior	69.96	1.43	1	40	9	10	6100	16	0.6	38408	1	0.25	1000
21	Dehradun	66.78	1.50	2	30	9	10	5400	13	0.3	72122	1	0.67	72
22	Agra	62.55	1.60	4	100	9	10	5250	38	0.75	11822	1	0.25	5000
23	Kanpur	62.55	1.60	3	100	9	10	5250	38	0.75	11822	1	0.25	5000
24	Prayagraj	62.55	1.60	3	50	9	10	5250	38	0.75	23644	1	0.25	5000
25	Varansi	62.55	1.60	3	50	9	10	5250	38	0.75	23644	1	0.25	5000
26	Ghaziabad	62.55	1.60	2	50	9	10	5250	38	0.75	23644	1	0.25	5000
27	Meerut	62.55	1.60	2	50	9	10	5250	38	0.75	23644	1	0.25	5000
28	Bareilly	62.55	1.60	2	25	9	10	5250	38	0.75	47288	1	0.25	5000
29	Moradabad	62.55	1.60	2	25	9	10	5250	38	0.75	47288	1	0.25	5000
30	Aligarh	62.55	1.60	4	25	9	10	5250	38	0.75	47288	1	0.25	5000
31	Jhansi	62.55	1.60	3	25	9	10	5250	38	0.75	47288	1	0.25	5000
32	Rajasthan (intercity)	53.70	1.86	1	50	12	10	18000	18	0.75	115992	0	1.50	5000
33	Goa (intercity)	78.87	1.27	1	50	12	10	6700	9	0	100000	0	0.10	0
34	Uttarakhand (intercity)	62.10	1.61	1	50	9	10	6000	11	0.75	200000	1	0.50	200

Table 5 presents the efficiency results of the CCR and BCC DEA models for nb and kpc. OTEs derived through the CCR model present the relative efficiencies between DMUs for their current size scale. PTEs derived through the BCC model measure how efficiently inputs are converted into output(s) regardless of the size of the State Transport Undertakings (STUs).Therefore, carrying out the BCC model along with CCR helps us to understand whether the reason for inefficiency of a DMU is inefficient contracting practices or unfavourable conditions displayed by its size.

Table 5.

Efficiency Results From Data Envelopment Analysis Models for Number of Bids and Kilometers per Cost

S. No.	Decision making unit (DMU)	Number of bids (nb)			Kilometer per cost (kpc) (km per INR 100)
		OTE (CCR model)	PTE (BCC model)	Scale efficiency (PTE/OTE)	OTE (CCR model)	PTE (BCC model)	Scale efficiency (PTE/OTE)
1	Delhi	1.00	1.00	1.00	0.92	0.96	0.95
2	Mumbai (9 m)	0.74	1.00	0.74	0.96	1.00	0.96
3	Mumbai (12 m)	1.00	1.00	1.00	0.75	0.78	0.97
4	Navi Mumbai (9 m)	1.00	1.00	1.00	1.00	1.00	1.00
5	Navi Mumbai (12 m)	1.00	1.00	1.00	0.68	0.75	0.91
6	Kolkata (12 m)	0.41	1.00	0.41	0.99	1.00	0.99
7	Pune	0.52	0.52	1.00	0.76	0.84	0.90
8	Ahmedabad	1.00	1.00	1.00	1.00	1.00	1.00
9	Surat	1.00	1.00	1.00	0.99	0.99	1.00
10	Jaipur	1.00	1.00	1.00	1.00	1.00	1.00
11	Bhopal	0.71	0.76	0.93	0.83	0.83	1.00
12	Indore	0.71	0.76	0.93	0.84	0.84	1.00
13	Bhubaneshwar	0.93	0.93	0.99	0.88	0.88	1.00
14	Nagpur	0.54	1.00	0.54	1.00	1.00	1.00
15	Patna	1.00	1.00	1.00	1.00	1.00	1.00
16	Rajkot	0.98	1.00	0.98	1.00	1.00	1.00
17	Lucknow	0.75	0.75	1.00	0.96	1.00	0.96
18	Jabalpur	0.48	0.48	1.00	0.78	0.78	1.00
19	Ujjain	0.48	0.48	1.00	0.76	0.76	1.00
20	Gwalior	0.51	0.52	0.99	0.76	0.76	1.00
21	Dehradun	1.00	1.00	1.00	1.00	1.00	1.00
22	Agra	1.00	1.00	1.00	0.96	1.00	0.96
23	Kanpur	0.75	0.75	1.00	0.96	1.00	0.96
24	Prayagraj	0.75	0.75	1.00	0.94	0.96	0.98
25	Varansi	0.75	0.75	1.00	0.94	0.96	0.98
26	Ghaziabad	0.50	0.50	1.00	0.94	0.96	0.98
27	Meerut	0.50	0.50	1.00	0.94	0.96	0.98
28	Bareilly	0.50	0.50	1.00	0.91	0.92	0.98
29	Moradabad	0.50	0.50	1.00	0.91	0.92	0.98
30	Aligarh	1.00	1.00	1.00	0.91	0.92	0.98
31	Jhansi	0.75	0.75	1.00	0.91	0.92	0.98
32	Rajasthan (intercity)	0.39	0.39	1.00	0.99	1.00	0.99
33	Goa (intercity)	1.00	1.00	1.00	1.00	1.00	1.00
34	Uttarakhand (intercity)	0.58	0.58	1.00	0.84	0.84	1.00
Mean efficiency		0.76	0.80	0.96	0.91	0.93	0.98
Number of efficient DMUs		12	16	na	8	14	na

