262

Chapter 9 Evaluation of Public Transportation Investments

263

Topics: Introduction to Urban Transit Performance Analysis, Performance Measures for Transit Evaluation, Models for Evaluating Transit Performance, Tools for Metropolitan Transportation Evaluation, Past Studies on Transit Evaluation.

9.1 INTRODUCTION TO URBAN TRANSIT PERFORMANCE ANALYSIS

Why consider performance?

• Private vs. Public Ownership

• Public Transit Goals

What are the possible uses?

• Establish funding requirements

• Operational improvement

• Regulatory compliance

• Performance audits

• Capital planning/programming

Who are the users?

• Federal agencies

• State/provincial agencies

• Local agencies

• Transit boards

• Operators

• Labor unions

• Professional organizations

• Research agencies/consultants

• Public

Definitions

Efficiency

1. Resource Utilization

A. Labor Utilization

B. Vehicle Utilization

2. Financial Performance

Efficiency – Service Output / Resource Input

Performance

Effectiveness – Achieving System Objectives

264

A. Cost Per Produced Output

B. Cost Per Consumed Output

3. Maintenance Performance

4. Safety Performance

Effectiveness

1. Service Supplied

A. Accessibility

B. Quality of Service

C. Service to Special Groups

2. Utilization of Service Passenger Utilization

3. Revenue Generation Capability

A. Revenue Per Produced Output

B. Revenue Per Consumed Output

Types of Performance Indicators

Efficiency

Operating Ratio

Operating Cost per Vehicle-Mile

Operating Cost per Vehicle-Hour

Operating Cost per Passenger-Mile

Effectiveness

Ridership Per Capita

Percent Population Served

Percent On-Time Arrivals

Percent Elderly and Handicapped Served

Passengers Per Route-Mile

Deficit Per Passenger.

Use of Performance Indicators for Operational Improvement

1. Develop Goals and Objectives

2. Conduct Baseline Inventory

3. Calculate Indicators -Systemwide/Route

4. Evaluate Performance -Targets and Standards

A. Time Series Comparison

B. Comparison with Peer Groups

5. Further analysis to explain why indicators are low or unfavorable, if necessary

265

6. Identify and evaluate possible corrective actions, if necessary

7. Select strategic actions

8. Implement through work programs, operational plans and budgets

9. Monitor results of actions in terms of goals

10. Repeat the process periodically.

Figure 9-1: Steps for Performance Evaluation

Service Variables Output Variables Resource Variables

• Speed

• No. of Stops

• Frequency

• Transfer

• Ridership

• Revenue

• Vehicle-miles

• Vehicle-hours

• Number of Buses

• Number of Drivers

• Operating Costs

• User Costs

Figure 9-2: Transit Performance Interactions

Transit system

establish goals and objectives

select functions to evaluate and indicate

collect and tabulate data

take corrective actions and monitor

analyze indicators

5

4

1

2

3

Policy Variables

Performance Measures

266

EXPENSE

Figure 9-3: Relationships between Expense Data, Indicators, and Corrective Actions

DATA INDICATORS CORRECTIVE ACTIONS

expenses + vehicle miles

expenses + vehicle hours

expenses + passengers

expenses + revenue hours

administrative expenses + total expenses

total expenses

vehicle hours

total passengers

administrative expenses revenue miles

revenue hours

decrease expenses

reroute service

expand ridership

decrease deadhead

modify fares

eliminate marginal routes

vehicle miles

267

REVENUE

Figure 9-4: Relationships between Revenue Data, Indicators, and Corrective Actions

DATA INDICATORS CORRECTIVE ACTIONS

revenue + revenue hours

revenue + revenue miles

revenue + passengers

passenger revenue + revenue miles

passenger revenue +

revenue hours

passenger revenue + passengers

passenger revenue +

expenses

fares + total revenue

passenger revenue + total revenue

passenger revenue

fare revenue

revenue miles

revenue hours

passengers

expenses

eliminate unproductiveroutes

increase service

increase speed

increase stop locations

decrease headways

increase fares

reduce administrative cost

increase fare payingpassengers

increase contract service

increase ancillary services

revenue

268

LEVEL OF SERVICE (LOS)

Figure 9-5: Relationships between LOS Data, Indicators, and Corrective Actions

DATA INDICATORS CORRECTIVE ACTIONS

revenue hours

vehicle miles

vehicle hours

decrease stop dwell

reroute/reschedule congested areas

reroute excessive turns and indirect

routes increase fare

collection speed increase stop

spacing enforce bus stop

no parking

revenue miles

revenue miles + revenue hours

vehicle miles per year

vehicle hours per year

monitor drivers

improve vehicle reliability

269

RIDERSHIP

Figure 9-6: Relationships between Ridership Data, Indicators, and Corrective Actions

DATA INDICATORS CORRECTIVE ACTIONS

passengers + vehicle miles

passengers + vehicle hours

fare passengers + total passengers

elderly passengers + total

passengers

1982 passengers + 1981 passengers

passengers

fare passengers

elderly passengers

current year passengers

previous year passengers vehicle miles

modify fare structure

improve cleanliness safety, reliability

fare incentives

alter routes and schedules

alter or stop marginal routes

increase vehicle speed

improve marketing

decrease deadheadvehicle hours

increase fare passengers

270

9.2 PERFORMANCE MEASURES FOR TRANSIT EVALUATION

Performance is defined as the execution of required function. Performance measures are assessment

data or techniques that directly or indirectly, quantitatively or qualitatively reflect the degree to which results

meet the needs and expectations. A variety of trends and forces in the field of public administration generally

have resulted in renewed interest in performance measurement. Many state transportation agencies are

concerned with assuring the availability of public transit in their local communities and accessibility to transit

for residents who either have to depend on it or who prefer to use it. Notwithstanding the issues of accessibility

and personal mobility, performance measurement in the transit industry has become fairly standardized, due in

part to longstanding federal reporting requirements for transit operators receiving financial assistance from

Federal Transit Administration (FTA). A detailed list of transit performance measures is given in Table…

Table 9-1: Summary of Transit Performance Measures

Goal Category Category Performance Measure Miles between road calls for transit vehicles

Age distribution System Preservation Transit vehicle

Capacity / remaining useful life index

Fare recovery rate of urban transit systems

Cost per PMT for urban transit systems

Cost per VMT for urban transit systems

Cost per revenue-mile for urban transit systems

Cost per PMT in rural areas

Cost per VMT in rural areas

Cost per revenue-mile in rural areas

Total transit operating expenditures per transit-mile

Financial

Grant dollars per transit trip

Transit ridership per capita

Transit ridership-to-capacity ratio

Transit ridership per VMT

Transit ridership per route-mile

Transit ridership per revenue-mile

Transit peak load factor

Ridership

PMT on intercity rail and bus service

Number of peak-period vehicles

Revenue vehicle hours per transit employee

Average wait time to board transit

Operational Efficiency

Operational

Ratio of number of transit incidents to investment in transit security

271

Table 9-1: Summary of Transit Performance Measures (continued)

Goal Category Category Performance Measure Percent of population with access to (or within "X” miles of) transit (or fixed-route transit) service Percent of urban and rural areas with direct access to bus service

Percent of workforce that can reach worksite in transit within “X” minutes

Access to and amount of transit

Access time to passenger facility

Route-miles (or seat-miles or passenger-miles) of transit service

Frequency of transit service

Route spacing Service

characteristics

Percent of total transit trip time spent out of vehicle

Transfer distance at passenger facility

Availability of intermodal ticketing and luggage transfer Facility characteristics

Existence of information services and ticketing v/c of parking spaces during daily peak hours for bus, or other passenger terminal lots Parking spaces per passenger

Parking spaces available loading / unloading by autos

Accessibility

Parking, pickup / delivery

Number of pickup / discharge areas for passengers

On-time performance of transit

Frequency of transit service

Average wait time to board transit

Number of public transportation trips

Passengers per capita within urban service area

Number of commuters using transit park-and-ride facilities

Number of demand-response trip requests

Mobility Transit

Percent of transit demand-response trip requests met

Economic indicator for people movement Percent of region’s unemployed or poor who cite transportation access as a principal barrier to seeking employment

Economic

Development Transit

Percent of wholesale and retail sales in the significant economic centers served by market routes Customer perception of satisfaction with commute time

Quality of Life Transit

accessibility, mobility Customer perception of quality of transit service

Transit collisions (injuries or fatalities) per PMT

Transit collisions (injures or fatalities) per VMT

Number of intercity bus collisions

Crimes per 1,000 passengers

Safety Transit

Ratio of number of transit collisions to investment in transit security

Tons of pollutants generated

Air quality rating

Number of days that Pollution Standard Index is in unhealthful range Air pollution

Customer perception of satisfaction with air quality

Fuel consumption per VMT

Environmental and Resource Conservation

Fuel usage Average mileage per gallon

272

9.3 MODELS FOR EVALUATING TRANSIT PERFORMANCE

9.3.1 Transit Economic Requirements Model.

The Transit Economic Requirements Model (TERM) consolidates existing engineering- based

evaluation tools and introduces a benefit/cost analysis to ensure that investment benefits exceed investment

costs. Specifically, TERM identifies the investments needed to replace and rehabilitate existing assets, improve

operating performance, and expand transit systems to address the growth in travel demand, and then evaluates

these needs on the basis of costs and benefits in order to select future investments. The TERM provides

estimates of the total annual capital expenditures required to maintain or improve the physical condition of

transit systems and the level of service they provide. The estimate represents the total urbanized area transit

investment required by all levels of government. The model also generates estimates of current transit

conditions and performance evaluates the impact of varying levels and types of investment on future conditions

and performance. TERM forecasts investment needs via four distinct modules: (1) asset rehabilitation and

replacement with reinvestments in existing assets to maintain and improve the assets' physical condition; (2)

asset expansion with investments in new assets such as vehicles and facilities to maintain operating

performance to meet forecasts of travel demand; (3) performance enhancement with investments in additional

transit capacity to improve operating performance; and (4) benefit-cost tests that analyze all investments

identified on a benefit-costs basis. The TERM modules are further subdivided by mode, asset type, and urban

area characteristics. In addition to investment estimates, TERM generates estimates of the physical condition of

the Nation's transit assets.

9.3.2 Personal Rapid Transit Operations Simulation Models

There are currently several computer-based simulation models that are capable of simulating the

operating performance of a personal rapid transit (PRT) system. They are briefly discussed as below:

(a) Taxi 2000 Corporation PRT System Simulator. This computer program simulates the motion of

passengers and vehicles in a PRT network of any desired configuration. The network is introduced by giving the

coordinates of the apexes of all curves, the connection between apexes, the position of the entry point of each

station from the nearest upstream apex, the line speed, desired curve radius at each apex, the desired level of

comfort (acceleration and jerk), and the desired length of each off-line station guideway. Any desired demand

matrix and demand-time distribution can be entered. The user specifies the estimated number of passengers per

vehicle. The output of each run gives the average, maximum and median wait times for each station and for the

network as a whole and a long list of performance parameters, including distances traveled, average vehicle

occupancy, average speed, average and peak power. The simulation is an accurate representation of a real PRT

system.

(b) Raytheon's NETSIM Simulation Program. The Raytheon Company has developed a PRT system called

PRT 2000 that offers the potential of complementing existing city infrastructure, reducing congestion and air

273

pollution while providing convenient origin-to-destination transportation. PRT 2000 is an automated transit

system which uses a series of loops to achieve maximum ridership over a large area. To support the application

of this new technology, Raytheon has developed the Network Simulator (NETSIM) to simulate the performance

of a PRT 2000 system in a particular environment and report its performance under varying passenger load

scenarios. The NETSIM can be used to construct a conceptual layout of a proposed PRT system; evaluate the

capacity of a PRT system including identifying potential bottlenecks; evaluate the performance of a PRT system

under various scenarios; evaluate traffic management algorithms; and size the station berth equipment.

(c) SIMAXAR. This program has been developed in France in support of work on the AXAR PRT technology.

It implements an event-type simulation. A set of data files is used to describe the geometry of the network, to

parameter the model of vehicle kinematics, to define passenger demand and the parameters of the decentralized

strategy for the management of empty vehicles. The results supplied by the program after a test are given per

time period and in a global form. Indicators such as the average and maximum waiting times per station, the

number of passengers conveyed and the average journey times are aggregated and reported. Global indicators

such as the total distance covered by occupied vehicles and empty vehicles and the total time spent in queues

are used to compare different routing strategies of empty vehicles.

(d) Princeton's PRT Simulation Program. This model has been developed at Princeton University, which is

useful for designing a PRT network as well as simulating its operation. It produces real time visual displays of

the components of the PRT network from any selected viewpoint. One can walk around the guideways and

stations, fly above them and view the network and surrounding buildings or ride in a vehicle down the

guideway. It represents a "virtual reality" concept and uses a powerful Silicon Graphics computing machine to

produce its graphics displays.

9.4 TOOLS FOR METROPOLITAN TRANSPORTATION EVALUATION

Introduction

Post-ISTEA flexibility in use of federal funds and emphasis on social, economic and environmental

objectives has increased the need for evaluation tools to supplement travel demand estimation tools. The US

DOT has recently developed several such tools to help planners estimate performance measures to assess travel

mode, congestion, air quality, equity and safety impacts, and to compare investment in alternative modes with

one another and with travel demand management strategies. The tools are categorized as follows:

• Tools for ITS Evaluation

• Tools for Rail and Highway-Rail Crossing Evaluation .Tools for Cross-Modal Evaluation

• Tools for Evaluation of Development Effects

• Tools for Equity Analysis, Financial Analysis, and Transportation Improvement Program Evaluation

This section summarizes the purpose, inputs and outputs of each tool, discuss appropriate use of each

tool using case study examples, and explain how the tools differ in their capabilities and results.

274

Tools for ITS Evaluation

Two tools for ITS evaluation. SCRITS and IDAS will be compared using case study application results

from the two models for ITS alternatives for Toledo, Ohio. SCRITS (SCReening for ITS) is a screening-level

spreadsheet analysis tool for estimating the user benefits of Intelligent Transportation Systems (ITS) strategies

at the corridor/subarea or system level. Each individual ITS application requires user input characterizing the

ITS alternative and hue travel demand. The primary measures of effectiveness include changes in vehicle hours

of travel (VHT), vehicle miles of travel (VMT), emissions (CO, NOx, HC), vehicle operating costs, energy

consumption, accidents, and economic benefit. SCRITS is available from the STEAM website at

www.fhwa.dot.gov/steam.

The ITS Deployment Analysis System (IDAS), like SCRITS, calculates the relative cost and benefit

of potential ITS investments either at the corridor/subarea level or the system (regional network) level. The

difference is that IDAS incorporates more detailed analysis of costs and benefits, and even includes a travel

demand model to account for effects of ITS not easily estimated by conventional four-step models. Inputs, as in

SCRITS, are ITS alternative characteristics. Demand data is provided using outputs from four-step models --

trip tables and loaded highway networks. IDAS is currently being beta-tested by four MPOs and will be

released in December 1999 through the McTrans Center, email: mctrans@ce.ufl.edu. Additional information

may be obtained from the IDAS website at www.cta.ornl.gov/cta/research/idas/index.htm.

Tools for Rail and Highway-Rail Crossing Evaluation

Tools for evaluation of rail and highway-rail crossing projects and strategies include RailDec 2.0 and

GradeDec2000. Appropriate use of these tools will be demonstrated using rail project data and travel demand

data for a case study in Toledo, Ohio. RailDec 2.0 forecasts transportation and non-transportation effects of rail

and rail-related intermodal projects for strategic planning and budgeting. Users provide data on rail and

highway conditions and a range of rail demand estimates reflecting uncertainty in market operating

characteristics. RailDec 2.0 estimates savings in costs for shipping, accidents, travel time, emissions, vehicle

operation, highway maintenance, and train delay. Outputs are benefit-cost estimates such as B/C ratio, and

financial results such as operating revenues and expenses. RailDec 2.0 is available from the Federal Railroad

Administration's web site at: www.fra.dot.gov.

GradeDec 2000 is a tool for benefit-cost evaluation of highway-rail grade crossing improvements at

the corridor level. Inputs are data on rail conditions, highway conditions, and life-cycle costs of the

improvement. The tool explicitly reports the results for each grade crossing and each benefit category (safety,

time savings, vehicle operating costs, reduced emissions, network and local benefits), and the present value of

benefits and costs of alternatives are compared. GradeDec 2000 is available from the Federal Railroad

Administration's web site at: www.fra.dot.gov.

275

Tools for Cross-Modal Evaluation

The FHW A has developed several tools to facilitate cross-modal evaluation: IMPACTS, SPASM and

STEAM. The appropriate use of each model and results from model application will be demonstrated using a

case study of alternatives for the I-15 corridor in Salt Lake City, Utah. IMPACTS was developed to help

screening-level evaluation of multi-modal corridor alternatives for a travel corridor, including highway

expansion, bus system expansion, light rail transit, HOV lanes, conversion of an existing highway facility to a

toll facility, employer-based travel demand management, and bicycle lanes. Inputs are corridor travel demand

estimates by mode for each alternative and unit costs. The impacts estimated include costs of implementation,

induced travel demand, trip time and out-of-pocket cost changes, other highway user costs such as vehicle

operation, parking and accident costs, transfer payments due to tolls, fares or parking fees, changes in fuel

consumption, changes in emissions, and net annual benefits. The IMPACTS spreadsheet and user guide is

available at www.fhwa.dot.gov/steam.

The "Sketch Planning Analysis Spreadsheet Model" (SPASM) is similar to IMPACTS in purpose, inputs

and outputs. The main difference is that it allows easier evaluation of improvements made at the same time to

more than one mode within the travel corridor. SPASM, its user guide and a paper describing its application in a

case study corridor evaluation are also available on the Related Links page at www.fhwa.dot.gov/steam.

The "Surface Transportation Efficiency Analysis Model" (STEAM) produces similar outputs as SPASM

and IMPACTS, but differs in the following ways: (1) it can be used not only for corridor analysis, but also for

systemwide analysis of multiple improvement projects and/or policies across the region; (2) it can be used for

screening-level as well as detailed analysis; and (3) travel demand inputs are needed for the entire region, even

if a corridor alternative is being evaluated, and demand inputs are more detailed -- a loaded regional highway

network is a required input, and zone-to-zone trips by mode are needed as input rather than "average" corridor

trip characteristics (as in IMPACTS and SPASM). STEAM, its user guide and a paper demonstrating its

application are available at www.fhwa.dot.gov/steam.

Tools for Evaluation of Development Effects

Two models, SMITE and SCALDS, have been developed by FHW A to address land use impacts of

transportation. The "Spreadsheet Model for Induced Travel Estimation" (SMITE) is a sketch-planning

model which estimates highway travel due to induced development and evaluates capacity expansion taking

into account travel that may be induced by new development. It is useful where MPO travel models have not

forecasted the full induced demand effects resulting from new development. Inputs are traffic volumes and

capacities on the improved facility and parallel facilities. Outputs are estimates of induced travel, highway

speeds, user benefits, external cost changes and net present value of benefits. SMITE and a paper demonstrating

its application are available on the Related Links page at www.fhwa.dot.gov/steam.

The "Social Cost of Alternative Land Development Scenarios" (SCALDS) model estimates the social cost

consequences of combinations of transportation and land use strategies at a sketch-planning level. Inputs are

276

existing and projected housing mix and regional employment by type, regional aggregate travel projections by

mode, and average local (or default national) infrastructure capital and operating unit costs. Its outputs are

monetary and non-monetary costs associated with urban land development at the metropolitan scale, including

costs for transportation, land consumption, water, sewer, schools, air pollution and energy. SCALDS and a short

paper and report documenting its procedures and demonstrating its application are available on the Related

Links page at www.fhwa.dot.gov/steam.

277

9.5 PAST STUDIES ON TRANSIT EVALUATION

CASE 1: STRATIFICATION APPROACH TO EVALUATION OF URBAN TRANSIT

PERFORMANCE (Sinha, Jukins, and Bevilacqua, TRR 761)

In a period of growing transit operating deficits, increasing attention and concern is being directed at

both the decreasing levels of productivity of transit systems in general and the broad difference in measured

service performance compiled for various transit systems. In making these performance assessments, analyses

have commonly relied on highly aggregated industrywide data and have not given adequate consideration to the

changing and unique operational context within which individual transit systems must function. This paper

presents a stratification approach to the evaluation of urban bus transit system performance. The stratification

scheme was used on the premise that there exist many environmental and policy factors outside the control of

the transit operator that constrain the performance of the transit system. Factors such as area population,

population density, union work rules, system configuration, fleet age, and operational forms have strong

influences on the productivity and efficiency levels of an individual transit service. By implementing the

stratification procedure and compiling temporal data pertaining to both environmental and policy influences and

system performance, the possible bias in making assessments and comparisons of existing transit systems can

be controlled, and changes in performance levels of a system in response to both external changes and

operational improvements can be predicted.

Performance measures based on available operating, financial, and ridership statistics have recently

been considered as criteria for the evaluation of public transit systems. Such measures can provide much insight

into the operation of a particular system. In addition, these measures can be used to examine the differences

among various transit systems and the changes that may occur from year to year. However, the injudicious

application of generic performance indicators in the direct comparison of systems can provide misleading

information about the relative effectiveness of the systems’ operation and service. To compare systems

adequately, it is necessary to adopt an approach that can allow for the unique local environments over which the

operator has limited influence.

Implicit in past research activities on transit performance has been the necessity of addressing the issue

of comparability of productivity elements and procedures by which comparable elements can be defined and

generated (Fielding et al., 1977; USARRS, 1976; UMTA, 1978). Yet little research has been done on the

examination of performance measures on a stratified basis. In general, the existing research focuses on the

performance evaluation in terms of such broad categories as the type of operation (e.g., fixed route versus

demand responsive) and organization (municipal service versus transit district) in order to facilitate the

comparability of system performance for use in managerial analyses (Fielding et al., 1977). The purpose of this

paper is to extend the issue of comparative evaluation of the systems and to examine the external factors that

cause the variations in performance from one system to another.

278

Stratification for Evaluation of Performance

If all bus systems operated in identical environments and under similar policy constraints, their

performance in different locations could then be explained only in terms of variations in level of service.

Presumably, under such conditions, performance differences would then be a function of operator-controlled

variables; thus the operator would have the potential to improve the system’s performance by increasing the

level of service provided.

However, environmental and policy factors external to the local transit operating decisions are not all

alike, and it is the major thesis of this research that some of the ways in which they differ affect the inherent

productivity of the bus systems. Population density, congestion, and network configuration are most often cited

as intruding environmental and policy effects (Fielding et al., 1977 and UMTA, 1978). These and other

elements outside the transit operator’s control can have a significant effect on certain performance indicators

that are used to describe system productivity. Their impact becomes apparent when vehicle mileage, for

example, is used in a productivity measure. Since vehicle mileage is affected by each of the above factors, there

is little doubt that a vehicle mile traveled in Milwaukee is not the same as a vehicle mile traveled in New York

City. The absence of congestion, for instance, may raise the vehicle’s average speed; the result is that more

vehicle miles are driven by a particular driver. The failure to consider such an effect will result in misleading

productivity measures. Consequently, several environmental and policy factors that appear to constrain

performance of transit systems are given below and discussed in the following sections.

Factors Type of Influence Population Environmental Population density Environmental Congestion Environmental Wage rate Policy Local transit policy Environmental and policy Organization type Policy Network configuration Environmental and policy System age Policy

Stratification Factors

The following discussion presents several factors that affect bus transit performance either directly or

indirectly. As such, these factors are potentially useful in establishing a stratification scheme that can be used to

explain the variation in the productivity observed among bus operations. In the discussion of each factor, a basis

for its use and possible variables used to quantify each factor are presented. Table 1 gives a list of stratification

factors and the variables by which the factors are measured. Also included are their respective data sources. For

the purpose of this discussion of an application of the stratification approach, only three factors were chosen,

those for which data are now available on a consistent basis so that the factors can be measured.

279

Table 1: Stratification factors, variables, and data sources

Stratification Factor Variable Data Source

Congestion Average vehicle operating speed System wide measure derived from 1975 American Public Transit Association (APTA) operating report; will also be available from Financial and Reporting Elements (FARE) system; calculated by dividing bus miles by bus hours

Wage rate Average wage per driver Systemwide measure; derived from APTA data and calculated by dividing compensation to operators by operator person hours; wage will be available from FARE

Population Number of people in urban area in which system operates

Obtained from 1975 APTA data; in FARE system this is reported by metropolitan planning organization (MPO)

Population density Population per square mile of land area in urbanized area, central city, and service area

MPO will provide these data to satisfy FARE requirements and will obtain necessary data from system route maps and available census population data

Organization type Qualitative distinction (e.g., municipal transit authority, contract management)

Obtained directly from transit management; not provided by FARE; highly susceptible to error due to varying definitions of management types

Network configuration Qualitative distinction (e.g., radial, grid, circumferential)

Determination made from route maps of transit systems; also susceptible to error due to definitional inconsistencies

Local transit policy Percentage of trips by elderly, percentage of work trips, percentage of elderly population

Elderly population available from census reports; distribution of trips by rider characteristics (age, sex, income, handicap) and by trip purpose (work, shop) will be provided by MPOs from transit-user surveys; available from FARE

System age Years transit system publicly operated; information available from transit systems

Congestion

Transit productivity is closely related to travel speed. As speed increases, a given driver can operate

over more route miles in a given period. To a certain extent, vehicle operating speed can be controlled by the

operator. For example, by increasing or decreasing the number of stops along a route, operating speeds will

decrease and increase, respectively. However, vehicle operating speeds are mainly determined by street-system

characteristics and local traffic policies and perhaps affected only marginally be transit operating policy. In

general, a bus transit management operates its vehicles as fast as traffic conditions permit in order to maximize

vehicle use. Thus it is felt that the average vehicle operating speed actually achieved by an urban transit system

reflects to a large extent the degree of congestion present in the urban area.

Since operating statistics are reported only on a systemwide basis, the measure of speed in necessarily

rough. Nevertheless, incorporation of systemwide average vehicle speed as a stratification variable is defensible

on the ground that if some of the detail is obscured, this is so for all systems in the sample.

Wage Rate

A major component of the operating cost is wages paid to drivers and support staff. In general, labor

costs are responsible for 50-60 percent of the total operating expenditure. However, since labor costs differ by

geographical area, this factor is important for system stratification. The variable used in this analysis is the

average wage per vehicle operator of a transit system.

280

Population

If an indicator of productivity such as passengers per revenue vehicle mile, for example, is used as a

measure of transit performance, consideration must then be given to the population characteristics of the

community the system serves. If levels of service are equal, passenger use has been found to vary directly with

population. Accordingly, total population of the urban area is the variable used in the system stratification

scheme.

Other Factors

As shown in the list above, there are several other factors that can be considered in the stratification

procedure. Such factors include population density, organization type, and network configuration. Local transit

policy is also a critical factor. Although these factors could not be included in the present analysis because of

the absence of relevant data on a consistent basis, a brief discussion of their general significance and possible

effects is presented in the following paragraphs.

Population Density

There is little disagreement that the population as well as the area of a city play some role in both the

provision and the consumption of transit service. Although size of the population alone can give some

indication of ridership levels, additional insight into transit performance can be gained if land area is integrated

into the analysis. In general, transit service is more efficiently and effectively provided in high-density area

(Fielding et al., 1977). Furthermore, transit operational and financial performance is affected not only by

density of residential population, but also by the density and size of nonresidential (i.e., industrial and

commercial) clusters in an urban area (Pushkarev and Zupan, 1977). The significance of such a relationship

permits estimation of the effect of different land-use policies on transit performance.

Organization Type

Identification of organizational structure and management can also contribute to the differences in

transit performance, since they vary with type of service, the area served, and the preferences of local

government. In a study by Bakr et al. (1975), the ability of various types of organizations to undertake tasks

commonly associated with the management of a transit system was examined. It was observed that government-

managed transit systems appeared to perform less effectively due to the lack of necessary support personnel

needed for operation and management responsibilities. On the other hand, transit authorities, by their nature,

were found much more efficient and effective in providing service, since they are much more flexible and

innovative in implementation of management and operational policy. The researchers also concluded that

contract management performs a justifiable role in the current state of development in the public transportation

sector. It can help improve overall performance, since contract management can provide standardized

procedures, planning, and scheduling techniques. Managerial performance can be improved, since expertise is

accumulated from years of experience, which includes extensive experience in labor negotiations.

281

Network Configuration

A bus system’s performance can be affected by the way in which certain geographical, topographical,

and governmental policy constraints force it to develop its network. Network configuration reflects the nature of

a city’s land-use and street-network patterns. Thus the consideration of bus system network layout can offer

insight into aspects of quality and accessibility of transit performance. For example, rectangular networks

generally do not follow desire lines as closely as do radial networks, and consequently more transfers may be

required, which affects the quality of the system.

Local Transit Policy

In addition to the factors described above, local transit policy is a critical determinant of the

performance of transit systems. Since public transportation may exist for different reasons in different cities, it

is only reasonable to consider various policy issues to see what goals and objectives the system is expected to

achieve. For example, it seems that most performance measures do not reflect the extent to which systems serve

special groups such as the elderly and the handicapped. Failure to recognize urban areas that have large

concentrations of such special population groups may result in inaccurate assessments of transit performance. In

general, it would seem reasonable to assume that those transit systems that provide a high level of service to the

elderly and the handicapped would incur higher unit operating costs than do those systems in cities with a small

elderly and handicapped population.

Similar distinctions can be made by disaggregating transit operations with respect to peak-period and

base-period service. In general, a system largely oriented toward commuters would appear to be much more

efficient and effective during peak hours than during the off-peak period. This situation is common for systems

that provide late-night (owl) service.

The number of years the system has been publicly owned can also reflect differences in operating and

financial performance. It might be reasonable to suggest that those systems that have recently made the change

from private to public ownership reflect poor service and passenger use together with low quality of service,

which might be indicated by limited accessibility to the service. Initially, such systems are likely to be

characterized by absence of the strong public and political support needed for successful development of good

management and high ridership.

Although lack of reliable and uniform data does not allow stratification of systems on such bases as

organizational type and network configuration, information on rider characteristics provide by user surveys

required by the Financial and Reporting Elements (FARE) system of the Urban Mass Transportation

Administration can make possible examination of performance on the basis of certain policy variables such as

the percentage of the elderly and the handicapped served.

Stratification Approach

Stratification can be accomplished in many ways. Techniques range from those schemes that stratify

by one criterion -- for example, whether a system is large or mail -- to a level of disaggregation that generates a

282

unique description for each transit system. The method developed in the current study seeks to achieve a level

of stratification that lies between these two extremes.

The approach taken is basically a classical taxonomic procedure. To begin, all systems are included in

one class. Based on the stratification factors selected, for which data are available, systems are stratified to form

cells, or groups of stratification-factor levels. The performance of transit systems within each cell can then be

explained by the stratification factors and factor levels.

The primary variables for the stratification approach presented here include average driver wage,

average vehicle operating speed, and total urban population of the area in which the system operates. These are

variables that not only are reflective of the operating environment under which the systems provide service, but

also are essentially independent of the systems operator’s influence on transit operations. In addition, measures

that reflect system performance are needed, since it is the purpose of stratification to explain variation of

performance from system to system. In the present analysis, resource use indicators were considered, such as

vehicle miles per vehicle, vehicle miles per driver, and revenue passengers per driver. These measures reflect

vehicle use, labor productivity, and labor use, respectively.

In general, each variable can be considered separately, in relation to another, or both to determine

appropriate stratification groups. For instance, in considering total population, the stratification can be initiated

by disaggregating the corresponding variable into two or three intervals. This stratification then results in a

small number of cells; each cell contains many systems. Alternatively, it is possible to stratify in such a manner

that each system defines its own unique cell. The objective in determining which stratification scheme is most

appropriate involves a trade-off between maximizing the extent to which the variability in system performance

can be explained through stratification and maintaining a manageable and practical stratification scheme.

Figure 1 illustrates a simplified schematic stratification scheme. For example, the shaded portion of

Figure 1 indicates those bus systems that can be characterized as having an operational service environment

with low population, high speed (low congestion), and low wage rate. Development of a final stratification

scheme involves the incremental modification of the initial stratification shown in Figure 1. For example, the

scheme may involve changing both the number of factors used and the factor levels (or one of these) until an

adequate stratification is achieved.

283

Figure 1: Schematic representation of stratification scheme.

In order to assess the adequacy of a proposed scheme, one must examine the stratification scheme with

respect to the performance measures; i.e., does the particular stratification scheme adequately explain the

variability in the performance measures? In other words, once cells have been obtained, a check is made to see

whether the measures (computed as the mean of each cell) vary with each cell. To make this determination,

univariate analysis is used to study the behavior of the mean values for the performance measures within each

cell of the stratification scheme. This particular statistical procedure is effective in establishing whether (a)

performance variation can be explained by environmental and policy factors or (b) performance variation

cannot be explained by using such factors.

Statistical Method for the Analysis

The analysis of variance (ANOVA) was applied in this study for development of the stratification

scheme for evaluation of transit system performance. ANOVA is merely a procedure by which the total

variation in the dependent variable is subdivided into meaningful components. For this application, the

dependent variables are the performance measures. The meaningful components are hypothesized as adequate

explanations of the variation in performance for the transit systems sampled.

The statistical criterion that is used to examine the adequacy of the scheme is the R2-value (coefficient

of determination). R2 is simply interpreted as the proportionate reduction of total variation associated with the

independent variables. As the R2-value increases, it will be stated that the variation of the selected performance

measure decreases. The change in the R2-value then becomes the basis for the introduction, modification, or

both of the stratification variables. The level of significance in explaining the total variation of a performance

measure by a stratification scheme was set at 5 percent. Furthermore, the Burr-Foster Q-test (USARRS, 1976)

was used to establish the homogeneity of population variances, which is required by the ANOVA technique.

WA

GE

RA

TE

POPULATION

LOW

H

IGH

HIGH LOW

SPEED

LOW

HIGH

284

Example of Stratification Approach

The three stratification (independent) variables used here include average driver wage, average vehicle

speed, and total urban area population. These variables were disaggregated into several intervals. For the initial

iteration, strata were formed as shown in Table 2. The initial choice of variables and class intervals is a

judgmental decision. However, the major requirement statistically for the formation of class intervals is that

variances among cells formed must be homogeneous.

Table 2: Initial Stratification of Wage, Speed, and Population Variables.

Level Average Wage ($/h)

Average Speed (mph)

Population (000 000s)

Low < 4.75 < 12.5 < 250 Medium 4.75 - 6.00 250 - 750 High > 6.00 >12.5 > 750

The stratification cells formed are shown in Table 3. By using the ANOVA technique, it was found

that the variability in each of the performance measures can be adequately explained by the proposed

stratification scheme. Table 4 gives the values for R2, the criterion used to evaluate the scheme. It can be noted

that this stratification scheme, which uses average driver wage, average vehicle speed, and total urban area

population as independent variables, explains 49 percent of the variation in vehicle use, 50 percent in labor

productivity, and 57 percent in labor use. On the basis of these results, it can be concluded that the stratification

scheme appears to be statistically acceptable. The variation that remains unexplained can be due to a variety of

factors, among which are differences in level of service and inaccuracies in data.

285

Table 3: Results of Stratification by Wage, Speed, and Population Variables

Stratification Variable Performance Measure

Transit Operation Wage ($/h)

Speed (mph) Population

Vehicle Miles per Vehicle

Vehicle Miles per Driver

Revenue Passengers per Driver

Cell 1: Low Wage, Low Speed, Low Population

Central WV TA, Clarksburg, WV 3.65 11.11 28 864 25 218 23 642 47 743 Broome County Transit, Binghamton, NY 4.21 12.46 167 224 35 262 21 322 39 291 Duke Power Company, Anderson, SC 4.38 11.44 27 556 24 116 14 470 20 011 Duke Power Company, Greensboro, NC 3.23 11.77 73 638 29 257 21 672 36 632 Montgomery Area Transit, AL 3.99 11.24 138 983 27 724 21 887 38 850 Mean 25 489 20 599 36 506

Cell 2: Low Wage, Low Speed, Medium Population

Metropolitan Transit Authority, Des Moines, IA 4.37 9.44 255 824 20 785 19 900 -

Cell 3: Low Wage, High Speed, Low Population

Corpus Christi Transit System, TX 3.08 13.42 212 820 56 933 31 882 37 257 City Utilities, Springfield, MO 4.19 12.81 121 340 18 216 21 528 22 211 Greenfield and Montague Transportation Area, MA 3.38 12.69 18 116 18 722 18 722 27 149 Bay County Metro TA, Bay City, MI 4.53 13.52 78 097 41 160 14 316 9 564 Monterey Peninsula Transit, CA 4.06 12.96 93 284 38 071 22 208 16 819 Hudson Bus Lines, Lewiston, ME 3.05 16.40 65 212 11 187 12 065 9 170 Mean 29 953 20 120 20 362

Cell 4: Low Wage, High Speed, Medium Population

Metropolitan Tulsa Transit Authority, OK 3.86 12.70 371 499 34 333 25 328 30 010 Metropolitan Transit Authority, Wichita, KS 3.53 12.99 302 334 37 687 27 961 32 064 Sun Tran-City, Tucson, AZ 4.50 13.01 297 451 40 953 25 173 33 671 Austin Transit Corporation, TX 4.57 13.61 264 499 46 607 21 909 30 425 Central Pinellas Transit Authority, St. Petersburg, FL 3.42 13.86 495 159 47 121 27 487 26 954 City Transit Division, Southeastern Pennsylvania, TA 4.32 15.10 685 942 41 721 16 688 16 954 Mean 39 924 24 091 28 346

Cell 5: Low Wage, High Speed, High Population

North Suburban Mass Transit District, Des Plaines, IL 4.69 14.59 6 714 578 14 409 14 409 17 187 Dallas Transit System, TX 4.68 13.62 1 338 684 31 584 24 076 44 625 Mean 22 997 19 242 30 906

Cell 6: Medium Wage, High Speed, Low Population

Kanawha Valley Regional Transit Authority, Charleston, WV 6.00 13.62 157 662 34 709 28 635 47 983 Lane County Mass Transit, Eugene, OR 5.25 14.49 139 255 48 067 22 531 22 375 South Carolina Electric, Charleston, SC 4.96 12.53 228 399 41 684 24 222 60 514 South Carolina Electric, Columbia, SC 4.89 12.76 241 781 38 982 24 892 53 728 Ft. Wayne Public Transportation Corporation, IN 5.73 13.45 225 184 29 882 24 902 29 898 Madison Metro, WI 5.00 14.02 205 457 22 030 25 041 74 885 Mean 35 634 25 037 48 230

Cell 7: Medium Wage, Low Speed, Medium Population

CNY Centro, Syracuse, NY 4.95 11.55 376 169 25 831 18 806 47 823 Calgary Transit, Alberta, Canada 5.58 11.90 403 319 29 518 - - Capital District TA, Albany, NY 4.80 11.78 486 525 26 975 20 036 43 203 Metro Regional Transit Authority, Akron, OH 4.77 11.96 542 775 29 890 14 945 24 221 Sandwich Windsor, Windsor, Ontario, Canada 5.12 10.17 258 643 28 507 20 332 57 935 Mean 27 113 18 530 43 295

Cell 8: Medium Wage, Low Speed, Medium Population

New Orleans Public Service, Inc., LA 4.72 10.27 961 728 23 792 15 023 59 092 Metropolitan Dade County TA, FL 5.66 10.24 1 219 661 - 21 127 65 794 Rhode Island Public Transit Authority 4.95 10.72 795 311 35 392 20 444 53 005 Mean 29 592 18 864 59 297

286

Table3: Continued.

Stratification Variable Performance Measure

Transit Operation Wage ($/h)

Speed (mph) Population

Vehicle Miles per Vehicle

Vehicle Miles per Driver

Revenue Passengers per Driver

Cell 9: Medium Wage, High Speed, Medium Population

City and County of Honolulu DOT Services, HI 5.45 14.44 442 397 44 490 21 902 66 154

Cell 10: Medium Wage, High Speed, High Population

City of Detroit DOT, MI 4.93 14.63 3 970 584 36 032 26 052 - Metropolitan Atlanta Rapid Transit Authority, GA 5.71 13.91 1 172 778 37 742 25 315 54 255 Metropolitan Transit Commission, St. Paul, MN 5.82 12.91 1 704 423 25 846 21 899 44 089 Metropolitan Transit Authority, Houston, TX 5.61 13.43 1 677 863 40 132 23 304 40 024 Southwest Ohio Regional Transit Authority, Cincinnati, OH 5.87 12.68 1 110 514 28 008 22 353 52 912 Tri-County Metropolitan Transportation District, Portland, OR 5.84 14.57 824 926 39 335 23 953 31 616 Mean 33 040 23 812 44 579

Cell 11: High Wage, High Speed, Medium Population

Tacoma Transit System, WA 7.54 12.60 332 521 27 473 21 105 43 229 Toledo Area Regional Transit Authority, OH 6.30 13.48 487 789 29 190 25 675 62 186 Mean 28 331 23 390 52 707

Cell 12: High Wage, Low Speed, High Population

Bi-State Development Agency, Alton, IL 7.10 12.50 2 987 850 26 081 19 971 35 032 City of Long Beach, NY 6.43 12.15 8 351 266 39 933 24 323 48 596 Milwaukee Transport Services, Inc., WI 7.28 11.76 1 252 457 34 608 19 748 50 412 Niagara Frontier TA, Albany, NY 6.23 10.79 1 086 594 - 17 158 51 600 Regional Transportation District, Denver, CO 6.41 11.80 1 047 311 29 000 16 162 25 252 Chicago Transit Authority, IL 7.70 8.84 6 714 578 36 204 16 383 - Montreal Urban Community Transit Commission 7.27 9.87 2 743 208 24 232 15 888 63 900 Toronto Transit Commission 7.02 12.15 2 628 043 37 921 - - Mean 29 858 18 519 45 799

Cell 13: High Wage, Low Speed, Medium Population

Ottawa-Contra Regional Transit Commission, Ontario, Canada 6.66 11.88 602 510 34 070 21 030 57 055 Winnipeg Transit System, Manitoba, Canada 6.25 10.96 540 262 30 516 17 147 71 752 TA of River City, Louisville, KY 7.41 12.33 739 396 22 141 15 599 34 914 Mean 28 093 17 925 54 574

Cell 14: High Wage, High Speed, High Population

Alameda-Contra Costa TD, Oakland, CA 8.56 14.41 2 987 850 30 003 18 527 33 960 Indianapolis Public Transportation Corporation, IN 6.43 12.86 820 259 22 809 18 200 34 960 San Diego Transit Corporation, CA 7.28 14.47 1 198 323 36 145 23 758 58 135 Southern California Rapid Transit District, Los Angeles, CA 6.28 13.22 8 351 266 30 775 18 576 45 354 Central Ohio Transit Authority, Columbus, OH 6.06 12.58 790 019 27 435 20 558 37 858 Southern Michigan TA, Detroit, MI 9.21 16.60 3 970 584 27 614 26 009 27 660 Transport of New Jersey, Trenton, NJ 7.11 14.24 7 168 164 37 451 26 423 41 812 Mean 30 289 21 722 38 980

Cell 15: Medium Wage, Low Speed, Low Population

Cumberland-Dauphin-Harrisburg TA, PA 5.22 11.41 240 751 17 315 20 595 33 561 Savannah Transit Authority, GA 5.73 11.88 163 753 33 922 22 869 44 974 Tri-State Transit Authority, Huntington, WV 4.96 12.06 167 583 24 491 25 190 36 463 Berks Area Reading TA, PA 5.49 10.00 167 932 27 397 20 091 37 882 Luzerne County TA, Kingston, PA 5.95 12.13 222 830 - 22 045 43 056 Mean 25 781 22 158 39 187

Note: TA = Transportation Authority; DOT = Department of Transportation; TD = Transportation District

287

Table 4: R2 and significance values for stratification evaluation

Stratification Variable Performance Measure R2 Significance

Average driver wage Vehicle miles per vehicle 0.489 0.013a

Average vehicle speed Vehicle miles per driver 0.500 0.012a

Total urban area population Revenue passengers per driver 0.570 0.001a aSignificance at the 0.05 level.

Implications of these results can now be examined in terms of the selected performance measures. For

illustrative purposes, the issue of system comparability is discussed in the following paragraphs.

The types of results shown in Table 3 can be used to make a specific evaluation of the performance of

systems that make up cell 14. Here the performance of an individual system (that of Indianapolis, for example)

can be compared with the mean performance of all the systems in that cell. While this is being done, however,

several questions may arise regarding the remaining variation in performance among systems in that cell. This

can be the result of policy or operational conditions.

If we look first at the vehicle-use measure (vehicle miles per vehicle), Indianapolis has a value (22809)

that falls below the mean (30289). Although the high average operating speed of the Indianapolis system

suggests that the ratio should be higher (since more vehicle miles would be generated at higher speeds), this is

not always the case. In fact, a lower-than-average value may indicate that the system is providing a higher level

of service by using a large fleet for the number of vehicle miles operated. On the other hand, service may be

characterized by low frequency; this results in fewer miles traveled by each vehicle. Consequently, for low

values of this measure, both effective and ineffective use of vehicles can be the case, depending on the

particular characteristics of the system.

Likewise, labor productivity, indicated by vehicle miles per driver, can also be interpreted in different

ways. For instance, due to certain policy constraints, run cuts and schedule problems may not permit efficient

allocation of operating personnel between peak and off-peak periods. A low value (18200) for Indianapolis

Public Transportation Corporation relative to the mean (21722) may suggest such a problem. On the other hand,

the high value for Transport of New Jersey (26423) suggests that this type of policy constraint has less influence

in this system. However, it also suggests that other policy constraints require a system that uses all drivers to

service areas.

Finally, revenue passengers per driver, a labor-use indicator, shows a wide range—from 27660 for

Southern Michigan Transportation Authority to 51835 for San Diego Transit Corporation. In this case, low

patronage may be the result of improper or inadequate route coverage; this may indicate that the system covers

a service area with low transit demand. Low patronage may also be the result of certain service0related policy

constraints. For example, if a system is mandated to provide an extensive service for the transportation

disadvantaged such as the elderly and the handicapped, it may not show a high value for revenue passengers per

driver. Other local transit policies, particularly those associated with fare levels and fare structures, can also

affect the patronage significantly.

288

Effect of Stratification by Wage, Speed, and Population on Other Performance Measures

Since stratification helps to explain the variation in performance among urban transit systems, there

may be certain performance measures that can be better explained by the stratification than those shown in the

previous section. The statistical results are given in Table 5. It can be seen that , for stratification on the basis of

wage, speed, and population, the variation in performance measures such as total operating cost per vehicle,

driver cost per vehicle hour, and revenue passengers per vehicle hour (among others) is explained reasonably

well by this scheme as indicated by R2-value of 0.548, 0.807, and 0.673, respectively. There are also certain

other performance measures (such as driver cost per total cost and operating ratio) that are not significantly

explained by the variables used in this stratification scheme.

Table 5: Effect of stratification by wage, speed, and population on selected performance measures

Performance Measure R2 Significance

Vehicle miles per employee 0.526 0.005a

Vehicle hours per bus 0.427 0.055

Percent peak vehicle use 0.516 0.012b

Total operating cost per vehicle mile 0.672 0.001a

Total operating cost per vehicle 0.548 0.004a

Driver cost per vehicle hour 0.807 0.001a

Driver cost per total cost 0.328 0.263 Total maintenance cost per passenger 0.469 0.032b

Total administrative cost per passenger 0.480 0.062 Total cost per passenger 0.534 0.003a

Percentage of population served 0.341 0.290 Percentage of transfers 0.366 0.248 Revenue passengers per vehicle mile 0.647 0.001a

Revenue passengers per vehicle hour 0.673 0.001a

Revenue passengers per population served 0.581 0.001a

Revenue per vehicle 0.483 0.038b

Revenue per revenue passengers 0.431 0.053 Operating ratio 0.277 0.502 Deficit per passenger 0.450 0.042b

aSignificant at the 0.01 level. bSignificant at the 0.05 level.

Although urban population, wage rate, and average vehicle speed have been identified as the basic set

of stratification variables, there are other variables that are potentially useful in establishing a stratification

scheme, as was indicated in Table 1. It is not possible at this time to consider all these variables, due to the lack

of data. However, similar analyses to this one should be undertaken once the appropriate data become available

on a consistent basis so that those factors that would facilitate the understanding of the variation in performance

form one system to another can be determined. As a preliminary indication of the effect that one variable has on

transit performance, the following section presents the stratification of 51 bus systems by the age of the systems

since they have been publicly operated.

289

Transit Systems Stratified By System Age

Inherent in this discussion is the fact that, depending on the adequacy of the data and the refinement

desired, there are some stratification variables that may be better than others in explaining variation in transit

performance. For example, knowledge of the age of a system since it became publicly operated can add insight

into transit operational performance, as discussed in the section that describes stratification variables.

Table 6: Results of Stratification by System Age

Mean Performance

Performance Measure Public After 1969

Public Since 1969 or Earlier All Systems R2 Significance

Number of observations 26 25 51 - - Vehicle hours per vehicle 2085 2462 2262 0.107 0.024a Operating expense per vehicle ($) 33 512 42 231 37 508 0.129 0.012a

Revenue per passenger ($) 0.38 0.30 0.34 0.104 0.028a

Deficit per capita ($) 4.38 9.54 6.68 0.118 0.018a

Revenue passengers per capita 17.2 36.2 26.1 0.157 0.025a aSignificant at the 0.05 level.

Table 6 gives the results of stratification by system age for the 51 bus systems for which data could be

obtained. The analysis covered the years up to 1975. It should be noted here that the analysis includes only

those systems that operate in the United States. Canadian systems were not considered, since Canada’s

governmental transit programs differ greatly from those under which U.S. systems operate. The results reported

are significant at the 95 percent confidence level. The results show that systems that have become publicly

owned after 1969 show significantly lower vehicle use (2085 vehicle-h/vehicle) than do systems that were

publicly owned in 1969 or earlier (2462 vehicle-h/vehicle). Similarly, passenger use appears lower for younger

systems, as indicated by the low value of 17.2 for revenue passengers per capita as compared to 36.2 revenue

passengers per capita for older system.

Consequently, systems that have been publicly owned since 1969 or earlier (six years or longer) appear

to be more efficient in terms of vehicle and passenger use. Because of strong and continuing political and public

support, such systems have been able both to make certain capital improvements and to implement policies that

provide increasing service levels. Purchase of new buses, for example, has allowed transit systems to provide

longer hours of service with fewer vehicles, whereas certain policy issues such as fare stabilization have helped

to assure continued patronage.

If we consider only the financial efficiencies, however, those systems that have been public since 1969

(less than six years) appear to be more cost efficient. Table 6 shows that operating expense per vehicle, on the

average, is lower ($33512) than it is for older systems ($42231). Not only can these systems operate at lower

unit costs, but they also appear to be more revenue efficient, revenue per passenger is $0.08 more than it is for

the older systems (Table 6). In general, the revenue efficiency of younger systems may be explained by a fare

policy that reflects the momentum of the profit-making objective of privately operated transit systems.

290

It should be noted that, for each of the performance measures presented in Table 6, the R2-value (which

reflect the extent of explanation of variation in transit performance) range between 0.104 and 0.157. These

values indicate, for example, that only 10.4 percent of the variation in vehicle use (as measured by vehicle hours

per vehicle) can be explained by stratifying 51 transit systems according to system age. These results are as

expected, and they suggest that other environmental and policy variables might be more useful in explaining

variation in urban transit performance when considered together with system age.

Conclusions

In this paper, a method has been developed by which certain environmental and policy variables have

been found useful in explaining the biases inherent in transit performance measurement. Through the example

presented, the extent of influence on transit resource use of the elements of wage rate, average operating speed,

and population has been identified.

Since such a procedure appears to explain performance variation adequately, its usefulness in

comparative evaluation is evident. Such evaluations lend themselves to direct comparison of systems within

their respective cells. Bus system performance can be compared against the mean cell values of performance

indicators of similar properties. These mean values constitute a par against which comparisons can be made,

primarily by managers of a transit property.

Stratification is therefore useful in explaining the possible bias in making assessments and

comparisons of bus transit systems. However, it is important to stress that the stratification scheme presented

here is only a beginning. Subsequent analyses that use additional environmental and policy factors will

undoubtedly improve the reliability and validity of the stratification scheme. The stratification approach to

comparing the performance of alternative systems holds promise of being a powerful program-analysis and

system-evaluation tool.

References 1. G.J. Fielding et al., Development of Performance Indicators for Transit: Interim Report. Univ. of California,

Irvine, July 1977.

2. USARRS (Uniform System of Accounts and Records and Reporting System)-Project FARE, Task V, Volume 1:

General Description. Urban Mass Transportation, July 1976.

3. UMTA, Proc., 1st National Conference on Improving Transit System Performance, Norfolk, Virginia. Urban Mass

Transportation Administration, U.S. Department of Transportation, Oct 1978.

4. B.S. Pushkarev and J.M. Zupan., Public Transportation and Land-Use Policy. Indiana Univ. Press, Bloomington,

1977.

5. M.M. Bakr et al., Role and Effectiveness of Contract Management in the Transit Industry. Urban Transportation

Program, Marquette Univ., Milwaukee, WI, 1975.

291

CASE 2: IMPACT OF SHORT-TERM SERVICE CHANGES ON URBAN BUS TRANSIT

PERFORMANCE (Bhandari and Sinha, TRR 1076)

This paper examines the impact on a fixed route of small changes in three operational policy variables:

frequency, number of bus stops, and fare. Analytical expressions are developed that trace the impact of each

variable on various other system variables, which leads to an assessment of changes in selected measures of

efficiency and effectiveness. The application of the methodology is demonstrated by a case study of a selected

bus route in a medium-sized Indiana city. Three specific options are evaluated in terms of alternative frequency,

number of stops, and fare policies. Since none of the options was actually implemented, the paper reports only

on a theoretical analysis of the changes that might be expected under each option. The results indicate that

significant improvements are possible in most of the efficiency and effectiveness measures under all three

options examined. The technique does not require an extensive amount of data or calibration effort; instead it

relies on information generally available from the records of a transit company and reasonable assumptions

where necessary. Much effort is currently being directed toward gaining a better understanding of urban transit

performance.

Under public ownership, transit systems are being subsidized heavily by federal, state, and local funds.

These subsidies are necessary if transit companies are to continue to provide service to the public even when

they cannot recover their operating costs from the farebox. Under these circumstances, if service improvements

are evaluated solely on the basis of cost recovery, few projects, if any, would be implemented. Previous studies

of short-term changes in service have concentrated on ridership, costs, and revenue impacts; little emphasis was

given to their impact on accepted measures of performance.

This paper presents a methodology for relating short-term service changes to changes in selected

measures of performance. Particular reference is made to bus transportation in medium-sized urban area.

Specifically, an examination is made of the effect of changes in three major operational policy variables along a

fixed bus route. These variables are (a) frequency of service, (b) spacing between stops, and (c) basic fare.

The emphasis is on the development of a systematic approach that traces the impact of each policy

variable on various other system variables, which will lead to an assessment of the appropriate performance

measures. The most important aspect is to establish reasonable impact relationships between the policy and the

impact variables as well as relationships among the impact variables themselves.

A number of factors were considered of prime importance and common to the development of the

specific relationships and the overall methodology. First, transit management and transportation planners should

find the procedure simple and quick to apply to provide a reasonable assessment of the impacts. Second, the

relationships developed should maintain a sound theoretical base, but they should not be unduly complex or

require a great deal of modeling and calibration effort. Third, the procedure should not be too restrictive in the

sense of being applicable only to unique situations. In other words, the methodology should be general and

adapt readily to different environments. Last, use of the procedure should not be very costly in terms of data

requirements. Most of the data required should be available from the usual records kept by the transit operators.

292

Identification of Impacts

On any given bus route, the entire spectrum of variables that can be affected directly or indirectly by

changes in the operational policy variables may be grouped as follows:

1. Service variables—Average operating speed, vehicle travel time, walking time, and waiting time;

2. Output variables—Ridership, passenger miles, vehicle miles, vehicle hours, and revenue;

3. Resource variables—Number of buses, number of drivers, operator costs, and user costs; and

4. Performance measures—Cost efficiency (operator and total cost per vehicle hour, operator and

total cost per vehicle mile, operator and total cost per passenger, and operator and total cost per

passenger mile), revenue efficiency (revenue per dollar of operating cost and revenue per vehicle

mile), driver utilization efficiency (vehicle miles per driver pay hour and passengers per driver pay

hour), vehicle utilization efficiency (annual vehicle miles per vehicle and annual passengers per

vehicle), user cost effectiveness (user cost per passenger and user cost per dollar of operating

cost), ridership effectiveness (passengers per vehicle mile, passengers per vehicle hour, passengers

per dollar of operating cost, and passenger miles per seat mile), and other measures (e.g., deficit

per passenger).

The operator costs are the direct cost of bus operation computed as a function of the total vehicle hours

and vehicle miles operated. Hourly costs include driver wages, fringe benefits, and advertising. Distance costs

include depreciation, maintenance, parts, fuel, oil, tires, insurance, tickets and timetables, and right-of-way

costs. The user costs consist of the value travelers place on their walking, waiting, and vehicle travel times.

The measures of performance selected here are those that are influenced most by changes in the policy

variables and are felt to cover adequately major areas of interest. A more complete treatment of performance

measures can be obtained by reference to other studies (Tomazinis, 1975; Fielding et al., 1977; Public

Technology Inc., 1977 and Gerleman, 1977).

Assessment of Impacts

The various linkages among the relevant variables are shown schematically in Figures 1, 2, and 3. The

most important outcome is the change in ridership due to changes in the operational policy variables. This

change occurs as a result of the inherent elastic nature of demand in response to changes in the level of service

characteristics.

Figure 1 shows, for example, that an increase in frequency will decrease waiting time, increase the

average speed, and decrease vehicle travel time, which will result in an increase in ridership. An increase in the

number of stops decreases walking time but also decrease the operating speed, which will cause an increase in

the vehicle travel time. The effect on ridership then depends on the relative elasticities and magnitudes of the

change of waiting time and vehicle travel time.

In the case of a fare change, an increase in fare, for example, will decrease ridership as a direct result

of the negative elasticity of demand with respect to fare. However, this decrease in ridership might improve the

average operating speed and cause a decrease in the vehicle travel time, thereby inducing an increase in

293

ridership. The net change in ridership may still be negative, depending, however, on the relative magnitudes of

these opposing changes. This reverse effect of the change in ridership on the average bus speed is also present

in the case of changes in frequency and number of stops, as shown by the dotted lines in Figure 1.

Figure 2 shows the linkages of the total system costs and revenues. Total costs are the sum of operator

costs and user costs. The change in operator costs is related directly to the change in vehicle hours and vehicle

miles of bus operation: however, change in user costs is a function of the change in the travel time components.

Eventually, interest lies in the effect these changes have on the performance of the system. This is

measured through changes in the appropriate performance indicators obtained via changes in variables such as

ridership, costs, revenues, vehicle miles, and vehicle hours, as shown in Figure 3.

294

Figure 1: Linkages and ridership.

Waiting Time

Ridership FREQUENCY

Average Speed

In-Vehicle Travel Time

Waiting Time

Ridership NO. OF STOPS

Average Speed

In-Vehicle Travel Time

FARE

Average Speed

In-Vehicle Travel Time

Ridership

295

Figure 2: Linkages and system costs and revenues.

Figure 3: Linkages and performance measures.

Ridership Revenue

FREQUENCY

NO. OF STOPS

FARE

Average Speed

Trip Times

Number of Buses

Vehicle-Miles

Vehicle-Hours Operator Costs

Value of Travel Time User Costs

Ridership

Number of Buses

Number of Drivers

Vehicle-Miles

Vehicle-Hours

Passenger-Miles

Operator Costs

User Costs

Revenues

Performance Measures FREQUENCY

NO. OF STOPS

FARE

296

Analytical expressions to represent the various linkages were developed as follows.

Average Operating Speed

The variables that are characteristic along a given bus route are defined below (SI units are not given

for the variables of this model because its operation requires that they be in U.S. customary units.):

L = round trop route length in miles;

Y = number of stops per mile;

Q = average hourly demand (i.e., the number of passengers served along the entire route per

hour);

M = average trip length per passenger in miles;

X = frequency of service in buses per hour;

S* = running speed of bus in miles per hour;

S = average operating speed over the entire route in miles per hour;

ε = time spent per passenger in boarding or alighting from a bus, converted to hours; and

δ = time spent in a topping and starting maneuver, converted to hours.

In addition, the following assumptions are made:

1. Origins and destinations are uniformly distributed along the route,

2. The probability distribution of the number of passengers that board a bus at a given stop follows a

Poisson distribution, and

3. Passengers are equally likely to get off at any stop, and they make their decisions to do so

independently of one another.

Mohring (1972) showed that under these assumptions the total round-trip time may be obtained as

SL / = ( ) ( ) ( )XYLQeYLXQSL /2* 1/2/ −−++ δε (1)

where

L/S = the round-trip time,

L/S* = the running time (i.e., the time spent when the bus is in motion),

2Qε/X = the time spent in loading and unloading passengers,

δYL = the time spent in staring and stopping maneuvers, and

(1-e-2Q/XYL) = the probability that a given stop is made.

Dividing Equation 1 by L,

S/1 = ( ) ( ) ( )XYLQeYXLQS /2* 1/2/1 −−++ δε (2)

297

which gives the desired expression for the average operating speed as a function of demand, frequency, number

of stops, and the running speed of the bus.

In-Vehicle Travel Time (IVTT)

This is simply the average trip length divided by the average speed.

IVTT = SM /60 , in minutes (3)

Walking Time (WKT)

In the absence of specific knowledge about the distribution of actual walking distance, we assume that

the maximum walking distance will be one-half of the distance between stops; therefore, 1/4Y miles can be

taken as the average walking distance. Since walking occurs at both ends of the trip, the total walking distance

per trip is 1/2Y.

If w is the walking speed in miles per hour,

WKT = wY/30 , in minutes (4)

Waiting Time (WTT)

For average waiting delays, the following relationships were used:

WTT = X/30 , for X > 2, in minutes (5a)

WTT = X/148 + , for X < 2, in minutes (5b)

The assumption is that the average waiting time will be equal to one-half the headway for headways

less than 30 min and vary linearly between 15 and 22 min for headways between 30 and 60 min. Headways

greater than 1 hour are not expected.

Ridership

If we assume that the demand function is of the product form with constant elasticities, the new

ridership level (Q1) after a small change (∆) in the service variables can be obtained from

1Q = ( )[ ] ( ) ( )[ ] ( ) ( )[ ]{ }00000 ///1 FAREFAREWTTWKTWTTWKTIVTTIVTTQ ∆+++∆+∆+ γβα (6)

where α, β, and γ are the demand elasticities with respect to vehicle travel time, excess travel time, and fare,

respectively. Subscript zero refers to the level before the change in service variables.

Any change in the operational policy variables (namely, X, Y, and FARE) is analyzed by sequential

solution of Equations 2-6. For greater accuracy, however, the change in X, Y, or FARE is divided into N

smaller increments (positive or negative) and the equations are solved N times.

298

The remaining impact variables are obtained as follows: Let there be n distinct periods during which

any of the variables such as ridership, fare, and frequency may be different, and let i denote the ith such period

where i = 1, 2, … n. The hourly impact variables in the ith period are then obtained as below:

Revenue

(revenue per hour)i = (ridership per hour)i × (FARE)i (7a)

or

REV/hi = Qi · FAREi (7b)

Vehicle Miles

(vehicle miles per hour)i = (frequency)i × (round trip length) (8a)

or

VMPHi = Xi · L (8b)

Vehicle Hours

(vehicle hours per hour)i = (frequency)i × (round trip time)i + layover (9a)

or

Vh/hi = Xi · (L/Si)(1 + LOF) (9b)

where, LOF = the layover time factor as a fraction of round trip time.

Passenger Miles

(passenger miles per hour) = (ridership per hour)i × (average trip length)i (10a)

or

PM/hi = Qi · Mi (10b)

To obtain the values on an annual basis, the hourly values are multiplied by the number of annual hours of the

respective period and summed over all the n periods.

Number of Buses

The number of buses required during any period i is computed as follows:

(number of buses)i = (frequency)i × (round trip time)i (11a)

or

NBUSi = Xi · (L/Si), rounded up to nearest whole number (11b)

299

Number of Drivers

The number of drivers required on any one day is largely a function of run cutting, labor rules, and the

peak to off-peak service ratios. However, a reasonable estimate may be obtained by making certain simplifying

assumptions. Assume, for example, a certain average ratio of the number of pay hours to platform hours

relevant to a particular situation. Let this ratio be denoted as R. Assume also that a driver is paid for an average

of N hours per day. Then, an estimate of the number of drivers required on any day can be obtained from

Number of drivers = (vehicle hours per day × R)/N (12a)

or

NDRVR = (Vh/D · R)/N (12b)

where Vh/D = total number of vehicle hours per day.

Operation Costs

The operating cost in the period i is obtained from

(operating cost per hour)i = a (vehicle hours per hour)i + b (vehicle miles per hour)i (13a)

or

OC/hi = a Vh/hi + b VMPHi (13b)

where a and b are the unit costs of bus operation per vehicle hour and vehicle mile, respectively, for a bus of a

particular size.

User Costs

This cost is taken as a function of the dollar value that users place on their travel time, obtained from

(UC/P)i = V(IVTT)i + ηV(WKT + WTT)i (14)

where

(UC/P)i = the user cost per passenger in the period i,

V = dollar value of vehicle travel time, and

ηV = dollar value of excess time (η ranges generally from two to three).

Hence, the user cost per hour (UC/h) in the period i is obtained as

(UC/h)i = (UC/P)i · Q (15)

300

Application of the Methodology

The application of the methodology to a case study is illustrated here by an examination of a typical

route selected from a transit system in a Midwestern city that has a population of 600 000. First, a comparison is

made of the results obtained by using the relationships developed above with those obtained from records of the

transit operator. Then, an analysis is presented of specific policy alternatives in terms of their impact on

performance.

Route Data and System Information

Most of the information required for the study was available form the transit corporation. The specific

information is given below:

Route selected — English Avenue, route number 10;

Route trip length (L) — 19.1 miles;

Number of stops (Y) — 9.11/mile;

Number of periods (N) — 4 (weekday peak and off-peak, Saturday peak and off-peak); and

Hours of service (weekdays and Saturdays) — peak, 7:00-9:00 a.m. and 3:30-6:00 p.m.; off-peak,

6:00-7:00 a.m., 9:00 a.m.-3:30 p.m., and 6:00-7:00 p.m.;

Running speed (S*) — 27.5 mph;

Average trip length (M) — 0.56 mile;

Fare — $0.50, all periods;

Average loading and unloading time (ε) (computed from small-scale, on-board survey) — 4.66

s/passenger;

Average stopping and staring time (δ) (computed from small-scale, on-board survey) — 19.29 s/stop;

Assumed walking speed = 3 mph;

Assumed value of vehicle travel time (V) — $2.00/h;

Assumed value of waiting and walking time (ηV) — $4.00/h;

Bus size — 47 seats;

Unit cost of bus operation — a = $10.5243/vehicle-h; b = $0.5646/vehicle mile;

Average ratio of pay hours to platform hours (R) — 1.20;

Average pay hours per driver — 9.25/day; and

Assumed demand elasticities — given below (Pratt et al., 1977; Yupo, 1977 and Charles River

Associates, 1968).

Variable Weekday Peak Saturday and Weekday Off-Peak

Vehicle travel time -0.35 -0.45

Excess time (waiting and walking) -0.70 -0.90

Fare -0.20 -0.40

The layover factor (LOF) is computed form 1 + LOF = NBUS/(X· L/S).

301

Headways, ridership, and hours of operation are as follows:

Measure Time Weekday Saturday

Headway (60/x) (min) Peak 20 40 Off-Peak 45 45 Ridership (Q) (passengers/h) Peak 77.11 20.67 Off-Peak 46.47 21.75

Peak 1147.5 234 Annual hours of operation (based on 255 weekdays and 52 Saturdays/year) Off-Peak 2167.5 442

Comparison of Route Performance with System Average

First the model was used to obtain the annual output and resource variables in each of the four periods

considered. The results are summarized in Table 1. The only route-specific data obtainable form the system

records for comparison with those shown in Table 1 were annual weekday and Saturday vehicle miles and

vehicle hours of operation. These values were found to differ by less than 10 percent, as given below.

Table 1: Summary of annual statistics for the base case generated by the model

Weekday Saturday Peak Plus Off-Peak Variable

Peak Off-Peak Peak Off-Peak Weekday Saturday Annual Total

Headway (min) 20 45 45 45

Number of buses 4 2 2 2 4 2 4

Layover factor 0.247 0.291 0.424 0.553

Number of drivers per day 5 4

Annual ridership 88 484 100 724 4 837 9 618 189 208 14 455 203 663

Annual passenger miles 49 551 56 405 2 709 5 386 105 956 8 095 114 051

Annual vehicle miles 65 752 55 198 6 704 11 256 120 950 17 960 135 910

Annual vehicle hours 4 590 4 335 468 884 8 925 1 352 10 277

Annual revenues ($) 44 242 50 362 2 416 4 809 94 604 7 227 101 831

Annual operator costs ($) 85 430 76 787 8 711 15 659 162 217 24 370 186 587

Annual user costs ($) 71 011 136 462 6 209 13 111 209 473 19 320 226 793

Annual total costs ($) 156 441 215 249 14 919 28 769 371 690 43 688 415 378

Annual deficit ($) 41 188 25 426 6 293 10 850 67 613 17 143 64 756

302

Period Annual Vehicle Miles Annual Vehicle Hours Weekday

Actual 125 460 9873 Model 120 950 8925 Difference (%) -3.6 -9.6

Saturday Actual 18 460 1381 Model 17 960 1352 Difference (%) -2.8 -2.1

We were able to obtain data on most systemwide performance measures. A comparison of these with

the route-specific values (obtained by using the model) is given in Table 2. The annual performance values

obtained with the model are in close agreement with the system averages. The difference is less than 15 percent

for all except the passengers per vehicle mile measure, which is about 33 percent below the system average.

Comparison of weekday ridership counts on routes that have comparable service levels showed route 10 to have

a much lower patronage per mile, which probably accounts for the lower route-specific passengers per vehicle

mile value.

Table 2: Comparison of route performance with system average

Route Specific Data Performance indicator Weekday Saturday Annual

System Average*

Efficiency Operating cost per vehicle hour ($) 18.18 18.03 18.16 20.51 Operating cost per vehicle mile ($) 1.34 1.36 1.34 1.60 Operating cost per passenger ($) 0.36 1.69 0.92 0.90 Operating cost per passenger mile ($) 1.53 3.01 1.64 1.62

Total cost per vehicle hour ($) 41.65 32.31 40.42 NA Total cost per vehicle mile ($) 3.07 2.43 2.99 NA Total cost per passenger ($) 1.96 3.02 2.04 NA Total cost per passenger mile ($) 3.51 5.40 3.64 NA Revenue per dollar operating cost ($) 0.58 0.30 0.55 0.55 Revenue per vehicle mile ($) 0.78 0.40 0.73 0.88

Vehicle miles per driver pay hour 10.26 9.34 10.13 NA Passengers per driver pay hour 16.04 7.51 14.85 NA Annual vehicle miles per vehicle 30 238 8980 34 728 36 771 Annual passengers per vehicle 47 302 7228 50 915 50 288

Effectiveness

User cost per passenger ($) 1.11 1.34 1.12 NA User cost per dollar operating cost ($) 1.29 0.79 1.23 NA Passengers per vehicle mile 1.56 0.81 1.47 2.19 Passengers per vehicle hour 21.20 10.69 19.82 23.17 Passengers per dollar operating cost 1.17 0.59 1.09 1.11 Passengers miles per seat mile 0.019 0.010 0.017 0.021

Other

Deficit per passenger ($) 0.36 1.19 0.42 0.41 *Numbers were obtained from a published report of the transit system.

303

An important result to note in Table 1 is the relatively high layover factor in each period. Since this

factor reflects the idle time between successive runs as a fraction of the total round trip, it seems that, if bases

adhere strictly to headways as scheduled, they spend a large fraction of the time laying over between runs—25-

30 percent on weekdays and 42-55 percent on Saturdays. Depending on individual labor contracts and

scheduling constraints, layover times should not be greater than 5-10 percent of the round-trip time for greater

performance efficiency.

Analysis of Specific Options

In order to demonstrate the possible use of the methodology by transit operators, a set of specific

service improvement options was evaluated. These alternatives were formulated as shown below, along with the

existing base case.

Headway (min)

Weekday Saturday

Alternative Peak Off-Peak Peak Off-Peak Number of Stops Fare (cents)

Base Case 20 45 40 45 9.11 50 Option 1 17 36 30 30 9.11 50 Option 2 24 36 30 30 9.11 50 Option 3 24 36 30 30 12.00 50, peak: 40, off-peak

and Saturday

Option 1 represents an improvement in the headways for all periods; the number of stops and fare are

unchanged. Option 2 is the same as option 1, bus headway is increased to 24 min in the weekday peak period.

Option 3 is the same as Option 2, but the number of stops is increased to 12.0/mile and fare is reduced to 40

cents during the weekday off-peak period and all day Saturday.

The results obtained for each option are summarized in Table 3. Option 1 results in a considerable

increase in annual ridership and vehicle miles operated, as well as corresponding increases in revenues and

operating costs. Although the operating deficit increases by $6328, the deficit per passenger decreases from

$0.416 to $0.398. Except for small increases in the operating cost and total cost per vehicle hour, the remaining

cost-efficiency indicators are generally improved and the driver and vehicle utilizations are increased

significantly. The option is also effective in reducing the user cost per passenger and user cost per dollar of

operating cost.

The service cutback in the weekday peak period in option 2 causes ridership to decline relative to

option 1, but it is still higher than the base-case ridership. The most significant impact is a reduction of one in

the number of buses and drivers required during weekdays. As a result annual operating costs are less, and the

deficit is reduced to $76 479 compared to the base-case value of $64 756. There is also further improvement in

304

the driver and vehicle utilization indicators and in the operating cost efficiencies, except for the cost per vehicle

mile.

The main effect of simultaneous reductions in fare and spacing between stops in option 3 is to increase

ridership relative to option 2. Operating costs remain the same due to no change in the number of buses;

however, revenues decrease due to the reduction in fare. As a result, total deficit increases relative to option 2,

but remains less than the base-case value. Option 3 is the most effective in terms of passengers per vehicle mile,

passengers per vehicle hour, passengers per dollar of operating cost, and passenger mile per seat mile. Values of

19.453 passengers/driver-h and 73 675 passengers/vehicle are also the highest under this option.

In general, all three options offer significant improvements in most of the performance indicators. If a

choice were to be made, it would have to be done with due regard to the relative importance of each

performance measure and the magnitude of the trade-offs available.

Conclusion

A relatively simple and quick technique for analysis and assessment of the impacts of major

operational policy variables has been presented in this paper. The technique involves identification of the

impacts and use of simple mathematical relationships to measure them; particular emphasis is on performance.

The applicability of the technique has been successfully demonstrated by a theoretical analysis of options for

transit service improvement in a specific route of a case-study area.

The technique does not require an extensive amount of data collection effort; most of the information

required is generally available from the records of a transit company. However, before it is applied, all of the

assumptions made in the procedure must be considered and modified to suit a specific situation.

305

Table 3: Comparison of alternatives

Impact Variables Base Case Option 1 Option 2 Option 3

Output Annual ridership 203 663 228 568 208 147 221 025 Annual passenger miles 114 051 127 998 116 562 123 774 Annual vehicle miles 138 910 172 176 149 612 149 612 Annual vehicle hours 10 277 10 277 9 129 9 129 Annual revenues ($) 101 831 114 284 104 073 96 308

Resources

Number of buses—weekday 4 4 3 3 Number of buses—Saturday 2 2 2 2 Number of drivers—weekday 5 5 4 4 Number of drivers—Saturday 4 4 4 4 Annual operator costs ($) 186 587 205 368 180 552 180 552 Annual user costs ($) 228 793 227 608 231 668 243 535 Annual total costs ($) 415 378 432 977 412 221 424 088

Efficiency

Operating cost per vehicle hour ($) 18.156 19.983 19.778 19.778 Operating cost per vehicle mile ($) 1.343 1.193 1.207 1.207 Operating cost per passenger ($) 0.916 0.898 0.867 0.817 Operating cost per passenger mile ($) 1.636 1.604 1.549 1.459

Total cost per vehicle hour ($) 40.418 42.131 45.155 46.455 Total cost per vehicle mile ($) 2.990 2.515 2.755 2.835 Total cost per passenger ($) 2.040 1.894 1.980 1.919 Total cost per passenger mile ($) 3.642 3.383 3.536 3.426 Revenue per dollar operating cost ($) 0.546 0.556 0.576 0.533 Revenue per vehicle mile ($) 0.733 0.664 0.696 0.644

Vehicle miles per driver pay hour 10.126 12.551 13.171 13.171 Passengers per driver pay hour 14.847 16.632 18.324 19.458 Annual vehicle miles per vehicle 34 728 43 044 49 871 49 871 Annual passengers per vehicle 50 916 57 142 69 382 73 675

Effectiveness

User cost per passenger ($) 1.123 0.996 1.113 1.102 User cost per dollar operating cost ($) 1.226 1.108 1.283 1.349 Passengers per vehicle mile 1.466 1.328 1.391 1.786 Passengers per vehicle hour 19.817 22.241 22.801 24.211 Passengers per dollar operating cost 1.092 1.113 1.153 1.224 Passengers miles per seat mile 0.017 0.016 0.017 0.018

Other

Annual deficit ($) 84 756 91 084 76 479 84 244 Deficit per passenger ($) 0.416 0.398 0.367 0.381

References 1. A. R. Tomazinis., Productivity, Efficiency and Quality in Urban Transportation Systems. D.C. Heath and Co.,

Lexington, MA, 1975.

2. G. J. Fielding et al., Applying Performance Indicators in Transit Management. Proc., 1st National Conference on

Transit Performance, Urban Mass Transportation Administration, Norfolk, VA, Sept. 1977.

306

3. Public Technology, Inc., Transit System Productivity. Urban Mass Transportation Administration, March 1977.

4. D. Gerleman., Transit Performance, Productivity, and Efficiency. Urban Mass Transportation Administration,

Draft Rept., 1977.

5. H. Mohring., Optimization and Scale Economies in Urban Bus Transportation,. American Economic Review, Vol.

62, No. 4, Sept. 1972, pp. 591-604.

6. R. H. Pratt, N. J. Pedersen, and J. J. Mather., Traveller Responses to Transportation System Changes: A handbook

for Transportation Planners. Federal Highway Administration, Feb. 1977. NTIS: PB 265 830/0G1.

7. C. Yupo., Review and Compilation of Demand Forecasting Experiences: An Aggregation of Estimation

Procedures. Pennsylvania Transportation Institute, Pennsylvania State University, University Park, June 1977.

8. Charles River Associates., An Evaluation of Free Transit Service. Aug. 1968. NTIS: PB 178 845.

SUMMARY OF THE BHANDARI AND SINHA TRANSIT EVALUATION MODEL

- Evaluates short-term impacts of urban bus transit performance

- Examines impact of small changes in 3 operational variables (Frequency, number of bus stops, fare)

- Traces the impact of each variable on other system variables

- Unlike other models, the Bhandari model emphasizes the impact of such changes on accepted

measures of efficiency & measures of effectiveness.

- Provides theoretical methodology behind the INDOT Transit Performance Software package

- Requires that the user be familiar with the following terms:

(a) Service variables: travel time, frequency, fare

(b) Resource variables: wages of drivers, number of drivers, number of buses

(c) Output variables: demand/ridership, number of routes served, revenue

(d) Performance variables: number of on-time trips, revenue/route, revenue/mile, cost/bus.

Model Equations

1. Total Round-Trip Time for a route, T

T = L/S* + (2Qε/X) + δYL(1 – e-2Q/XYL)

where

Service variables Average operating speed Vehicle travel time Walking time Waiting time Bus frequency etc.

Resource variables Number of buses Number of drivers Number of bus stops etc.

Output variables Ridership (demand) Passenger-miles Vehicle-miles Vehicle-hours Revenue User costs Agency costs, etc.

Performance variables Cost efficiency (e.g. $ per vehicle) Revenue efficiency (e.g. annual income

per vehicle) Driver utility efficiency (e.g. vehicle-

mile per driver) Vehicle utility efficiency (e.g. annual

passengers per vehicle) User cost effectiveness Ridership effectiveness

307

L/S* = Total running time 2Qε/X = Time spent in loading and unloading passengers δYL(1 – e-2Q/XYL) = Time spent in starting and stopping total times probability of a given stop L = Round trip length of route (miles) S* = Running speed of the vehicle Q = Average hourly demand ε = Average boarding/alighting time per passenger X = Frequency of service (buses/hr) Y = Number of bus-stops per mile δ = Time spent in a single start/stop 1 – e-2Q/XYL = probability that a given stop in made

2. Average Operating Speed, S

S = ( ) ( )[ ]XYLQeYXLQS /21/2*/1

1−−++ δε

3. In-Vehicle Travel Time

IVTT = ( )( )SspeedoperatingAverage

MlengthtripAverage

4. Average Walking Time (WKT)

WKT = W

Y30

Y = distance between stops

W = walking speed, mph

Assumption:

Maximum walking distance = ½ × distance between stops

= 1/2Y

Average walking distance = ½(½Y) = ¼ Y

Total average walking distance = walking to board + walking after alighting

= ¼ Y + ¼ Y

= ½ Y

Time = D/S = ½ Y/W

5. Average Waiting Time

30/X if X ≥ 2

WTT =

8 + 14/X if X < 2

308

where

WTT is average waiting time in minutes

X is the frequency of service (buses/hr)

6. Demand (Ridership)

The new demand after small changes in the values of the service attributes is:

Q1 = ( )

∆⋅+

++∆

⋅+

∆⋅+⋅

00000 1

FAREFARE

WTTWDTWTTWKT

IVTTIVTTQ γβα

α = the demand elasticity with respect to vehicle travel time β = the demand elasticity with respect to excess travel time γ = the demand elasticity with respect to vehicle fare X0 = the initial values of service variable Xc

7. Hourly Impact Variables

The hourly impact variables in the ith period are obtained as follows:

• (Revenue per hour)i = (ridership per hour)i × (fare)i

• (Vehicle miles per hour)i = (frequency)i × roundtrip length

• (Vehicle hours per hour)i = (frequency)i × roundtrip time + layover

• (Passengers miles per hour)i = (ridership per hour)i × (average trip length)i

• (Number of buses needed)i = (frequency)i × round trip time

• (Number of drivers needed)i = (vehicle hours per day × R)/N

where R = the ratio of pay hours to platform hours

N = average period over which a driver is paid every day, in hours

• (Operating cost in period)i = a(vehicle hours per hour) + b(vehicle miles per hour)

where a & b are the unit costs of bus operation per veh-hour & veh-miles respectively

• (User cost per passenger)i = c(IVTT)i + ηc(WKT + WTT)i

where c = the dollar value of vehicle travel time

ηc= the dollar value of excess travel time (typically η is between 2-3)

309

CASE 3: DEVELOPMENT AND IMPLEMENTATION OF A PERFORMANCE-BASED STATE

TRANSIT SUBSIDY ALLOCATION FORMULA FOR INDIANA

(Sinha, Parsons, Barnum, Sharaf, TRR 1076)

Abstract

This paper presents the findings of a study to develop and evaluate an equitable procedure for

allocating Public Mass Transportation Fund (PMTF) monies in Indiana. Background information, the pre-1985

allocation method, the recommended formula, and the strategy for implementation of the formula are described.

The recommended formula allocates 50 percent of the funds on the basis of a sustenance criterion and the other

50 percent according to performance criteria. The sustenance funds are distributed in proportion to service area

population. The performance funds are allocated using a two-step process. First, all systems are clustered into

four groups, and funds are allocated to group’s funds are allocated to each system within the group on the basis

of its operating ratio, passengers per capita, and passengers per revenue vehicle-mile. Each of the three

performance factors is weighted by the ratio of the system’s locally derived income to the group total locally

derived income. The recommended formula was adopted by the Indiana Department of Transportation (IDOT)

and became effective July 1, 1985. The study was strongly oriented to implementation and therefore

emphasized a close interaction with transit operators at every phase. The key elements of the formula are system

clustering, use of performance indicators, and consideration of locally derived income.

Introduction

As of 1984 there were a total of 30 publicly assisted transit systems in Indiana, including 29 fixed-

routes or demand-responsive systems and 1 commuter rail system, the Northern Indiana Commuter

Transportation District (NICTD). Five of these systems serve areas with populations above 200,000, 12 are in

areas with between 50,000 and 200,000 people, and another 13 are in areas with populations below 50,000.

NICTD provides commuter service to Lake, Porter, La Porte, and St. Joseph Counties along the northern

corridor of the state. A new system, which provides demand-responsive service in Madison County, was added

in 1985. Four of the bus systems now essentially serve counties, and the remaining 26 serve various cities and

some surrounding areas.

In 1984 provision of public transit service in Indiana cost a total of $55.5 million, an increase of 0.8

percent over 1983 levels. The cost of providing a vehicle=mile of service in 1984 increased from $2.85 to $2.94

while the cost per passenger declined from $1.59 to $1.57. The statewide fare recovery ratio, which remained

unchanged from 1982 to 1983, increased form 0.34 to 0.35. The average fare revenue collected per passenger,

however, remained constant at $0.54 (AASHTO, 1984). In 1984 federal operating assistance increased slightly

to $16.8 million while state funding at $10.3 million fell slightly from previous levels. The operating subsidy

per passenger dropped from $0.98 to $0.90 from 1983 to 1984, while the subsidy per revenue-mile increased

from $1.75 to $1.77.

310

Description of the Public Mass Transportation Fund

The Public Mass Transportation Fund (PMTF) is a special revenue fund created by the 1980 Indiana

General Assembly to assist public transportation in the state. The PMTF evolved form a state grant program

established by the legislature in 1975. Part of the 1975 program called for annual general fund appropriations

designed to aid local units of government in matching public transportation grants provided under the federal

Urban Mass Transportation Act of 1964, as amended. It was strictly a program to augment local matching funds

at a time when municipalities were under financial constraints imposed by a property tax control program.

In creating the PMTF in 1980, the Indiana General Assembly changed the funding source form a

general fund appropriation to a dedicated 0.76 percent of the state’s 5 percent general sales and use tax. In

addition to creating a dedicated source of funds, the General Assembly also increased the state’s maximum

proportion of the local share of a federal grant from 1/2 to 2/3. (It should be noted that the PMTF was originally

set at 0.95 percent of the 4 percent general sales and use tax, 100 percent of the 1983 penny increase in the sales

tax was dedicated to the state’s general fund. This requited the PMTF percentage to drop from 0.95 to 0.76

percent.)

The PMTF allows any municipal corporation that receives a federal mass transit grant to apply for state

assistance. Until mid-1985, the Indiana Department of Transportation (IDOT) used and annual list of proposed

projects to gauge total demand for state assistance and made actual allocations based on certain criteria. Service

area population was defined either as the population within the municipal corporation’s taxing unit or as the

urbanized area as defined by the Census Bureau in the current decennial census.

State assistance can be used for both capital and operating purposes. However, because state funds are

limited and localities can bond for the local share of capital projects, most systems opt to use their state

assistance to offset operating costs. For example, in FY 1983 only about 0.3 percent of available state funds was

used to match federal capital grants (AASHTO, 1984).

Problems with the Population-Based Formula

When the state program began in 1975, service area population appeared to be the only generally

agreed on basis for allocating state funds. However, IDOT found that a primary attribute of this factor, stability,

is also its major drawback. Using a formula that relies solely on population rendered the IDOT unable to

respond to changes in the operation and financing of local transit service.

One of the most significant changes was the influence of federal transit assistance. What started out as

a federal formula program that provided modest capital and operating assistance grants to Indiana’s larger urban

transit systems grew into a $30 million annual federal appropriation that now includes the state’s small urban

and rural transit systems.

When, in 1978, congress made federal assistance available to small urban and rural transit systems and

increased assistance to larger systems, the state began to experience increasing demand for state assistance. This

growth in federal assistance caused the state program to expand from 14 to 30 systems within a 5-year period.

These “new” systems lacked the characteristics shared by the earlier recipients. This new group consisted of

311

rural counties and smaller cities providing demand-responsive and low-frequency, fixed-route transit service.

Lacking a suitable alternative for allocating state funds to the new transit systems, the IDOT applied the same

state allocation formula, service area population. Under the formula, numerous small systems were allocated

funds they could not use or legally spend because their allocations were in excess of the 2/3 match limitation

imposed by state statute. On the other hand, other systems, especially the largest ones, experienced immediate

shortfalls.

In 1982 Congress voiced its concern over escalating operating costs by cutting operating assistance to

pre-1982 levels. This meant that many systems were forced for the first time to look more closely at their costs

and revenues. To accommodate the cuts, most of the state’s transit systems raised their fares and reduced

service. Others attempted more sophisticated marketing and service strategies to increase ridership and maintain

control over escalating costs.

In both cases the population-based state allocation method failed to address the changes occurring

locally as a result of reductions in federal operating assistance. Because it merely compared each system with

every other system on the basis of service area population, the state formula did not encourage systems to

improve their performance. The population-based formula perpetuated the role of the state as a passive funding

agent instead of an active participant in defining state objectives that could lead to improvements in transit

performance. Contributing to the magnitude of this problem was the growing importance of state funding in

local transit budgets. When the state program changed from a periodic appropriation to a dedicated fund, annual

receipts jumped from $4.25 million to $9.5 million. After the change, state participation grew to the point where

in 1983 state funds provided 19.4 percent of total operating revenue, eclipsing local tax revenue by 27.9

percent. These developments, coupled with diminishing federal operating funds, created a unique opportunity

for the state to use its resources to leverage improvements in local productivity and play a more significant role

in the provision of public transit service.

Because of the problems associated with a 100 percent population-based formula, a study was

undertaken to examine alternative allocation strategies. The purpose of the study was to identify an allocation

method that would

• Provide an incentive for systems to improve performance,

• Encourage local governments to increase their financial assistance to transit,

• Improve the efficiency of the allocation process by reducing the need to administratively

reallocate funds from systems that could not use the money to those that could, and

• Provide some minimal level of support to all systems.

Study Methodology

In the study, particular emphasis was placed on interacting with transit systems at every phase of the

research. Figure 1 shows the major elements of the study and the points of interaction with transit operators.

312

At the initiation of the study an organizational meeting was held with representatives from IDOT and

general managers and other representatives of transit systems in Indiana. In this meeting issues involving the

state transit subsidy and possible approaches to an equitable allocation scheme were discussed. The objectives

underlying the Indiana PMTF were reviewed and the need for increased transit performance and accountability

was emphasized. The general agreement was that the current state transit subsidy allocation procedure was not

related to either need or performance. It was believed that an improved procedure would be desirable; however,

the procedure had to be equitable.

The next phase of the study involved an examination of the current subsidy allocation procedures at

state and federal levels. A review was also made of possible approached that Indiana could adopt in allocating

state transit funds, including the incorporation of a range of performance indicators.

A questionnaire survey of the transit operators in Indiana was then conducted over the telephone. The

basic purpose of this survey was to identify the perception of transit managers concerning the missions, goals,

and objectives of their systems as well as to determine their thoughts about the effectiveness of the current

allocation procedures and the possible inclusion of performance measures in allocation formulas.

The next phase consisted of the development of a set of criteria along with the identification of

appropriate operating characteristics that could be used in an allocation scheme. A set of fund allocation

formulas based on the information collected in this phase was developed so that the effects of various

combinations of factors in the allocation formulas could be evaluated.

Also considered in this phase was the stratification of transit operators. A major difficulty is that the

operating performance of one system cannot be directly compared with that to other systems. There are certain

biases that are inherent in transit performance measurements because of the effect of several environmental

variables beyond the control of transit operators. Therefore a statistical analysis was employed to subdivide

transit systems into a number of groups so that the systems within a group would have substantially similar

operating environments.

A large number of possible options were developed and evaluated in consultation with the personnel of

the IDOT. A set of 14 options and the list of possible evaluation criteria were then presented to a working

committee comprised of IDOT personnel, the study team, and a representative group of transit managers. The

committee reviewed the proposed operations and developed an evaluation procedure that could be used to select

a formula for adoption. A ranking procedure based on an iterative Delphi approach was used to determine the

specific evaluation criteria to be included and their weights. The committee reduced the number of feasible

options to four. It was decide to survey all of the transit systems about the relative desirability of the four.

A mail-back questionnaire survey of transit systems, in which they were asked to evaluate the relative

merit of four selected options according to the list of evaluation criteria, was conducted. The transit systems

were also encouraged to provide comments and suggestions about the proposed options.

A modeling framework based on goal programming was simultaneously developed to create an

optimization approach to subsidy allocation. The model’s objective was to determine how much to allocate to

313

each transit system, given the constraints of available resources and the need to satisfy stated subsidy allocation

goals. The results of the optimization program were also used to evaluate the four options.

On the basis of the responses of the transit systems and other input, one option was selected. A set of

guidelines was developed, including the criteria, data type, system grouping procedure, and allocation scheme.

The recommended approach was presented to the IDOT and then to the Indiana Transportation Coordinating

Board (TCB). The TCB approved the recommended approach at its meeting of April 11, 1985, and the new

formula took effect on July1, 1985. The aspect of the preceding overview of the development process.

314

Figure 1: Schematic diagram of the major elements of the study.

Organizational Meeting of the Study Team and

Transit Operators

Review of Current Practices

Survey of Transit Operators

System clustering Performance Indicators

Development of Allocation Criteria and Options

Working Session with a Representative Group of Transit Operators

Selection of Evaluation Criteria and Definition of Desirable Options

Survey of Transit Operators

Use of Optimization Model

Selection of Recommended Approach

Monitoring and Feedback

Evaluation of Impacts of Recommended Approach

Recommendation Implementation

315

State Transit Subsidy Allocation Methodologies

There is little uniformity among the states in methods used to allocate transit funds. Judd and Spies

(1983) provide a summary of the state transit aid allocation methodologies. In general, state capital assistance

based on a proportion of the nonfederal share, a proportion of total costs, or both, is given to transit systems.

Methods of allocating operating assistance can be grouped into four major categories: (a) deficit based—a

proportion of costs remaining after revenue and nonstate aid; (b) cost based—a flat proportion of costs, usually

disregarding other aid; (c) multiple factor based—formula using population, ridership, costs, and so on; and (d)

discretionary.

A few states have either used or recently considered using transit performance as one of the factors in

subsidy allocation. In California, transit operators that use state funds made available under the California

Transportation Development Act (TDA) of 1971 are required to have mandated triennial performance audits,

but no penalties or bonuses are involved (Fielding and Glauthier, 1976). In 1980 Pennsylvania enacted

legislation (Act 101) that established a state funding formula that incorporates financial need and system

performance (Miller, 1980). The key performance indicators include ridership increase, revenue increase, cost

increase, and the revenue-to-expense ratio. These system performance indicators are not evaluated by peer

group analyses buy by annual changes in performance of individual transit systems. In Iowa a performance-

based allocation formula was adopted in February 1982. This formula is a modification of the proposed formula

developed in 1978 using ridership and revenue-miles as the basis for transit fund allocation (Forkenbrock and

Dueker, 1979). New York started monitoring the performance of local transit systems in 1979 to carry out a

legislative mandate that required economy, efficiency, and effectiveness on the part of local transit operators

receiving state assistance (New York DOT, 1982).

There are several other states that include some system performance factors in state transit fund

allocation formulas (AASHTO, 1984). However, the current state of the practice of transit subsidy allocation

relies heavily on population-based criteria. Few states include performance as a criterion for subsidy allocation.

Most states that do consider performance allocate only a small portion of their funds on this basis. The major

obstacles to the use of performance as a criterion are the lack of consensus about the type of indictors to be

used, data reliability, and the problem of comparability of performance of systems in diverse operating

environments.

Factors Considered In the Allocation Formula

A review of the various allocation criteria indicated that the two major considerations in the

development of an effective allocation formula that were most applicable to the situation Indiana were

sustenance and performance. Sustenance refers to the broad objective of the state PMTF to provide sufficient

support to individual systems to allow them to maintain a reasonable level of service. On the other hand, the

criterion of performance includes service efficiency, effectiveness, and equity considerations. The way these

two major factors are combined in a formula determines the extent of incentive and innovation that a subsidy

allocation procedure encourages.

316

The size of a service area population, in an aggregated sense, indicates the volume of total service in an

area and recognizes the obligation that the system has to serve the residents within its base. Furthermore,

service area population can be easily computed on the basis of census data. Therefore, service area population

can be used as a reasonable indicator of the sustenance criterion.

Much work has already been done on system performance. Fielding et al. (1977), Sinha et al. (1978),

and Gleason and Barnum (1982) have conducted extensive research on definition and measurement of transit

system performance. Several studies have also been conducted on the use of performance measures in subsidy

allocation. The studies done by Drosdat (1977), Keck and Schneider (1979), Miller (1980), Underwood (1979),

Fielding and Glauthier (1976), Forkenbrock (1979), Crider and Sinha (1982), and Barnum and Gleason (1979)

are examples of some of these effects.

After a series of discussions and evaluations in consultation with IDOT personnel and representative of

transit operators, the following three performance indicators were selected:

1. Operating ratio,

2. Passengers per capita, and

3. Passengers per revenue vehicle-mile.

The operating ratio is the ratio of locally derived income to operating expense. Locally derived income

consists of operating revenue and local assistance including both local government subsidy and private sector

assistance, if any. This factor indicates the self-sufficiency of a system as well as the extent of local

commitment to the provision of transit service.

Passengers per capita is a factor that measures the degree to which residents in a particular area

patronize their transit service. It also reveals the quality of service in the sense that a higher per capita ridership

is an indicator of effective delivery of service.

The third factor, passengers per revenue vehicle-mile, measures service utilization more specifically. It

also represents the efficiency of the operation because it indicates passenger utilization throughout the system.

All data needed to compute these factors are included in the current data reporting system. Operating

revenue and expense data are directly auditable, and the reliability of the ridership and vehicle-mile data can be

readily checked against historical records and records of on-site inspection, and by cross-checking with other

financial data.

Clustering Of Transit Systems

Indiana has a diverse group of transit systems that range from a regional commuter railroad to many

small demand-responsive systems. The operating characteristics of these systems are widely different. The

environment in which these systems operate also varies greatly in terms of land use, street network, existence of

large traffic generators, and other factors. These factors are mostly beyond the control of transit management.

Obviously, all 30 constituent systems cannot be treated as a single group for performance evaluation. An effort

317

was therefore made to subdivide the systems into a number of groups of systems with similar characteristics.

Such a stratification scheme would allow systems of similar types to be compared with each other.

The clustering or grouping of the systems was done by a statistical correlation matrix. Transit systems

were placed in clusters such that elements within a cluster would be homogeneous (similar) and elements

between clusters would be heterogeneous (dissimilar). The variables considered in the clustering analysis

include type of service, service area population, peak-hour fleet size, average system speed, and employee wage

rage.

On the basis of the clustering analysis a stratification scheme consisting of four groups was achieved.

Group 1 includes the relatively large systems in Indiana. Group 2 consists of medium-sized systems, and Group

3 is a collection of small fixed-route, fixed-schedule systems. Group 4 is made up of all demand-responsive

systems, some of which are countywide and serve primarily elderly and handicapped passengers.

Alternative Allocation Formulas

The approach used in developing alternative allocation formulas was to incorporate population and

performance factors in several combinations. Another dimension of these combinations was clustering; that is,

several alternatives were developed with clusters and without clusters. In addition, factors such as deficit,

passenger trips, locally derived income, or population, used to allocate funds among four clusters, provided yet

another dimension. Furthermore, several alternative formulas were developed with factors such as population

density as well as with various weights for each of the factors considered. Again, some alternatives included

different definitions of deficit, locally derived income, and operating ratio. These combinations produced a

large number of alternatives that were screened for reasonableness. The primary parameter used to do the first

screening was the deviation of the individual system’s allocations from that which resulted form the population-

based allocation formula. In addition, the resulting group total of allocations by cluster was also used as a

control for assessing the reasonableness of an alternative. Another consideration was total allocation by

geographic area, particularly for the six systems located in northwestern Indiana.

Evaluation of Proposed Options

A mail-back questionnaire survey of transit operators was conducted to assess the evaluation of the

four final options. Each system was requested to complete a separate evaluation from for each of the dour

options. There were 22 responses received from the 30 operators in the study, a sample of about 70 percent. The

respondents were asked to score each option as “desirable,” “neutral,” or “undesirable” on the basis of the

following criteria:

1. Sustenance: The higher the proportion of subsidy that is based on population, the higher the

guaranteed amount;

2. Administrative practicality: An option’s impact on data collection costs for the systems,

administrative costs for the state, auditing costs, consistency of variable definitions, and simplicity of

procedure for collecting the data, and simplicity of the formula.

318

3. Fund reallocation: The present formula requires a high degree of fund reallocation because it is not

closely related to the needs of the systems;

4. Comparability: The extent to which similar systems are grouped together;

5. Controllability: The degree to which system can be made to operate in a particular manner.

6. Subsidy predictability and variability: The ease of predicting future funding levels and potential

fluctuations in the annual subsidy;

7. Impact on service: Possible impact of an option on such variables as passenger trips, passenger-miles,

percentage of population served, revenue vehicle-miles, maintenance of fare levels, and service to the

transportation disadvantaged;

8. Operational efficiency: Possible impact on such ratios as cost per passenger trip, cost per passenger-

mile, cost per revenue vehicle-mile, operating ratio, and cost per transportation-disadvantaged trip; and

9. Equity: The fairness of the allocation of state funds, considering measures of equity such as subsidy

per passenger trip, subsidy per revenue vehicle-mile, (state subsidy) / (local subsidy + operating

revenue), and subsidy per capita.

A detailed statistical analysis of the responses was performed. The appropriate tests used were the

distribution-free test and the distribution-free multiple comparisons based on Friedman rank sums (Sinha et al.,

1985 and Hollander and Wolfe, 1973). The favored option involved allocating 50 percent of the funds on the

basis of population before clustering and the rest on the basis of performance factors after clustering the

systems. Under this option, group amounts would be determined according to total group need measured in

terms of the difference between operating expense and operating revenue. After analyzing the responses and

further evaluation of the options in terms of the stated criteria, the option discussed in the following section was

considered the optimal solution.

Recommended Formula for PMTF Funds Distribution

The following approach was recommended for distribution of PMTF funds among constituent systems:

1. Distribute 50 percent of the total available PMTF funds directly to each transit system according

to the service area population of each system.

2. Divide the remaining 50 percent of the funds into four group amounts according to the subsidy

needs of each group in relation to the total statewide subsidy requirement (measured by the group

summation of operating expense minus operating revenue):

Group allocation = 0.5 × PMTF × (Group deficit/Total statewide deficit)

3. Suballocate each group amount among constituent systems within a group equally in proportion to

the following three factors:

319

∑ ××n

LDIOR )LDI OR(/)(

∑ ××n

LDIPC )LDI PC(/)(

∑ ××n

LDIPRVM )LDI PRVM(/)(

where OR = operating ratio = locally derived income + operating expense,

LDI = locally derived income = operating revenue + local government assistance + private contribution, PC = unlinked passenger trips per capita of service area population, PRVM = unlinked passenger trips per revenue.

Results of Recommended Approach

Because the 1984 operating data were available in May 1985, the recommended approach was applied

to allocate PMTF funds for FY 1986 ($12,399,870) on the basis of the 1984 data. The input variables for each

system are given in Table 1. Service area populations were determined by examining the actual area served by a

system. Other information was obtained from the 1984 Annual Report of the IDOT (1985). A new transit

system in Madison County was initiated in 1985. Because no operating data were available, only service area

population was used to allocate FY 1986 funds for this system.

The results of the application of the recommended approach to 1986 PMTF allocation are given in

Table 2. In addition to the total allocated amounts, the percentage of funds assigned to each system on the basis

of population and performance indicators is also presented. For example, in FY 1986, the total allocation for

Indianapolis is $3,784,814 of which $1,742,907 (46 percent) is derived on the basis of population and

$2,041,907 (54 percent) is the result of performance. On the other hand, only 36 percent of Gary’s total

allocation of $1,027,107 is derived on the basis of population and 64 percent results from performance.

In Table 3 are given the equity values that result from the recommended formula. It can be seen that

the recommended formula succeeds in achieving an overall parity in equity factors within each group. Although

there are still some discrepancies, the wide variation that existed earlier has been significantly reduced.

In 1986 the systems in Group 1 will receive 62 percent of the total funds, and the systems in Groups 2,

3, and 4 will receive 23, 8, and 7 percent, respectively. The allocation using only population figures for Groups

1, 2, 3, and 4 would have been 56, 26, 10, and 8 percent, respectively. The use of performance measures thus

would move about 6 percent of the total funds from medium-sized and small urban and rural systems (Groups 2,

3, and 4) to the state’s five largest systems (Group 1). It should be recognized, however, that many of the

smaller systems could not use their full allocation in the past because of lack of need or because they reached

the maximum 2/3 limitation on PMTF participation.

Although the recent removal of the 2/3 maximum limitation by P.L. 372 of the 1985 Acts of the

Indiana General Assembly coincided with the completion of the study, the new formula would have greatly

320

streamlined reallocation needs. However, the most important aspect of the new formula is that the PMTF

allocation is now tied to actual needs and system performance.

The new formula satisfies the state objective of providing a minimum level of sustenance while

encouraging performance improvements. Furthermore, by placing added emphasis on locally derived income

within the formula, the new formula should have a positive effect on reversing the recent trend of declining

local participation in operating costs and on encouraging transit systems to seek out private-sector involvement.

321

Table 1: Transit System Characteristics, 1984

System Population Passenger Trips

Revenue Vehicle- Miles

Passenger Trips per Capita

Passenger Trips per RVM

Operating Ratio

Local Assistance ($)

Fare Box + Other Revenue ($)

Locally Derived Income ($)

Deficit ($)

Total Operating Expenses ($)

Group 1 Fort Wayne 236479 1770200 1245551 7.49 1.421 0.52 1161676 1040017 2210693 3191561 4231578 Gary 151953 3823782 1492864 25.16 2.561 0.65 321132 2708798 3029930 1926797 4635595 Indianapolis 711539 15493382 6204178 21.77 2.497 0.57 1874497 8567417 10441914 9708994 18276411 South Bead 149928 4456216 1727939 29.72 2.579 0.52 1239933 1222910 2462843 3527343 4750253 NICTD 171371 2248795 1526032 13.12 1.474 0.53 0 5472680 5472680 4780746 10253426

Group Total 1421270 27792375 12196564 4597238 19011822 23618060 23135441 42147263 Group average 284254 5558475 2439313 19.56 2.279 0.56 919448 3802364 4721812 4627088 8429453

Group 2 Anderson 66910 339185 359285 5.07 0.944 0.28 181325 99480 280805 886650 986130 Bloomington 52044 308455 282324 5.93 1.093 0.44 234660 114653 349313 684718 799371 Evansville 130496 1540797 687678 11.81 2.241 0.49 135423 510176 645599 812541 1322717 Lafayette 91380 1147401 858369 12.56 1.337 0.46 326326 425350 751676 1219968 1645318 Muncie 77216 1376901 721978 17.83 1.907 0.50 374726 500718 875444 1260887 1761605 Terre Haute 63931 500360 492887 7.83 1.015 0.37 104689 198024 302713 624244 822268 Hammond 93714 355822 297198 3.80 1.197 0.31 142918 130018 272936 746248 876266 Southern Indiana 73487 184165 182325 2.51 1.010 0.27 87982 71057

159039 527890 598947

Group total 649178 5753086 3882044 1588049 2049476 3637525 6763146 8812622 Group average 81147 719136 485256 8.86 1.482 0.41 198506 256185 454691 845393 1101578

Group 3 Columbus 30614 179264 233483 5.86 0.768 0.32 36935 49889 86824 221607 271496 East Chicago 39787 409252 149219 10.29 2.743 0.17 98870 0 98870 593223 593223 La Porte 21796 113826 211831 5.22 0.537 0.35 38468 65279 103747 230808 296087 Marion 35874 134923 142580 3.76 0.946 0.27 40677 34104 74781 244062 278166 Michigan City 36850 219150 176818 5.95 1.239 0.33 48631 71074 119705 291788 362862 Richmond 41349 164647 218747 3.98 0.753 0.44 32662 95178 127840 195972 291150 Washington 11325 24481 32223 2.16 0.760 0.39 3707 8071 11778 22241 30312 New Castle 20056 98138 122545 4.89 0.801 0.24 45073 26440 71513 270440 296880 Bedford 14410 35591 90940 2.26 0.358 0.25 21489 14220 35709 128936 149156

Group total 252061 1379272 1378386 366512 364255 730767 2199077 2569332 Group average 28006 152919 153154 5.46 0.998 0.29 40724 40473 81196 244342 284815

Group 4 Goshen 19665 8812 21763 0.45 0.405 0.36 3212 5979 9191 19272 25251 Kosciusko 29778 77051 140028 2.59 0.550 0.29 47094 46940 94034 282565 329505 LCEOC 25711 161732 481552 6.29 0.336 0.37 64985 178471 243456 486135 664606 Monroe County 25557 38229 163196 1.50 0.234 0.22 43354 18947 62301 260122 279069 Trade Winds 25711 108861 367057 4.23 0.297 0.23 20609 70610 91219 318105 388715 KIRPC 38119 56640 197864 1.49 0.286 0.23 49048 26120 75168 294288 320408 Union County 3430 9508 34857 2.77 0.273 0.24 8731 4912 13643 52384 57296 Mitchell 4641 9004 13681 1.94 0.658 0.25 6140 4239 10379 36843 41082 Madison County 36213

Group total 208825 469837 1419998 243173 356218 599391 1749714 2105932 Group average 23203 58730 177500 2.53 0.331 0.28 30397 44527 74924 218714 263242

Grand total 2531334 35394570 18876992

322

Table 2: 1986 PMTF Allocations Using the Recommended Formula

Distribution of PMTF by 1986 PMTF Performance

Allocation ($)

Percentage

Population (%)

OR (%)

PT/Capita (%)

PT/RVM (%)

Total Performance (%)

Group 1 Fort Wayne 837,403 6.75 69.17 14.58 5.98 10.27 30.83 Gary 1,027,107 8.25 36.24 20.45 22.55 20.77 63.76 Indianapolis 3,784,814 30.52 46.05 16.77 18.24 18.94 53.95 South Bead 900,731 7.26 40.77 15.16 24.68 19.39 59.23 NICTD 1,168,897 9.43 35.91 26.46 18.66 18.97 64.09

Group Total (average) 7,718,897 62.22 45.10 18.30 18.30 18.30 54.90

Group 2 Anderson 218,926 1.77 74.86 9.44 6.72 8.99 25.14 Bloomington 217,582 1.75 58.59 18.56 9.83 13.02 41.41 Evansville 588,883 4.75 54.28 14.11 13.37 18.24 45.72 Lafayette 486,779 3.93 45.98 18.66 20.03 15.32 54.02 Muncie 589,314 4.75 32.09 19.51 27.36 21.03 67.91 Terre Haute 233,310 1.88 67.11 12.61 10.49 9.78 32.89 Hammond 286,736 2.31 80.05 7.75 3.74 8.46 19.95 Southern Indiana 207,322 1.67 86.82 5.44 1.99 5.75 13.18

Group Total (average) 2,828,853 22.82 (56.20) (14.60) (14.60) (14.60) (43.80)

Group 3 Columbus 119,707 0.97 62.64 13.70 14.18 9.49 37.36 East Chicago 187,509 1.51 51.97 5.29 18.10 24.64 48.03 La Porte 102,375 0.83 52.15 20.93 17.64 9.28 47.85 Marion 121,218 0.98 72.49 9.83 7.74 9.95 27.51 Michigan City 162,605 1.31 55.51 14.34 14.61 15.55 44.49 Richmond 167,837 1.35 60.34 19.78 10.11 9.77 39.66 Washington 32,822 0.26 84.51 8.26 2.59 4.65 15.49 New Castle 80,674 0.65 60.89 12.55 14.46 12.10 39.11 Bedford 45,435 0.37 77.68 11.60 5.92 4.80 22.32

Group Total (average) 1,020,183 8.24 (60.04) (13.32) (13.32) (13.32) (39.96)

Group 4 Goshen 52,212 0.42 92.25 3.79 0.35 3.61 7.75 Kosciusko 126,176 1.02 57.80 12.94 8.49 20.77 42.20 LCEOC 225,669 1.82 27.91 23.89 29.85 18.35 72.09 Monroe County 82,302 0.66 76.06 9.97 5.00 8.98 23.94 Trade Winds 106,208 0.86 59.29 11.82 15.98 12.90 40.71 KIRPC 119,537 0.96 78.11 8.66 4.12 9.12 21.89 Union County 13,908 0.11 60.40 14.09 11.95 13.55 39.60 Mitchell 17,265 0.14 65.84 8.99 5.13 20.04 34.16 Madison County 88,696 0.72 100.00 0 0 0 0

Group Total (average) 831,974 6.72 (61.84) (12.84) (12.84) (12.84) (38.52)

All groups (average) 12,399,870 100.00 (50.00) (16.67) (16.67) (16.67) (50.00)

323

Table 3: Equity Values Resulting from the Recommended Formula, 1986

Equity Factora 1 2 3 4 5 Group 1

Fort Wayne 0.4731 0.6723 0.6026 0.1979 3.5411 Gary 0.2686 0.6880 0.6290 0.2216 6.7594 Indianapolis 0.2443 0.6100 0.5793 0.2071 5.3192 South Bead 0.2021 0.5213 0.4974 0.1896 6.0078 NICTD 0.5198 0.7659 0.7326 0.1140 6.8205

Group 2 Anderson 0.6454 0.6093 0.5351 0.2220 3.2719 Bloomington 0.7054 0.7707 0.7371 0.2722 4.1807 Evansville 0.3822 0.8563 0.8458 0.4452 4.5127 Lafayette 0.4242 0.5671 0.5630 0.2959 5.3270 Muncie 0.4280 0.8162 0.8012 0.3345 7.6320 Terre Haute 0.4663 0.4734 0.4231 0.2837 3.6494 Hammond 0.8058 0.9648 0.8683 0.3272 3.0597 Southern Indiana 1.1257 1.1371 1.0826 0.3461 2.8212

Group 3 Columbus 0.6678 0.5127 0.5006 0.4409 3.9102 East Chicago 0.4582 1.2566 1.1596 0.3161 4.7128 La Porte 0.8994 0.4833 0.4742 0.3458 4.6970 Marion 0.8985 0.8502 0.8264 0.4358 3.3970 Michigan City 0.7420 0.9196 0.9024 0.4481 4.4126 Richmond 1.0194 0.7673 0.7513 0.5765 4.0590 Washington 1.3407 1.0186 1.0186 1.0828 2.8982 New Castle 0.8220 0.6583 0.6384 0.2717 4.0224 Bedford 1.3941 0.4996 0.4955 0.3174 3.1530

Group 4 Goshen 5.9251 2.3991 2.3991 2.0677 2.6551 Kosciusko 1.6376 0.9011 0.5853 0.3829 4.2372 LCEOC 1.3953 0.4686 0.4532 0.3396 8.7771 Monroe County 2.1529 0.5043 0.4236 0.2949 3.2203 Trade Winds 0.9756 0.2893 0.2604 0.2732 4.1308 KIRPC 2.1105 0.6041 0.5093 0.3731 3.1359 Union County 1.4628 0.3991 0.2930 0.2427 4.0548 Mitchell 1.9175 1.2620 1.2464 0.4203 3.7202

aEquity factors: 1 = state subsidy per unlinked passenger trip, 2 = state subsidy per revenue vehicle-mile, 3 = state subsidy per revenue vehicle-hour, 4 = state subsidy/total revenue, and 5 = state subsidy per capita.

324

Implementation

The recommended formula was presented to the Indiana Transportation Coordinating Board on

February 14, 1985. The board approved the formula at its meeting on April 11, 1985. The study team, in

cooperation with the IDOT, held a 1-day workshop on May 22, 1985, for transit operators and local

transportation officials to disseminate the essential elements of the study and to discuss the details of the

recommended formula and how it would be implemented. Discussion was also held on what transit systems

can do to maximize their shares of state subsidy as well as to optimize their overall financial situation.

The IDOT intends to use the approved formula to allocate PMTF funds received after July 1,

1985. The formula would incorporate operating and financial data from the most recently completed

calendar year. Under the IDOT’s current reporting system, each transit system is required to complete a

detailed annual report questionnaire by February 13. In most cases, the annual report is based on unaudited

information. However, each report is subjected to a comprehensive review and any significant variations

between current and past reports and any significant inconsistencies within the report are noted and

resolved with the transit system.

IDOT would make the actual allocation no less than 1 month before the beginning of the affected

state fiscal year. Each system would have the option of using its allocation for capital or operating

purposes. However, to maintain the integrity of the performance allocation, any excess funding available

for reallocation would not be made available for operating purposes.

Because the recommended formula represents a significant departure from the existing method,

the IDOT will conduct an evaluation before allocating state FY 1987 funds. This will enable the IDOT to

gauge the relative effect of any major changes in the formula’s data inputs on the allocation process and, if

necessary, propose formula modifications to the Transportation Coordinating Board. Particular attention

will be given to the group allocation of the PMTF. If the deficit figures show fluctuations, alternative

procedures will be examined to determine the group amounts, including the assignments of group

percentages, on an administrative basis.

Conclusions

The recommended formula became effective July 1, 1985. A successful implementation will

require efforts from both the IDOT and the individual operators. The IDOT needs to develop an improved

procedure to check the reliability of the reported annual operating data. For that, each variable must be

uniformly defined and measured. There exist sufficient historical data to perform longitudinal analysis of

an individual system’s reported data for a particular year. Nevertheless, it may be desirable to actually audit

individual systems.

The IDOT must also carefully monitor the impact of the recommended formula. If the results do

not indicate that the intended objectives are being achieved, necessary modifications should be made. An

important aspect of this monitoring process should be an evaluation of what adjustments individual systems

undertake to achieve greater operating performance and to seek out increased local assistance.

325

The IDOT should provide technical assistance to individual systems in their efforts to improve

their operation. Much help can be rendered, particularly to small systems that generally do not have access

to a variety of management tools that can be used in making decisions on operating improvements.

The individual systems should, at the same time, seize the opportunity offered by the change in the

state subsidy allocation formula to improve their operation. Much can be achieved in the area of service

innovation, and transit systems must seek out alternative ways to improve service and reduce costs. Private-

sector participation in transit service delivery may be one such possibility. Most important, management

decisions must be evaluated in terms of sound financial considerations and the goal of being as self-

sufficient as possible must be set.

With continuation of the mutual cooperation that has existed between the IDOT and individual

transit operators in Indiana during the past decade, it can only be expected that the recommended change in

the state subsidy allocation procedure will point a positive direction for the transit industry in Indiana. In

addition to achieving the intended objectives of improved transit service and increased financial efficiency,

an important effect of the new subsidy allocation procedure should be to make the Indiana transit subsidy

program more accountable and thus in the long run legislatively more viable.

References 1. 1984 Survey of State Involvement in Public Transportation. AASHTO, Washington, D.C., Sept 1984.

2. L. Judd and R. Spies. State Urban Transit Assistance Programs. Division of Planning and Budget and

Division of Transportation Assistance, Wisconsin Department of Transportation, Madison, March 1983.

3. G.J. Fielding and R.E. Glauthier. Distribution and Allocation of Transit Subsidies in California. Transit

Management Program, School of Sciences and Institute of Transportation Studies, University of California,

Irvine, Sept. 1976.

4. J.H. Miller. The Use of Performance-Based Methodologies for the Allocation of Transit operating Funds.

Traffic Quarterly, Vol. 34, Oct. 1980, pp. 555-585.

5. 1982 Iowa Administrative Code. Transportation (820), 09 Public Transit Division, Articles A and B, March

1982.

6. D.J. Forkenbrock and K.J. Dueker. Transit Assistance Allocation. Transportation Research, Vol. 13A, No. 5,

1979, pp. 317-327.

7. Report on Transit Operating Performance in New York State. Transit Division, New York Department of

Transportation, Albany, 1982.

8. G.J. Fielding, R.E. Glauthier, and C.A. Lave. Development of Performance Indicators for Transit. DOT

Report CA-11-0014. University of California, Irvine, Dec. 1977.

9. K.C. Sinha and A.S. Bhandari. A Comprehensive Analysis of Urban Bus Transit Efficiency and Productivity,

Part 3: Analysis of Options to Improve Urban Transit Performance, CE-TRA-78-2. School of Civil

Engineering, Purdue University, West Lafayette, Ind., Dec. 1978.

10. J. Gleason and D. Barnum. Toward Valid Measures of Public Sector Productivity: Performance Indicators in

Urban Transit. Management Science, Vol. 28, No. 4, April 1982, pp. 379-386.

11. H.A. Drosdat. Transit Performance Measures: Their Significance in Local Funding Allocation. Ph.D.

dissertation. University of Washington, Seattle, 1977.

326

12. C.A. Keck and N.R. Schneider. Efficiency, Economy and Effectiveness: The Development and Application

of Multimodal Performance Measures for Transit Systems in New York State. Preliminary Report 22. State

Planning and Research Section, Planning and Research Bureau, New York State Department of

Transportation, Albany, Nov. 1979.

13. W. Underwood. Performance Indicators: A Necessary Management Tool? Transit Journal, Vol. 5, No. 1,

Winter 1979.

14. D.J. Forkenbrock. A Method for Distributing Federal Small Urban and Rural Transit Assistance at the State

Level. Technical Report 121. Institute of Urban and Regional Research, University of Iowa, Iowa City, June

1979.

15. D.A. Crider and K.C. Sinha. Use of Performance Case Study. Presented at the 61st Annual Meeting of the

Transportation Research Board, Washington, D.C., Jan. 1982.

16. D. Barnum and J. Gleason. Measuring the Influence of Subsidies on Transit Efficiency and Effectiveness.

NTIS PB-80-189269. National Technical Information Service, Springfield, Va., 1979.

17. K.C. Sinha et al. Development of an Approach for the Allocation of the Public Mass Transportation Fund of

Indiana. CE-TRA-85-3. School of Civil Engineering, Purdue University, West Lafayette, Ind., July 1985.

18. M. Hollander and D.A. Wolfe. Nonparametric Statistical Methods. John Wiley and Sons, Inc., New York,

1973.

19. 1984 Annual Report: Indiana Public Transportation. Draft. Indiana Department of Transportation,

Indianapolis, May 1985.