140
626

 

.THS.

 

 

LIBRARY
Michigan State
University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/Dat90ue.p65«p. 15

 

STATEWIDE MULTI-MODAL TRANSPORTATION MODELING:
AN EVALUATION OF THE STANFORD RESEARCH INSTITUTE'S

INTERCITY DEMAND/MODAL SPLIT MODEL

BY

Sam L. Wallace

A PLAN B PAPER

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER IN URBAN PLANNING
School of Urban Planning and Landscape Architecture

1979

W' , ‘ .
{DAJE fiii U A;

M ti tiiAL
EH BA CKOF BOOK

TABLE OF CONTENTS

PAGE

ACKNOWLEDGEMNTSOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOCOOOOOOij—i

INTRODUCTION...‘0.00.00.00.00...OOOOOOOOOOOOOOOOOOOOOOOO l

I. THE EVOLUTION OF STATEWIDE MULTI-MODAL MODELING..... 3

II. THE STANFORD RESEARCH INSTITUTE'S INTERCITY
DEMAND/MODAL SPLIT MODELCCOCOCOOCOOCOOOCOIO00.......11

III. THE MICHIGAN DEPARTMENT OF TRANSPORTATION'S
STATEWIDE MODELING SYSTEMOOOOOOOOOOOO0.0.0.00000000017

IV. MODELING PROCEDURES AND RESULTS

1.
2.
3.

4.

Documentation of Procedures and Analyses........23
Rail Results and Analyses.......................26
Bus Results and Analyses........................41

Aviation Results and Analyses...................53

V. SUMMARYOOOOOIOOOOOOO000......0.0.0.000...0.0.0.0....65

CONCLUSIONS...’OOOOOOOOOOOOOOOO0.0.00.0000000000000066

FUTURE RESEARCH NEEDSOOOOOO0.0...0.0.00.00000000000067

FOOTNOTESOOOOOOOOCOCOOOOOOOOOOOOOOOOOOOOOOOOOOO000......69

SELECTED BIBLIOGMPHYOOOOOCOOOOOOO0.0.0.000000000000000071

APPENDICES

HIGHWAY COST CAIICULATIONSOO00.00.000.000...oooooooooA-l

DATA SET FOR MIL MODELINGOOOOOOOOOOOOOO0.0.0.0...COB-1

DATA SET FORBUS MODELINGOOOOOOOOOOIOOO0.0...OOOOOOOB-4

DATA SET FOR AVIATION MODELING......................B-5

AVIATION SURVEY DATAoooooooooooooooocoo-000.000....oC-l

mmNETWORKOOOOOOOCOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOD

BUS ORIGINS AND DESTINATIONS . O C O O O O O O O O O O ..... O O O O O O D

AVIATION ORIGINS AND DESTINATIONS ...... ............. D

TABLE

10

ll

12

l3

14

15

LIST OF TABLES
PAGE
CITY SIZE CATEGORIESOOOCCOOOCOOOOOOOOOOOOOOOOOOOOOZS

DISTANCE SEGMENTATION ANALYSIS WITHOUT A WAIT-
TIME FACTOR ON RAIL MODELING RESULTS..............28

CITY SIZE ANALYSIS WITHOUT A WAIT-TIME FACTOR
ON MIL MODELING MSULTS....OOOOOOCOCCOOOCO0.0.0.031

DISTANCE SEGMENTATION ANALYSIS WITH A lZ-HOUR
WAIT-TIME FACTOR ON RAIL MODELING RESULTS.........34

CITY SIZE ANALYSIS WITH A lZ-HOUR WAIT-TIME
FACTOR ON MIL MODELING MSULTSOOOOOCOOO00.00.00.037

EXAMPLES OF INDUCED VERSUS DIVERTED DEMAND FROM
MIL MODELING RESULTSOOOOOOO....00000000000000000042

DISTANCE SEGMENTATION ANALYSIS WITHOUT A WAIT-
TIME FACTOR ON BUS MODELING RESULTS...............44

CITY SIZE ANALYSIS WITHOUT A WAIT-TIME FACTOR
ON BUS MODELING RESULTSICCCOOOOOOOOOOCOOOOOC0.0.0.46

DISTANCE SEGMENTATION ANALYSIS WITH A lZ-HOUR
WAIT-TIME FACTOR ON BUS MODELING RESULTS..........48

CITY SIZE ANALYSIS WITH A lZ-HOUR WAIT-TIME
FACTOR ON BUS MODELING RESULTSOOOCOOOOCCOO0.0.0.0050

EXAMPLES OF INDUCED VERSUS DIVERTED DEMAND FROM
BUS MODELING ESULTSOCOOOOOOOIOOOOOOOO00.00.....0054

INTRASTATE AVIATION TRAVELERS SURVEYED
JANUARY24-3O’19720000000000000000.0000000000000055

DISTANCE SEGMENTATION ANALYSIS WITHOUT A WAIT-
TIME FACTOR ON AVIATION MODELING RESULTS..........59

DISTANCE SEGMENTATION ANALYSIS WITH A l6-HOUR
WAIT TIME FACTOR ON AVIATION MODELING RESULTS.....61

EXAMPLES OF INDUCED VERSUS DIVERTED DEMAND FROM
AVIATION MODELING RESULTS...0.000000000000000...0.64

ACKNOWLEDGMENTS

Special thanks are offered to Richard E. Esch, Manager
and Joyce A. Newell, Transportation Planner, of the Statewide
Planning Procedures Section, Highway Planning Division,
Michigan Department of Transportation. Mr. Esch and Ms.
Newell offered the author technical advisement and the com-
puter resources of the Section. Computer files and informa-
tion generated from this research have been used by the State-
wide Planning Procedures Section.

The author is currently employed with the Michigan Depart-
ment of Transportation, Environmental and Community Factors

Division.

iii

INTRODUCTION

The purpose of this research is to examine on a preliminary
basis the performance of the Stanford Research Institute's
Intercity Passenger Demand/Modal Split Modell prepared for the
Michigan Department of Transportation in 1971 for use in
Statewide multi-modal planning. At the onset of this research
minimal investigation and testing of the model had been performed
at the agency. The model will be tested for its prediction per-
formance of common carrier modes by comparing generated results
from the model with three separate sources of actual modal origin-
destination data.

Section 1 of this paper discusses the need and evolution
of statewide multi-modal planning and its application toward
use as a policy tool. Attention is given to the contribution
of the Northeast Corridor Research Project in upgrading the
state-of-the-art in intercity transportation modeling.

Section 2 gives a detailed explanation and analysis of the
Stanford Research Institute Model.

The statewide transportation modeling system used for this
research and for planning purposes at the Michigan Department of
Transportation is discussed in Section 3.

Section 4 has four main parts. In Part One is background
information on modeling procedures used in analysis of each

modal component of the SR1 Model. Parts Two, Three, and Four

analyze the results of tests performed on the model for rail,
bus, and air modes, respectively.

A summary of the results of this research and conclusions
are given in Section 5. The remainder of the chapter gives
attention to future research needs in statewide intercity model-

ing.

I. THE EVOLUTION OF STATEWIDE MULTI-MODAL MODELING

The role of transportation as a major infrastructure
component has evolved into a highly complex and delicate one.
This observation reflects both the variety of policy instruments
transportation involves (i.e., the existence of various modes)
and the myriad of consequences transportation contributes to
society. There is a growing need for statewide multi-modal plan-
ning. In broad terms statewide planning implies a concern for
large facilities serving long-distance movement. Regional and
urban planning is concerned with shorter distance movements and
with the specific placement of terminals or route locations of
roadways. The pressures for statewide transportation planning
derive from the fact that a lack of a comprehensive and coordi-
nated approach to transportation imposes real costs upon society.
Statewide planning can be seen as a process to aid in investment
decision-making. In a comprehensive fashion it should define
transportation needs, clarify problems and issues and give assist-
ance in predicting the impacts of alternative policies. In order
to successfully use statewide multi—modal planning, it is impor-
tant to have an idea of the range of conditions which the modal
systems may have to serve. This point reflects the concern
that there is significant uncertainty regarding the future and
thus, there are a series of different public actions, which, if
taken, would result in different impacts upon such variables

as the socio-economic and natural environment. Statewide

modeling is a tool to test the effects of alternative policies.
For example, what would be the effect if a state were to decide
to promote vast fiscal and institutional support for mass trans-
it and limit its support for highway development? Or, what
changes in travel habits would occur given different levels of
fuel supply cutbacks? In order to answer questions such as these,
a statewide modeling approach is necessary and the data for this
research must be comprehensive in nature, inclusive of all modes
and must be statewide in scope.

The current practice of statewide planning is an outgrowth
of highway planning at the urban level. Though statewide plan-
ning has been discussed over the last decade in professional cir-
cles, little attention has really been directed toward signifi-

cantly upgrading the state-of-the-art.2

In general, transporta-
tion planning has devoted much of its resources towards research-
ing urban and regional concerns. The current status of statewide
planning has been aptly reflected by the noted scholar, Peter
Stopher, Director of the Transportation Center at Northwestern
University:

With the exception of work done in the 19603 for the

Northeast Corridor (by Quandt and Baumol, and McLynn

and Watkins) there has been very little work done in

attempting to develop sophisticated procedures for

travel forecasting between cities.

Current efforts at statewide planning are the by-product of
highway studies conducted in the last 40 years. It is useful
to consider statewide planning in a continuum, as part of the
historical evolution of planning. Phillip Hazen, in his unpub-

lished thesis, A Comparative Analysis of Statewide Transporta-

tion Studies, delineates three distinct periods of transportation

 

planning: 1916-1935, 1936-1955 and 1955-1975.4

He sees the
fourth period, 1975-1995, as an evolution of the previous three
where emphasis turns to multi-modal planning and the need for
coordination of the various modes at the state level.

1916-1935: The Federal-aid Road Act of 1916 first provided
federal funds for building highways. The Federal-aid Highway
Act of 1921 provided for the selection of the federal—aid road
system connecting all important population centers.

1936-1955: This second period was marked by the concept of
providing a better constructed highway built to handle the in-
creasing volumes of automobile and truck traffic and to provide
a secondary level of highways serving places of lesser importance.
The Hayden-Cartwright Act of 1934 authorized expenditures, not
to exceed one and one half percent of federal funds to each state
for making surveys, plans and engineering investigations of projects
for future construction. From this legislation there evolved a
rather institutionalized series of highway planning surveys which
provided information such as traffic volumes, vehicle speeds,
truck weights and origin and destination information. Planning
techniques in this period rarely considered analysis of inter-
dependencies among various highway links, but rather emphasized
planning to relieve the most pressing of currently observed prob-
lems. In most cases this involved investment in urban-oriented
roads.

1956-1975: The third period has been characterized as the
"interstate system era." With the Federal-aid Highway Act of

1956, this country engaged in massive investment in interstate

highway development. In total, 41,000 miles of interstate
highways were scheduled to be built. A key development in this
period was the extensive utilization of origin-destination sur-
veys of existing travel patterns. As travel patterns became
more dispersed, emphasis turned from an almost complete urban
orientation to a more balanced approach in consideration of
rural or intercity travel. With this type of information, tech-
niques which evaluate a state highway network as a system of
interdependent links became implementable.

1975-1995: With the completion of the interstate system in
the early 19803, a new period in planning will evolve. In the
past ten years or more growing emphasis has been placed on the
importance of non-highway passenger modal planning. Most states
have now created Transportation Departments from their previous
Highway Departments. Added responsibilities have been assumed in
planning for rail, bus and aviation needs. It becomes increas-
ingly important that efficient investment decisions are made
as budget constraints and energy supply limitations place addi-
tional problems on state agencies. Statewide planning's role,
in testing alternative policies, is increasingly important.5

The stimulus for research in intercity multi-modal modeling
came from the Northeast Corridor Project. Most recent research
is, in fact, based on work performed during this study. In 1965
the High Speed Ground Transportation Act was passed and some
$148 million was appropriated over the following six years for
research in multi-modal planning. The first major demonstration
project of the Act was the Northeast Corridor Project which

encompassed the pOpulation corridor stretching between Boston

and Washington, D.C. The project was generated by a sense that
' the modes providing for corridor movement in the Northeast were
overcrowded, unable to expand their capacities adequately and
subject to decision-making that was not sufficiently centralized
to yield solutions. It was felt by some that the greatest lack
of balance and coordination of investment strategy was found in
urban areas and regions in the Northeast. In order to overcome
these perceived imbalances, the approach of the project was to
start with some desired level of transportation service, consider
ways to provide it most effectively by whatever mode, and then
simulate the play of demand against resource availability to
reach the most effective system as a whole. Explicit from the
beginning was the idea that the project would pose alternative
transportation system patterns for the corridor from among which
decision makers could choose. Nine widely different alternatives
were developed. Different mixes of short-haul air and high speed
ground modes in combination with conventional means of passenger
transport were produced. The need for simulation of various
alternatives in the corridor resulted in considerable funding of
research money into computer modeling of the various modes.6
Two major models produced from this research are the Quandt and
Baumol Abstract Model and the McLynn Cross-Elasticity Model.
The SR1 model is a derivative of these models. The models are
descendents of the basic gravity model which hypothesizes trip
demand between a pair of cities as proportional to their popu-

lation size and inversely related to some impedance factor, such

a. as distance between them.

w.“

The Quandt-Baumol model assumes that demand is characterized
by the values of service variables exhibited by the various modes.
Examples of service variables are travel time, travel cost, and
departure frequency. The model presupposes that individuals
are characterized by modal neutrality. A person thus chooses
among modes on the basis of their characteristics rather than
on the basis of what they are called. Modal competition is in-
troduced into the demand equation by causing the predicted de-
mand for the given mode of travel to depend on the price of the
cheapest mode and the time of the fastest competing mode as well
as the service characteristics of the given mode. The number
of passengers-7;](who travel between city 3 and city 5 by way

of mode}( is estimated as follow

7
773k- PI’P *(‘H’jp Y j; amatebflo)
where; ' J
{MM-(#9, )8" ‘;'(”"")?I
F (<3: (<2: 8%) 6’4“ “)7

£3(o\=( :56") (0'5 06'

T5 k-v Aenmvk Lajwoezy 647 :' 9:03 €475 \J'm Mo):
p

city pOpulation

y = city income (per capita)
#4 = travel time

C, = travel cost

D = departure frequency

Superscript 5 indicates best for that city pair,
therefore (Fin b) is the travel time of the mode
serving cities; and 5 which has the shortest
travel time.

are parameters to be estima ed.

One of the virtues of the model is its ability to predict
demand for new types of modes insofar as the new mode can be
described by a new set of values for the service variables.
Another advantage is the relative simplicity of construction.
The major criticism of the model is that it often fails causality
tests; i.e., it is not based on behavioral characteristics of
passengers but is extracting temporal or structural correlation
in the data base.

The McLynn Cross-Elasticity Model is also a gravity model
utilizing an abstract model construction. The major addition
made to the abstract model is the cross-elasticity concept, de-

noted by the following ratio.

Wk“ if“: Inf”
2 cs'U—‘Kf"

4" ‘ J
thrcl

 

é

- for example, is the cost variable in the modal
split

calibration coeffieicnts

transportation cost variable

index identifying a mode

index identifying a modal attribute

b. -.‘7< 1‘

McLynn's formula is based on the idea that the rate of change

in modal share with respect to each variable can be measured by
the elasticity of modal share. Elasticity of modal share is
defined as the percentage change in modal share resulting from

a percentage change in a given modal attribute. Cross-elasticity
then, is the change in a mode's share resulting from a change in

another mode's attributes. The cross-elasticity concept can be

applied to all considered variables. In the Northeast Corridor
Project research, time and frequency elasticities, besides that
for cost, per mode, were calculated.8

As with all models, the Quandt-Baumol and McLynn models are
an approximation of reality. The real life demand/modal split
process involves a large number of complex and changing patterns
of subjective relationships. In contrast, the models employ very

simple relationships shown empirically to be most significant.

10

II. THE STANFORD RESEARCH INSTITUTE'S INTERCITY DEMAND/MODAL

SPLIT MODEL

In 1970 the Michigan Department of Transportation contracted
with the Stanford Research Institute to produce a series of com-
puter programs to conduct transportation modeling. One of the
models was to be an intercity passenger model that would perform
demand and modal split for four modes: automobile, rail, bus
and air. In June, 1971 a report was prepared by John W.

Billheimer documenting the model, The Michigan Intercity Pas-

 

senger Demand Model.9

 

The model is a descendant of the gravity model which can be

stated as:

d~
D ‘3 [([O/‘ia'uv Rva‘k’uI x Ded’in/Q'HOV POFOLJ’IOQ
A‘J‘Hcch p

 

Where:

k’d- ”3F: calibration coefficients

SR1 investigated a number of current intercity models before
choosing the McLynn model as the basic type of model to be used
for Michigan analysis. Due to the diversity of populated areas
in the state, ranging from the heavily industrialized Detroit
area to isolated rural hamlets, some modifications were made to
the model.

The SR1 model uses the number of families in each state zone
whose income exceeds $10,000 as a trip generation characteristic.

The impedence function is a composite of the time, cost and

11

frequency of service experienced on each interzonal mode of
travel. These three measures reflect a sum of the access, line-
haul and egress portions of a trip. Based on the zone-to-zone
data, the model calculates percentages of trips using each mode
and uses these percentages in combination with automobile trip
tables and income data to generate trips by mode for each zone
pair.

The model is defined by the following relCationships:

10
w .. «AM-W04 H37fil-ufflk-Ffljb‘ (M: “gas m

M " ( 7
7‘~ “‘nc (I '77“ ‘ “it.“
’0‘ (A?
w: % wM F
p.74
30-) 372‘) a' 0
[309365. F- j w (3)
. . 4L
[3 703pr 313 INC!) 3‘37 RF: "6
V (4)
tzu'3 [)/n//1A/
Where:
Vfln = a modal travel conductance
\wv’ = total travel conductance
D = total predicted travel demand
Dwx = daily one-directional modal demand
qM = common carrier conductance multiplier
+X~ = total (:-,3]; i.e., origin-destination
pair travel time for the n-th
mode (hours)
CM= total (: 43) out of pocket per capita

cost (dollars)

frequency of C.’-y5)service (trips per

day)

F: == number of families with annual incomes
exceeding $10,000 (families x 105) in
the origin or destination zone.

(I == specified value used to segment pairs

0L

3'!)

having larger population products from
those having smaller products

= weightings for the impedance measures
to account for the traveler's perceived
importance of each measure

12

The E coefficients are zone specific constants.
They are included to compensate for factors that
are not explicitly included in the model.
A< = modal level of service conductance
multiplier
The following bounds were imposed on the model parameters
in advance of the calibration process:
\
0‘3. )3 {034;}?(0‘)
\
o {2/3 mépme 1.!
O iprme; /
"5 & oLCs‘) E. O J="‘5‘*:5
oa.(3') :: .3247
[(20.11

01‘; emf-.5

The bounds were found to be necessary to maintain the model's
consistency of behavior. If ;?fi),for example, was allowed to
exceed 1.1, population increases would have a disproportionate
effect on predicted demand. Likewise, should. FTi)exceed unity,
a small change in time or cost by one mode could cause excessive
increases in travel over competing modes. The d~exponents are
held to be negative so that small changes in time or cost will
not have a disprOportionate effect on demand. ck? and Kvalues
are simply those set by McLynn in his studies. An upper bound-
ary was placed on the common carrier conductance multiplier
because it was felt that larger values would create unrealistic

imbalances between common carrier traffic and automobile traffic.

13

The model was tested by SR1 and calibrated by 1967 data from
20 city pairs, eight of which were intrastate pairs. The cali—
bration places resulted in the identification of the following

parameter values:11

074 : 1.5" #4: «EA
0.73" .M: bx”, fin”

0&0) 7-6—07 2 “4'5
cc. (3} = 0. 3347, k = 0.1.;
our“? 2 d—f‘p‘ ‘-'-' -L8
,8 (03 2 347000.) fi’rojzallroo
E 01 ‘-‘- 1.04 [3‘01 -: O.(

[3 Q33 1: CD-Q‘
(3: : (3.0'75”

As can be seen, the parameters whose changes have the great-
est potential impact on demand are the time and cost components
63"”;an (“3 0W; \L-“qand the conductance exponent/5’3 . Ser-
vice frequency v/S'?‘ is the least effective of the input variables
in terms of its ability to influence sizeable demand changes.
Since an increase in time and price for a given mode implies that

the mode has become relatively less attractive (if all other

modes do not change), the size of the calibrated exponents

vLIKW and $§;)must be negative to ensure that increases in time

and price for the mode decrease that mode's share. If 953‘ the
measure of modal frequency is a transformation which increases
with increased frequency, then an increase in frequency signals
an improvement in a model's competitive position and the sign of

djfl

the must be positive.

14

because of its service characteristics. In the Northeast
Corridor studies, induced demand made up approximately 85 percent
of increases in volume resulting from improvements in service.12
The SR1 model is evaluated by Bennett, Ellis and Prokopy of
Peat, Marwick, Mitchell and Company (PMM) in a paper performed

13 The

for the United States Department of Transportation.
authors selected seven intercity modal split models and tested
their prediction powers against Northeast Corridor data as well
as non-Northeast Corridor data. All of the models were either
tested or derived from research done for the Northeast Corridor
Project. The researchers concluded that all models tended to
overestimate bus and rail traffic and underestimate air traffic.
The models were found to overestimate low volume traffic on all
modes and tended to compensate by underestimating automobile and
air traffic at high volume levels. In comparing the SR1 model
against Northeast Corridor data it was found that the model over—
estimated bus traffic and underestimated air and rail traffic,
though the air estimates were very close to the observed. In
non-Northeast Corridor data, the SR1 model again was found to
overestimate in bus volumes, compare well in air traffic, and
overestimate rail. Thus, from these findings, it seems clear
that,in the PMM tests, the model consistently overestimated bus
traffic, tended to predict air volume quite well, and had mixed
results in the rail mode. Compared to the other models, accord—
ing to ability to replicate observed volumes, in Northeast Corri-
dor pairs, SR1 ranked sixth, second, and fifth, respectively

as to bus, air and rail and fifth, first, and fifth, respectively

according to non-Northeast Corridor pairs.

16

The elasticity concept of McLynn is reflected in the value
of these parameters. For example, the cost parameter 0693-1 is
set at -l.5. Hence, a one percent increase in price would re-
sult in a 1.5 percent decrease in the number of trips demanded.
This price elasticity assumption follows that changes in price
have a slight disproportional change in the volume of demand.

1

Likewise, the time elasticity parameter 661 also exceeds this
change function. It is somewhat unusual that time and price
elasticities are the same in that some studies have concluded

time to be a more significant change variable than price.

SR1 concluded that the model tends to underestimate long-
distance trips (defined as over 600 miles) and overestimate
traffic involving short distances. Another problem discovered
was the relationship between induced versus diverted demand.

The model tends to overstate induced demand at the expense of
diverted demand. When improvements in a single mode cause an
incremental increase in the number of travelers using that mode,
the travelers can be assumed to come from one of two sources:

(1) Other modes (diverted demand); (2) the pool of potential
travelers who currently are not included in total intercity de—
mand (induced demand). There are two potential types of induced
demand, assuming that modal choice remain constant. The travelers
could be induced to change their previous chosen destinations
for various trip purposes and thus go to other destinations
served by the corridor because of its attractive service charac-
teristics. Secondly, travelers could maintain their destination

choices but select a different routing to get there; thus aban—

doning their old routing and choosing the subject corridor

15

III. THE MICHIGAN DEPARTMENT OF TRANSPORTATION'S STATEWIDE
MODELING SYSTEM
The transportation modeling system used for this research
is the system devised and Operated by the Statewide Transporta-
tion Planning Procedures Section of the Michigan Department of
Transportation in Lansing, Michigan. Modeling in Michigan can
either be performed on a 547 zone or a 2300 zone classification.
The 547 zone system was chosen for this research. Michigan is
divided into 508 of these zones. Zone boundaries coincide with
political boundaries. Major cities are each one zone with the
exception of Detroit which is three zones. Some of the smaller
cities also are one zone. In rural areas, the size of a zone may
vary from one to several townships. Besides the zones in Michigan
there are 39 other zones which are divided into 32 for neighbor—
ing states and Canada and an outer ring of seven zones. The
outstate zones are never smaller than a county and the seven
outer zones may be several states. The zonal system is shown in
Maps 1 and 2. Each zone has a "centroid" or center of population.
This is a given point within the zone at which all travel is
assumed to originate or terminate. This paper will consider
only intrastate zones in that socio—economic data for outstate
zones are not currently available.14
The basic element of the statewide highway network is a
"link", a small segment of highway approximately 1-5 miles in
length. Each link is uniquely identified by a pair of numbers

called nodes, designating its end points. A node number is

17

found at each intersection and often at county lines. Thus,

a link is generally a segment of highway between two consecutive
intersections. Other links, called "access" or "centroid" links
are included which connect the centroids to the highway system.
Links and centroid links are shown in Map 3.

The highway network is composed of three major data compo-
nents: The Statewide Socio-economic Data File; the Statewide
Transportation Network and the Statewide Public and Private Facil-
ity File. Each of these files provides information which is sum-
marized into the 508 intrastate analysis zones.

The Socio-economic File contains 888 pieces of selected cen-
sus information concerning the overall population characteristics
within each zone. The data is from the 1970 Census of Population
and Housing.

The Transportation Network File contains the physical descrip-
tion of each highway link such as average speed, distance and
annual daily traffic volumes.

The Public and Private Facility File contains information
pertinent to the man-made, physical aspects of the environment,
such as the location of airports and major commercial centers.

Using these three files it is possible to accomplish a very
useful process called proximity analysis which analyzes the
relationship between facilities and various socio-economic charac-
teristics. A computer program accumulates selected socio-economic
data based on driving time bands from the zone of the selected
facility under study. The driving times between each of the

508 zones is derived from "skimmed trees". Before discussion of

18

J

MAPl

MlCI-IIGAN’S TRANSPORTATION MODELING SYSTEM

547 ZONE
OUTSTATE ANALYSIS ZONES

ONTARIO
SI I

 

Ear—2:,

     
   
 

 

 

Ea
WISCONSIN
52 I
524

new vom:

t 523 mcmaAN 5 547

A

IOWA E!

.4 j) I) 5
IL

 

 

 

 

 

 

 

° "' N 8 3'5
3 EL: '9 I . penvaVANIA
538 546
LINOIs 533
526
W omo
INDIANA 54°
[ 534 WEST
VIRGINIA
lesoum 527 541
543
KENTUCKY
542

 

 

19

 

 

 

 

so

00,30

("HANNEI

 

Q

547 ZONE
STATEWIDE TRANSPORTATION ‘

MODELING SYSTEM ,

INSTATE ZONE MAP
MAY |974

(\

ill!

  
 

5' film

 
 
   

,A NA DA

20

MAP3

MICHIGAN STATEWIDE HIGHWAY NETWORK PLOT (547 ZONE SYSTEM)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

21

proximity analysis and its use in the research in this paper, a
brief explanation of the skim tree process is presented.

The basis of the skim tree process is the analysis used to
choose the "path of least resistance" from each zone to every
other zone. In this research the average driving time, as deter-
mined by the distance and speed information coded in each link,
is used to select the minimal paths. The time between two zones
is assumed to be that time required to travel between the two
zone centroids. The centroid of a zone is a given time from the
centroid of another zone so that all persons residing in one zone
are assumed to live within that traveling time of all persons in
the other zone, although portions of the zone may be closer
or further apart. Since the total pOpulation of a zone is as-
sumed to reside at the centroid, no travel time within a zone
can be calculated. Once the paths between all desired zones have
been completed, the paths or trees are "skimmed" to select the
zone-to-zone travel times. These times are then used in proxim-
ity analysis.

Proximity analysis searches in the selected time bands of
each facility for zone centroids. The value assigned to the zone
centroid within the desired time radius of the facility is accumu-
lated. Since the facility is located at the zone centroid, the
value assigned to that zone is included in the final sum for that
facility. The output from this process summarizes for each analy—
sis zone and for each time band: 1) The total socio-economic
statistic occurring in the band, and 2) a list of zones in the

band.

22

IV. MODELING PROCEDURES AND RESULTS
1. Documentation of Procedures and Analyses

The research performed was conducted on the Michigan
Department of Transportation's Burroughs 7700 Computer. The
automobile was assumed to be the dominant mode. The highway
network and planning model used in this research was developed
in 1966 by the Statewide Studies Unit of the Michigan Department
of Transportation with assistance from the consultant firm of
Arthur D. Little. The demand model is a gravity type model and
has been recalibrated according to traffic volumes recorded con-
tinually by the Department's traffic monitoring system. The SR1
model compares the characteristics of the highway mode for a
given pair against the service characteristics of a common carrier
mode. The generated volume for that mode then results from di-
verted demand from the automobile mode and from induced demand.

Highway cost calculations used in all modeling processes are
shown in Appendix A.

Two variations of the basic SR1 model are used: Proximity
analysis and a wait-time factor. As previously mentioned the
market area for a given modal service can be varied based on cal-
culated highway driving time bands. For each of the modes tested,
experimentation is performed with this function until the best
prediction for a given pair is obtained.

A wait-time factor is incorporated into the model because

of its tendency to overestimate demand. This factor, in effect,

23

augments the din \parameter. The calculation is based on the
length of a normal service day and the frequency of daily serv-

ice for a given pair.

Wait Time = L“? +1I‘:‘F' SIR/Vice 0‘7 X (0

 

‘F’lgowv

\
A:
If, for example, the service day for flights from Grand

Rapids to Detroit is 16 hours and the daily frequency is nine

trips. the wait-time per trip can be calculated as:

Wait Time = H X é: : 53 Irwin:
9 . 5
Ya

 

Two different categories of analysis are applied to the
model's modal prediction: A city size combination analysis and
a distance segmentation analysis. Because of the wide range of
city sizes in Michigan, a city size segmentation function, G,
is in the model. The purpose of the city size analysis is to
determine whether the G factor is properly accommodating city
size variations. Thus, if certain city size categories reveal
a consistent pattern of overestimation or underestimation a
preliminary assumption can be made that recalibration of this
function may be desirable. In order to perform this type of
analysis, all the various cities used in model analysis are com-
bined and the associated population statistics are accumulated
using 1970 census estimates. The cities are then clustered into
five categories based on the array of sizes. City size categories

are shown in Table 1.
24

TABLE 1

CITY SIZE CATEGORIES

 

SIZE 1970 SIZE 1970
CATEGORY CITY POPULATION CATEGORY CITY POPULATION
A Detroit 1,514,063 E Lowell 3,068

Gaylord 3,012
B Grand Rapids 197,649 Clare 2,639
Flint 193,317 L'Anse 2,538
Lansing 131,403 Brighton 2,457
Ann Arbor 100,035 Imlay City 1,980
Saginaw 91,849 Fowlerville 1,978
Pontiac 85,279 Pellston 469
Kalamazoo 85,035 New Hudson N.A.
C Jackson 45,484
Muskegon 44,631
Battle Creek 38,931
Port Huron 35,749
Midland 35,176
D Ypsilanti 29,538
Holland 26,479
Marquette 21,907
Mt. Pleasant 20,504
Traverse City 18,048
Owosso 17,179
Benton Harbor 16,481
Escanaba 15,368
Sault Ste. Marie 15,136
Alpena 13,805
Niles 12,988
Albion 12,112
Grand Haven 11,844
Menominee 10,748
Farmington 10,328
Cadillac 9,990
Coldwater 9,232
Ironwood 8,711
Iron Mountain 8,702
Charlotte 8,244
Manistee 7,723
Marshall 7,253
Tecumseh 7,120
South Haven 6,471
Lapeer 6,341
Mason 5,468
Hancock 4,830
Durand 3,678

25

The second category of analysis is distance segmentation,
In previous evaluations of the model, conclusions were made
that it tended to overestimate short distance trips and under-
estimate long distance trips. In order to test for this, city
pairs for all modes are categorized into nine segments, accord-

ing to pair distance:

Trips 5:. 40 miles

7 4o 5:. 60 miles

760 ‘5'... 80 miles
7780 Ei.l30 miles
7130 .4. 190 miles
7190 $1 250 miles
77250 .~ 350 miles
7350 ‘53. 450 miles
~7450 miles

II

2. Rail Results and Analyses

The data used to test the SR1 model for the rail mode is
from the Amtrak Origin-Destination records. Ms. Joyce A. Newell
of the Statewide Transportation Planning Procedures Section
collected information from the station manager of the Amtrak
terminal in East Lansing, Michigan. Portions of daily and
monthly traffic for nine months in 1974 and for the entire year
of 1975 were collected. The author then coded this data for
keypunching and transfer to a computer disc. Trip tables for
intrastate travel were formed and the data was analyzed. Origin-
destination data for stations in Albion and Ypsilanti were
incomplete and the stations were omitted from consideration.

It was determined that an average daily trip table from the
months of May and October, 1975 would provide the best data for
comparison with model predictions. Input data for the model

was obtained from the Amtrak Fare Guide supplied by the Amtrak

26

Nut

District Offices in Chicago, Illinois. The Amtrak network is
shown in Appendix D. A data set for all the origins and destina-
tions is given in Appendix B.

Initial model calculations were performed by experimentation
with various combinations of market areas for each pair. Mar-
ket bands resulting in the best prediction for each origin-
destination were chosen and are shown in Appendix B. In general,
the market area for pair distances under 100 miles was set at
(0,0). It is logical that travelers will not drive much distance
to board a train for a destination under approximately 100 miles
in that the highway travel time may require only 2 to 2% hours.
Likewise pair distances 150 miles and over required market area
augmentation. This varied from (20,20) to a maximum of (30,30).
Niles-Port Huron, the longest pair distance, was set at (40,40).
In cases where city size varied significantly, the selected
market area was disprOportional; e.g., Niles-Detroit (30,0);
Kalamazoo-Lapeer (10,20).

Analysis based on an absolute error calculation was performed
on the model results according to distance segmentation and is
given in Table 2. Examination of Table 2 shows the model severely
overestimates demand in the first distance category and incre-
mentally lessens this tendency in the next two categories. The
last two categories reveal the opposite tendency and are also
closest approximations to the survey data. City size analysis
is shown in Table 3. No discernible pattern is evident from this
analysis.

A wait-time factor based on a 12-hour service day was included

in the calculations. Analysis on the results according to

27

TABLE 2

DISTANCE SEGMENTATION ANALYSIS
WITHOUT A WAITbTIME FACTOR
ON
RAIL MODELING RESULTS

 

 

 

 

 

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR
City Pair Distances ii 40 Miles
Detroit-Ann Arbor 303 45.3 +257.7
Ann Arbor-Detroit 303 34.3 +268.7
Kalamazoo-Battle Creek 82.1 12.4 + 70.2
Battle Creek-Kalamazoo 68 13.7 + 54.3
Ann Arbor-Jackson 41.6 6.2 + 35.4
Jackson-Ann Arbor 41 4.9 + 36.1
Lansing-Durand 14.2 0.3 + 13.9
Durand-Lansing 14.2 2 + 12.2
Flint—Durand. 42.2 0.5 + 41.8
Durand-Flint 42.3 0.4 + 41.9
Flint-Lapeer 11.9 0.1 + 11.8
Lapeer-Flint 11.9 0.03 + 11.9
Durant-Lapeer 0.2 .06 + .1
Lapeer-Durand 0.3 .03 + .3

Total Absolute Error = 711.8%
City Pair Distances 740 5.. 60 Miles
Lansing-Flint 81.9 1.1 + 80.7
Flint-Lansing 81.9 2.4 + 79.5
Lansing-Battle Creek 17.7 12.5 + 5.2
Battle Creek-Lansing 17.0 2.9 + 14.1
Kalamazoo-Niles 6.6 4.2 + 2.4
Niles-Kalamazoo 6.6 5.2 + 1.4
Jackson-Battle Creek 9.3 5.5 + 3.8
Battle Creek-Jackson 8.3 5.6 + 2.7
Port Huron-Lapeer 1.1 0.03 + 1.1
Lapeer-Port Huron 1.1 0.1 + 1
Total Absolute Error = 483.4%

 

28

'63

Table 2 (cont'd.)

 

 

 

ABSOLUTE

CITY PAIR GENERATED ACTUAL ERROR
City Pair Distances 7 60$ 80 Miles
Detroit-Jackson 28 8 +20
Jackson-Detroit 26.5 5.2 +21.3
Lansing-Kalamazoo 13.9 3.3 +10.6
Kalamazoo-Lansing 13.9 4.1 + 9.8
Kalamazoo-Jackson 7 2.1 + 4.9
Jackson-Kalamazoo 7.1 1.7 + 5.3
Flint-Port Huron 11.9 0.1 +11.8
Port Huron-Flint 13.7 0.4 +13.3
Lansing-Lapeer l 0.8 + 0.2
Lapeer-Lansing l 0.7 + 0.3
Battle Creek-Niles 0.9 1.2 - 0.3
Niles-Battle Creek 0.8 1.1 - 0.3
Battle Creek-Durand 0.4 0.3 + 0.1
Durand-Battle Creek 0.1 0.2 - 0.1

Total Absolute Error = 342.5%

 

City Pair Distances 7’80f2130 Miles

 

Kalamazoo-Ann Arbor 11.9 9.5 + 2.4
Ann Arbor-Kalamazoo 5.9 9.4 - 3.5
Flint-Kalamazoo 2.6 1.6 + l
Kalamazoo-Flint 2.6 2.4 + .2
Ann Arbor-Battle Creek 5.6 4- + 1.6
Battle Creek-Ann Arbor 5.2 2.5 + 2.7
Lansing-Port Huron 3.2 2.6 + 0.6
Port Huron-Lansing 3.5 2.5 + l
Flint-Battle Creek 2.6 1.1 + 1.5
Battle Creek-Flint 2.9 0.7 + 2.2
Kalamazoo-Durand 0.4 0.5 - 0.1
Durand-Kalamazoo 0.4 0.3 + 0.1
Kalamazoo-Lapeer 0.6 0.8 - 0.2
Lapeer-Kalamazoo 0.6 0.6 0
Lansing-Niles 2.7 2.9 - 0.2
Niles-Lansing 3.7 3.7 0
Battle Creek-Detroit 8.7 4.3 + 4.4
Detroit-Battle Creek 7.9 4.8 + 3.1
Jackson-Niles 1 1.5 - 0.5
Niles-Jackson l 1.4 - 0.4
Durand-Port Huron 1.4 1.6 - 0.2
Port Huron-Durand 1.5 1.5 0
Lapeer-Battle Creek 0.4 0.2 + 0.2
Battle Creek-Lapeer 0.4 0.2 + 0.2
Total Absolute Error = 42.2%

 

29

in»

Table 2 (cont'd.)

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR

 

 

Citnyair Distances'7130 Miles

 

Detroit—Kalamazoo
Kalamazoo-Detroit
Detroit-Niles
Niles-Detroit

Port Huron-Kalamazoo
Kalamazoo-Port Huron
Ann Arbor-Niles
Niles-Ann Arbor
Flint-Niles

Niles-Flint

Battle Creek-Port Huron
Port Huron-Battle Creek
Port Huron-Niles
Niles-Port Huron
Lapeer-Kalamazoo
Kalamazoo-Lapeer
Durand-Niles
Niles-Durand
Lapeer-Niles
Niles-Lapeer

FJH
was
PIA
o<3c>o<3c>o<3FACIAPnbmswraeupuan

anwmoooxkooxmmbm
+

o o o o O o o
l—lmONNNUJQ

b0\\lU1U1\DI§W\I

NI—‘bl—‘QQKDQWWhCDb-bl—‘I—‘wm

+
CDOCDCDOWDCDOCDPJOCDCDOCDCDOIOCDH
namIahamId Iduaw

OOOOOOOOI—‘I—‘I—‘ONWI—‘I—‘wm

Total Absolute Error = 17.8%

 

30

as»

TABLE 3

CITY SIZE ANALYSIS
WITHOUT A WAIT-TIME FACTOR
ON
RAIL MODELING RESULTS

 

 

 

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR
City Size Analysis with Category A Origin
Detroit-Ann Arbor 13. 14.7 - 1.7
Detroit-Kalamazoo 303 45.3 +257.7
Detroit—Jackson 28 8 + 20
Detroit-Battle Creek 7.9 4.8 + 3.1
Detroit-Niles 6.6 4.2 + 2.4

CityiSize Analysis with Category B Origin

 

 

Kalamazoo-Detroit 13 13.3 .3
Ann Arbor-Detroit 303 34.3 +268.7
Kalamazoo-Ann Arbor 11.9 9.5 + 2.4
Ann Arbor-Kalamazoo 5.9 9.4 - 3.5
Lansing-Kalamazoo 13.9 3.3 + 10.6
Kalamazoo-Lansing 13.9 4.1 + 9.8
Lansing-Flint 81.9 1.2 + 80.7
Flint-Lansing 81.9 2.4 + 79.5
Flint-Kalamazoo 2.6 1.6 + l
Kalamazoo-Flint 2.6 2.4 + 0.2
Total Absolute Error 553.7%
Kalamazoo-Battle Creek 82.6 12.4 + 70.2
Kalamazoo-Jackson 7 2.1 + 4.9
Kalamazoo-Port Huron 1.1 1.6 - 0.5
Ann Arbor-Jackson 41.6 6.2 + 35.4
Ann Arbor-Battle Creek 5.6 4 + 1.6
Lansing-Battle Creek 17.7 12.5 + 5.2
Lansing-Port Huron 3.2 2.6 + 0.6
Total Absolute Error 286.9%

 

31

Table 3 (cont'd.)

 

 

 

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR
City Size Analysis with Category B Origin (cont'd.)

B-D Kalamazoo-Niles 6.6 4.2 2.4
Kalamazoo-Durand 0.4 0.5 - 0.1
Kalamazoo-Lapeer 0.6 0.8 - 0.2
Ann Arbor-Niles 3.4 4.1 - 0.7
Lansing-Niles 2.7 2.9 - 0.2
Lansing-Durand 14.2 0.3 + 13.9
Lansing—Lapeer l 0.8 + 0.2
Flint-Niles 0.8 1.2 - 0.4
Flint—Durand 42.2 0.5 + 41.8
Flint-Lapeer 11.9 0.1 + 11.8

City Size Analysis with Category C Origin

C-A Jackson-Detroit
Battle Creek-Detroit

C-B Jackson-Battle Creek

1 Jackson-Ann Arbor
Battle Creek-Ann Arbor
Battle Creek-Kalamazoo
Battle Creek-Lansing
Battle Creek-Flint
Port Huron-Flint
Port Huron-Lansing
Port Huron-Kalamazoo

C-C Jackson-Battle Creek
Battle Creek-Jackson
Battle Creek-Port Huron
Port Huron-Battle Creek

C-D Jackson-Niles
Battle Creek-Niles
Battle Creek-Durand
Battle Creek-Lapeer
Port Huron-Niles
Port Huron-Durand
Port Huron-Lapeer

Total Absolute Error = 432.7%

 

26.5
8.7

9.3
41

5.2
68
17

1

I-‘OOLON
I—‘U'IQKO

5.2 + 21.3
4.3 + 4.4
5.5 + 3.8
4.9 + 36.1
2.5 + 2.7
3.7 + 54.3
2.9 + 14.1
0.7 + 2.2
0.4 + 13.3
2.5 + l

1.6 - 0.5

Total Absolute Error = 207.7%

 

I-‘I---‘<DO<:>Ol'--l HI—‘mko
O O O O O O O
l—‘UIQWWO wwww

l-F+-+
OI—‘Nw
WI-‘Qm

oracM3c>HId Iac>mcn
ocnoxnmokam
I
PHDCDOCDCDO
O O O O O
H PaNIauam

3 +

Total Absolute Error = 44.2%

 

32

Table 3 (cont'd.)

 

 

 

ABSOLUTE

CITY PAIR GENERATED ACTUAL ERROR
City Size Analysis with Category D Origin

Niles-Detroit 3.3 4.2 - 0.9
Niles-Kalamazoo 6.6 5.2 + 1.4
Niles-Lansing 3.7 3.7 0
Niles-Ann Arbor 3.4 4 — 0.6
Niles-Flint 3.7 3.7 0
Durand-Kalamazoo 0.4 0.3 + 0.1
Durand-Lansing 14.2 2 +12.2
Durand-Flint 42.3 0.4 +4l.9
Lapeer-Kalamazoo 0.7 0.6 + 0.1
Lapeer-Lansing 1 0.7 + 0.3
Lapeer-Flint 11.9 0.03 +1l.9

Niles-Battle Creek
Niles-Jackson
Niles-Port Huron
Durand-Battle Creek
Durand-Port Huron
Lapeer-Battle Creek
Lapeer-Port Huron

Niles-Durand
Niles-Lapeer
Durand-Niles
Durand-Lapeer
Lapeer-Durand
Lapeer-Niles

Total Absolute Error = 370.3%

 

 

0.8 1.1 - 0.3
l 1.4 - 0.4
0.9 0.9 0
0.1 0.2 - 0.1
1.4 1.6 - 0.2
0.4 0.2 + 0.2
1.1 0.1 + 1
Total Absolute Error = 39.3%
0.4 0.3 + 0.1
0.2 0.5 - 0.3
0.1 0.2 - 0.1
11.9 0.1 +11.8
0.3 0.03 + 0.3
0.1 0.6 - 0.5
Total Absolute Error = 741.2%

 

33

“6»

TABLE 4

DISTANCE SEGMENTATION ANALYSIS
WITH A lZ-HOUR WAIT-TIME FACTOR
ON
RAIL MODELING RESULTS

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR

 

City Pair Distances£40 Miles

Detroit-Ann Arbor 58.8
Ann Arbor-Detroit
Kalamazoo-Battle Creek 13.3
Battle Creek-Kalamazoo
Ann Arbor-Jackson
Jackson-Ann Arbor
Lansing-Durand
Durand-Lansing
Flint-Durand
Durand-Flint
Lapeer-Flint
Flint-Lapeer
Durand-Lapeer
Lapeer-Durand

NH

U'l

00

O

CD
I—‘WJ-‘n
NIbU'I

O
nwmwwhwmw

l-t+-t+-t+-t+

IbkabUIWWWUI-‘KOUIW

h

+-+
OOOOI—‘OHOOOOOhw

H
OOOI—‘OWGOI—‘Odfl
l—l
OOOONWU‘IOOOON

0 0
(D

Total Absolute Error = 40%

 

City Pair Distances 7 40 f; 60 Miles

Lansing-Flint 7.2 1.2 + 6
Flint-Lansing 7.2 2.4 + 4.8
Lansing-Battle Creek 1.7 12.5 -10.8
Battle Creek-Lansing 1.7 2.9 - 1.2
Kalamazoo—Niles 2.2 4.2 - 2
Niles-Kalamazoo 2.2 5.2 - 3
Jackson-Battle Creek 2.0 5.5 - 3.5
Battle Creek-Jackson 1.9 5.7 - 3.8
Port Huron-Lapeer 0.1 0 + 0.1
Lapeer-Port Huron 0.1 0.1 0

Total Absolute Error = 89%

 

34

Table 4 (cont'd.)

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR

 

City Pair Distances 760 £80 Miles

Detroit-Jackson
Jackson-Detroit
Lansing-Kalamazoo
Kalamazoo-Lansing
Kalamazoo-Jackson
Jackson-Kalamazoo
Flint-Port Huron
Port Huron-Flint
Lansing-Lapeer
Lapeer-Lansing
Battle Creek-Niles
Niles-Battle Creek
Battle Creek-Durand
Durand-Battle Creek

I+-+
'° In?»

NNl—‘NflmnhQQI—‘HNN
I +-I+-+ I

(DOOOOCDI-‘OOOIm‘II-‘Ihl—l

hnkaHLflUIDd>quO\m

OOOOOOI—‘l—‘NNNNDKO
O O O O O O O

OOI—‘I—‘OOOOI—‘thmm

I—‘NQmeI-‘mflwm

H

Total Absolute Error = 45%

 

City Pair Distances 7 80 £ 130 Miles

 

I_I

Kalamazoo-Ann Arbor
Ann Arbor-Kalamazoo
Flint-Kalamazoo
Kalamazoo-Flint

Ann Arbor-Battle Creek
Battle Creek-Ann Arbor
Lansing-Port Huron
Port Huron-Lansing
Flint-Battle Creek
Battle Creek-Flint
Kalamazoo-Durand
Durand-Kalamazoo
Kalamazoo-Lapeer
Lapeer-Kalamazoo
Lansing-Niles
Niles-Lansing

Battle Creek-Detroit
Detroit-Battle Creek
Jackson-Niles
Niles-Jackson
Durand-Port Huron
Port Huron-Durand
Lapeer—Battle Creek
Battle Creek-Lapeer

. . O O
thONIbU'I
I

NNl—IHQO‘QQDKOWWAIb

NNU'IONIbU'ICDwml—‘mmLOU'IQI-‘UIQUI
I

NNNwKDHHIbml-‘IbChNbI-‘UIGDkDONI-‘I-‘Q

OOOOOOWWI—‘NOOOOOOOOI—‘I—‘NNWU‘I

OOI—‘l—‘I—‘I-‘DhOJIbOOOOOI-‘NNNIbNI-‘QKO
I

OOHI—‘OOOONNOOOOOOI—H—‘ONOOh-b

Total Absolute Error = 43%

 

35

*9

Table 4 (cont'd.)

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR

 

 

City Pair Distances 77130 Miles

 

Detroit-Kalamazoo
Kalamazoo-Detroit
Detroit-Niles
Niles-Detroit

Port Huron-Kalamazoo
Kalamazoo-Port Huron
Ann Arbor-Niles
Niles—Ann Arbor
Flint-Niles

Niles-Flint

Battle Creek-Port Huron
Port Huron-Battle Creek
Port Huron—Niles
Niles-Port Huron
Lapeer-Kalamazoo
Kalamazoo-Lapeer
Durand-Niles
Niles-Durand
Lapeer-Niles

Niles Lapeer

Hrs

o<3c>o<3c>huohaowahanun+4Pnba>uIn

o o o o
tthNN
o o o o o o
I—lmquwq
I
o o o O 0
HM” qu

O O O O O O O
NNIwab U1II>
O O O O O
mmNmN
l
O O O O O O O O O
00

IbUINNthIhWNN

OOCOOOI—‘OOOOONNOOWWQQ
I
OOOOOOOOI—‘OI—‘ONNI—‘I—‘I—‘I—‘mfl

O O
HIAPJ
O O
O O

Total Absolute Error = 51%

 

36

TABLE 5

CITY SIZE ANALYSIS
WITH A lZ-HOUR,WAIT-TIME FACTOR
ON
RAIL MODELING RESULTS

 

 

 

 

ABSOLUTE

CITY PAIR GENERATED ACTUAL ERROR
City Size Analysis with Categgry A Origin

Detroit-Ann Arbor 58.8 45.3 +13.5
Detroit-Kalamazoo 7 14.7 - 7.7
Detroit-Jackson 9.6 8 + 1.6
Detroit-Battle Creek 3.8 4.8 - l
Detroit-Niles 3.2 4.7 - 1.5
City Size Analysis with Category B Origin

Kalamazoo-Detroit 13.3 - 6.3
Ann Arbor-Detroit 58.8 34.3 +24.5
Kalamazoo-Ann Arbor 5.4 9.5 — 4.1
Ann Arbor-Kalamazoo 5.4 9.4 - 4.1
Lansing-Kalamazoo 2.3 3.3 - 1
Kalamazoo-Lansing 2.3 4.1 - 1.8
Lansing-Flint 7.2 1.2 + 6
Flint-Lansing 7.2 2.4 + 4.8
Flint-Kalamazoo 2.3 1.6 + 0.7
Kalamazoo-Flint 2.3 2.4 - 0.1

Total Absolute Error = 69%

 

Kalamazoo-Battle Creek 13.3 12.4 + 1.1
Kalamazoo-Jackson 2.4 2.1 + 0.3
Kalamazoo-Port Huron 0.4 1.6 - 1.2
Ann Arbor-Jackson 6.3 6.2 + 0.1
Ann Arbor-Battle Creek 1.9 4 - 2.1
Lansing-Battle Creek 1.7 12.5 -10.8
Lansing-Port Huron 0.7 2.6 - 1.9

Total Absolute Error = 42%

 

37

Table 5 (cont'd.)

 

 

 

ABSOLUTE

CITY PAIR GENERATED ACTUAL ERROR
City Size Analysis with Category B Origin (cont'd.)
Kalamazoo-Niles 2.2 4.2 - 2
Kalamazoo-Durand 0.1 0.5 - .4
Kalamazoo-Lapeer 0.2 0 8 - .6
Ann Arbor-Niles 2 4.1 - 2.1
Lansing-Niles 2 4.1 - 2.1
Lansing—Durand 0.6 0.3 + 0.3
Lansing-Lapeer 0.1 0.8 - 0.7
Flint-Niles 0.4 1.2 - 0.8
Flint-Durand 1.3 0.5 + 0.8
Flint-Lapeer 0.4 0.07 + 0.3

City_Size Analysis with Category C Origin

Total Absolute Error = 61%

 

 

Jackson-Detroit
Battle Creek-Detroit

Jackson-Battle Creek
Jackson-Ann Arbor
Battle Creek-Ann Arbor
Battle Creek-Kalamazoo
Battle Creek-Lansing
Battle Creek-Flint
Port Huron-Flint

Port Huron-Lansing
Port Huron-Kalamazoo

Jackson-Battle Creek
Battle Creek-Jackson
Battle Creek-Port Huron
Port Huron-Battle Creek

Jackson-Niles
Battle Creek-Niles
Battle Creek-Durand
Battle Creek-Lapeer
Port Huron-Niles
Port Huron-Durand
Port Huron—Lapeer

(J00
O O
\003

H
OOl-‘Ol—‘LAH—me
o o o

nqmmqwow

5.2 + 4.4
4.3 - 0.4
5.5 - 3.5
5 + 1.3
2.5 - 0.6
3.8 - 0.5
2.9 - 1.2
0.7 - 0.1
0.4 + 1.1
2.5 - 1.8.
1.6 - 1.2

Total Absolute Error = 32%

 

OOI—‘N
o o o o o
9‘.” IbIbNO

O
LOW

OOOOOOO

5.5 - 3.5
5.5 - 3.6
0.2 + 0.2
1.6 - 1.2
1.5 - 1

1.2 - 0.8
0.2 - 0.2
0.2 - 0.2
0.6 - 0.3
1.5 - 1.2
0 0

Total Absolute Error = 71%

 

38

Table 5 (cont'd.)

 

 

 

ABSOLUTE
CITY PAI R GENE RATED ACTUAL ERROR
City Size Analysis with Category D Origin
Niles-Detroit 3.2 4.2 - 1
Niles-Kalamazoo 2.2 5.2 - 3
Niles-Lansing 1 3.8 - 2.8
Niles-Ann Arbor 2.4 4 — 2
Niles-Flint 0.5 1.5 - l
Durand-Kalamazoo 0.1 0.3 - 0.2
Durand-Lansing 0.6 .2 + 0.4
Durand-Flint 1.3 0.4 + 0.9
Lapeer-Kalamazoo 0.2 0.6 - 0.4
Lapeer-Lansing 0.1 0.7 - 0.6
Lapeer-Flint 0.4 0.03 + 0.37

Total Absolute Error = 63%

 

 

Niles-Battle Creek 0.4 1.1 - 0.7
Niles-Jackson 0.5 1.4 - 0.9
Niles-Port Huron 1.4 l + 0.4
Durand-Battle Creek 0.1 0.2 - 0.1
Durand-Port Huron 0.3 1.6 - 1.3
Lapeer—Battle Creek 0 0.2 - 0.2
Lapeer—Port Huron 0.1 0.1 0
Total Absolute Error = 64%
Niles-Durand 0.1 0.3 - 0.2
Niles-Lapeer 0.1 0.5 - 0.4
Durand—Niles 0 0.2 - 0.2
Durand-Lapeer 0 0.1 0
Lapeer-Durand 0 0 0
Lapeer-Niles 0.1 0.6 - 0.5
Total Absolute Error = 76%

 

39

distance segmentation is given in Table 4. Examination shows
that the absolute error per distance category has been signifi-
cantly reduced by the waitrtime factor. In the first category
demand is still overestimated but in the next two categories a
mixture of overestimation and underestimation is evident. As in
the analysis performed without a waitetime factor, the two long
distance categories are characterized by consistent underestima-
tion by the model.

City size analysis is shown in Table 5. Again, no discern-
ible pattern is evident.

Comparison of the results from these analyses reveals that
the wait¥time factor greatly improves model predictive ability
for the short distance pairs, but tends to add to the tendency
for underestimation of longer distance pairs. Thus the non-wait-
time calculations prove a better predictor for city pairs over
80 miles apart.

In analyzing the results for the Kalamazoo-Battle Creek ser-
vice it was found without the wait-time factor, that the service
characteristics of rail and the highway mode were very close,
given that the frequency variable has minimal change impact.

The rail mode was only 10 minutes longer and 0.28 higher than
the highway mode. The wait-time factor in this case introduced
the inconvenience factor of the mode and therefore provided a
more realistic prediction.

Based on these results, it can be concluded that the model
has a greater tendency to overestimate demand between short

distance pairs than to underestimate demand between long distance

40

pairs. Considering only these results, it seems advisable that
a permanent distance segmentation factor should be added to the
model.

The results from these analyses compares favorably with
conclusions made by Billheimer and by Peat, Marwick and Mitchell
that the model clearly overstates short distances and under-
states long distance pairs.

Preliminary consideration was given to whether generated
demand was diverted or induced. In that the highway model is
very finely tuned, the generated results and observed results
for this mode were compared against generated and observed rail
results when the model was run. In almost all cases very little
change occurred in generated highway traffic even though rail
traffic accounted for ten or more passengers. Examples of
this analysis are shown in Table 6. These results appear to
coincide with Billheimer and Peat, Marwick and Mitchell's con-
clusions that the model is failing to consider diverted demand

and almost exclusively generates induced demand.
3. Bus Results and Analyses

The data used to test the SR1 model for the bus mode is
from a ticket survey at the Lansing-East Lansing terminals con-
ducted April 6, 1977. The survey was administered by Dennis
Hill of the Mass Transportation Planning Section, Michigan Depart-
ment of Transportation. A total of 211 ticket stubs (119 sold
at Lansing, 92 at East Lansing) were obtained from management

at the end of the service day. One hundred sixty-eight tickets

41

b’

TABLE 6

EXAMPLES OF
INDUCED VERSUS DIVERTED DEMAND

FROM RAIL MODELING RESULTS

 

CITY PAIRS MODE
Battle Creek-Detroit Highway
Rail
Jackson-Detroit Highway
Rail
Lansing-Kalamazoo Highway
Rail

42

 

GENERATED ACTUAL
9.329 9.709
3.784 4.250

46.866 48.260
9.553 5.226
58.157 58.470
2.287 3.290

. -4-
.-’

(79.6%) of the tickets sold were for intrastate travel. Bus
origin-destinations are shown in Appendix D. A data set was

constructed using Russell's Official National Motor Coach Guide,

 

April, 1977. The data set is given in Appendix B. Cost data
was obtained from station managers at the two terminals. Cost
information for city pairs served by more than one bus line was
averaged. The two terminals, for calculation purposes, were
consolidated in that all buses service both stations. All city
pair data was combined and input variables were adjusted accord-
ingly.

Initial model calculations were performed by experimentation
with various combinations of market areas for each city pair.
Market bands resulting in the best prediction for each origin—
destination were chosen and are shown in Appendix B. Due to the
generally short distances traveled on bus only a minimum amount
of driving bands was established.

Analysis based on an absolute error calculation was performed
on the model results according to distance segmentation and given
in Table 7. In accordance with the rail analyses, the model
severely overestimates demand. This tendency decreases incre-
mentally as distance decreases but unlike the rail results, con-
sistent underestimation of longer distance is not evident. City
size analysis is shown in Table 8. No discernible pattern is
evident and all error magnitudes are assumed to relate to the
distance function.

A wait-time factor based on a 12-hour service day was in-

cluded in further calculations. Analysis on the results according

43

TABLE 7

DISTANCE SEGMENTATION ANALYSIS
WITHOUT A WAIT-TIME FACTOR

ON

BUS MODELING RESULTS

 

 

 

 

 

 

 

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR
City Pair Distances €140 Miles
Lansing/East Lansing-
-Owosso 25.3 6 +19.3
-Charlotte 29.6 2 +18.7
-Fowlerville 6.7 2 + 4.7
-Mason 29.9 1 +28.9
-Albion 1 l 0
Total Absolute Error = 661.3%
City Pair Distances 7’40 £160 Miles
Lansing/East Lansing-
-F1int 63.4 5 +58.4
-Marsha11 4.4 2 + 2.4
-Battle Creek 2.4 2 +22.8
-Lowe11 1.4 l + 0.4
~Farmington 2.5 4 + 2.1
-New Hudson 11.6 1 +10.6
-Brighton 12.7 1 +1l.7
Total Absolute Error = 783.1%
City Pair Distances 760 5:. 80 Miles
Lansing/East Lansing-
-Mt. Pleasant 7.3 11 - 3.7
—Grand Rapids 69.7 13 +56.7
-Saginaw 10.6 8 + 2.6
-Ypsilanti 2.3 3 - 0.7
-Ann Arbor 17.6 10 + 7.6
-Kalamazoo 19.8 3 +16.8
-Pontiac 4.1 2 + 2.1
-C1are 1.2 2 - 0.8
-Tecumseh 2.3 1 + 1.3
-Coldwater l 1 0
Total Absolute Error = 170.9%

 

44

Table 7 (cont'd.)

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR

 

 

City Pair Distances 7 80 52.130 Miles

 

Lansing/East Lansing-
-Detroit 6
-Mid1and
-Muskegon
-Port Huron
-Cadillac
-Grand Haven
-Holland
-South Haven
-Benton Harbor
-Imlay City

\D-bI-J

KO
l+++|++l+
OOOI—‘I—‘OI—‘OOIb
O \O\II-'

\O \O

ox

OONNNONwa
\O
I-‘

mmwm
HHNHHHwam

Nowoo
.5

Total Absolute Error = 170.9%

 

City Pair Distances 77130 Miles

 

Lansing/East Lansing-

-Traverse City 1.1 l + 0.1
-Gaylord 0.17 1 - 0.23
-L'Anse 0 l - 1

Total Absolute Error = 64.3%

 

45

4‘)

'0'

A

B

C

D

TABLE 8

CITY SIZE ANALYSIS

WITHOUT A WAIT-TIME FACTOR

BUS MODELING RESULTS

CITY PAIR

ON

 

Lansing/East Lansing-

-Detroit

Lansing/East Lansing-

-Grand Rapids
-Saginaw

-Ann Arbor
-Kalamazoo
-F1int
-Pontiac

Lansing/East Lansing-

-Jackson
-Mid1and
-Muskegon
-Port Huron
-Battle Creek

Lansing/East Lansing-

-Mt. Pleasant
-Ypsilanti
-Owosso
—Charlotte
-Marsha11
-Cadi11ac
-Mason
-Traverse City
-Tecumseh
-A1bion

-Grand Haven
-Holland
-Farmington
-South Haven
-Benton Harbor
-Coldwater

 

 

 

ABSOLUTE
GENERATED ACTUAL ERROR
63.1 59 + 4.1
69.7 13 +56.7
10.6 8 + 2.6
17.6 10 + 7.6
19.8 3 +16.8
63.4 5 +58.4
4.1 2 + 2.1
Total Absolute Error = 351.7%
25.3 6 +19.3
2.4 3 - 0.6
4.99 4 + 0.99
2 l + 1
24 82 +22.8
Total Absolute Error = 279.3%
7.3 11 - 3.7
2.3 3 - 0.7
20.7 2 +18.7
29.6 2 +27.6
4.4 2 + 2.4
0.91 1 - 0.09
29.9 1 +28.9
1.1 1 + 0.1
2.3 1 + 1.3
1 l 0
2 1 + 1
2.8 1 + 1.8
25.4 4 +21.4
2.3 2 + 0.3
0.96 1 - 0.04
l l 0
Total Absolute Error = 450.3%

 

AR

Table 8 (cont'd.)

 

 

ABSOLUTE
CITY PAI R GENE RATED ACTUAL ERROR
Lansing/East Lansing-

E -C1are 1.2 2 - 0.8
-Im1ay City 0.8 1 - 0.2
-Fow1ervi11e 6.7 2 + 4.7
-Lowell 1.4 1 + 0.4
-Gaylord 0.17 1 - 0.83
-L'Anse 0 1 - 1
-New Hudson 11.6 1 +10.6
-Brighton 12.7 1 +ll.7

Total Absolute Error = 302.3%

 

47

TABLE 9

DISTANCE SEGMENTATION ANALYSIS
WITH A lZ-HOUR WAIT-TIME FACTOR
ON
BUS MODELING RESULTS

 

 

 

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR
City Pair Distances €340 Miles
Lansing/East Lansing-
-Jackson 8.3 6 + 2.3
-Owosso 5.3 2 + 3.3
-Charlotte 5.1 2 + 3.1
-Fow1erville 0.8 2 - 1.8
-Mason 1.1 1 + 0.1
-A1bion 0.3 2 - 1.7

Total Absolute Error = 82%

 

City Pair Distances ’740 {'1 60 Miles

Lansing/East Lansing-
-F1int 3
-Marshall
-Battle Creek
-Lowe11
-Farmington
-New Hudson
-Brighton

4.
w

wrouwowoc>m
wId~Jw
dedhtkoNLn
+-+-+I + I
NrdkuaxJHFH
O O O O
W\O\JQ

Total Absolute Error = 285%

 

City Pair Distance 7760 £180 Miles

Lansing/East Lansing-
-Mt. Pleasant
-Grand Rapids
-Saginaw
-Ypsilanti
-Ann Arbor
-Ka1amazoo
-Pontiac
-C1are
-Tecumseh
-Coldwater

11
13

U)
0 O

OOONNQI—‘OAU'IQ
II-+
N

[—l
O
owmwwmmmwb

I ++ l
OOI—J \DNI—‘I-‘Nm

o o o
o o o o o o o
\oqmwwI—Ioooowm

I—‘I—‘NNUJOUJCD

1

Total Absolute Error = 87.87%

 

48

Table 9 (cont'd.)

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR

 

City Pair Distances ‘780 £L130 Miles

 

Lansing/East Lansing-
-Detroit 4
-Midland
-Muskegon
-Port Huron
-Cadillac
-Grand Haven
-Holland
-South Haven
-Benton Harbor
-Imlay City

I
'—-I

I
COOOOOOI—‘I-‘w
O O O O O O O O O O

\IwmI—‘wammb

O O O O
FJHIOFJHF‘PMbUJO
I

OOI—‘I—‘OOONI—‘U‘l
WQNHWONODNNm

Total Absolute Error = 27.3%

 

City Pair Distances 77130 Miles

 

Lansing/East Lansing-

-Traverse City 0.7 l - 0.3
-Gaylord 0.6 1 — 0.94
—L'Anse 0.003 1 - 0.997

49

TABLE 10

CITY SIZE ANALYSIS
WITH A 12-HOUR WAIT-TIME FACTOR
ON
BUS MODELING RESULTS

 

 

ABSOLUTE
CITY PAIR GENERATED ACTUAL ERROR
Lansing/East Lansing-
A -Detroit 45.6 59 -13.4
Lansing/East Lansing-
B -Grand Rapids 35.3 13 +22.3
-Saginaw 6.2 8 - 1.8
-Ann Arbor 7.9 10 - 2.1
-Ka1amazoo 12.3 3 + 9.3
-F1int 36 5 +31
—Pontiac 2.3 2 + 0.3

Total Absolute Error = 163%

 

Lansing/East Lansing-

C -Jackson 8 3 6 + 2.3
-Midland 1.2 3 - 1.8
-Muskegon 2.2 4 - 1.8
-Port Huron 0.8 l - 0.2
-Battle Creek 9.7 2 + 7.7

Total Absolute Error = 86.3%

 

Lansing/East Lansing—

D -Mt. Pleasant 4.4 11 - 6.6
—Ypsi1anti 1.2 3 - 1.8
-Owosso 5.3 2 + 3.3
-Charlotte 5.1 2 + 3.1
-Marsha11 0.3 2 - 1.7
-Cadi11ac 0.6 1 - 0.4
-Mason 0.1 1 + 0.1
-Traverse City 0.7 1 - 0.3
-Tecumseh 0.3 1 - 0.7
-A1bion 0.3 2 — 1.7
-Grand Haven 0.3 1 - 0.7
-Holland 1.1 1 + 0.1
-Farmington 5.3 4 + 1.3
-South Haven 1.2 2 - 0.8
-Benton Harbor 0.7 l - 0.3 .
-Coldwater 0.01 1 - 0.99

Total Absolute Error = 66.4%

 

50

Table 10 (cont'd.)

 

 

ABSOLUTE
C ITY PAI R GENE RATE D ACTUAL E RROR
Lansing/East Lansing-

E -C1are 2.3 2 - 1.5
-Im1ay City 0.3 l - 0.7
-Fowlerville 0.8 2 - 1.8
—Lowell 0.1 l - 0.9
-Gaylord 0.06 l - 0.94
-L'Anse 0.003 1 - 0.99
-New Hudson 2 1 + 1
—Brighton 3 l + 2

Total Absolute Error = 98.3%

 

51

to distance segmentation is given in Table 9. Examination
shows that the absolute error per distance category has been
significantly reduced. City size analysis is shown in Table 10.
Again, absolute error functions do not appear to correlate to
city size categories.

Comparison of the results reveals that the wait-time factor
greatly improves model predictive ability for distances less
than 80 miles. The model's tendency to reduce overestimation
at long distances resulted in better predictions over 80 miles
without the wait-time factor, as likewise in the rail analysis.

In analyzing the results from a specific pair, it was found
without the wait-time factor, the service characteristics of
rail and highway mode were very close. Between Lansing and
Grand Rapids the bus mode was only 23 minutes longer and $1.50
higher than the highway mode. The waitrtime factor in this case
introduced the inconvenience factor of the node and, therefore,
provided a more realistic prediction.

The results from bus analyses do not differ in general from
those of the previous node and, thus, the recommendation for a
permanent distance segmentation factor in the model still seems
advisable. The results also coincide with the findings of Peat,
Marwick and Mitchell in that consistent overestimation is evi-
dent in most city pair calculations.

Preliminary consideration was given to the source of bus
passengers; i.e., whether travelers were diverted or induced
to the mode. Examples of this analysis are shown in Table 11

and coincide with rail findings. The model appears to

52

overestimate induced demand at the expense of diverted demand

from the dominant mode, the automobile.

4. Aviation Results and Analyses

The data used to test the SR1 model for the air mode is

from an airline passenger survey conducted by the Michigan Aero-
nautics Commission in conjunction with SR1. The survey was
conducted January 24-30, 1972. Ticket accounts were accumulated
at the end of the week by the State Airport System Planning
Section. Mr. Edward Mellman of the Planning Section supplied
to the author the survey data. Three Michigan airports were
selected to be surveyed:

Lansing Capitol City Airport

Flint's BishOp Airport

Grand Rapids' Kent County Airport

Airlines included in the survey were:

Lansing - United
North Central

Flint - United
North Central

Grand Rapids - Allegheny
United
North Central
Table 12 shows the number of intrastate travelers during the sur-
vey week per airport. Intrastate aviation movements are shown
per airport in Appendix D.
The data was coded for keypunching and transferred to a com-
puter disc. Trip tables were formed and the data was analyzed.

Input variable data was then constructed utilizing the Official

Airline Guide, North American Edition, September, 1973. This

53

TABLE 11

EXAMPLES OF
INDUCED VERSUS DIVERTED DEMAND
FROM BUS MODELING RESULTS

 

 

 

CITY PAIR MODE GENERATED ACTUAL
Lansing-East Lansing-
-Detroit Highway 306.793 321.738
Bus 45.56 59
-Grand Rapids Highway 111.844 115.202
Bus 35.2 13
-Saginaw Highway 39.1 39.7
Bus 6.236 8

54

TABLE 12

INTRASTATE AVIATION TRAVELERS
SURVEYED
JANUARY 24-30, 1972

 

 

 

AIRPORT PASSENGERS
Lansing 1,218
Flint 111
Grand Rapids 3,151
Total 4,480

55

edition was the closest available information to the actual
survey data. Information for both direct and indirect flights
was obtained. In the creation of input variables for flights

with connections not stated in the Airline Guide frequencies for

 

a given origin and destination were calculated by first deter-
mining the shortest travel time and then only connections rela-
tively close to this time were considered for input. In calcu-
lating travel time, care was taken to account for the two different
time zones in Michigan. Cost figures were cross-checked on some
routes with information obtained from Ms. Kay Lund, Director of
Consumer Affairs, United Air Lines District Office, Chicago,
Illinois. It is not unlikely, however, that connecting flights
that were created but were not available for cross-checking may
be slightly higher than the actual ticket price. A data set for
considered origins and destinations is given in Appendix B.

Upon analyzing the trip tables it was noticed that many pairs
had unexpected volumes; for example, Lansing to Marquette recorded
576 tirps. The survey data was cross-checked with the closest
available origin-destination data for the subject pairs. Aver—
age weekly travel for the survey week and for 1975 is shown in
Appendix C. Observation of the data reveals that flights to
destinations in the Upper Peninsula are from two to six times
higher than the weekly average. Given the calendar time of the
survey, these volumes probably reflect ski trips to winter re-
sorts. Another unexpected pair volume occurred from Grand
Rapids to Detroit. The survey data is five times the weekly
average. This variance may be due to conventions or other irreg-

ular events.

56

The survey data was received in a weekly aggregate per
airport. The data was converted into daily volumes in that
the SR1 model is designed for daily calculations. A slight
amount of error was introduced into the analysis in that daily
flights per a given pair on a weekend versus a weekday basis
may differ. This, however, was not felt to unreasonably distort
the data.

Initial model calculations were performed by experimentation
with various combinations of market areas for each pair. Driv-
ing bands resulting in the best prediction for each origin-
destination pair were chosen and are shown in Appendix B. In

general, time bands for city pairs were distributed as follows:

 

 

City Pair Distance Market Area
<80 miles (0.0)

780 é; 130 miles (10,10)
7130 £250 miles (20,20)
7250 - 500+ miles (30,30)

It is logical for distances under 2-28 hours auto driving times
that travelers will not drive very far to get to an airport
and that distances requiring from 5-10 hours driving time trav-
elers will drive up to 30 minutes to board a plane.

Analysis based on absolute error calculations was performed
on the model results according to distance, segmentation and
is given in Table 13. Examination shows the model is predicting
very poorly. For distances under 130 miles, volumes are severely
underestimated. Because Cd'the magnitude of error, individual

distance segment error was not calculated. City pair analysis

57

(not shown) reveals that city size interactions do not
significantly contribute to error.

A waitstime factor based on a l6—hour service day was in-
cluded in the model calculations to correct for overestimations.
Results according to distance segmentation are shown in Table 14.
Overestimation of volumes for pair distances under 130 miles
were significantly reduced by the wait-time factor, but predic-
tions still differ significantly from the observed.

Based on the results from these analyses, model adjustments
are advisable. A permanent distance segmentation factor should
be attached to the model to modify the tendency to overestimate
short distance pairs and underestimate long distance pairs.

Many of the high volume destinations, particularly in the
Upper Peninsula, do not reflect the socio-economic characteris-
tics of the inhabitant but rather recreational attractions.

The model as currently constructed is unable to accommodate such
considerations. Moreover, to attempt to calibrate the current
model construction to fit such variations would destroy its
predictive capabilities for non-resort destinations. It is ad-
visable that an additional variable sensitive to resort destina-
tion volumes be added to the model. In general, the special
amenities of many Michigan cities in the northern Lower Peninsula
and the Upper Peninsula introduce many complications in the
modeling effort. This unique factor is particularly evident in
the air mode due to the attractiveness of air travel in winter

months.

58

TABLE 13

DISTANCE SEGMENTATION ANALYSIS
WITHOUT A WAIT-TIME FACTOR
ON
AVIATION MODELING RESULTS

CITY PAIR GENERATED ACTUAL

 

City Pair Distances 440 Miles

 

 

 

 

Flint-Saginaw 45.7 0.43
Lansing-Jackson 0.38 0
Grand Rapids-Muskegon 8.6 0
City Pair Distances 740.5 60 Miles
Flint-Lansing 27.6 0
Flint-Detroit 61.8 0.3
Lansing-Flint 29.7 0
Grand Rapids-Kalamazoo 0.46 0
City Pair Distances 760 f". 80 Miles
Flint-Jackson 1.77 0
Lansing-Grand Rapids 6.9 0.14
Lansing-Saginaw 2.3 0
Lansing-Kalamazoo 0.8 0
Grand Rapids-Lansing 50.9 0.3
City Pair Distances '780 5 130 Miles
Lansing-Muskegon 34.1 0
Lansing-Detroit 79 1.6
Lansing-Benton Harbor 2 0
Grand Rapids-Saginaw 17.4 3.4
Grand Rapids-Manistee 1.4 0
Grand Rapids-Benton Harbor 9 0
Grand Rapids-Flint 2 0
Grand Rapids-Jackson 0.12 0
Flint-Kalamazoo 3.4 0.14
Flint-Grand Rapids 6.9 0.4
City Pair Distances ”7130.£:190 Miles
Flint-Muskegon 6 0
Flint-Alpena 0.15 0.3
Flint-Traverse City 0.8 5.1
Flint-Benton Harbor 0.7 0
Flint-Manistee 0.1 0
Lansing-Traverse City 0.6 1.7
Lansing-Manistee 0.2 0
Grand Rapids-Traverse City 4.5 0
Grand Rapids-Detroit 170.4 333.7

)1
D

ABSOLUTE
ERROR

+45.3
+ .38
+ 8.6

+-I+-+-I

O O O O O O O
WU'lNI-‘I-‘QLAJP-‘ON
U'I

O‘ncahdococsoco

¢++I++II+

Table 13 (cont'd.)

£9

 

 

 

 

 

 

ABSOLUTE

CITY PAIR GENERATED ACTUAL ERROR
City Pair Distances 7190 5:. 250 Miles

Flint-Pellston 0.1 0 + 0.1
Lansing-Alpena 0.2 0 + 0.2
Lansing-Pellston 0.1 0 + 0.1
Grand Rapids-Pellston 0.02 0 + 0.0
Grand Rapids-Alpena 0.2 0 + 0.2
City Pair Distances '7250 ‘5‘ 350 Miles

Flint-Sault Ste. Marie 0.2 8.6 - 8.4
Lansing-Escanaba 0.6 19.3 -18.7
Lansing-Sault Ste. Marie 0.13 10 - 9.9
Grand Rapids-Sault Ste. Marie 0.43 2.1 - 1.7
City Pair Distances 7350 £- 450 Miles

Flint-Marquette 0.3 0.6 - 0.3
Flint-Menominee 0.1 0 + 0.1
Flint-Iron Mountain 0.2 0 + 0.2
Flint-Escanaba 0.2 0 + 0.2
Lansing-Menominee 0.1 2 - 1.9
Lansing-Marquette 0.6 82.3 -81.7
Lansing-Iron Mountain 0.3 33 -32.7
Grand Rapids-Escanaba 0.9 15.4 -14.5
Grand Rapids-Marquette 1 46.3 -45.3
Grand Rapids-Menominee 0.5 5.1 - 4.6
Grand Rapids-Iron Mountain 0.6 15.4 -14.8
City Pair Distances 77450 Miles

Flint-Hancock 0.1 0 + 0.1
Flint-Ironwood 0.1 0 + 0.1
Lansing-Hancock 0.1 18 -17.9
Lansing-Ironwood 0.1 6 - 5.9
Grand Rapids-Hancock 0.2 28.3 -28.1
Grand Rapids-Ironwood 0.2 0 + 0.2

60

n?

TABLE 14

DISTANCE SEGMENTATION ANALYSIS
WITH A l6-HOUR WAIT TIME FACTOR

ON

AVIATION MODELING RESULTS

CITY PAIR

 

GENERATED

 

Citngair Distances ‘540 Miles

Flint-Saginaw

8.7
Lansing-Jackson 0.04
0.2

Grand Rapids-Muskegon

City Pair Distances ;740 Ei60 Miles

Flint-Lansing 0.6
Flint-Detroit 15.3
Lansing-Flint 0.6
Grand Rapids-Kalamazoo 0.1

City Pair Distances 760 6 80 Miles

Flint-Jackson 0
Lansing-Grand Rapids 5
Lansing-Saginaw 0
Lansing-Kalamazoo 0
Grand Rapids-Lansing 7

City Pair Distances '780 £5130 Miles

Lansing-Muskegon

Lansing-Detroit 3
Lansing-Benton Harbor

Grand Rapids-Saginaw

Grand Rapids-Manistee

Grand Rapids-Benton Harbor

Grand Rapids-Flint

Grand Rapids-Jackson
Flint-Kalamazoo

OOOI—‘OOOOU’I

ONONQIbONIbUI

City Pair Distances '7130 £3190 Miles

 

Flint-Muskegon
Flint-Alpena
Flint-Traverse City
Flint-Benton Harbor
Flint-Manistee
Lansing-Traverse City
Lansing-Manistee

Grand Rapids-Traverse City
Grand Rapids-Detroit

a o o O O O O O
moon—Imow

WOOOOOOOO

\O

51

NNIbN

DJ
U)

ACTUAL

COO

OOOOO COCO
O
U)

o
(h

OOOOONOI—‘O
O
.b

o
'—l
b

o 0
HO)

WOOI-‘OOU'IOO
O
\I

o
\I

ABSOLUTE
ERROR

I + +-I+-+-+I +-+4- +-+-I+-+ +-I+-+ -+4-+

+-+I +-+I

I
N
(A)

0000

q<5c>m<3
O O O O O
enunappu

OOOI—‘ONOWU‘I
U'lowdhmubko

O O O O O
\INOIbOI-‘Koww
Nth

mOOI-‘OOIbOO
N

Table 14 (cont'd.)

 

 

 

 

 

 

ABSOLUTE

CITY PAIR GENERATED ACTUAL ERROR
City Pair Distances 7190 ‘5; 250 Miles
Flint-Pellston 0.02 0 + 0.0
Lansing-Alpena 0.3 0 + 0.3
Lansing-Pellston 0.2 0 + 0.2
Grand Rapids-Pellston 0 0 0
Grand Rapids-Alpena 0.1 0 + 0.1
City_Pair Distances ‘7250 £1350 Miles
Flint-Sault Ste. Marie 0 8.6 - 8.6
Lansing-Escanaba 0.2 19.3 - 19.1
Lansing-Sault Ste. Marie 0.1 10 - 9.9
Grand Rapids-Sault Ste. Marie 0.1 2.1 - 2
City Pair Distances 7350 5.450 Miles
Flint-Marquette 0.1 0.6 - 0.5
Flint-Menominee 0 0 0
Flint-Iron Mountain 0 0 0
Flint-Escanaba 0 0 0
Lansing-Menominee 0 2 - 2

I Lansing-Marquette 0.2 82.3 - 82.1
Lansing-Iron Mountain 0.1 33 - 32.9
Grand Rapids-Escanaba 0.4 15.4 - 15
Grand Rapids-Marquette 0.4 46.3 - 45.9
Grand Rapids-Menominee 0.1 5.1 - 5
Grand Rapids-Iron Mountain 0.1 15.4 - 15.3
City Pair Distances 7'450 Miles
Flint-Hancock 0 0 0
Flint-Ironwood 0 0 0
Lansing-Hancock 0.1 18 - 17.9
Lansing-Ironwood 0 6 - 6
Grand Rapids-Hancock 0 28.3 - 28.3
Grand Rapids-Ironwood 0 0 0

Q

62

£9

The results from these analyses are more difficult to
interpret due to the large error factor. It is apparent, how-
ever, that short distance pairs are overstated and long dis-
tance pairs understated. This conclusion is in concurrence with
findings from previous modes. These results, however, differ
from the conclusions of Peat, Marwick and Mitchell who found
the air mode to be the most precise prediction of the three
modes.

Table 15 shows the model, as discussed with previous modes,

appears to attribute disprOportionate values to induced demand.

63

«:1!

TABLE 15

EXAMPLES OF

INDUCED VERSUS DIVERTED DEMAND
FROM AVIATION MODELING RESULTS

 

CITY PAIR MODE

Flint-Saginaw Highway
Air

Flint-Detroit Highway
Air

Lansing-Grand Rapids Highway
Air

Lansing-Detroit Highway
Air

64

 

GENERATED ACTUAL
1467.827 1470.688
8.7 3
101.2 102.5

15.3 2
135.9 136.6
5.1 1
356.2 368.4
30.5

11

at

("1,1

V . S UMMARY

This research has evaluated and tested, using Michigan-based
data, the Stanford Research Institute's Intercity Passenger De-
mand/Modal Split Model. Unfortunately only limited comparison
of these results with other intercity modeling research was pos-
sible. Much transportation literature only discusses, in theory,
statewide modeling. Research concerned with multi-modal model-
ing is usually characterized by a small data base. These re-
sults do compare favorably with two sources of published research
on the model. It has been shown the model:

1. Overestimates demand for short-distance city pairs.
2. Underestimates demand for long-distance city pairs.

3. City size differences do not significantly affect
the model's performance.

4. Bus demand is consistently overestimated.

5. Induced demand is overestimated to the detriment of
derived demand.

Two variations of the SR1 model were used which served to aug-
ment it. Driving time bands increase the market area for a given
model terminal and provide a more realistic measure of the at-
tractiveness of the service to surrounding p0pu1ations. A wait
time factor based on the length of a common carrier mode's ser-
vice day, tends to compensate for the model's tendency to over-
estimate demand and introduces into the model the inconvenience

factor of the mode.

65

a?

R

The model aggregates "quality" variables of the common
carrier modes, such as comfort, safety, reliability, into one
parameter in the formula. This aggregation may be too gross
to reasonably reflect reality. The model considers both time
and cost as input variables. Some degree of predictive ability
is lost because of multi—collinearity; however, for policy-
testing purposes, it may be necessary to retain both variables.
Time and cost are treated as equal variables in terms of ability
to influence changes in demand. Some studies, however, have
shown time to be a significantly more important variable.

Finally, the socio-economic statistic used in demand fore-
casting by the model was families with incomes of $10,000 or
more in 1970. This measure needs to be updated to reflect more

current per capita income levels.

CONCLUSIONS

The conclusions found through this research may be summarized
as follows:

1. The model needs a permanent distance segmentation
function. In the results from all three modal model-
ing efforts, it was found that short distance pair
demand was overestimated and that demand between
long distance pairs was underestimated.

2. A measure to more adequately distribute induced
versus derived demand is needed. Again, in all
three modal results, it was discovered almost all
the generated demand for non-highway modes come from
induced demand. A minimal amount of demand was di-
verted from generated automobile demand to common
carrier modes.

3. An additional variable sensitive to resort areas in

Michigan needs to be augmented, especially with ref-
erence to air travel. Many of the high volume

66

destinations in air travel were found not to be

a reflection of the socio-economic characteris-
tics of the inhabitant, as the model presupposes
but rather due to the special amenity factors of
the area, such as ski facilities, water recreation
Opportunities, etc.

4. The city size adjustment factor, G, which segments
pairs having large pOpulation products from those
having smaller products, appears to be functioning
adequately. In city size tests performed on all
modal results, no consistent pattern of error was
evident.

5. With apprOpriate adjustments the model can be used
to forecast horizon-year modal volumes. The ease
of changing levels of the input variables makes

the model especially attractive for testing policy
alternatives.

FUTURE RESEARCH NEEDS

In the conduct of this research several areas requiring addi-
tional investigation were discovered. Further work on this model
should use as recent data as possible. Data in this research
ranged from 1972 aviation data to 1975 rail data. It should not
be difficult to cull recent rail data. Up-to-date aviation and
bus data, however, may require time consuming passenger surveys.
Effort should be exerted to correlate the time periods of the
data as close as possible. A further refinement of the research
herein would be to run the computer program simultaneously for
all four modes so that more acceptable multi-modal comparisons
can be performed. Experimentation with the adding of "quality"
modal attributes, such as comfort, safety, time dependability,
should be explored. This may result in the addition of other
parameters to the equation. The use of a constant price elastic-

ity in the model needs further research confirmation. The model

67

assumes that a doubling of ticket price will affect all income
groups similarly. This assumption does not appear to be reason-
able. Finally, more evaluation on statewide modeling procedures
is necessary so that more precise results are available as to
wiether statewide aggregate modeling sufficiently reflects behav-
ioral characteristics of the pOpulation. Disaggregate modeling
prOponents argue that much accumulated error is contained in ag-
gregate modeling and that modeling results reflect peculiarities
of a particular data set and prediction equations may not be

transferred effectively to different data sets.

68

4.

FOOTNOTES

1John W. Billheimer, Stanford Research Institute, The
Michigan Intercity Passeoger Demand Model, June, 1971.

 

2Transportation Research Board, Special Repot 146, Issues
in Statewide Transportation Planning, (Washington, D.C.;
National Research Council, 1974).

 

3Peter R. Stopher and Joseph N. Prashker, "Intercity Pas-
senger Forecasting: The Use of Current Travel Forecasting
Procedures", Transportation Research Forum, 1976.

 

4Philip I. Hazen, A Comparative Analysis of Statewide
Transportation Studies (Evanston, Illinois: Northwestern Uni-
versity, 1971). An unpublished M.S. Thesis.

 

 

5Ibid., and Transportation Research Institute—-Carnegie
Mellon University and Pennsylvania Transportation and Traffic
Safety Center--Pennsylvania State University; Methodological
Framework for Comprehensive Transportation Planning, pp. 90-92.

 

 

6Robert A. Nelson, Paul W. Shuldiner, Myron Miller, Miller
Stinchcombe and Robert L. Winestone, Northeast Corridor Trans-
portation Project Report 209 (Washington, D.C.: U.S. Government
Printing Office, April, 1970), pp. 8-20.

7David Arthur Brown, An Intercity Passenger Transportation
Demand Model (Stanford, California, 1969), pp. 19-20.

 

 

3a. A. Josephs, D. M. Hill, N. A. Irwin, J. M. McLynn,
R. H. Watkins and Arrigo Mongini, Northeast Corridor Transpor-
tation Project Technical Paper No. 7, Approaches to the Modal
Split: Intercity Transportation (Washington, D.C.: U.S. Govern-
ment Printing Office, February, 1967), pp. 32-33.

9Billheimer, loc. cit.

10;p;g., pp. 7—8; 11-12.

llipig., p. 15.

12H. C. W. L. Williams, "Travel Demand Models, Duelity

Relations and User Benefit Analysis," Journal of Regional
Science, 1976, p. 310.

 

69

Footnotes (cont'd.)

13John C. Bennett, Raymond H. Ellis and John C. ProkOpy,
Peat, Marwick, Mitchell and Co., A Comparative Evaluation of
Intercity Modal Split Model (date not available): U.S. Depart-
ment of Transportation.

 

 

14The following discussion derives from the seven listed
publications of Richard E. Esch listed in the bibliography.

70

,T‘x

SELECTED BIBLIOGRAPHY

Books and Published Reports

 

Amtrak. Midwest Corridors Train Timetables Effective October 31,
1976.

 

Brown, David Arthur. An Intercity Passenger Transportation
Demand Model; Stanford, California; An unpublished PhD
Thesis for the Department of Regional and City Planning,
Stanford University, 1969.

 

 

Creighton, Roger. Michigan Scheduled Air Service Study, Final
Technical Report, September, 1977.

 

Department of Transportation, State of New York. Statewide
Master Plan for Transportation (date not available).

 

 

Donaldson, John. Northeast Corridor Transportation Project,
Report 213. Washington, D.C.: U.S. Government Printing
Office, December, 1969.

 

Donnelley, Reuben H., Inc. Official Airline Guide, North American
Edition, September, 1973.

Esch, Richard E., and Statewide Transportation Planning Procedures
Section, Michigan Department of Transportation.

Statewide Socio-Economic Data File, March, 1973.

 

Statewide Public and Private Facility File, January, 1974.

 

Michigan Goes Multi-Modal, July, 1974.

 

Statewide Travel Impact Analysis Procedures, October, 1974.

 

Multi-Modal Mobility and Accessibility Analysis, November,
1974.

Statewide Socio-Economic and Transportation Resources and
Their Role in Intercity Transportation Decisions, November,
1974.

Amtrak Market Area Analysis System Application, October,
1976.

71

Selected Bibliography (cont'd.)

Fertal, Martin J. and Ali F. Sevin. Estimatipg Transit Useage:
Modal Split. Washington, D.C.: U.S. Government Printing
Office, 1967.

 

Fertal, Martin J. and Edward Weiner, Miller Stinchcombe and Ali
F. Sevin. Modal Split: Documentation of Nine Methods for
Estimating Transit Useage. Washington, D.C.: U.S. Govern-
ment Printing Office, 1966.

Hazen, Philip I. A Comparative Analysis of Statewide Transpor-
tation Studies. Evanston, Illinois: Northwestern Univer-
sity, 1971. An unpublished Masters of Science Thesis for
the Department of Civil Engineering.

 

Highway Research Board, Highway Research Record Number 264:
Statewide TranSportation Plannipg. Washington, D.C.:
National Research Council, 1969.

Highway Research Board, Highway Research Record Number 401:
Intermodal Transportation Planning of the State, Multi-
State and National Scale. WaSHington, D.C.: National
Research Council, 1972.

 

Highway Research Board, Highway Research Record Number 422:
Land Use and Transportation Planning. Washington, D.C.:
National Research Council, 1973.

 

Highway Research Board, National COOperative Highway Research
Program, Synthesis of Highway Practice 15, Statewide Trans-
portation Planning: Needs and Reguirements. Washington,
D.C.: National Research Council, 1972.

 

 

Josephs, J. A., D. M. Hill, N. A. Irwin, J. M. McLynn, R. H.
Watkins and Arrigo Mongini. Northeast Corridor Transpor-
tation Project Technical Paper No. 7, Approaches to the
Modal Split: Intercipy Transportation. Washington, D.C.:
U.S. Government Printing Office, February, 1967.

 

 

McLynn, J. M., J. R. Endriss, R. H. Watkins, and D. E. Smith.
Northeast Corridor Transportation Project, Analysis and
Calibration of a Modal Allocation Model (Revised).
Washington, D.C.: U.S. Government Printing Office, June,
1967.

 

McLynn, J. M. and T. T. Woronka. Northeast Corridor Transpor-
tation Project Report 230: Passenger Demand and Modal
Split Models Calibration and Preliminary Analysis. Washington,
D.C.: U.S. Government Printing OffICe, December, 1969.

 

 

72

Selected Bibliography (cont'd.)

Michigan Aeronautics Commission, Department of Commerce in
conjunction with Stanford Research Institute. Airline

Passenger Surve at Selected Michigan Airports. Lan51ng,
Michigan, June, I972.

National Railway Publication Company. The Official Railway
Guide, North American Passenger Travel Edition, October,
1975.

Nelson, Robert A., Paul W. Shuldiner, Myron Miller, Miller
Stinchcombe and Robert L. Winestone. Northeast Corridor
Transportation Project Report 209. WashIngton, D.C.:
U.S. Government Printing Office, April, 1970.

New York State Department of Transportation. IntercityoPas-
senger Demand Models: State-of-the;Art, Volume 1.
Washington, 575.: U.S. Government Printing Offlce,
May, 1977.

 

Richards, Martin G. and Moshe E. Ben-Akiua. AoDisaggregate
Travel Demand Model. Boston: Lexington Books, 1975.

Russell's Official National Motor Coach Guide, Cedar Rapids,
Iowa, April, 1977.

Stanford Research Institute. Analysis of Alternative Rail

Passen er Routings in tho Detroit-Chicago Corridor.
MenIo Park, CaIifornia, JuIy, 1971.

Stanford Research Institute. The Michigan Intercit Passen-
ger Demand Model. Menlo Park, California, June, I971.

Transportation Research Board Special Report 146, Issues in
StatewideITransporEation Planning. Washington, D.C.:
NationaI Research Council, 1974.

Transportation Research Institute--Carnegie Mellon University
and Pennsylvania Transportation and Traffic Safety Cen-
ter--Pennsylvania State University. Methodolo ical
Framework for Comprehensive Tpansportatlon_Plann1ng.
PreparedEfor the Governor's Committee for Transportation,
Commonwealth of Pennsylvania (no date available).

Periodicals
Bennett, John C., Raymond H. Ellis and John C. Prokopy, Peat,

Marwick, Mitchell and Co. A Comparative Evaluation of
Intercitngodal Split Models, Tdate not available).

73

Selected Bibliography (cont'd.)

Dobson, Ricardo. "Data Collection and Analysis Techniques for
Behavioral Transportation Planning," Traffic Quarterly
(date not available).

 

Drake, Joseph S. and Lester A. Hoel. "Issues in Statewide
Transportation Planning," Transportation Engineering
Journal of the ASCE, Proceedings of the AmeriCan Sociepy
of C1V1I Engineers, 96:TE37 August, 1975f

 

 

 

 

Ellis, Raymond H. "Reexamination of Transportation Planning,"
Tranoportation Engineering Journal of the ASCE, Proceed-
ings ofithe American Society ofIClV1I Engineers, 99:TE2,
May, 19731

 

 

Quandt, Richard E. "The Theory of Travel Demand," Transporta-
tion Research, December, 1976.

 

StOpher, Peter R. and Joseph N. Prashker. "Intercity Passen-
ger Forecasting: The Use of Current Travel Forecasting
Procedures," Tranoportation Research Forum, 1976.

Wegman, F. J. and E. G. Carter, "Statewide Transportation Plan-
ning," Transportation Engineering Journal of the ASCE,
Proceedings of the American Society of CiviI Engineers,
May, 1973.

 

Williams, H. C. W. L., "Travel Demand Models, Duality Relations
and User Benefit Analysis, Journal of Regional Science,
1976.

74

APPENDIX A

 

Hlé

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.mes .e .zIemus wanna lose .3
.Hlm manna .amma swcmmfioo Moonuxme mcowumc
numuqH ”0H:m>ahmnnwnI.coucmuom .m ~33 Hm :oHHmw\om. u mcwaommo Am
you mammaqu owfiocoom .mmanom .mmumcwzs mpmno Hm>wq AH ”chHumESmmm
mh.mm H.mm mm.Hm mw.HH no.m Hm.H NH.¢ mm
hm.¢m m.hm om.m~ mo.~H ma.m m¢.H mm.m om
no.~m «.mm om.m~ mv.~H vm.m hm.H mo.m mm
mm.m¢ m.m¢ om.vm mm.ma ma.m mv.H Hw.m om
h~.m¢ m.mv o¢.mm ~m.ma hh.b mm.H mm.m ms
oa.>¢ >.¢v mm.m~ Hm.MH hm.» mm.H om.a ov
m¢.o¢ m.mv mo.am vm.vd hm.m mm.H oo.H mm
h¢.mv m.m¢ o¢.Hm mm.ma om.m om.a mm.a om
hm.h¢ m.mv mh.am hm.oa mm.o «m.a om.a mm
BmOU omzomzou Amom onHmHommmmn muz¢szszz qu mmmHB mm:
AdBOB mzonqdo HZszm Qmmmm
ZHBH mmm BmOU OZHZZDm

 

 

.mmawfi waowsm> ooo.H Hmm mumaaoo «was:

«nmmmm MZHA mwdmm>¢
ZO Qmmdm
mZOHfiquuqdu BmOU Hmzmem

fl NHDzmmmfl

 

APPENDIX B

APPENDIX B

DATA SET
FOR RAIL MODELING

MARKET AREA

 

 

 

 

COST TIME (DRIVING TIME DISTANCE
CITY PAIR FREQUENCY (DOLLARS) (MINUTES) BANDS) (MILES)
NILES TO:
Kalamazoo 4 2.75 60 (0,0) 47
Battle Creek 4 4.00 100 (0,0) 71
Detroit 3 10.50 230 (30,0) 190
Lansing l 6.50 190 (20,20) 128
Flint l 8.75 265 (15,15) 178
Lapeer l 10.00 290 (20,20) 198
Port Huron 1 12.50 345 (45,45) 229
Ann Arbor 3 8.25 175 (25,25) 154
Durand 1 8.00 215 (20,20) 148
Jackson 3 6.50 140 (10,10) 116
KALAMAZOO TO:
Niles 4 2.75 60 (0,0) 47
Durand l 5.50 155 (10,10) 101
Lapeer l 7.50 230 (20,20) 138
Battle Creek 4 1.50 40 (0,0) 24
Jackson 3 3.75 85 (0,0) 69
Detroit 3 7.75 160 (10,10) 143
Ann Arbor 3 5.75 120 (15,15) 127
Flint l 6.50 205 (15,15) 127
Lansing 1 3.75 130 (0,0) 72
Port Huron l 9.75 285 (30,30) 182
BATTLE CREEK TO:
Niles 4 4.00 100 (0,0) 71
Kalamazoo 4 1.50 40 (0,0) 24
Jackson 3 2.50 55 (0,0) 45
.Ann Arbor 3 4.50 90 (0,0) 83
Port Huron 1 8.50 235 (30,30) 158
Lapeer l 6.25 190 (0,10) 114
Flint l 5.25 165 (0,0) 94
Durand 1 4.25 125 (0,0) 77
Detroit 3 6.50 145 (0,0) 119
Lansing 1 2.50 80 (0,0) 48

Appendix B (cont'd.)

MARKET AREA
(DRIVING TIME DISTANCE

 

 

 

CITY PAIR FREQUENCY (DOLLARS) (MINUTES) BANDS) (MILES)
JACKSON TO:

Niles 3 6.50 140 (10,10) 116
Kalamazoo 3 3.75 80 (0,0) 69
Battle Creek 3 2.50 50 (0,0) 45
Ann Arbor 3 2.25 35 (0,0) 38
Detroit 3 4.00 90 (0,0) 74
LANSING TO:

Niles 1 6.50 190 (15,15) 128
Kalamazoo 1 3.75 130 (0,0) 72
Battle Creek 1 2.50 80 (0,0) 48
Durand 1 2.00 40 (0,0) 29
Flint 1 2.75 75 (0,0) 46
Lapeer l 3.75 105 (0,0) 66
Port Huron 1 6.25 155 (0,0) 110
ANN ARBOR TO:

Niles 3 8.25 175 (25,25) 154
Kalamazoo 3 5.75 120 (15,15) 127
Battle Creek 3 4.50 90 (0,0) 83
Jackson 3 2.25 35 (0,0) 38
Detroit 3 2.25 55 (0,0) 36
DETROIT TO:

Niles 3 10.50 230 (0,30) 190
Kalamazoo 3 7.75 160 (10,10) 143
Battle Creek 3 6.50 145 (0,0) 119
Jackson 3 4.00 90 (0,0) 74
Ann Arbor 3 2.25 55 (0,0) 36
PORT HURON TO:

Flint 1 3.50 85 (0,0) 64
Lapeer 1 2.50 50 (0,0) 48
Durand 1 4.50 125 (20,20) 81
Lansing 1 6.25 155 (0,0) 110
Battle Creek 1 8.50 235 (30,30) 158
Kalamazoo l 9.75 285 (30,30) 182
Niles 1 12.50 345 (40,40) 229

Appendix B (cont'd.)

MARKET AREA

 

 

 

 

COST TIME (DRIVING TIME DISTANCE

CITY PAIR FREQUENCY (DOLLARS) (MINUTES) BANDS) (MILES)
FLINT TO:

Niles 1 8.75 265 (10,20) 178
Kalamazoo 1 6.50 205 (15,15) 127
Battle Creek 1 5.25 165 (0,0) 94
Lansing 1 2.75 75 (0,0) 46
Durand 1 1.00 30 (0,0) 17
Lapeer 1 1.50 35 (0,0) 21
Port Huron l 3.50 85 (0,0) 64
DURAND TO:

Niles 1 8.00 215 (15,15) 148
Kalamazoo l 5.50 155 (10,10) 101
Battle Creek 1 4.25 125 (0,0) 77
Lansing 1 2.00 40 (0,0 29
Flint 1 1.00 30 (0,0) 17
Port Huron 1 4.50 125 (20,20) 81
Lapeer 1 1.50 65 (0,0) 37
LAPEER TO:

Niles 1 10.00 290 (20,10) 198
Kalamazoo 1 7.50 230 (20,10) 138
Battle Creek 1 6.25 190 (10,0) 114
Lansing 1 3.75 105 (0,0) 66
Flint l 1.50 35 (0,0) 21
Durand l 1.50 65 (0,0) 37
Port Huron l 2.50 50 (0,0) 48

APPENDIX B

DATA SET
FOR BUS MODELING

MARKET AREA

 

 

 

 

COST TIME (DRIVING TIME DISTANCE

CITY PAIR FREQUENCY (DOLLARS) (MINUTES) BANDS) (MILES)
LANSING/EAST LANSING TO:

Detroit 10 5.20 165 (0,0) 85
Mt. Pleasant 6 5.20 120 (10,30) 658
Grand Rapids 6 4.15 90 (0,0) 65
Saginaw 5 4.90 155 (0,0) 70
Jackson 4 2.95 70 (0,0) 38
Ypsilanti 4 5.50 145 (0,0) 77
Ann Arbor 4 5.10 115 (0,0) 63
Midland 2 6.85 240 (10,10) 86
Muskegon 3 6.35 145 (0,0) 104
Kalamazoo 7 4.60 125 (0,0) 73
Flint 7 3.70 100 (0,0) 50
Owosso 4 2.50 50 (0,0) 31
Charlotte 4 1.55 30 (0,0) 20
Pontiac 4 6.00 220 (0,0) 69
Clare 3 5.40 135 (20,20) 88
Port Huron 2 8.40 220 (0,0) 119
Imlay City 2 6.00 175 (15,15) 84
Fowlerville 2 2.30 50 (0,0) 27
Marshall 1 3.00 60 (0,0) 45
Battle Creek 5 3.20 70 (0,0) 49
Cadillac 4 8.80 225 (30,30) 127
Lowell 1 4.00 80 (0,0) 53
Mason 1 1.25 35 (0,0) 10
Traverse City 3 12.30 280 (30,30) 171
Tecumseh l 5.10 120 (0,0) 65
Albion 2 4.35 145 (0,0) 40
Gaylord 1 11.90 270 (30,30) 168
Grand Haven 1 6.10 130 (0,0) 96
Holland 3 5.90 135 (0,0) 88
L'Anse l 31.00 1080 (45.45) 457
Farmington 2 4.45 85 (0,0) 59
South Haven 2 7.75 197 (30,30) 111
North Hudson 2 3.65 70 (0,0) 59
Brighton 3 2.90 60 (0,0) 42
Benton Harbor 5 7.85 220 (0,0) 123
Coldwater l 6.15 120 (0,0) 69

APPENDIX B

DATA SET
FOR AVIATION MODELING

MARKET AREA

 

 

 

 

COST TIME (DRIVING TIME DISTANCE
CITY PAIR FREQUENCY (DOLLARS) (MINUTES) BANDS) (MILES)
GRAND RAPIDS TO:
Saginaw l 22 25 (10,10) 116
Traverse City 1 24 30 (20,20) 139
Lansing 4 16 20 (0,0) 65
Benton Harbor 4 19 25 (10,10) 83
Muskegon 1 16 20 (0,0) 40
Manistee l 21 30 (10,10) 119
Escanaba 3 35 130 (30,30) 368
Marquette 3 38 135 (30,30) 387
Menominee 2 29 105 (30,30) 423
Flint l 38 130 (20,20) 104
Detroit 9 27 45 (20,20) 149
Alpena 2 55 190 (20,20) 247
Pellston 1 43 175 (10,10) 195
Jackson 1 49 270 (10,10) 98
Kalamazoo l 39 150 (0,0) 50
Sault Ste. Marie 1 48 90 (30,30) 278
Iron Mountain 3 46 110 (30,30) 420
Hancock 1 42 240 (30,30) 492
Ironwood 1 52 140 (30,30) 532
LANSING TO:
Flint l 19 20 (0,0) 50
Grand Rapids 3 16 20 '(0,0) 65
Muskegon 3 17 25 (20,20) 104
Escanaba 2 38 150 (30,30) 327
Marquette 2 42 165 (30,30) 391
Menominee l 35 195 (30,30) 426
Detroit 8 17 25 (0,0) 83
Saginaw 1 36 95 (10,10) 70
Alpena l 45 145 (20,20) 211
Pellston l 53 240 (30,30) 212
Traverse City 4 45 140 (20,20) 171
Jackson 1 36 80 (0,0) 138
Kalamazoo 4 38 140 (0,0) 73
Benton Harbor 2 35 95 (20,20) 120
Manistee 1 37 125 (20,20) 171

Appendix B (cont'd.)

MARKET AREA

 

 

 

 

 

COST TIME (DRIVING TIME DISTANCE
CITY PAIR FREQUENCY (DOLLARS) (MINUTES) BANDS) (MILES)
LANSING TO: (cont'd.)
Sault Ste. Marie 3 58 230 (30,30) 282
Iron Mountain 2 49 135 (30,30) 423
Hancock 2 56 230 (30,30) 485
Ironwood 1 55 175 (30,30) 535
FLINT TO:
Saginaw 6 16 20 (10,10) 36
Jackson 1 22 75 (0,0) 80
Lansing 1 20 20 (0,0) 50
Kalamazoo 2 26 90 (10,10) 123
Grand Rapids 1 22 90 (10,10) 104
Muskegon 3 42 210 (10,10) 144
Escanaba l 65 235 (30,30) 369
Marquette l 61 210 (30,30) 388
Menominee l 60 215 (30,30) 424
Detroit 5 19 25 (0,0) 60
Alpena 1 41 165 (10,10) 173
Pellston l 55 210 (20,20) 210
Traverse City 2 47 100 (20,20) 182
Benton Harbor 1 41 215 (20,20) 170
Manistee l 43 270 (20,20) 182
Sault Ste. Marie 1 60 210 (30,30) 279
Iron Mountain 1 57 215 (30,30) 421
Hancock 1 71 300 (30,30) 483
Ironwood 1 75 255 (30,30) 533

APPENDIX C

APPENDIX C

COMPARISON
OF
AVIATION SURVEY DATA
WITH
1975 ORIGIN-DESTINATION AVERAGES

 

 

SURVEY O-D 1975*
VOLUME WEEKLY

CITY PAIR (WEEK) AVERAGE
LANSING TO:
Flint 0 0.38
Escanaba 135 67.5
Menominee 14 17.3
Benton Harbor 0 8.6
Detroit 11 149
Hancock 126 72.7
Muskegon 0 3.7
Grand Rapids 1 3.7
Ironwood 42 14.6
Iron Mountain 231 44.2
Sault Ste. Marie 70 3.5
Traverse City 12 17.5
Manistee 0 0.4
Pellston 0 0.96
Marquette 576 91.7
FLINT TO:
Lansing 0 0.38
Escanaba 0 0.96
Menominee 0 0.2
Detroit 2 57.3
Hancock 0 1.7
Marquette 4 2.5
Muskegon 0 5
Grand Rapids 3 80.4
Iron Mountain 0 0.96
Kalamazoo 1 0
Saginaw 3 0.38
Alpena 2 1.9

Appendix C (cont'd.)

CITY PAIR
GRAND RAPIDS TO:

Lansing
Escanaba
Menominee
Benton Harbor
Detroit
Hancock
Marquette
Ironwood

Iron Mountain
Sault Ste. Marie
Traverse City
Pellston
Saginaw

Flint

*Creighton, Roger.

SURVEY O-D
VOLUME
(WEEK)

 

108
36

2336
198
324
108

15

24

1975*
WEEKLY
AVERAGE

Michigan Scheduled Air Service

Study, Final TechnicaI Report, September, 1977.

 

APPENDIX D

' "Zr-1' I!

 

II

I
I

III

I

M
U!
The.”
A

I

H! I

1293 02645 9

WI

3

 

 

 

I!‘ ll