Note: OTE = overall technical efficiencies; PTE = pure technical efficiencies; CCR = Charnes et al. ( 18 ) model; BCC = Banker et al. ( 19 ) model; na= not applicable.

Results from OTE analysis: The average OTE scores for nb and kpc are 0.757 and 0.912 respectively. This reveals that, on average, e-bus contracts in India can increase their nb and kpc outputs by 24.3% and 8.8% respectively with the same input variables. Of the 34 DMUs, 12 are relatively efficient for nb (OTE = 1) while the remaining 22 are relatively inefficient. These 12 DMUs: Delhi, Mumbai (12 m), Navi Mumbai (9 m), Navi Mumbai (12 m), Ahmedabad, Surat, Jaipur, Patna, Dehradun, Agra, Aligarh, Goa (intercity) form the efficiency frontier for the remaining DMUs to emulate. Ghaziabad, Meerut, Bareilly and Moradabad are the cities with the least efficiency for nb. In the case of kpc, only eight DMUs are relatively efficient while the remaining 26 are relatively inefficient. However, 20 of these had OTE greater than the mean OTE of all DMUs.

Results from PTE analysis: BCC efficiency (PTE) is always greater than or equal to CCR efficiency (OTE). Therefore, the number of DMUs on the frontier under the BCC model is always greater than or equal to the number of DMUs on the frontier under the CCR model. As a result, 16 DMUs are relatively efficient for nb and 14 are relatively efficient for kpc. The average PTEs for nb and kpc are 0.800 and 0.928, indicating scope for increasing output by 20% and 7.2% given the scale of the DMUs.

In the case of Mumbai (9 m), Kolkata (12 m), Nagpur and Rajkot we observe that the PTE for nb is one but the OTE is less than one. This indicates that these DMUs are able to 100% covert their available inputs to outputs; however, their OTE is low because of scale-size disadvantages (low scale efficiency). Mumbai (9 m), Kolkata (12 m), Lucknow, Agra, Kanpur and Rajasthan (intercity) have a PTE of one but OTE of less than one.

Results from scale efficiency analysis: Scale efficiency assumes each DMU is efficient and tells us if it is possible to increase the output with the same level of inputs. A comparison of the results for CCR and BCC gives an assessment of whether the size of the DMU has an influence on its OTE. If the value of scale efficiency is one, then the DMU is apparently operating at optimal scale. If the value is less than one, then the DMU is either small or big relative to its optimum scale-size and has scope to improve efficiency by varying its scale of operations through alternatives ( 25 ).

Ten DMUs show scope for scale efficiency improvement for nb: Mumbai (9 m), Kolkata (12 m), Bhopal, Indore, Bhubaneshwar, Nagpur, Rajkot, Jabalpur, Ujjain and Gwalior. Similarly, 16 DMUs show scope for scale efficiency improvement for kpc: Navi Mumbai (9 m), Ahmedabad, Surat, Jaipur, Bhopal, Indore, Bhubaneshwar, Nagpur, Patna, Rajkot, Jabalpur, Ujjain, Gwalior, Dehradun, Goa (intercity) and Uttarakhand (intercity).

Conclusion

This study is a first of its kind empirical and quantitaive analysis of e-bus contracts to identify the key drivers of their costs and overall procurement efficiency. The paper presents a detailed analysis of 36 e-bus procurements conducted across 33 Indian cities to understand their key cost drivers and the current level of efficiency. The input variables of the contracts with significant correlation with the number of bids attracted and cost of service delivery are derived along with their strength of association using correlation analysis. A DEA based approach is used to derive the relative procurement efficiencies of cities and the scale efficiencies of each city in maximizing the number of bids and minimizing the cost quoted per kilometer.

The number of bids attracted by a city is a function of the overall size of the contract, measured through variables like number of buses, contract duration and assured payment-kilometers. The cost quoted by operators depends more on the cost input specifications like type of bus and the financial specifications of the contract such as payment for additional kilometers and performance guarantee to be furnished by the operator. Additionally, results from the correlation analysis show that the assured kilometers of payment and specification of daily operational kilometers are significant factors in both number of bids and costs. These results underline the need for effective service planning in preparation of e-bus procurement as the kilometers and hours of service volume needed to be delivered by the fleet is determined while planning for the service.

The DEA based efficiency analysis shows that the average procurement efficiency is 0.80 for number of bids attracted, indicating significant scope for improvement in selecting the values of input variables correlated with this output. The average efficiency of cost per kilometer is already at 0.93 and has limited scope to improve further. The average scale efficiency of both the output variables is 0.96 and above, implying that they are already deriving optimal outputs from the available inputs.

Findings from this study have global relevance given the universal nature of the e-bus contract elements. Given the increasing ambition of inducting e-bus fleets globally, these findings identify the least cost path of electrification for cities. This study may be extended further by incorporating e-bus results from geographical areas outside India to identify other location-specific input variables. Moreover, learnings from cost drivers of GCC-based e-bus procurement analyzed here can be applied to predict the cost efficiency of emerging business models like unbundling of asset ownership and operations.

Key Takeaways for Decision Makers

• Number of bids received and the cost quoted by the bidders, the two key outputs of an e-bus procurement process, are influenced by a wide range of technical and financial input variables, some of which are common between the two variables while the rest influence only one of the outputs.

• Input variables of e-bus procurement can be classified under categories such as eligibility criteria of bidders, fleet specifications, bid process management, payments, penalties, operational readiness and obligations on the operator and authority.

• Number of bids received is positively correlated with variables associated with increasing financial returns to the operator such as number of buses being procured, their contract duration and preparedness of the contracting authority. At the same time additional obligations on the operator in the form of technical and financial eligibility criteria can reduce the number of bids received.

• Cost quoted by bidders is likely to reduce with increased financial returns, measured through variables like contract duration, assured kilometers of payment, incentive for additional kilometers of operation, level of capping on penalties and annual indexation on payments. It is positively correlated with items increasing the cost of operations, such as the length of the buses and performance guarantee amounts to be pledged by the operator.

• Cities need to strike the right balance in setting stringent eligibility criteria and obligations on the operator. Focus on improved infrastructure and operational preparedness of the contracting authority along with subsequently clarity on the returns to the operator and its indexation with changing operational costs is also crucial to increasing number of bids and reducing their quoted costs.

Supplemental Material

sj-xlsx-1-trr-10.1177_03611981221088593 – Supplemental material for Cost Drivers of Electric Bus Contracts: Analysis of 33 Indian Cities

Supplemental material, sj-xlsx-1-trr-10.1177_03611981221088593 for Cost Drivers of Electric Bus Contracts: Analysis of 33 Indian Cities by Ravi Gadepalli, Sushmitha Gumireddy and Prateek Bansal in Transportation Research Record

Footnotes

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: R. Gadepalli; data collection: R. Gadepalli, S. Gumireddy; analysis and interpretation of results: R. Gadepalli, S. Gumireddy, P. Bansal; draft manuscript preparation: R. Gadepalli. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Ravi Gadepalli

Sushmitha Gumireddy

Prateek Bansal

Supplemental Material

Supplemental material for this article is available online.

References

Global EV Outlook. Entering the Decade of Electric Drive. International Energy Agency, Paris, France, 2020.

UITP. Large Scale Bus Electrification: The Impact on Business Models. International Association of Public Transport, Brussels, Belgium, 2021. https://cms.uitp.org/wp/wp-content/uploads/2021/07/Large-scale-Bus-Electrification-KB-Final.pdf. Accessed July 31, 2021.

ADB. Sustainable Transport Solutions: Low Carbon Buses in the People’s Republic of China. Asian Development Bank, Manila, Philippines, 2018. https://doi.org/10.22617/TCS189646-2.

ZEBRA. E-bus Radar: Electric Buses in Latin America, 2021. https://www.ebusradar.org/en/. Accessed July 31, 2021.

Sustainable Bus. E-bus Radar: Electric Buses in Latin America, 2020. https://www.sustainable-bus.com/electric-bus/electric-bus-public-transport-main-fleets-projects-around-world/. Accessed July 31, 2021.

UITP. Electric Bus Procurement Under FAME II: Lessons Learnt and Recommendations for Phase-II. International Association of Public Transport, Brussels, Belgium, 2020. https://cms.uitp.org/wp/wp-content/uploads/2020/05/Draft-Electric-bus-procurement-under-FAME-II_Final-Report.pdf. Accessed July 31, 2021.

Currie

Introduction to the Handbook of Public Transport Research. In Handbook of Public Transport Research ( Currie

, ed.), Edward Elgar Publishing, Cheltenham and Camberley, 2021.

Sheng

Meng

Public Bus Service Contracting: A Critical Review and Future Research Opportunities. Research in Transportation Economics, Vol. 83, 2020, p. 100938.

Wong

Y. Z.

Hensher

D. A.

The Thredbo Story: A Journey of Competition and Ownership in Land Passenger Transport. Research in Transportation Economics, Vol. 69, 2018, pp. 9–22.

10.

Pedro

M. J.

Macário.

A Review of General Practice in Contracting Public Transport Services and Transfer to BRT Systems. Research in Transportation Economics, Vol. 59, 2016, pp. 94–106.

11.

Currie

Fournier

Good Practice Public Transport Concessions: The Cases of London and Melbourne. International Transport Forum, 2021. https://www.itfoecd.org/sites/default/files/docs/public-transport-concessions-london-melbourne.pdf.

12.

Hensher

D. A.

Knowles

Efficient Contracting and Incentive Agreements Between Regulators and Bus Operators: The Influence of Risk Preferences of Contracting Agents on Contract Choice. Transportation Research Part A: Policy and Practice, Vol. 87, 2016, pp. 22–40.

13.

Aarhaug

Fearnley

Gregersen

F. A.

Norseng

R. B.

20 Years of Competitive Tendering in the Norwegian Bus Industry–An Analysis of Bidders and Winning Bids. Research in Transportation Economics, Vol. 69, 2018, pp. 97–105.

14.

Castellanos

Maassen

Emerging Trends and Innovations for Electric Bus Adoption—A Comparative Case Study of Contracting and Financing of 22 Cities in the Americas, Asia-Pacific, and Europe. Research in Transportation Economics, Vol. 69, 2018, pp. 470–481.

15.

Kotsemir

Measuring National Innovation Systems Efficiency–A Review of DEA Approach. Research Paper No. WP BRP 16/STI/2013. 2013. National Research University Higher School of Economics (HSE), Moscow, Russia, 2013.

16.

Saxena

R. R.

Measuring Efficiencies in Indian Public Road Transit: A Data Envelopment Analysis Approach. OPSearch. Vol. 47, No. 3, 2010, pp. 195–204.

17.

Sherman

H. D.

Zhu

Service Productivity Management: Improving Service Performance Using Data Envelopment Analysis (DEA). Springer Science & Business Media, Boston, MA, 2006.

18.

Charnes

A. C.

Cooper

W. W.

Rhodes

Measuring the Efficiency of Decision Making Units. European Journal of Operational Research, Vol. 2, No. 6, 1978, pp. 429–444.

19.

Banker

R. D.

Charnes

Cooper

W. W.

Some Models for Estimating Technical and Scale Inefficiencies in Data Envelopment Analysis. Management Science, Vol. 30, No. 9, 1984, pp. 1078–1092.

20.

Jarboui

Forget

Boujelbene

Public Road Transport Efficiency: A Literature Review via the Classification Scheme. Public Transport, Vol. 4, No. 2, 2012, pp. 101–128.

21.

Agarwal

Yadav

S. P.

Singh

DEA Based Estimation of the Technical Efficiency of State Transport Undertakings in India. Opsearch, Vol. 47, No. 3, 2010, pp. 216–230.

22.

Boussofiane

Dyson

R. G.

Thanassoulis

Applied Data Envelopment Analysis. European Journal of Operational Research, Vol. 52, No. 1, 1991, pp. 1–15.

23.

Bowlin

W. F.

Measuring Performance: An Introduction to Data Envelopment Analysis (DEA). The Journal of Cost Analysis, Vol. 15, No. 2, 1998, pp. 3–27.

24.

Dyson

R. G.

Allen

Camanho

A. S.

Podinovski

V. V.

Sarrico

C. S.

Shale

E. A.

Pitfalls and Protocols in DEA. European Journal of Operational Research, Vol. 132, No. 2, 2001, pp. 245–259.

25.

Daraio

Diana

Di Costa

Leporelli

Matteucci

Nastasi

Efficiency and Effectiveness in the Urban Public Transport Sector: A Critical Review With Directions for Future Research. European Journal of Operational Research, Vol. 248, No. 1, 2016, pp. 1–20.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB