1,: 0.

4a

grin. . ”5.5.?
fixgflgm. « "Eda. _
. 35"..

v 1
, . .
~ id“.-
- A
(1

‘. ‘1...

 

 

 

 

 

r.’

r
“wig :J
I 31%“

m

6
-

 

 

;‘.}B: 'Lll.
.i Inkliul

~
yeah“... a“. {37.1. I
. ‘Iélt ‘II
Aukakfl.» Sunk-u
fr £9-33 .1.
01..

 

 

 

in“.
givlzrfiuln

; .17. up. I.

inf... .

S‘S

llBRARY
Michigan State
University

This is to certify that the

thesis entitled

A METHOD FOR IMPROVING THE USEFULNESS OF HIGHWAY TRAFFIC
DATA IN TOURISM STUDY: A MICHIGAN CASE STUDY

presented by

SINJI YANG

has been accepted towards fulfillment
of the requirements for

 

 

 

Doctoral . Park, Recreation & Tourism‘
degree m W3
7
EMA/WK...
Major professor
Date—NWL

0-7639 MS U i: an Affirmative Action/Equal Opportunity Institution

 

 

 

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 cJClRC/DatoOmpGS-pJS

 

A METHOD FOR IMPROVING THE USEFULNESS OF HIGHWAY TRAFFIC DATA
IN TOURISM STUDIES: A MICHIGAN CASE STUDY

By

Sinji Yang

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Park, Recreation, and Tourism Resources

2001

ABSTRACT
A METHOD FOR IMPROVING THE USEFULNESS OF HIGHWAY TRAFFIC DATA
IN TOURISM STUDIES: A MICHIGAN CASE STUDY
By

Sinji Yang

Although the highway traffic data collected by the Department of Transportation
is a set of comprehensive recordings of highway travel activities, the inherent problems in
the data have deterred tourism researchers from using the data up to their full potential for
many tourism studies and planning purposes. In addition, using untreated data, which
contain tourism and non-tourism traffic, in tourism reports can produce misleading
results.

This exploratory study was launched to examine the characteristics of highway
traffic data, to enhance connections between highway traffic data and tourism studies, and
to demonstrate that valuable tourism information can be derived from underutilized
highway traffic data. These goals were achieved by the development of a new data
processing method called the Removal of Routine Traffic Method. The method is
designed to mitigate the problem of separating tourism traffic from non-tourism traffic
thereby facilitating greater and more meaningful use of highway traffic data in the field of
tourism.

Based on the patterned behavior of highway travelers, a conceptual model was

developed to link highway traffic data to tourism studies. Traveler behavior theory

suggests that removing routine traffic from total traffic can improve data relevancy for
tourism studies.

As a measurement of tourism traffic, the Removal of Routine Traffic Method
provides face and construct validity in estimating tourism traffic. In a nutshell, the
method functions as a filter screening out non-tourism traffic from total traffic and
leaving the residual as an improved estimate of tourism traffic. Although the concept is
relatively straightforward, it has been proved to be powerful. Using 1998 Michigan
highway traffic data as an example, the method improved the overall data relevancy to
tourism by 364%. Even simply performing the removal of truck traffic (i.e., non-
recreational type vehicle traffic) can improve the overall data relevancy to tourism by
1 2%.

With the Removal of Routine Traffic Method, researchers not only can better
understand the behavior of highway travelers and tourism traffic flows but also know how
to utilize the extensive highway traffic data with the confidence that the estimates they
derive are closer to their true values. Regional tourism planners or business operators can
promptly estimate tourism traffic flow on a specific day or period of time if traffic
counters are installed on vicinity highways. With only a small amount of initial
investment in data storage and database programming, the removal of routine traffic
operation can be highly automated. That is, the method is efficient and economical
compared to other methods used in tourism studies. Hopefully, this study will encourage
more researchers to use highway traffic data for regional tourism studies and planning,

and to build upon this research to further improve tourism traffic volume estimates.

Copyright by
SINJI YANG

2001

iv

ACKNOWLEDGEMENTS

This research is the result of efforts and encouragement provided by many
individuals. Expression of appreciation here is limited to only a few of them. I am
deeply grateful to my advisor and the chairperson of the dissertation committee, Dr.
Donald F. Holecek for his insights, guidance, support, and persistent revision throughout
completion of this research. If it were not for his assistance and patience, the completion
of this dissertation would have been much more difficult.

A special thank you is extended to Dr. Edward M. Mahoney for his valuable
suggestions. The author is truly grateful for his guidance and support on this research. I
also wish to acknowledge the other members of my committee, Dr. Larry Leefers and Dr.
Richard Paulsen, for their constructive ideas and criticisms. The direction and support

received are very much appreciated.

A very special thank you to my ex-committee member, Dr. Daniel M. Spencer,
who left Michigan State University to pursue other personal goals. Words cannot express
my gratitude for his assistance and guidance on the initial development of this research. I
would also like to thank Mr. David Schade, the Director of Transportation Planning
Services Division, Michigan Department of Transportation, and his staff for providing
valuable information, assistance, and traffic data. I thank them wholeheartedly.

The deepest measure of appreciation goes to my family and friends for their
constant encouragement, support, and love. I am most grateful to my wife for sharing the
exhaustion, frustration, and joy that the research and my academic work produced. I
would especially like to thank Mrs. Keturah Thunder Haab and Dr. Jakob Heckert for
editing this dissertation. I would like to thank my parents. This doctoral program has

been completed through their economic and spiritual support.

TABLE OF CONTENTS

LIST OF TABLES _______________________________________________________________________________________________________________________ x
LIST OF FIGURES ..................................................................................................................... xiii
CHAPTER 1
INTRODUCTION _______________________________________________________________________________________________________________________ 1
1.1 Highway System and Tourism ___________________________________________________________________________ 1
1.1.1 Highway Traffic Data Collection ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 2
1.1.2 Current Uses of Highway Data _________________________________________________________________ 4
1.2 Problem Statement ............................................................................................. 4
1.2.1 Problems in Using Highway Traffic Data ________________________________________________ 6
1.2.2 Need for a New Method to Resolve Problems ________________________________________ 8
1.3 Purposes and Objectives of the Study _______________________________________________________________ 9
1.4 Definitions of Terms __________________________________________________________________________________________ 10
1.5 Organization of the Study __________________________________________________________________________________ 14
CHAPTER 2
REVIEW OF LITERTURE _________________________________________________________________________________________________________ 16
2.1 Study Area--Michigan ________________________________________________________________________________________ 16
2.2 Related Literature _______________________________________________________________________________________________ 17
CHAPTER 3
METHODS ___________________________________________________________________________________________________________________________________ 2 1
3.1 Data Collection ___________________________________________________________________________________________________ 22
3.1.1 Permanent Traffic Recorder Data _____________________________________________________________ 26
3.1.2 Classified Permanent Traffic Recorder Data ___________________________________________ 32
3.2 Identifying Characteristics of Highway Traffic Data _______________________________________ 40

vi

3.2.1 The Random Process of Traffic Time Series 41

3.2.2 Daily Traffic Distributions ________________________________________________________________________ 4 3
3.2.3 Seasonality ................................................................................................. 4 8
3.2.4 Vehicle Type Distributions _______________________________________________________________________ 49
3.2.5 Geographical Locations of Traffic Counters ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 50

3.3 A Conceptual Model Linking Highway Traffic Data to Tourism Studies 51

3.3.1 The Conceptual Model ______________________________________________________________________________ 51
3.3.2 Patterned Behaviors and Routine Traffic _________________________________________________ 54
3.3.3 Method for Estimating Routine Traffic ____________________________________________________ 57
3.4 Estimating Tourism Traffic ............................................................................... 7 0
3.4.1 Step One—-Data Filtering and Transposing ______________________________________________ 70
3.4.2 Step Two-Data Grouping ________________________________________________________________________ 73
3.4.3 Step Three—Estimating and Removing Routine Traffic ,,,,,,,,,,,,,,,,,,,,,,,, 75
3.4.4 Estimating Tourism Traffic _______________________________________________________________________ 79
3.5 Validation of the Removal of Routine Traffic Method ____________________________________ 82
3.5.1 Validation Designs .................................................................................... 82
3.5.2 Hypotheses _________________________________________________________________________________________________ 83
3.6 Measurement of Data Improvement __________________________________________________________________ 86
CHAPTER 4
RESULTS ..................................................................................................................................... 89
4.1 Results from the Data Processing Phase ___________________________________________________________ 89
4.1.1 Results from Data Preparation __________________________________________________________________ 89
4.1.2 Results fiom the Removal of Routine Traffic Operation ........................ 100

vii

4.2 Results from the Method Validation Phase ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
4.2.1 Results from Paired Samples T Test on Null Hypothesis 1 .....................
4.2.2 Results from Regression Analyses on Hypothesis II to V ......................
4.3 Results fiom Data Improvement Measurement ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
4.4 Example Results from Using Daily Tourism Estimates (DTEs) .....................
CHAPTER 5
SUMMARY AND CONCLUSIONS .........................................................................................
5.1 Summary of the Study ________________________________________________________________________________________
5.2 Major Findings and Implications ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
5.3 Applications ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
5.3.1 Site-Specific Tourism Application ...........................................................
5.3.2 Regional Tourism Application ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
5.3.3 Applications of Tourism Dominance ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
5.4 Projected Cost for Using the Removal of Routine Traffic Method ,,,,,,,,,,,,,,,,
5.5 Limitations ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
5.6 Recommendations ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
5.6.1 Recommendations to the Department of Transportation ........................
5.6.2 Recommendations for Future Research ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
APPENDICES ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
APPENDD( A

viii

110

110

111

118

119

125

125

128

131

131

132

133

134

136

137

137

138

140

141

142

APPENDIX B

ooooooooooooooooooooooooooooooooooooooooo

1998 Calendar and Major Holidays .........................................................................

APPENDD( D

Population Estimates of Michigan Counties, 1998 _________________________________________________

APPENDIX E

VOOCDOICOOODIQQQUQODIOOIOIOQOIIOOOOOOOOOOCU...ODI....‘ODOOIOODOIIOOOODO0.....-IIOCOOOIOOOOOOOOOOOIICIOOOI...-IOIOOC...

ix

143

144

145

146

147

148

149

150

152

Table 3-1

Table 3-2

Table 3-3

Table 3-4

Table 3-5

Table 3-6

Table 3-7

Table 3-8

Table 3-9

Table 3-10a

Table 3-10b

Table 3-11

Table 3-12

Table 3-13

Table 3-14

Table 3-15

LIST OF TABLES

Regional Distribution of Michigan Permanent Traffic Recorders ,,,,,,,,,,,,,,,,,,, 24
Directional Distribution of Michigan Permanent Traffic Recorders _______________ 28

Locations of Michigan Permanent Traffic Recorders and Number of
Available Records from 1995 through 1998 _____________________________________________________ 29

Michigan Permanent Traffic Recorders which Provided Vehicle
Classification Information in 1998 38

Examples of Percentages of Daily Traffic and Percentages of Lowest Daily

Traffic Volume in a 7-Day Period (4049 North, October 1998) ...................... 61
Average Percentage of Lowest Daily Traffic Volume in a Seven-Day

Period (in 1998) __________________________________________________________________________________________________ 63
An Example of Coarse Estimation of Weekly Tourism Traffic Based on
Average Percentage of Lowest-Traffic Day in a 7-Day Period ________________________ 65
Separating NRV Traffic from RV Traffic (4049 North, 10/1/1998). ,,,,,,,,,,,,,, 71
Transposing RV Traffic Data Points (4049 North, 10/1/1998) ________________________ 72

Example of Data Grouping--Weekend Groups (4049 North, 10/ 1998) _________ 74
Example of Data Grouping--Weekday Groups (4049 North, 10/ 1998) ,,,,,,,,, 74

Example of Estimating Hourly Routine Traffic for 4049 North, Weekends

in October 1998 ___________________________________________________________________________________________________ 76
Examples of Estimating Hourly Tourism Traffic for 4049 North, Weekend
Days in October 1998 _________________________________________________________________________________________ 78
Example of Difference Matrix Derived from Removal of Routine Traffic
Operation (4049 North Weekend Days in October 1998) ________________________________ 80
Daily Tourism Estimates (DTEs) for 4049 North, Weekend Days in

October 1998 _______________________________________________________________________________________________________ 81
A Chart of Data Improvement at Hypothesized Percentage Points ,,,,,,,,,,,,,,,,, 88

Table 4-1

Table 4-2

Table 4-3

Table 4-4

Table 4-5

Table 4-6

Table 4-7

Table 4-8

Table 4-9

Table 4-10

Table 4-11

Table 4-12

Table 4-13

Table 4—14a

Table 4.14b

Table 4-14c

Table 4-15a

Table 4-15b

Table 4-15c

Table 4-16

Table 4-17

Traffic Distributions by Vehicle Types on Different Routes (°/o) ,,,,,,,,,,,,,,,,,,,,
Correlations between Direction and Percentage of RV Traffic .......................
Summary of Data Preparation ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Percentages of RV Traffic in Total Traffic .......................................................
Paired Two-Sample T-Test: Weekend Day VS. Weekday ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Estimated Average Daily Tourism Traffic Volumes (ADTT) .........................

Percentages of Tourism Traffic ..........................................................................

A Paired Two-Sample T-Test on Tourism Traffic: Weekend Day vs.
Weekday ______________________________________________________________________________________________________________

Summary of Traffic Components after the Removal of Routine Traffic
Operation _____________________________________________________________________________________________________________

Summary of Tourism Traffic Percentages by Weekend Day, Weekday, and
Traffic Directions

Correlation of Weekend Day and Weekday Tourism Traffic Volumes ,,,,,,,,,,
Correlation of Weekend Day and Weekday Tourism Traffic Percentages _____
Paired Samples T Test ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Summary of Weekend Tourism Traffic Regression Models ,,,,,,,,,,,,,,,,,,,,,,,,,,,,
AN OVA of Weekend Regression Models ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Coefficients of Weekend Tourism Traffic Regression Models ,,,,,,,,,,,,,,,,,,,,,,,
Summary of Weekday Tourism Traffic Regression Models ,,,,,,,,,,,,,,,,,,,,,,,,,,,,
AN OVA of Weekday Regression Models ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
Coefficients of Weekday Tourism Traffic Regression Models ,,,,,,,,,,,,,,,,,,,,,,,

Improvement Measures Using Methods Developed in This Stild)’ ,,,,,,,,,,,,,,,,,

Example of Using the Removal of Routine Traffic Method to Estimate
Tourism Traffic on Memorial Day Weekend ____________________________________________________

xi

93

95

96

98

101

104

106

107

109

109

111

113

113

114

116

116

117

119

120

Table 4-18 Averaged Weekly Tourism Traffic Flows around Station 4049 ,,,,,,,,,,,,,,,,,,,,,, 121
Table 4-19 Averaged Weekly Tourism Traffic Flows around Station 9369 ,,,,,,,,,,,,,,,,,,,,,, 123

Table 5-1 Projected Cost and Time Investments Required to Implement the Removal
of Routine Traffic Method. 135

xii

Figure 3-1
Figure 3-2

Figure 3-3

Figure 3-4

Figure 3-5

Figure 3-6

Figure 3-7

Figure 3-8

Figure 3-9

Figure3-10

Figure 3-11

Figure 3-12

Figure 3-13
Figure 4-1
Figure 4-2

Figure 4-3

LIST OF FIGURES

Michigan Department of Transportation Highway Monitoring Regions ________ 23
Vehicle Classifications-—Typical Vehicle Silhouettes ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 35

Geographical Locations of Permanent Traffic Recorders which Provided
Vehicle Classification Information in 1998 ...................................................... 39
An Example of Autocorrelation Functions of Traffic Time Series (4049
North) ___________________________________________________________________________________________________________________ 43

An Example of Weekday Daily Traffic Distributions in a Rural Area
(4049 North) ....................................................................................................... 44

An Example of Weekday Daily Traffic Distribution in an Urban Area
(9999 Northeast) _________________________________________________________________________________________________ 45

An Example of Three Clusters of Daily Traffic Distributions
(9969 East, October 1998) ................................................................................. 47

A Conceptual Model Linking Highway Traffic Data to Tourism Studies ...... 53

A Hypothetical Graph Showing the Separation of Hourly Traffic Volume
into Three Components ______________________________________________________________________________________ 56

Examples of Hour-Column Distributions of Traffic Data (4049 North,

Weekdays in October 1998) ............................................................................... 58
A Hypothetical Example Showing an Outlier's Effect on the Estimation of
Routine Traffic ____________________________________________________________________________________________________ 68
Example of Using the 10th Percentile Traffic Flow to Estimate Routine
Traffic (4049 North, Weekdays in October 1998) ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 69
Routine Traffic for 4049 North, Weekends in October 1998 .......................... 76
Distribution of Traffic by Vehicle Types __________________________________________________________ 90
Volumes of RV Traffic: Weekend Day vs. Weekday ...................................... 99
Percentages of RV Traffic: Weekend Day vs. Weekday __________________________________ 99

xiii

Figure 4-4a
Figure 4-4b
Figure 4-4c
Figure 4-Sa
Figure 4-5b

Figure 4-5c

Weekly Tourism Traffic Flow around Station 4049, All Weeks ,,,,,,,,,,,,,,,,,,,,,
Weekly Tourism Traffic Flow around Station 4049, Regular Weeks ,,,,,,,,,,,,,
Weekly Tourism Traffic Flow around Station 4049, Weeks with a Long

Weekly raggata'f'sa'fiss $51153 saga; s35; Elissa;11111112122211:
Weekly Tourism Traffic Flow around Station 9369, Regular Weeks .............

Weekly Tourism Traffic Flow around Station 9369, Weeks with a Long
Weekend

xiv

122

122

122

124

124

124

CHAPTER 1

INTRODUCTION

1.1 Highway System and Tourism

In the early 19808, Francis B. Francois, Executive Director of the American
Association of State Highway and Transportation Officials, considered the US. highway
systems to be "the most awesome highway networks anywhere" in the world
(Transportation Research Board, 1982). In 1996, the total length of public roads was 3.9
million miles. On each square mile of land, there was 1.05 miles of public road (F HWA,
1997). Although air travel has become an increasingly popular means of transportation,
U.S. highway systems remain intensively used in the United States. According to the
US. Travel Data Center (1996), the trend in use of personal vehicles and highways in the
United States remains stable. This is because the need for personal vehicles has become
indispensable for most travel activities, especially outdoor recreation, in the United
States. Furthermore, for ahnost all air travelers, air carrier services are simply a part of
their total travel services itinerary. Ground vehicles are ahnost always needed for air
travelers to move from point to point within both origin and destination areas.

In the United States, 2,423 billion vehicle miles were registered in 1995. There
are about 128 million registered automobiles in the United States, about 0.48 automobiles
per capita (U .8. Bureau of Census, 2000).

About 83% of person-trips were taken by auto (including light trucks, RVs) and
buses. Pleasure and business travelers accounted for 63% and 22% of auto person-trips

during 1995, respectively. An inclusive definition of tourism recommended by the World

Tourism Organization (1991) counts both pleasure and business trips as tourism. (The
WTO's definition of tourism will be discussed in the section of "Definitions of Terms")
If this definition of tourism is used, 85% of auto person-trips can be considered tourism
related. According to the US. Bureau of Census, air travel only accounted for 16% of
person-trips in the United States in 1995. The rest of person-trips (i.e., 84%) were almost
all taken on highways (U .S. Bureau of Census, 2000). Since, as stated before, most air-
trips also involve some kind of ground transportation, the actual percentage of auto
person-trips would be much higher than that estimated by the US. Bureau of Census.

As one can see, almost every tourism activity involves using ground vehicles and
highway systems. Therefore, highway traffic data are potentially an important source of

information for those who engage in regional tourism studies.

1.1.1 Highway Traffic Data Collection

In the 19303, statewide highway planning surveys were established to collect
highway traffic characteristic information (FHWA, 1995). Since then, highway traffic
data have been continuously collected by the Departments of Transportation of both state
and local governments under the guidance of the Federal Highway Administration
(FHWA) of the US. Department of Transportation (DOT). For example, the Michigan
Department of Transportation (MDOT) is responsible for collecting and preparing
highway traffic data in Michigan and making them available to the general public.

To increase reliability of data collection and efficiency of highway monitoring, the
Federal Highway Administration has continuously made efforts to improve the

methodologies applied in highway performance monitoring. According to the FHWA

(1995), its current highway performance monitoring procedures emphasize the use of a
stratified random sampling method to produce reliable estimates, and integrated data
collection processes to minimize data collection efforts, so that duplications in data
collection can be reduced.

The integrated Highway Performance Monitoring System is designed to collect
highway traffic volume, vehicle classifications, and truck weight data in a sequential
format. Specifically, traffic volumes are sampled from the existing Highway
Performance Monitoring System, vehicle classification information is derived from the
traffic volume samples, and truck weight data are derived from the vehicle classification
samples. When traffic volume data are collected, vehicle classification and truck weight
data are also collected. Recently vehicle speed has been experimentally added to the data
collected by the integrated Highway Performance Monitoring System, and, as new
monitoring equipment is installed across the system, vehicle speed data will become more
widely available.

Since highway traffic monitoring demands a significant amount of financial and
human resources, the locations of traffic monitoring stations are generally determined by:
(1) available funding to state and local governments, (2) perceived highway traffic
characteristic information needs, and (3) previous efforts and commitment to data
collection (FHWA, 1995). Thus, traffic data are usually collected in more populated
areas where traffic volumes are relatively high. Regional tourism development may not

be a major factor that influences the locations of permanent traffic monitoring stations.

1.1.2 Current Uses of Highway Data

Information about highway traffic can be used in a variety of ways by both public
and private sector planners. The public sector uses traffic data in highway construction
and re-construction planning, routing and detour design, economic benefit studies,
transportation growth forecasting, and highway traffic control policies and regulations.
The private sector uses traffic data to select business service areas and store locations, to
plan parking and shuttle connections, and to develop marketing strategies (FHWA, 1995;
US. Bureau of Census, 1999).

All levels of governments which spend a great amount of effort and money in
collecting highway traffic data naturally remain the major users of these data. In the
private sector, oil companies, restaurants, banks, real estate developers, and outdoor
advertisers are the major users of highway traffic data (Cuyahoga County Engineer's
Office, 1999). Although highway traffic data are readily available to the general public,
the private sector has made less use of them than the public sector; moreover, the data are

usually used in the forms that they are obtained.

1.2 Problem Statement

Chadwick (1994) points out that, because of the special nature of the tourism
industry, it lacks distinct products and services. Therefore, economists generally don't
consider travel and tourism to be an industry. Thus, any attempt to account for tourism in
the gross national product (GNP) (i.e., the National Account) is liable to lead to double

counting because economic activities of all travel and tourism establishments are already

allocated to existing industries according to the Standard Industrial Classification system
in use. A great amount of research effort has been focused on distinguishing tourism
from non-tourism to avoid the double counting problem. For example, a Satellite
Account, which uses a more sophisticated approach to distinguish tourism from other
industries, can measure tourism more precisely than a National Account (Chadwick,
1994; Statistics Canada, 1991a). However, there is a continuing need to search for better
methods to separate tourism form non-tourism. A similar problem is faced by tourism
researchers using highway traffic data as a source of information for tourism studies.

As previously mentioned, a great portion of tourism trips are taken on highways in
personal vehicles. Thus, highway traffic data have the potential for becoming a central
source of information that researchers can utilize in tourism studies. Examining highway
traffic data by can help tourism researchers better understand the tourism phenomenon,
namely, its directional flow over periods of time, peaking tendencies, spatial distribution,
and so on. F urtherrnore, estimates of tourism volume from highway traffic data can be
used independently or in combination with other tourism data.

Although highway traffic data could be a valuable source of information for
tourism studies, they are sometimes ignored and rarely, if ever, used to their full potential.
Even governments, which recognize the important contributions of tourism to local
economies, have failed to facilitate the use of these data fully in regional tourism planning
studies. Even though a great number of tourism studies have been conducted, there has
been little effort expanded to make highway traffic data more useful in the field of
tourism. Reviewing published travel and tourism research in journals, one can rarely find

articles or reports focusing on highway traffic and tourism. In the few instances where

highway traffic data have been used, they are presented in their raw formats (e. g., the
Average Daily Traffic) prepared by the Department of Transportation. Taking the
Michigan Travel Indicators Reports (Yang and Holecek, 1996-99) as an example,
Average Daily Traffic (ADT) volumes from a set of traffic counters over a year were
aggregated and compared to their levels in the previous year. The derived percent
changes were used as indicators of increase or decrease in tourism trips during the year.
This method of estimating tourism growth is correct only when the growth of tourism
traffic is in the same proportion as growth of total traffic. The fact is that the relation
between tourism traffic and total traffic is generally unknown, and that reported traffic
data include both tourism and non-tourism traffic. Thus, using them in tourism studies is
always problematic.

To date, little scientific attention has focused on increasing understanding of the
relationship between highway traffic and tourism trips. Traffic data in raw form have
utility in some applications, but even in these cases, only insubstantial conclusions can be
drawn. Given the costs and complexities of generating reliable tourism data, it seems
worthwhile to explore ways by which readily available traffic data can be modified to

improve their usefulness.

1.2.1 Problems in Using Highway Traffic Data
Researchers may fail to fully utilize highway traffic data in tourism studies due to
the following problems with the data:
1. Traffic data are collected for non-tourism purposes. The main purpose for

collecting highway traffic data is not for tourism studies but for general

purposes such as highway planning and re-constructions. Useful
information for tourism and private business is neglected in the
government traffic data reports. While more and more state and local
governments have started to publish Average Daily Traffic data on the
Internet for public access, it is still not easy for the private sector to
analyze the data and sort out meaningful information.

. Untreated traffic data are imperfect indicators of tourism. To private
business operators, traffic flow is considered a business opportunity
passing through a region or a site. Whatever deters traffic flow going
through a region may generate opportunities to do business with those
traveling through the region. For example, traffic flows may be deterred
by severe weather conditions. Travelers may stop in a region for special
scenery, events, stores, or restaurants on the way to their primary
destination if the special attractions generate sufficient interest. While
highway traffic data implicitly contain information on the size of potential
tourism business opportunities, the size of vehicle count includes people
traveling for a wide range of purposes. The inherent problem in using
untreated highway traffic data in tourism studies is that traffic counting
devices indiscriminately count all vehicles without recording whether a
passing vehicle is driven by a pleasure traveler, a commuting worker, or a
business traveler. Despite significant advances in traffic monitoring
technology that allows separating car traffic from truck traffic, it is still

very difficult to capture the trip purposes of highway travelers using

remote counting devices. There are ways to develop trip purpose
information. For example, vehicles can be randomly stopped and their
occupants surveyed. However, this method is rarely used today because it
is costly. It also disrupts traffic and no longer is legal. Information on trip
purposes, however, is what tourism researchers use to distinguish tourism
from other forms of travel. Without knowing the trip purposes of
travelers, it is theoretically impossible to separate tourism traffic from
non-tourism traffic. Although highway traffic data are a set of
comprehensive recordings of highway travel activities, using untreated

highway traffic data in tourism studies may yield misleading results.

1.2.2 Need for a New Method to Resolve Problems

The Department of Transportation’s current data processing method, which is not
designed for tourism studies and unable to distinguish tourism from non-tourism traffic in
highway traffic data, has limited use in the field of tourism. If highway traffic data can be
separated into tourism and non-tourism related traffic, they will be much more valuable to
tourism researchers as well as to business operators. Information on the potential size of
tourism traffic on highways is also valuable to regional tourism planners, such as local
Conventions and Visitors Bureaus (CVBs). To date, however, very little effort has been
expanded in developing an appropriate data processing method to make traffic data more
useful for tourism researchers, planners, and business operators. Thus, there is a need for
new highway traffic data processing methods to mitigate the inherent problems in

highway traffic data and to make them more useful for tourism studies.

1.3 Purposes and Objectives of the Study

The purposes of this study are: (1) to improve the relevancy of remotely Sensed
highway traffic data to tourism studies, and (2) to develop sound procedures for removing
non-tourism elements from highway traffic data sets in order to enhance their
applicability. Hopefully, this effort will stimulate greater use of these data as well as
research that will further improve the utility of these data in tourism research
applications.

The main objective of this study is to develop a methodology for refining highway
traffic data for tourism research applications. It is designed to mitigate the inherent
problem of distinguishing tourism from non-tourism traffic in remotely sensed highway
traffic data. Using this new method, researchers will be able to estimate tourism from
total traffic flow.

The purposes of this study will be facilitated and achieved by accomplishing the
following objectives:

1. Identify the essential characteristics of highway traffic data and tourism
related information that help link highway traffic data to tourism.

2. Construct a theoretical (conceptual) model to guide the development of a
data processing method toward achieving the main objective described
above and in the next step.

3. Develop a data processing method that mitigates the problem of separating
tourism traffic from non-tourism traffic and which facilitates the use of

highway traffic in the field of tourism.

4. Demonstrate how to estimate tourism traffic from total traffic flow by
using the new method and determine the tourism relatedness of different
highway routes.

5. Assess the validity of the method based on estimated tourism traffic and
theoretical criteria.

6. Measure the degree to which the newly proposed method improves the

relevancy of traffic data to tourism.

1.4 Definitions of Terms
The following definitions of terms are provided to avoid misunderstanding and
confusion and will be used throughout this report.
Tourism—While there are different definitions of tourism, the term generally
means pleasure travel in its daily usage. In 1991, the World Tourism Organization
(W TO) recommended the following inclusive definition:
Tourism: The activities of a person traveling outside his or her usual
environment for less than a specified period of time and whose main
purpose of travel is other than exercise of an activity remunerated from the
place visited, where:
1. "usual environment" is intended to exclude trips within the area of usual
residence and also frequent and regular trips between the domicile and

workplace and other community trips of a routine character;

2. "less than a specified period of time" is intended to exclude long-term
migration; and

3. "exercise of an activity remunerated from the place visited" is intended
to exclude only migration for temporary work. (WTO, 1991)

10

In accordance with WTO's defrrrition, Chadwick (1994) suggests the term "travel
and tourism" used in combination to describe human and business activities associated
with the aspects of temporary movement of people away from their home and work
environments for pleasure, business, and personal reasons. In this study, the WTO's
definition of tourism is used. Further, the term "tourism" is used as an equivalent to

"travel and tourism" as suggested by Chadwick (1994).

Tourism Traffic—In this study, the term "tourism traffic" means traffic generated
by persons who travel for pleasure, visiting fiiends or relatives, staying in seasonal
homes, attending business conferences, or for other reasons related to/or associated with
leisure. These trips can be either long or short distance day trips as well as overnight
trips. There are many kinds of tourism trips. The most common tourism trips are listed
below:

I Multipurpose vacation trips

I Business trips that include some leisure opportunities and associations

I Weekend trips to a recreation destination or to visit friends or relatives

I Trips to a second home or time-share

I Day trips to recreation destinations, visit friends or relatives, or leisure

shopping

I Day trips primarily for driving itself (Kelly, 1996)

The main characteristic of these tourism trips is that they are not routinely taken by most

people.

11

Non-Tourigrn Rafi—Although it may not be possible to distinguish clearly
tourism from non-tourism traffic, generally non-tourism traffic includes: 1) truck traffic
generated by cargo deliverers who use large size trucks to accomplish these tasks, and 2)
routine traffic generated by commuters and local residents whose purposes of trips do not
fit into the WTO's definition of tourism. For example, routine trips would include going
to and from work, school, grocery shopping, seeing a doctor, etc.

Traffic Volume—Traffic volume is a count of passing vehicles at a specific site
on a highway over a period of time. While mathematically highway segments are one-
dimensional lines on a surface, traffic volumes are point data with no dimension. The
Department of Transportation’s concept of highway performance monitoring is basically
using point data to infer traffic flow on a line and then possibly in an area.

Directional and Non-Directional Traffic Dag—Most highway routes are bi-
directional. Thus, traffic volumes are collected according to the directions of traffic
flows. Usually, the traffic volume in one direction does not equal that in the other
direction. Non-directional traffic data are derived by aggregating directional traffic data
fi'om both directions.

Toufim Dominflg—Tourism dominance is determined by the percentage of
tourism traffic on a route. There are two kinds of tourism dominance—absolute and
relative. A highway route is "absolutely" tourism dominant if tourism traffic is
proportionally greater than non-tourism traffic on the route. A highway route is
"relatively" tourism dominant if tourism traffic on the route is proportionally greater than

that on other routes. The tourism relatedness of highway routes can be determined by

12

their tourism dominance. A highway route is more tourism related if it is more tourism
dominant.

TourrfirLRelevalrgy-Tourism relevancy of traffic data can be determined by the
percentage of tourism traffic in total traffic. A higher percentage of tourism traffic in data
sets implies that their relevancy to tourism is also higher.

Wd Uwype Tourism--The term rural-type tourism is used to describe
trips whose destinations are in rural areas. Urban-type tourism is used to describe trips
whose destinations are in urban areas.

Average Dafiy Tra_ffi_c (ADT)—According to DOT, Average Daily Traffic is the
weighted averages of traffic volume occurring on weekdays, Saturdays, and Sundays
within a given period of time. The following formula is used by DOT to calculate ADT:

(average weekday l. x 5) + average Saturday , + average Sunday i
7

 

ADE:

where i is a period of time such as a month or a year. ADTs negate disparities in the
number of weekdays and weekend days in a month or a year and thus facilitate comparing
traffic volumes over different time periods.

Weekday—The term "weekday" means any day from Monday through Friday.
Although, Friday evening may be included as a part of weekend in some research, in this
study all of Friday is considered a weekday.

Weekend day—The term "weekend day" is used to describe a day being either a
Saturday or a Sunday. While a weekend includes both Saturday and Sunday, a weekend

day is only one of the two days.

13

Four Seasons—The Travel Industry Association (TIA), Travel Michigan
(previously known as the Michigan Travel Bureau), and the Michigan Travel, Tourism
and Recreation Resource Center at Michigan State University have been using the
following definition of the four seasons. Winter is from December to February, spring is
from March to May, summer is fiom June to August, and fall is from September to

November. This definition of the four seasons is also used in this study.

1.5 Organization of the Study

A review of literature pertaining to the subject of this study is presented in
Chapter 2. Chapter 3, the core of this study, contains relevant theory development and
methodological details. Specifically, the first section of Chapter 3 contains descriptions
of data collection and types of data used in this study. The second section of Chapter 3
contains the results of some initial data analyses and findings used in identifying the
characteristics of highway traffic data. The conceptual model and theory development in
the third section is based on the initial analyses and findings in the second section. The
detailed layout of a new highway data processing method is presented in the fourth
section of Chapter 3. The fifth section contains a discussion on how to assess the validity
of the newly proposed method. Finally, the chapter concludes with a discussion on how
to measure the degree to which the method improves the relevancy of traffic data to
tourism (i.e., a measurement for data improvement).

Chapter 4 presents a variety of descriptive and inferential statistical test results
from data preparation and a demonstration of the new method. Results of method

validation and data improvement evaluation are also included in Chapter 4.

14

Chapter 5 includes summaries, major findings, and implications of the study. The
applications and limitations of the new method are also discussed. Finally, the study
concludes with recommendations to the Department of Transportation for future data

collection and to tourism researchers for further studies on using highway traffic data.

15

CHAPTER 2

REVIEW OF LITERATURE

The first section of this chapter reviews tourism literature pertaining to the
geographical area of this study--its tourism activities, seasons, and characteristics. While
little literature focuses on highway tourism traffic, the second section of this chapter
includes a review of literature that inspired the author’s thoughts on the subject of the
study that highway traffic data can and should be more effectively used in tourism

studies.

2.1 Study Area—Michigan

Michigan’s highway traffic data were used as an example in this study because in
Michigan nearly 90% of tourism involves the use of personal vehicles (cars, RVs, etc). In
Spotts’ (1991) study of Michigan’s tourist attractions, he pointed out that visiting tourist
attractions is one of the major reasons for pleasure travel in Michigan. The Upper and
Northern Lower Peninsula are important natural resource-based tourism areas in
Michigan and are heavily visited by tourists.

Most of Michigan’s population resides in the Southern Lower Peninsula. In 1983-
84, about 90% of trips to or through Michigan originated from either large or small
metropolitan areas in Michigan and other states (Holecek, 1991). Attendance at
attractions in Michigan’s Upper Peninsula and Northern Lower Peninsula areas is mostly
generated from residents of Michigan's Southern Lower Peninsula and neighboring states.

Since Michigan has a relatively long snow season, many outdoor recreation and tourism

16

activities take place during warmer periods in late spring, summer, and early fall. On the
other hand, winter sports lovers generate a substantial amount of tourism traffic during
the long snow season. In addition, since a large number of second homes (seasonal
homes) are located in Michigan’s Upper and Northern Lower Peninsulas, a significant
amount of north-southbound traffic is generated by second home owners during

weekends and holidays.

2.2 Related Literature

Reviewing published travel and tourism research, one can hardly find any articles
or reports focusing on highway traffic and tourism. Among the few examples of existing
published research associating tourism measurement with highway traffic, the Michigan
Travel, Tourism and Recreation Resource Center's Michigan Travel Indicators serial
reports, are the most continuous one in using highway traffic data. Another example is
Yang and Holecek’s (1997) investigation of the effectiveness of using Average Daily
Traffic (ADT) in monitoring travel activity in Michigan.

Although very few tourism studies using highway traffic data were found, there
are some intriguing published research articles that can be considered highway and
tourism related. Langer (1996) examined the relationship between traffic noise and
profits of hotels that are located close to highways. Steward et a1. (1993) used a high-
speed-shutter camcorder to capture license plate numbers of vehicles on Texas highways.
Based on license plate information, Steward et a1. then retrieved vehicle registration
information to study the differences in trip characteristics between those who stop at

highway welcome centers and those who don't.

l7

While Gunn (1979) stressed the importance of attractions, services, infiastructure,
and information in tourism development and planning, Roehl, Fesenmaier, and
Fesenmaier (1993) investigated the relationship between highway accessibility
(infi‘astructure) and tourism expenditures.

Roehl, F esenmaier, and Fesenmaier (1993) used cluster, factor, and path analyses
to examine the contribution of highway infrastructure to regional economic impact of
tourism. In their study, the accessibility of a region was measured by the miles of
highways in the region. They discovered that, when natural and man-made resources
were held constant, regions with more miles of highways received more tourist
expenditures. Further analysis of rural area highway infrastructure showed that the
highway system alone explained a significant proportion of spatial variation in tourist
expenditures. Spotts (1997) used factor analysis to examine tourism resources of each
county in Michigan. In his study, tourist spending was regressed on the factor scores of
tourism resources in a multiple regression analysis to determine the extent to which factor
scores relate to tourism spending. His study results indicate that the factor scores of
tourism resource common factors explained 64% of spatial variation in aggregate tourist
spending in Michigan.

Other tourism infrastructure research shows that counties with interstate highways
have a distinct advantage over other counties, which results in employment and
population growth (Briggs, 1981; Lichter and Fuguitt, 1980; Kuehn and West, 1971).
Further, highway development has different influences in rural and urban areas. An
improved highway infrastructure has no identifiable effects on growth for places more

than 25 or 30 miles from a metropolitan area (Stephanedes and Eagle, 1986; Humphrey

18

and Sell, 1975; Kuehn and West, 1971). Nevertheless, Huddleston and Pangotra (1990)
point to a consensus among scholars that a good highway infrastructure system is a
necessary while not a sufficient condition for regional economic growth. Tourists
businesses especially benefit from the increased volume of travelers due to improved
highway accessibility.

In the reviewed regional analyses of tourism resources by the author, researchers
have all taken a supply-side point of view to the study of tourism. While highway
infrastructure represents accessibility in a passive, static, and supply-side sense, highway
traffic, which takes place on highway infrastructure, represents accessibility of a region in
an active, dynamic, and demand-side sense. Although tourism business operators may
passively benefit from improved highway infrastructure, they are more interested in
knowing the potential size of business opportunities due to the improvement in highway
accessibility. Based on the potential size of business opportunities, tourism business
operators can plan their marketing strategies. While mileage of highway will not provide
them this information, highway traffic data have the potential of making available to them
this kind of information.

Cross-sectional supply-side studies of tourism, such as Roehl, Fesenmaier, and
F esenmaier’s (1993) and Spotts’ (1997), often encounter the problem of using relatively
static tourism resource inventory data to explain a dynamic tourism phenomenon. While
tourism resources and highway infrastructure often remain unchanged overtime, tourism
demand and the trend of participation in tourism activities fluctuate constantly.
Analogous to inflation-adjusted price index, researchers can adjust tourist expenditures

according to regional price differences. Generally speaking, the same bundle of tourism

l9

goods and services costs more in urban than in rural areas. However, regional (spatial)
price differences in tourism goods and services often are not accounted for in regional
(cross-sectional) analysis of tourism. The finding of a strong relationship between
highway infrastructure (or tourism resources) and tourist expenditures often becomes less
significant if researchers regress regional price difference adjusted tourist expenditures on
highway infrastructure (or tourism resources).

The results of previous regional tourism and tourism infrastructure studies, which
established a relationship between highway infrastructure and tourism, did not include
traffic volume on highways as a potential explanatory variable. However, they shed light
on the idea that highway traffic volume like highway infi'astructure and tourism resources
data should have explanatory power in regional tourist expenditures. The following

chapters explore the possible utility of highway traffic data in tourism studies.

20

CHAPTER 3

METHODS

The study design and objectives are discussed in this chapter. The first section
describes the types of data used in the study. The second section identifies the useful
characteristics of highway traffic data for tourism studies (Objective # 1). The third
section presents a theoretical model for linking highway traffic data to tourism studies
(Objective # 2). The fourth section details the proposed new traffic data processing
method to mitigate the problem of separating tourism from non-tourism traffic described
in Chapter 1. Examples of data processing procedures are also provided (Objective # 3
and 4). The fifth section provides a proposed design to assess the validity of the new
method (Objective # 5). The sixth section discusses how to measure the degree to which
the newly proposed method improves the tourism relevancy of original traffic data. A
measurement formula is also derived for this evaluation (Objective # 6).

A basic understanding of the concept of database is important for readers to
follow subsequent discussions. Therefore, a short summary of the concept is presented
here. A database is a collection of tables. Each table contains rows and columns. At the
intersection of a row and a column is a data point. Usually a row is referred to as a piece
of record which contains a number of related data points. A data point is referred to as a
cell in a table or a row. A column is a collection of data points whose attributes (or
characteristics) are the same across rows. A table can be referred to as a record set. The

above database terminology should be sufficient to limit subsequent confusion.

21

3.1 Data Collection

This study is a secondary research design using highway traffic data collected by
the Michigan Department of Transportation (MDOT). The acquired highway traffic data
were recorded by a set of so-called "Permanent Traffic Recorders" (PTR). While there
are temporary counters set up to record traffic information for a short period of time, this
study uses data from Permanent Traffic Recorders only. Also, the Permanent Traffic
Recorders are not truly permanent. The continuing existence of a counter depends on the
needs for traffic information around its location; therefore, the exact number of
permanent traffic recorders changes slightly from year to year. Between 1995 and 1998,
there were 179 permanent traffic recorders distributed in nine MDOT highway
monitoring regions in Michigan (Figure 3-1).

Approximately 47% of counter stations are located in the Southeastern and
Southwestern Lower Peninsula and urban areas (i.e., Region 7, 8, and 9). A detailed
regional distribution of these counter stations is presented in Table 3-1. Of the 179
counter stations in operation during the study period, not all remained operational
continuously. According to the Michigan Department of Transportation, most non-
operating counters are due to equipment failures and the difficulties in repairing the
devices without closing highways. Thus, when there are serious equipment failures of
traffic recorders, traffic data from failed counters will not be available for months or years

until the next scheduled highway re-construction is completed.

22

    
 
   
   
 

Upper Peninsula

Region 1 Region 2

D

 

Northern
Lower
Peninsula
11: 111 ill Iii 141 II in 1111 III 111' 111 ill‘
Region 5
® Region 9 includes:

I Battle Creek, Southern C9 "‘““‘" "“1"”
I Bay City, Lower
. Detroit, .
. Grand Rapids Peninsula
I Kalamazoo,
. Lansing,
I Saginaw

 

 

 

Figure 3-1. Michigan Department of Transportation Highway Monitoring
Regions.

23

Table 3—1. Regional Distribution of Michigan Permanent Traffic Recorders.

 

 

 

MDOT Traffic .
Region Number of 0
Monitoring . . . A
R . * Descrrptron Statrons
egron
1 West Upper Peninsula 21 12
2 East Upper Peninsula 10 6
3 Northwestern Lower Peninsula 14 8
4 Northeastern Lower Peninsula 11 6
5 Middle-western Lower Peninsula 16 9
6 Middle-eastem Lower Peninsula 23 13
7 Southwestern Lower Peninsula 13 7
8 Southeastern Lower Peninsula 23 13
9 City Areas—Detroit, Lansing,
Grand Rapids, Kalamazoo, Battle 48 27
Creek, Saginaw, Bay City
Total: 179 100

 

* MDOT changed its highway monitoring region scheme in 1999. In this study, the old
scheme is used.

24

In 1998, 41 out of 179 counters were installed with the capacity to provide vehicle
type information for the purpose of traffic classification. Counters with this capacity
collect two sets of traffic data. One is Permanent Traffic Recorder (PTR) data, and the
other is Classified Permanent Traffic Recorder (CPRT) data. Older counters without
vehicle classification function record only PTR data.

The daily total traffic from PTR and CPTR usually do not equal each other. In
most cases, the daily traffic volumes from PTR are greater than fiom CPTR. This is
caused by the counters' wheel axle sensors failing to identify correctly the types of
approaching vehicles under the following conditions:

1) extremely high vehicle speed,

2) sudden lane or speed change, and

3) unclassifiable vehicle types.

Traffic data are recorded according to traffic direction in hourly intervals. The
daily recording period is from 00:00 to 23:59. It is an all year round, 24 hours a day, and
7 days a week recording system. Each day at midnight, data are downloaded through
phone lines from traffic counters and then converted into database files. When a whole
month’s data are collected, they are sent to MDOT for highway performance monitoring.

Upon receiving raw traffic data, MDOT performs some basic data cleaning on the
data sets. The criteria that MDOT uses in preparing the data are as follows. Abnormal
traffic data points, which might have been subjected to problems such as partial
equipment failure, highway construction, car accidents, and other unknown events, are
identified. If more than five data points are missing in a given record, an indicator

variable of a problematic record in the day's record will be labeled "TRUE" (The name of

25

this variable is "DEL_FLAG" which means problems exist in this record and the record
can be deleted. See Appendix A.); otherwise, the variable is labeled "FALSE" (which
means there is no problem with the record). If fewer than fiveconsecutively missing
values are recorded, those missing values will be filled in by the average value of traffic
volumes at the same hour interval and the same day of week within the month. For
example, if the traffic volume of first Monday 9 AM in January is missing, then the
missing value will be filled in by the average value of the second, third, and fourth
Monday 9 AM traffic volumes in January. If some data points in a record are dubious
(e. g., outliers), MDOT will try to check the data and determine their acceptability. Ifnot
acceptable, MDOT will replace the dubious data points with appropriate average values.
Ifestirnating a missing value or accepting a dubious data point is deemed problematic, the
whole record will be discarded (i.e., the indicator of a problematic record is labeled as

'OTRU'E'I).

3.1.1 Permanent Traffic Recorder Data

Most counters record data fi'om both traffic directions. If a counter functions
properly, the counter will record 24 data points for each traffic direction daily or 17,520
data points a year (i.e., 24 data points x 2 directions x 365 days). The PTR data used in
this study were collected during January 1, 1995 through December 31, 1998. The
MDOT Permanent Traffic Recorder (PTR) data sets come in both directional and non-
directional format. The non-directional highway traffic data are not suitable for this
study; therefore, non-directional traffic data were not used. Instead of being recorded

hourly, bridge crossing data are only available in daily total volume. Not being

26

compatible with other PTR data, the bridge crossing data were also excluded from this
study.

Since MDOT has already performed basic data cleaning on the PTR data sets,
they are ready for data analysis. Missing records in the data set were left alone without
further treatment. Records with the problematic record indicator coded "TRUE" were
excluded from this study. After removing problematic, non-directional, and bridge
crossing traffic records, the acquired PTR data set comprised 242,979 usable directional
records or 5,831,496 traffic data points. The successful rate of data recording is about
60%, which is considered adequate in sampling studies (Babbie, 1994).

In addition to traffic volumes, each record also contains information on its counter
ID number, traveled lane, direction of traffic flow, year, month, date, day of week of the
recording, total traffic volume of the day, indicator of problematic record, and a note
about the problem. In the acquired PTR data sets, "Traveled lane" is always equal to
zero. This means that the value is a summation of traffic volumes from all of the same
direction lanes on a highway. Although the traffic volume on each lane is not provided,
this information is not needed for this study. A typical example of PTR data format is
displayed in Appendix A for the readers’ reference.

Most major highway corridors in Michigan are either north-south or east-west
oriented. F ifty-percent of counters are located on north-southbound highways and 43%
are located on east-westbound highways. Counters located on northeast-southwest bound

and northwest-southeast bound highways only account for 7% of all counters (Table 3-2).

27

Table 3-2. Directional Distribution of Michigan Permanent Traffic Recorders.

 

 

 

Highway Direction Count %
North-South Bound 88 49.16
Northeast-Southwest Bound 5 2.79
East-West Bound 74 41.34
Southeast-Northwest Bound 6 3.35
North Bound only 1 0.56
South Bound only 1 0.56
East Bound only 2 1.12
West Bound only 2 1.12

Total 179 100.00

 

There were 136 operating Permanent Traffic Recorders during the 1995 through
1998 time period. Their ID numbers, associated highways and locations, directions of
traffic, and available number of records are listed in Table 3-3. Note that the first digit of
a counter ID represents the MDOT highway monitoring region (See Table 3-1 for region

descriptions).

28

Table 3-3.

Locations of Michigan Permanent Traffic Recorders and Number of

Available Records from 1995 through 1998.

 

 

. Number of
Station ID Location County Records*
1019 US-14l Covington, N-S Bound Baraga 898
1029 M-28 Bruce Crossing, E-W Bound Ontonagon 1391
1039 US-2 Wakefield/Bessmer, E-W Bound Gogebic 998
1049 US-2 Iron River, E-W Bound Iron 2500
1069 US-41 Carney, N-S Bound Menominee 2695
1089 US-41 Skandia, NW-SE Bound Marquette 2624
1109 US-41, M-28 Champion, E-W Bound, W. of Jet. M-95 Marquette 2545
1149 US-41, M-28 Champion, E-W Bound, E. of Jet. M-95 Marquette 1960
1189 M-95 Champion, N-S Bound, S. of Jet. US-41, M-28 Marquette 2139
1309 US-45 Land O' Lakes, N-S Bound Gogebic 2361
1449 US-2 Powers, E-W Bound, E of County Road 557 Menorrrinee 2061
1529 US-2 Norway, E—W Bound Dickinson 1357
2029 - US-2 Brcvort, E-W Bound Mackinac 2054
2049 I-75 St. Ignace, N-S Bound Mackinac 2466
2089 I-75 Mackinac Bridge, N-S Bound Mackinac 2922
2109 I-75 International Bridge, N-S Bound Chippewa 2922
2189 M-28 Raco Corners, E-W Bound Chippewa 2344
2209 M-28 Deerton, E-W Bound Alger 2448
3029 M-115 Farwell, NW-SE Bound Clare 2360
3049 M-61 Harrison, E-W Bound Clare 1656
3069 US-131, M-66 Kalkaska, N-S Bound Kalaska 2121
3079 M-72 Kalkaska, E—W Bound Kalaska 60
3089 M-66 Sears, N-S Bound, S. of US-lO Osceola 2706
3109 M-37 Baldwin, N-S Bound Lake 2146
3129 M-37 Traverse City, N-S Bound, S of US-3l Grand Traverse 2715
3149 US- 10 Sears, E-W Bound, W. of M-66 Osceola 2704
3189 US-lO Sears, E-W Bound, E. of M-66 Osceola 2263
3229 M-66 Sears, N-S Bound, N. of US-lO Osceola 2674
3249 US-lO, US-27 Clare, N-S Bound at Travel Info. Center Clare 2558
3269 US-lO Branch, E-W Bound Mason 2362
3289 US-lO Farwell, W Bound Clare 755
4029 US-23 Alpena, N-S Bound Alpena 2742
4049 I-75 Vanderbilt, N—S Bound Otsego 1531
4069 Old M-76 Sterling, NW-SE Bound Arenac 2610
4089 M-33 Rose City, N-S Bound Ogemaw 2788
4129 US-27 Houghton Lake, N-S Bound Roscomrnon 2252
4149 I-75 Prudenville, N-S Bound, at Maple Valley Roscomrnon 2442
4229 US-23 Au Gres, E-W Bound Arenac 2558
5029 US-27 St. Johns, N-S Bound Clinton 1142
5039 US-27 By-Pass, St. Johns, S Bound Clinton 581
5049 I-69 Dewitt, E-W Bound Clinton 1990
5059 I-l96 Hudsonville, NE-SW Bound Ottawa 1516
5069 US-l3l Wyoming, N-S Bound Kent 2166
5109 Washington Road, Ithaca, E-W Bound Gratiot 2778
5169 M-57 Perrinton, E-W Bound Gratiot 2677

 

29

 

 

Table 3-3. (cont’d)
Number of
Station ID Location County Records*
5189 Jordan Lake Road, Lake Odessa, N-S Bound Ionia 2474
5229 I-96 Grand. Rapids, E-W Bound Kent 2243
5249 US—131 Morley, N-S Bound Mecosta 1524
5269 US-3 I Pentwater, N-S Bound Oceana 2512
5289 US-3 I Muskegon, N-S Bound Muskegon 2368
5299 1-96 Ionia, W Bound Ionia 500
5309 US-l3l Big Rapids, N-S Bound Mecosta 654
6049 M-25 Port Sanilac, N-S Bound Sanilac 2610
6069 I-69 Lansing, NE-SW Bound Shiawassee 2531
6109 [-94 Blue Water Bridge, E-W Bound St. Clair 2922
6129 I-75, US-lO, US-23 Birch Run, N-S Bound Saginaw 2674
6149 I-75, US-lO, US-23 Carrollion, NW-SE Bound, SE of I-675 Saginaw 611
6169 M-53 Hemans, N-S Bound, N of Jet. M-46 Sanilac 1966
6189 I-675 Carrollton N-S Bound, S of I-75 Saginaw 183
6209 M-46 Hemans, E-W Bound, E of Jet. M-53 Sanilac 641
6229 I-75 Carrollion NW-SE Bound, NW of I-675 Saginaw 222
6249 M-53 Hemans, N-S Bound, S of Jet. M-46 Sanilac 1848
6269 1-475 Mt. Morris, E—W Bound Genesee 2555
6289 M-46 Hemans, E-W Bound, W of Jet. M-53 Sanilac 1925
,_ 6309 M-57 Clio, E-W Bound Genesee 1846
6319 M-83 Frankenmuth at Cass River Bridge, N-S Bound Saginaw 1383
6369 1-69 Capac, E Bound St. Clair 648
6389 I-69 Capac, E-W Bound, E of Capac Road St. Clair 2463
6429 I-75 Kawkawlin, N-S Bound Bay 2526
6449 I-69 Swartz Creek, E-W Bound Genesee 2293
6469 I-94 Port Huron, E-W Bound St. Clair 2169
6479 US-lO, Bay City, E-W Bound Bay 792
7029 1-94 Grass Lake, at The Truck Jackson 1791
7069 M-60 Homer, E-W Bound Calhoun 2342
7109 US-l31 Schoolcraft, N-S Bound Kalamazoo 2006
7129 Niles-Buchanan Road, Buchanan, E-W Bound Berrien 2642
7159 I-94 Battle Creek, E-W Bound Calhoun 158
7169 I-94 Marshall, E-W Bound Calhoun 2098
7179 I-94 Coloma, E-W Bound Berrien 196
7189 [-94 New Buffalo, N-S Bound Berrien 1639
7269 I-69 Coldwater, N-S Bound Branch 2050
7289 M43 Bangor, E—W Bound Van Buren 2257
7309 I-196 Glenn, N-S Bound at the 114'h St. Allegan 2764
7329 US-12 White Pigeon, E-W Bound St. Joseph 1340
8029 US-127 Mason, N-S Bound Ingham 2219
8049 I-96 F owlerville, E-W Bound at The Truck Livingston 2214
8129 US-12 Jonesville, E-W Bound Hillsdale 1822
8169 US-24 Erie, N-S Bound, N of Lakewood Road Monroe 2262
8209 I-96 New Hudson, E-W Bound Oakland 1961
8219 I-96 Howell, E-W Bound Livingston 92

 

30

 

 

Table 3-3. (cont’d)
. . Number of
Station ID Location County Records“

8229 US-23 Brighton, N-S Bound, S of M-59 Livingston 2053
8249 I-75 Luna Pier, N-S Bound, S of Luna Pier Rd. Monroe 2174
8269 I-75 Luna Pier, N-S Bound, N of Luna Pier Rd. Monroe 296
8409 M-59 Troy, E-W Bound at John Road Oakland 2554
8629 I-75 Clarkston, N-S Bound, S. of M-15 Oakland 1893
8649 I-75 Clarkston, N-S Bound, N. of M-15 Oakland 409
8669 M-15 Clarkston, N-S Bound, Over [-75 Oakland 2583
8689 US-23 Dundee, N-S Bound at the Travel Info. Center Monroe 1612
8709 US-223 Dundee, E-W Bound, W. of US-23 Monroe 2705
8729 US-23 Lambertville, N-S Bound Monroe 1659
9029 I496 Lansing, E-W Bound at Clemens Ingham 1969
9049 US-127 Lansing, N-S Bound, N. of Grand Ingham 2112
9069 I-496 Lansing, E-W Bound at Everett Ingham 1914
9089 Clemens St Lansing, N-S Bound at Ingham 2430
9109 M-lO Detroit, N-S Bound at Milwaukee Wayne 1126
9189 I-275 Romulus, N-S Bound Wayne 280
9199 I-275 Canton Twp., N-S Bound Wayne 1931
9209 1-275 N—S Bound at Cherry Hill Road Wayne 45
9219 I-675 Saginaw, N-S Bound at Saginaw River Saginaw 744
9369 I-94 Kalamazoo, E-W Bound Kalarrrazoo 1191
9419 I-94 Detroit, E-W Bound at Brush St. Wayne 1390
9449 [-75 Detroit, N-S Bound at 12th St. Wayne 250
9489 I-94 Detroit, E-W Bound at Central St. Wayne 219
9499 1-94, E-W Bound at Trumbull Wayne 855
9529 I-94 (Dickrnan) Battle Creek, E-W Bound Calhoun 2782
9629 Liberty Bridge Bay City, E-W Bound Bay 2408
9649 Independence Bridge Bay City, N-S Bound Bay 2748
9669 M-25 (Vet. Bridge) Bay City, E-W Bound Bay 2205
9689 M-l3/M-84 Bridge, Bay City, E-W Bound Bay 2766
9709 I-75 Taylor, N-S Bound Wayne 1464
9729 I-196 Grand Rapids, E-W Bound Kent 2053
9749 M-ll (28th St) Gd. Rapids, E-W Bound Kent 2643
9769 US-l31 Grand Rapids, N-S Bound Kent 2168
9789 M-39 Dearborn, N-S Bound Wayne 805
9809 M-39 Detroit, N-S Bound Wayne 1117
9829 1-696 Southfield E-W Bound, E of Southfield Road Oakland 727
9839 1-696, E-W Bound at Schoenheer Macomb 928
9849 M-lO, Detroit, N-SW Bound at 8 Mile Road Wayne 412
9869 M-lO Detroit, N-S Bound Between 7 Mile Wayne 162
9889 I-75 Detroit, N-S Bound On Rouge River Wayne 1165
9939 M-8 Davison, NE-SW Bound at John Road Wayne 439
9959 I-75 N-S Bound at Mack Ave. Wayne 990
9969 I-94 Detroit, E-W Bound at Dickerson St. Wayne 2382
9979 1-75, N-S Bound at Wattles Road Oakland 703
9989 1-75 Royal Oak, N-S Bound Oakland 1429
9999 M-8 Davison, NE-SW Bound at 2"“ St. Wayne 688

 

*A complete data set should contain 2,920 records for each counter during the 4-year

period.

31

3.1.2 Classified Permanent Traffic Recorder Data

The same traffic counters that provide PTR data can also detect vehicle types if

they were installed with a wheel-axle sensor "loop" under the highway surface. Vehicles

are classified according to the FHWA’s classification method. FHWA classifies vehicles

into the following 13 classes.

"Type Name and Description

1.

Motorcycles (Optional)--All two-or three-wheeled motorized vehicles.
Typical vehicles in this category have saddle type seats and are steered
by handle bars rather than [by steering] wheels. This category includes
motorcycles, motor scooters, mopeds, motor-powered bicycles, and
three-wheel motorcycles. This vehicle type may be reported at the
option of the State.

Passenger Car--All sedans, coupes, and station wagons manufactured
primarily for the purpose of carrying passengers and including those
passenger cars pulling recreational or other light trailers.

. Other Two-Axle, Four-tire Single Unit Vehicles--All two-axle, four

tire vehicles, other than passenger cars. Included in this classification
are pickups, panels, vans, and other vehicles such as campers, motor
homes, ambulances, hearses, carryalls, and minibuses. Other two-axle,
four-tire single unit vehicles pulling recreational or other light trailers
are included in this classification. Because automatic vehicle
classifiers have difficulty distinguishing class 3 from class 2, these two
classes may be combined into class 2.

Buses--All vehicles manufactured as traditional passenger-carrying
buses with two axles and six tires or three or more axles. This
category includes only traditional buses (including school buses)
functioning as passenger-carrying vehicles. Modified buses should be
considered to be a truck and be appropriately classified.

NOTE: In reporting information on trucks the following criteria should
be used:
a. Truck tractor units traveling without a trailer will be considered

single unit trucks.

32

10.

11.

12.

b. A truck tractor unit pulling other such units in a "saddle moun "
configuration will be considered as one single unit truck and will
be defined only by the axles on the pulling unit.

c. Vehicles shall be defined by the number of axles in contact with the
roadway. Therefore, "floating" axles are counted only when in the
down position.

(1. The term "trailer" includes both semi- and full trailers.

Two-Axle, Six-Tire, Single Unit Trucks-~All vehicles on a single
frame including trucks, camping and recreational vehicles, motor
homes, etc., having two axles and dual rear wheels.

Three-Axle Single Unit Trucks--All vehicles on a single frame
including trucks, camping and recreational vehicles, motor homes, etc.,
having three axles.

Four or More Axle Single Unit Trucks--All trucks on a single frame
with four or more axles.

Four or Less Axle Single Trailer Trucks--All vehicles with four or less
axles consisting of two units, one of which is a tractor or straight truck
power unit.

Five-Axle Single Trailer Trucks--All five-axle vehicles consisting of
two units, one of which is a tractor or straight truck power unit.

Six or More Axle Single Trailer Trucks--All vehicles with six or more
axles consisting of two units, one of which is a tractor or straight truck
power unit.

Five or Less Axle Multi-Trailer Trucks--All vehicles with five or less
axles consisting of three or more units, one of which is a tractor or
straight truck power unit.

Six-Axle Multi-Trailer Trucks--All six-axle vehicles consisting of three
or more units, one of which is a tractor or straight truck power unit.

33

13. Seven or More Axle Multi-Trailer Trucks--All vehicles with seven or
more axles consisting of three or more units, one of which is a tractor
or straight truck power unit." (F HWA, 1995)

Among the 13 vehicle classes, Classes 1 (motorcycle), 2 (passenger car), and 3
(pickup) are generally used for multiple purposes which include leisure and recreation.
Although Class 3 is labeled as pickup in Classified Permanent Traffic Recorder data (see
Appendix B), sport utility vehicles (SUVs) and mini vans are also included in Class 3
according to the F HWA’s classification. (Note that a pickup pulling a recreational trailer
is classified as Class 3.) Class 4 (bus) is used for tourist group transportation. Most of
the large-size recreational vehicles (RV in its common usage) are classified into one of
the Classes in 4, 5, or 6.

Since Classes 1, 2, 3, 4, 5, and 6 are more likely used for leisure and recreation
purposes, they are considered "recreational type vehicles" (RV in the usage of this study)
in this discussion. And Classes 7-13 will be categorized into "non-recreational type
vehicles" (NRV) because these are large trucks for delivering cargo and not likely to be
used for leisure purposes. The typical vehicle silhouettes of these 13 vehicle classes are
exhibited in Figure 3-2. (Note that a silhouette in Figure 3-2 does not imply that all
vehicles classified into the class would look like the silhouette provided in the Figure.)
The FHWA's vehicle classification is based on the number and distance of wheel axles
that are non-floating (i.e., wheels have to touch ground). It is not based on the shape of

vehicles.

34

 

Recreational Type Vehicles

 

 

 

 

 

 

 

 

Class 1. Motorcycles Class 2. Cars Class 3. Two Axle, Four Tire
Single Unit Trucks
c..\
95;)! -15!
Class 4.Buses Class 5. Two Axle, Six Tire Class 6. Three Axle Single Unit
Single Unit Trucks Trucks

 

 

Non-Recreational Type Vehicles

 

Class 7. Four or More Axle Single Unit Trucks

w

Class 9. Five Axle Single Trailers

Class 11. Five or More Axle Multi-Trailers
I l
0 W’

Class 13. Seven or More Axle Multi-Trailers

__I la

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Class 8. Four or Less Axle Single Trailers

L—eam

Class 10. Six or More Axle Single Trailers

raise—ml

Class 12. Six Axle Matti-Trailers

[stale—44m

 

 

 

 

 

 

 

 

 

 

 

 

Source: FHWA

Note: The FHWA'S vehicle classification is based on the number and distance of wheel
axles that are non-floating (i.e., wheels have to touch ground). Separation of recreational
type from non-recreational type vehicles is based on the possible recreational use of

vehicles.

Figure 3-2.

35

Vehicle Classifications—Typical Vehicle Silhouettes.

Installing new equipment and retiring the old is a gradual process. To date, not all
traffic counters are installed with the capacity to provide vehicle type information. Also,
Classified Permanent Traffic Recorder (CPTR) data are currently available only for 1998.
In 1998, only 41 Permanent Traffic Recorders had the capacity to provide classified
traffic data. Among these 41 recorders, only 24 stations have the hourly CPTR data
available. The 17 stations that have only daily CPTR data available are less suitable for
the purposes of this study.

CPTR data are recorded hourly. Instead of generating one data point (i.e., one
traffic volume from all vehicle types) per hour as PTR, CPTR generate 13 data points
(i.e., 13 traffic volumes from 13 vehicle types) per hour. Thus, if it functions properly, a
CPTR can generate 624 data points per day (i.e., 13 classes x 24 hours x 2 directions) and
227,760 data points per year (i.e., 624 data points per day x 365 days). Expressed in
terms of a record set, a CPTR counter produces 48 records per day (i.e., 24 hours x 2
directions) and 17,520 records per year (i.e., 48 records per day x 365 days). The
acquired hourly CPTR data from MDOT contain 214,319 useful records, which is about
51% of a complete data set derived fiom 24 CPTR stations (i.e., 24 stations X 17,520
records per year = 420,480 records).

Along with the 13 classified traffic data points, each CPTR record also includes
information about its classification code, state code, county code, station number,
direction of traffic flow, traveled lane, year, month, date, and hour of the recording, and
total traffic volume of the day. A typical example of CPTR data format is provided in

Appendix B for the reader's reference.

36

Stations that provide hourly CPTR data are listed in Table 3-4 with information
on the number of available records. Their geographical locations are displayed in Figure
3-3. Note that there were no CPTR devices installed in the Upper Peninsula of Michigan
during 1998. Since the acquired hourly CPTR data were not cleaned by MDOT, data
cleaning was preformed according to the criteria and procedures used by MDOT
described in the previous section of this chapter. Despite the fact that the acquired CPTR
data are relatively incomplete in terms of available numbers of counter stations compared
to PTR data, hourly CPTR data provide more detailed and valuable information for
tourism studies than PTR data. This is because using hourly CPTR data allows
researchers to remove traffic generated by non-recreational type vehicles (NRV) from
total traffic within each hour interval. Thus, developing of new data processing method
will be based on the assumption that hourly CPTR data are available. In this study, only
the 24 sets of hourly CPTR data were used to demonstrate how highway traffic data are
connected to tourism studies. PTR data were primarily used in the following section

which concerns theory development.

37

 

 

 

Table 3-4. Michigan Permanent Traffic Recorders which Provided Vehicle
Classification Information in 1998.

13:21:; Direction Location 1:23:33!-
3069 N—S US-131, M-66 Kalkaska 13,441
4049 N-S I-75 Vanderbilt 10,170
4129 N-S US-27 Houghton Lake 3,768
5029 N-S US-27 St. Johns 7,083
5039 S US-27 By-Pass, St. Johns 6,452
5059 NE-SW I-196 Hudsonville 15,362
5249 N—S US-131 Morley 14,320
5299 W [-96 Ionia 4,417
5309 N-S US-l31 Big Rapids 10,710
6369 E l-69 Capac 4,705
7029 E 1—94 Grass Lake 2,522
7109 N-S US-13l Schoolcraft 9,263
7159 E-W I-94 Battle Creek 3,824
7179 E-W I-94 Coloma 4,176
8219 E-W I-96 Howell 2,186
8229 N-S US-23 Brighton, 8 of M-59 8,248
8249 N-S I-75 Luna Pier, S of Luna Pier Road 12,353
8689 NS US-23 Dundee 15,454
8729 N-S US-23 Lambertville 11,136
9049 N-S US-127 Lansing, N. of Grand River Ave. 15,730
9369 E-W I-94 Kalamazoo 8,197
9829 E-W I-696, E of Southfield Road 8,121
9959 N-S I-75, at Mack Avenue 11,585
9979 N-S I-75, at Wattles Road 11,096
Total 214,319

 

* A complete CPTR data set should contain 17,520 records in 1998.

38

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5059

 

 

 

9 8229

8 9
21 9829

7159 7029 I
9' 69 ,V, 9959
7109' l 8689
I 8729 3249

Figure 3-3. Geographical Locations of Permanent Traffic Recorders which
Provided Vehicle Classification Information in 1998.

 

 

 

7179

 

 

 

 

 

 

39

3.2 Identifying Characteristics of Highway Traffic Data

As an exploratory step toward understanding highway traffic and creating a
connection to tourism studies, this study begins with identification of highway traffic data
characteristics. This is the first step toward generating basic information about highway
traffic data and discovering its possible uses in tourism studies. Once this basic
information on highway traffic characteristics is explained, the author will use these
characteristics to develop a conceptual model that links highway traffic data to tourism.
The model will become one of the supports for the author's postulation that highway
traffic data is valuable for tourism studies, because traffic data can provide time series
information about traffic volumes with characteristics that implicitly reflect the collective
behavior of travelers.

Both PTR and CPTR data were used in identifying traffic data characteristics.
The major analysis technique involved in these initial examinations of highway traffic
data characteristics was time series analysis, which includes series data plotting, classical
decomposition (i.e., separating time series data into trend, seasonal, and random
components), the method of moving average, and analyzing the autocorrelation function
of the random components. In addition to time series analysis, univariate analysis,
bivariate analysis, pattern recognition, and tests of significance (such as one-way analysis
of variances and two-sample t tests) were also applied in the initial analyses.

After carefully examining the traffic data, the author identified the following
special characteristics of highway traffic data that are usefirl for tourism studies and

worthy of being studied and documented.

40

(1) The random process of highway traffic time series--Taking a micro-view of
the data, the author examined the autocorrelation functions and random
processes of traffic data.

(2) Daily traffic distributions--Taking a 24-hour view of the data, the author
examined the daily variations in traffic flows and their patterns.

(3) Seasonality—-Taking a macro-view of the data, the author examined the
seasonal variations of traffic flows.

(4) Vehicle type distributions--Using CPTR data, the author examined the
distribution of vehicle types in traffic flows.

(5) Geographical locations of traffic counters--Locations of traffic counters
largely determine what will be recorded since the characteristics of traffic vary

widely across a highway system.

3.2.1 The Random Process of Traffic Time Series

Highway traffic data are a set of discrete time series. Each highway traffic data
point is recorded at a specified time interval, and each recorded data point is a realization
of the random process that generates the traffic time series. As previously mentioned,
time series usually can be decomposed into trend, seasonal, and random components.
Here, the author focuses on examining the random component of the traffic time series. It
is a micro-view of the data after removing the trend and seasonal components from the
series.

After performing classical decomposition on some randomly selected sets of PTR

time series, the sampled autocorrelation functions (ACFs) of these randomly selected

41

highway traffic time series display a clear pattern of cyclical fluctuation, which reflects a
periodic behavior in the "random" components. The cyclical period measured from one
peak to the next peak is an exact 24-hour period (see Figure 3-4 for an example). As
most people would expect, the cyclical pattern of highway traffic is influenced by the
alternation of day and night. However, contrary to what one might expect, the random
components are not as random as the ones spawned primarily by random forces. The
cyclical pattern of autocorrelation functions indicates that the random components are not
individually independently distributed white noises. A series of individually
independently distributed white noises should yield a total randomness of ups and downs
without any identifiable patterns, which means each data point has no relationship with
any other data points. However, the ACFs of PTR data demonstrate that highway traffic
data points are highly autocorrelated, and the strength of autocorrelation slowly subsides
after a period of long lags (Figure 3-4). Thus, the ACFs of the random components of
highway traffic data imply that highway travelers’ behavior is not totally random as in a
chaotic situation but closely follows distinct patterns. It follows that travel behavior is
quite predictable as suggested by the observed autocorrelation functions. Travelers are
likely to repeat the same travel behavior over any given periods of 24 hours, because 24-
hour defines a cyclical period for traffic flow. Having made this observation about ACFs,
the author looked further into daily traffic flow distributions in the traffic data to identify

travel behavior for each cyclical period.

42

 

\
\

 

 

 

 

 

   

 

 

 

 

 

::
.2
E
i 0.4
8
.9.
2 0.2
o UtlililrfiillilitfifITYfiIlUIII
-O.2
1 3 5 7 9 111315171921232527293133353739
Legs (in hour)
—0— Auto-Correlation Function

Note: A set of time series { X , ,t = 1,2,... }, the autocorrelation function of X, is defined
Co var iance(X L X H h )
Variance(X,)

random component of traffic data; h is the lag period. For example, p(1) means the
correlation of 1AM with 2AM, and p(2) is the correlation of 1AM and 3AM, and so on.

 

as: p(h) = , h = l to n. Here, {X,,t = 1,2,... } represents the

Figure 3-4. An Example of Autocorrelation Function of Traffic Time Series (4049
North).

3.2.2 Daily Traffic Distributions

The hourly directional PTR data from each counter station can be plotted to reveal
traffic volume distribution over a 24-hour interval. In this section, the author examines
the daily traffic distributions, and plots them for various periods of time (e. g., a month).

Daily traffic distributions reveal the patterns of travelers’ collective travel
behaviors during a period of 24 hours. Different patterns between daily traffic

distributions on certain days of week indicate that travelers demonstrate different travel

43

behaviors across the days of the week. Thus, differences in daily traffic distributions may
help to separate non-discretionary types of travel (e.g., commuting to work) fiom
discretionary types of travel (e.g., leisure). The results of the plotting indicate that the
patterns of weekday daily traffic distributions are very different from weekends. Also,
counters on tourism routes exhibit significant differences in traffic distribution patterns
between Friday and other weekday afternoon hours. This may indicate that a substantial
amount of tourism trips takes place during Friday afternoon and evening hours. For
example, in Figure 3-5, Friday afternoon northbound traffic on highway I-75, a major
highway linking the populated Southeast Michigan area to popular northern Michigan

tourist destinations, increased significantly compared to other weekdays.

 

900

 

800

 

700

 

600

 

 

500

 

 

400

 

300

Average Traffic Volume

 

 

200

 

 

100 a

 

 

 

 

I

T U I U l r U 1

 

O I I—I_I I I I I I I I I I I
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

+ Monday —0— Tuesday + Wednesday + Thursday “'O-Friday

Figure 3-5. An Example of Weekday Daily Traffic Distributions in a Rural Area
(4049 North).

44

The results from comparing daily traffic distributions of different counters
indicate that the patterns of these distributions differ significantly between counters along
tourism dominant and less tourism dominant routes (see Figure 3-5 and 3-6 for example).

Daily traffic distributions of counters on more populated areas frequently exhibit
bimodal patterns on regular weekdays. The two peak points of the bimodal type
distributions appear during morning and afternoon rush hours. Also for counters located
in urban areas, the pattern of distributions indicates that the daily traffic distributions
during weekdays are very similar to each other, and the curves cluster closely such that

weekdays’ distribution curves are almost identical to one another (see Figure 3-6).

 

 

 

 

 

 

 

 

 

 

Average Traffic Volume

 

 

 

 

 

12 3 4 5 6 7 8 9101112131415161718192021222324
Hour

+ Monday -D—Tuesday + Wednesday +Thursday -0-Friday

Figure 3-6. An Example of Weekday Daily Traffic Distributions in an Urban
Area (9999 Northeast).

45

Monthly compilations of daily traffic distributions often reveal that there are three
or four distinct clusters of daily traffic distributions. Usually, these clusters are Friday,
Saturday, Sunday, and the other weekdays. Sometimes, Saturday and Sunday
distributions are similar, but frequently the plots indicate that they do not cluster into a
single group. Similar observations also occur on Friday and the other weekday clusters.

Moreover, for northbound traffic, the daily traffic distributions of the last day of
long weekends (such as Memorial Day, Independence Day, and Thanksgiving) and
regular Sundays frequently stay at lower positions (less traffic) compared to other
distributions. This may imply that there is less northbound tourism traffic on regular
Sundays and the last day of long weekends. For southbound rural traffic, the daily traffic
distributions of the first day of long weekends and regular Saturdays frequently stay at
lower positions compared to other daily distributions. Daily traffic distributions on New
Year’s Day, Halloween, Thanksgiving, and Christmas frequently stay at lower positions
compared to the distributions for other days of the month. For traffic in urban areas, the
daily traffic distributions of regular Sundays frequently stay at lower positions compared
to the distributions for other days of the week. An example of three clusters of daily
traffic distributions is exhibited in Figure 3-7. In Figure 3-7, the thin black curves
clustering at higher positions are weekday traffic (including Friday), the thin gray curves
are Saturday traffic, the heavy gray dash curves are Sunday traffic, and the black dash

curve at the lowest position is the traffic on Halloween.

46

8000

 

 

 

 

 

7000 i

6000 a
O ..
E 5000
2
3
2
E 4000 ~
g.
2*
h
3
O
:c 3000 -

2000 ~

s.
1000 - - . \
-—
\ .
0 T I I I I I I I I I I I I T ‘I r I I I I I I I
12 3 4 5 6 7 8 9101112131415161718192021222324
Hour
Weekday Saturday -- —Sunday ----------- Halloween
Figure 3-7. An Example of Three Clusters of Daily Traffic Distributions (9969

East, October 1998).

47

 

It appears that collective changes in travel behavior of highway travelers are
reflected in the different patterns of daily traffic distributions. This characteristic of
traffic data definitely connotes valuable information that is not regularly exploited by
most tourism researchers. Therefore, the differences in daily traffic flow distributions

could be used to derive information for tourism studies.

3.2.3 Seasonality

Seasonality is one of the most familiar topics to tourism researchers. Highway
traffic data also display seasonal variations. To study highway traffic seasonality the
author examined traffic data within a year as a whole. It is a macro-view of traffic data.

Highway traffic data exhibit three kinds of periodicity:

1) 24-hour daily period,

2) 7-day weekly period, and

3) 12-month (or 4-season) annual period.
In the previous sections, daily and weekly periodicities have been considered in the
classical decomposition operation when the random process of the data were examined,
and in the analysis of daily traffic distributions when the differences in daily traffic flows
were compared. In this section, the author focuses on the 12-month (4-season) annual
periodicity. In the following discussion, seasonality is referred to as the 12-month annual
periodicity.

One would expect that non-tourism traffic flow would be relatively more stable

over time than tourism traffic flow and that tourism traffic varies widely due to the

48

seasonal variations of tourist visitations. Thus, traffic data exhibiting significant seasonal
variations are likely to contain a greater preportion of tourism traffic.

Observation of the traffic data reveals that seasonality is much less significant in
urban areas than in rural areas. For example, rural north-southbound highways exhibit
significant increases in traffic volumes at the beginning of the summer season, usually in
late May or early June. The Michigan highway traffic peak season is during the summer
and early fall. Since seasonality is more significant in rural highway traffic, annual
seasonality is a useful characteristic of traffic data for deriving rural tourism related

information.

3.2.4 Vehicle Type Distributions

The Federal Highway Administration (1995) has pointed out that "[t]rucks and
other commercial vehicles serve different purposes and may have travel patterns which
differ from those of automobiles." Thus, one could expect that the vehicle type
distributions on more tourism-dominant routes would be different fiom those on less
tourism-dominant routes. CPTR data were used in this section of study, because CPTR
data sets provide all information included in the PTR data sets plus vehicle type
information.

To derive more useful tourism related information, it is better if one removes
traffic volumes generated by non-recreation type vehicles from overall traffic data set.
An initial analysis of vehicle type distribution reveals that, on average, about 90% of
traffic volume is generated by recreation type vehicles, that is F HWA vehicle Classes 1-6.

Variations occur across regions, seasons, directions, and days of week (i.e., weekday vs.

49

weekend day). Generally speaking, the percentage of traffic volume generated by
recreation type vehicles is higher on weekend days than during weekdays. One-way
analysis of variances (ANOVA) was applied in testing the significance of differences
between weekdays and weekend days and also in testing the significance of differences
across seasons. Information derived from vehicle type distribution also suggests which
highway routes and which days of the week are more tourism dominant. More details

about tourism dominance are presented in Chapter 4.

3.2.5 Geographical Locations of Traffic Counters

Location can be the most important factor in determining what kinds of traffic
data are recorded. For example, traffic counter stations located in urban areas record a
higher proportion of commuter, service and freight traffic than rural traffic counter
stations; and rural traffic counters record a higher proportion of tourism traffic than urban
traffic counters. However, this does not necessarily mean that more tourism activities (in
quantity) take place in rural areas than in urban areas. It simply means that tourism traffic
flow may be a more dominant component in total traffic flow in rural areas than in urban
areas. The previously identified highway traffic data characteristics are strongly affected
by the locations of traffic counters. Theoretically, a counter's location should be an
important factor in determining the tourism dominance of a highway route. Later in this

study, this postulation will be tested.

50

3.3 A Conceptual Model Linking Highway Traffic Data to Tourism Studies
Based on observed highway traffic characteristics, the author developed the
conceptual model introduced in this section. It depicts a simple theory of the collective
behaviors of highway travelers that can be used to link highway traffic data to tourism

studies. The model also suggests possible uses of highway traffic data in tourism studies.

3.3.1 The Conceptual Model

In a prirrritive society, people’s daily life follows nature’s schedule closely. They
work and collect foods during the day and rest during the night. They plant during the
spring and harvest during the fall. As human societies became civilized, institutions and
laws become the norm and regulate people’s behaviors. In an industrialized society, most
people’s daily lives closely follow an 8-to-5 (i.e., 8 work hours and one lunch hour),
weekday-and-weekend, workday-and-holiday calendar schedule. Therefore, people’s
behaviors are influenced by the institutional (social) factors and natural settings of their
societies. Naturally, their travel behaviors are also under the influence of these
institutional and natural settings. As a matter of fact, there is a school of scholars that use
institutional settings of a society and property rights to explain human social and
economic behaviors (e.g., Eggertsson, 1990; Furubotn and Richter, 1998)

While tourism studies are about travelers’ behaviors, traffic data characteristics,
such as the autocorrelation functions, daily traffic distributions, and seasonality of traffic
time series, provide information about the collective behaviors of highway travelers for

tourism studies. Since not all types of vehicles are used for leisure and recreation, vehicle

51

classification information allows researchers to derive even more relevant information
from traffic data for tourism studies.

However, these special characteristics of highway traffic data are not only
determined by peoples' travel behaviors under the influence of institutional and natural
settings of society but are also determined by the geographical location (environmental
factors) of traffic counters that record the data. Thus, conceptually, the random processes
of highway traffic time series, daily traffic distributions, seasonality, vehicle type
distributions, and geographical locations of traffic counters provide concordant evidence
about tourism traffic flow.

In the model, the connection between the concept of patterned travel behaviors
and tourism studies is based on highway traffic data characteristics. Therefore,
information about traffic data characteristics will be used to develop a new data
processing method to estimate tourism traffic for tourism studies. The derived tourism
traffic information can be more confidently used (i.e., data relevancy to tourism is
improved) in tourism research and planning in either a site-specific or a regional tourism

monitoring system. The conceptual model in a graphic format is presented in Figure 3-8.

52

 

Characteristics of Highway
Traffic Data

 

Series

 

T

 

Geographical
Location of
Counter

 

 

 

 

0 Random Processes of Traffic Time

Daily Traffic Distributions
Seasonality

 

_> Vehicle Type

Distributions

 

 

 

 

 

 

Institutional and
Natural Settings of
a Society

 

 

 

 

Collective

— Behaviors of

Highway Travelers

 

 

 

 

 

New Method in Highway Traffic Data
Processing and Refinement
for Tourism Studies

 

 

Data Usage

 

 

Site-Specific Tourism
Applications

0 Monitoring

0 Planning

 

 

 

 

Regional Tourism
Applications
Monitoring
Planning

 

 

Figure 3-8. A Conceptual Model Linking Highway Traffic Data to Tourism

Studies.

53

3.3.2 Patterned Behaviors and Routine Traffic

For purposes of tourism studies and planning, traffic data need to be divided into
their tourism and non-tourism components. As previously mentioned, it is technically
impossible to precisely separate tourism traffic from non-tourism traffic, because no
researcher knows the trip purposes of all vehicles running on highways. However, a
possible way to mitigate the problem of separating tourism from non-tourism traffic is to
remove what it is believed to be the most unlikely tourism-related traffic from the total
traffic. Thus, the remaining traffic flow would be more relevant for tourism studies than
the original untreated data sets. It has already been shown that we can remove traffic
generated by vehicles that are not likely to be used for leisure and recreation. Further, the
concept of patterned behaviors of highway travelers can help us in the data refinement
process.

Based on the conceptual model and initial observations of the random process, daily
traffic distributions, and seasonality of highway traffic time series, the author assumes that
there is a necessary amount of highway traffic that people have to take in order to satisfy
society’s institutional settings and individuals’ societal functions. For example, peeple have
to travel to work, to attend schools, to do grocery shopping, to see doctors, to deliver goods
and services. Most public and private offices operate between 8AM and 5PM (i.e., 8 work
hours and one lunch hour), and most stores are closed after 10PM. These scheduled and
patterned collective behaviors are necessary so that the entire society can function normally.
To perform each individual’s societal functions, most people follow a fixed work schedule
(for example an 8-to-5 and 40-hour workweek). Those who don’t have to follow a fixed

work schedule most often adhere to the majority’s schedule, because the services they need

54

are often not readily available at other times (i.e., the majority rules). Collectively, random
behaviors of individuals in performing their roles in the society and to take care of their own
personal business as well public affairs lead to relatively stable traffic patterns. In other
words, as the result of people’s patterned travel behaviors, this necessary amount of traffic
can be characterized as routine. Although, routine traffic may vary across places and
seasons and differ between weekday and weekend, it is often non-discretionary and less
likely to be considered tourism-related.

Based on this observation, the author developed a new data refinement method and
calls it the Removal of Routine Traffic Method in this study. The method uses the
differences among daily traffic distributions to remove routine traffic. Mathematically, it

performs the following Operation:

Tourism traflic = Total traffic — NR V traflic — Routine traffic ,

where NRV traffic refers to traffic generated by non-recreation type vehicles (i.e., the
FHWA vehicle Classes 7-13), easily abstracted from Classified Permanent Traffic Record
data. Since routine traffic is unknown, it has to be estimated. Tourism traffic is the residual
after removal of routine traffic. In Figure 3-9, each vertical bar represents the total traffic
voltune within an hour. Each vertical bar is comprised of three parts———the gray-colored
portion represents the traffic generated by non-recreational type vehicles (NRV), the black-
colored portion represents routine traffic, and the white-colored portion represents estimated
tourism traffic during each hour interval. Estimated tourism traffic for a day is the sum of

white-colored bars across a 24-hour period.

55

 

1400

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1200
1000
it
a A
g .i ,.
3 800 ‘ l
e ,
> i
i- 1%
E‘ 600 .4
3
o
I II I
o 4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
I Routine Traffic Cl Tourism Traffic [21 NRV Traffic

Figure 3-9. A Hypothetical Graph Showing the Separation of Hourly Traffic
Volume into Three Components.

56

3.3.3 Method for Estimating Routine Traffic

Unlike NRV traffic, routine traffic needs to be estimated, because there are no
statistics available for this kind of traffic. In order to estimate routine traffic, we need to
select an appropriate cut-off point to separate routine from tourism traffic for each hour's
traffic data. The following is the author’s rationale and suggestion for how to select this
cut-off point. First let us look at a special distribution of traffic data.

So far, we have examined the traffic data in a longitudinal way, that is from 1AM
to 12PM, a 24-hour period, period by period. We have called each of the 24—hour's
distribution "daily traffic distribution". We can also look at the data in a cross-sectional
way, that is to examine traffic data recorded at the same hour point across different days.
These hourly traffic data points from different days form a distribution. Let us call this
kind of hourly traffic distribution the "hour-column distribution". Since there are 24
hours in a day, there are 24 hour-column distributions in any group of daily traffic
distributions. For example, Figure 3-10 displays these 24 hour-column distributions from
October's traffic data on highway I-75 around Station 4049 North. Each hour-column has
its own distribution. Some distributions are wider and some distributions are narrower.
For example, in Figure 3-10, the distribution range of hour-column at 9AM is much
narrower than that at 5PM. A narrower range of hour-column distribution suggests that
the traffic registered at that hour is primarily routine traffic, because traffic volumes vary
slightly across days. A wider distribution suggests that there is more variation which may

be due to non-routine traffic at that hour.

57

1 000

 

 

 

 

 

 

 

 

 

 

 

 

Hour-column 0
distribution at 5PM, / ii ,. .
with a wider range of . : .
. distribution. 0 O
900 8 \ . .. e
. 1 9
800 -
~O
‘0
700 - l .
e ’ 0 . . a
. a “ ‘
E 600 o" “I 7 .‘ . t
.2 1 . 0“ e . ° .
g i ‘l ‘ . .
2 I . . l
3: 500 -
E u . . . C . . i
'- 0
>0 . . . . . l‘
'C
:l .1 Hour-column . O . . . .. L
3‘3 400 distribution at 9AM, .‘ g: g . . .e . 0
withanarrower . z . . _ ‘ ‘.
range of distribution. 0, . . '. 0 g ‘ .. O . . . 3
300- 3 9' ! I " I ' 7 9 .e
\ ' . a ' ' 0 .» \
" U '1 f ’ . ' . w . ‘uz. . i, '
fl , . . " , ll ' C ‘ . .
200 ' 1 rl I -‘ . . . ‘5‘ '. .
" ' . . . e“
. . O 13 . -‘ .
I .0, 0.“
100 " . l~ ‘! ‘1 in. g
‘ ‘ ’ mil
3 _ Ml
I I I I I I I

 

 

 

12 3 4 5 6 7 8 9101112131415161718192021222324
Hour

Note: Each point in the figure represents the hourly traffic volume for one day.
Each vertical stack of points is an hour-column distribution (across days). Hourly traffic
data points on the same day are connected by the gray-colored lines (i.e., daily traffic
distributions). Most of the outliers in this figure are Fridays' traffic, which may lead to
the proposition that Friday noon is the beginning time of a weekend. However this kind
of observations often associated with traffic on rural highways. This proposition may not
be true for traffic in urban areas (see Figure 3-6 for an example).

Figure 3-10. Examples of Hour-Column Distributions of Traffic Data (4049 North,
Weekdays in October 1998).

58

Now let us consider the following scenario. Generally speaking, most people's
lives and activities are scheduled on a weekly basis (i.e., weekdays and weekend)
throughout the calendar year. Therefore, a period of seven days (i.e., one week) is a
logical time flame to study people's travel behaviors and patterns. If traffic flow during
the seven days distributed evenly across the period (i.e., everyone follows exactly the
same travel pattern day by day), then each day's traffic volume should account for 14.3%
(i.e., 100% + 7 = 14.3%) of the total traffic volume during the 7-day period. However, in
reality, traffic volumes do not distribute evenly across any 7-day periods. There are high
and low traffic days. Therefore, the percentage of traffic volume accounted for by the
lowest traffic day will be less than 14.3%, and the percentage of traffic volume accounted
for by the highest traffic day will be more than 14.3%. The percentage of daily traffic in a
7 -day period (DP) is calculated as:

DV

I

DP, =
Sum(DV. +DVM +-~+DV...)

 

where DP, = Percentage of Day t traffic volume in a 7-day period, and

DV, = Daily traffic volume on Day I.

For example, the third column in Table 3-5 exhibits the percentages of daily
traffic in the 7-day period (DPS) 10/1/1998 to 10/31/1998 (using data flom 4049 North).
It is very likely that, on the lowest traffic day, there is more routine traffic but little
tourism traffic. Thus, the traffic on lowest traffic day may suggest a cut-off point to

estimate routine traffic.

59

We can also calculate the percentage of lowest daily traffic volume in a 7-day
period (LP) as:

_ Minimum(DV, ,DV,+1 ,...,DV,+6)
I sum(DVt +DVIH +"'+DVt+6)

 

where LP, = Percentage of lowest daily traffic volume in a 7-day period for Day I, and

DV, = Daily traffic volume on Day t.

The fourth column of Table 3-5 exhibits the percentages of lowest daily traffic volume in
the 7-day period (LPs) 10/1/1998 to 10/31/1998.

During a period of time (e.g., a month or a year), the percentages of daily traffic in
any 7-day periods (DPS) form a probability distribution. Theoretically, the lowest
possible percentage is 0% when the day has no traffic on the monitored highway. This
may happen when all traffic is stopped due to some catastrophe. On the other hand, the
highest possible percentage of daily traffic in a 7-day period can come close to 100%.
This happens when the other 6 days have no traffic at all. Although these extreme cases
usually do not happen, theoretically they are possible. Therefore, the theoretical
distribution range of DPs is flom 0% to 100%, and the percentages of lowest daily traffic

in a 7-day period (LPs) are points within the distribution range of DPS.

60

Table 3-5. Examples of Percentages of Daily Traffic and Percentages of Lowest
Daily Traffic Volume in a 7-Day Period (4049 North, October 1998).

 

Daily Traffic Percentage of Darly Traffic Percentage of Lowest Daily Traffic

 

 

Date 2‘23? Volume in ($23)” Period Volume in a 7-Day Period (LP,)
(1) (2) (3) (4)
10/1/98 6537 15% 10%
10/2/98 1 1492 27% 10%
10/3/98 6591 15% 10%
10/4/98 5679 13% 9%
10/5/98 4673 10% 9%
10/6/98 4159 9% 9%
10/7/98 4258 9% 9%
10/8/98 6424 14% 9%
10/9/98 1 1950 27% 9%
10/10/98 7753 18% 9%
10/11/98 6053 14% 9%
10/12/98 4795 12% 10%
10/13/98 3907 10% 10%
10/14/98 4364 11% 9%
10/15/98 5707 15% 9%
10/16/98 1 1348 30% 9%
10/17/98 6239 18% 10%
10/18/98 4695 14% 10%
10/19/98 3922 12% 10%
10/20/98 3402 10% 10%
10/21/98 3640 11% 10%
10/22/98 4672 14% 10%
10/23/98 8242 25% 1 1%
10/24/98 5322 17% 11%
lO/25/98 4567 16% 11%
10/26/98 3766 13%" 12%*
10/27/98 3479 13%"' 11%"
10/28/98 3688 14%"' 10%*
10/29/98 3967 15%* 1 l%"'
10/30/98 6322 25%* 11%*
10/31/98 3315 13%* ll%"‘
DV, _ Minimum(DV,,DVM,...,DV,+6)

 

 

Note: DP,= , ,— ,
.6) Sum(DV,+DV +...+DV,+6)

(+1

Sum(DV, + DV + ...+ DV

Hi I
where DP, = Percentage of Day t traffic volume in a 7-day period,
DV, = Daily traffic volume on Day t ,
LP, = Percentage of lowest daily traffic volume in a 7-day period on Day t .
* Some data flom November were used in the calculation in order to form a period of 7
days.

61

Using CPTR data, the author calculated the percentages of daily traffic (DPS) and
lowest daily traffic (LPs) in a 7 -day period throughout the entire year for each CPTR
counter. Then, he averaged the derived percentages (LPs) flom a counter according to
traffic direction. The results are exhibited in Table 3-6. The overall average of these
percentages (LPs) flom all CPTR data sets is about 11%. Since it is possible that, on the
lowest traffic day, some tourism trips are taken by people, the amount of routine traffic,
on average, would most likely be a bit less than the amount of traffic represented by the
11% derived here. The author therefore made a decision to lower the 11% to 10% and
then move on in the processes of estimating routine traffic. (The 10% position is a point

in the distribution of DPS whose range is flom 0% to 100%.)

62

 

lal

 

 

 

Table 3-6. Average Percentage of Lowest Daily Traffic Volume in a Seven-Day
Period (in 1998).

Station Direction Location Average Percent
3069 N US-13l, M-66 Kalkaska 10.71%
3069 S 12.13%
4049 N I-75 Vanderbilt 9.88%
4049 S 10.28%
4129 N US-27 Houghton Lake 10.17%
4129 S 10.43%
5029 N US-27 St. Johns 10.50%
5029 S 11.74%
5039 S US-27 By-Pass, St. Johns 11.75%
5059 NE I-l96 Hudsonville 9.75%
5059 SW 10.04%
5249 N US-131 Morley 10.85%
5249 S 11.32%
5299 W I-96 Ionia 11.41%
5309 N US-l3l Big Rapids 10.88%
5309 S 11.68%
6369 E I-69 Capac 12.15%
7029 E I-94 Grass Lake 11.85%
7109 N US-l31 Schoolcraft 10.81%
7109 S 11.26%
7159 E I-94 Battle Creek 12.00%
7159 W 12.61%
7179 E I-94 Coloma 12.06%
7179 W 11.79%
8219 E I-96 Howell 11.84%
8219 W 11.68%
8229 N US-23 Brighton, S of M-59 10.39%
8229 S 12.36%
8249 N I-75 Luna Pier, S of Luna Pier Rd. 11.74%
8249 S 11.94%
8689 N US-23 Dundee 12.33%
8689 S 12.22%
8729 N US-23 Lambertville 11.79%
8729 S l 1.75%
9049 N US-127 Lansing, N. of Grand 10.08%
9049 S 11.09%
9369 E I-94 Kalamazoo 11.13%
9369 W 11.00%
9829 W 1-696, E of Southfield Rd. 9.54%
9829 E 9.51%
9959 N 1-75, at Mack Ave 10.73%
9959 S 10.40%
9979 N I-75, at Wattles Rd. 9.08%
9979 S 9.38%

Overall Average 11.09%

 

63

Although one could use the 10% position to estimate weekly (i.e., 7-day) tourism
traffic, this method would be too general and look coarse, since the percentage of tourism
traffic would then be consistently greater than 30% of each week’s total traffic (see Table
3-7).

In Table 3-7, a day's tourism traffic percentage (based on 7-day's total traffic) is
equal to DP, - 10%. If the subtraction is less than 0%, a 0% is used as the day's tourism
traffic. Adding these daily tourism traffic percentages in a group of 7 days, we can see
that the estimated weekly tourism traffic percentages are consistently greater than 30%.
The same problem arises on each CPRT data set (i.e., on each highway route).

The use of the 10% position flom the distribution of DPS is therefore a very
coarse way to estimate weekly tourism traffic. Further, while traffic data are collected in
an hourly basis, the above method utilizes daily total traffic data only, that is, it fails to
utilize hourly (more detailed) information from traffic data sets. Based on these concerns,
it appears better to find a method which utilizes the information flom hourly traffic data
to estimate routine traffic and tourism traffic.

Now, let us explore the "hour-column distribution" which is a distribution of
hourly traffic volume at the same hour across different days. Conceptually, an hour-
column distribution is similar to the distribution of DPS, because both are distributions of
traffic across days. The difference is that the former is on an hourly basis and the latter is

on a daily basis.

64

Table 3-7. An Example of Coarse Estimation of Weekly Tourism Traffic Based
on Average Percentage of Lowest-Traffic Day in a 7-Day Period.

 

 

 

 

 

 

Percentage of One Day 10% position as Estimated Tourism Percentage of
Date Traffic Volume in a 7-Day Routine Traffic Traffic Tourism Traffic
Period (DP,) for a 7 Day Period
(2) (3) (4)* = (2) - (3) (5)“
10/1/98 15% 10% 5%
10/2/98 27% 10% 17%
10/3/98 15% 10% 5%
10/4/98 13% 10% 3% 30%
10/5/98 10% 10% . 0%
10/6/98 9% 10% 0%*
10/7/98 9% 10% 0%*
10/8/98 14% 10% 4%
10/9/98 27% 10% 17%
10/10/98 18% 10% 8%
10/11/98 14% 10% 4% 36%
10/12/98 12% 10% 2%
10/13/98 10% 10% 0%
10/14/98 11% 10% 1%
10/15/98 15% 10% 5%
10/16/98 30% 10% 20%
10/17/98 18% 10% 8%
10/18/98 14% 10% 4% 40%
10/19/98 12% 10% 2%
10/20/98 10% 10% 0%
10/21/98 11% 10% 1%
10/22/98 14% 10% 4%
10/23/98 25% 10% 15%
10/24/98 17% 10% 7%
10/25/98 16% 10% 6% 37%
10/26/98 13% 10% 3%
10/27/98 13% 10% 3%
10/28/98 14% 10% 4%
10/29/98 15% 10% 5%
10/30/98 25% 10% 15%
10/31/98 13% 10% 3%

 

Note: This table is using the same data as Table 3-5.
*The values in Column (4) = (2) — (3). If (4) is less than zero, zero is used.
"The values in Column (5) are sum of every 7 values in Column (4).

65

The distribution of DPs is within a range flom 0% to 100%, which matches the
hour-column distribution whose range is flom 0 to some larger number of traffic count
(i.e., the maximum traffic volume that a highway can carry in one hour). Conceptually, it
should be feasible to apply the 10% position in the distribution of DPS (0% to 100%) to
hour-column distributions. Practical difficulties arise when the 10% position in a range
of 0 to some larger number is selected. If the choice of the 10% position in an hour-
column distribution were based on the maximum traffic volume in the distribution (i.e.,
maximum traffic volume X 10%), a problem would occur if there was an extremely high
volume (outlier) in the distribution. The estimated routine traffic would become much
higher when there is an outlier than when there is no outlier. Since routine traffic is
generally more stable than tourism traffic, it is desirable that the effects of an outlier are
on tourism traffic rather than on routine traffic. Using the 10% position to estimate
routine traffic does not obviate this problem; using a "percentile" position on the other
hand can help avoid this difficulty.

Ifthere are 100 data points in a distribution which are ordered flom minimum to
maximum according to their values, the data point in the middle of the rank is called
median. A median point is also called the 50th percentile point, because 50% of data
points in the distribution are smaller than the median and 50% of the data points are
greater than the median. The most commonly used percentile points are the 25th, 50th,
and 75th percentiles. A median is not equal to the mean value, unless the data are
normally or evenly distributed. So, if traffic data points are evenly distributed, the 10%

position is also equal to the 10th percentile position in the distribution.

66

There are two benefits to using the 10th percentile position as compared to using
the 10% position in a distribution like the hour-column distribution:

1) It is easier to locate in an hour-column distribution. Since the calculation is not
based on the maximum value in the distribution, it is not necessary to know
what the maximum value is, and

2) It can prevent the distorting effects of an outlier on estimating routine traffic, so
that there is no need to be concerned about how large the maximum value is.

In this approach, outliers which may be caused by tourism activities will not affect
the estimation of routine traffic. They instead contribute to the estimation of tourism
traffic since a percentile is used as the cut-off point. For example, Figure 3-11
hypothetically displays two hour-column distributions. The two distributions are almost
identical except that there is an outlier in the distribution of Hour 2. Using the 10th
percentile position to estimate routine traffic, we will derive exactly the same estimates
for both hour-column distributions. However, using the10% of the maximum value in the
distribution, the estimate routine traffic in hour-column 2 is greater than that in hour-
column 1. The rationale for selection of the 10th percentile in this case is analogous to
the common practice by analysts that the median rather than the mean is used as
descriptor of central tendency of a distribution when outliers exist.

In view of the above demonstration, the author recommends the use of the 10th
percentile position in each hour-column distribution to estimate routine traffic. Further,
he would like to point out that this selection is reasonable since, on average, the estimated
tourism traffic is close to what the Federal Highway Administration has estimated by

using telephone surveys (see Chapter 4).

67

45-

 

 

 

 

 

 

40 . e
o 35 . o Hourly
E Traffic
2 3o . Volume
0
>
2 25 - El 10th
51: Percentile
E 20 -
>.
E 15 « A 10% Position
0
I.
10 a
5'1
0 . .
Hour 1 Hour 2
Hour

Note: The 10% and 10th percentile positions as the estimates of routine traffic volumes
overlap in Hour-Column 1.

Figure 3-11. A Hypothetical Example Showing an Outlier's Effect on the
Estimation of Routine Traffic.

Therefore, in an hour-column distribution, each data point can be ranked flom
minimum to maximum and referred to by a percentile position in the distribution. When
connecting the Nth (N = 1 to 100) percentile points across these 24 hour-column
distributions, a curve similar to daily traffic distribution is obtained and the curve is
associated with the Nth percentile. Figure 3-12 depicts an example of using the 10th

percentile daily traffic flow to estimate routine traffic.

68

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

e ' .
3 e
900 ._L4_..T.—
e
800 '
e
o
a
o 700 . . . .
E e
'2 600 I . .
0 Fr—
> . . e g
e . c
E 500 1 e ' O f
'; t . - . .
400 a
'5 . rI—l—i
O . e 9 e e
I 0 e U
300 . I r
3 , . II
I , 1 .
200 , . fl —
V .
e
100 w: r . _‘ ‘ O
0 " — _ _ ‘ I I I I r I I a I I I I I I I T I r
1 2 3 4 5 6 7 8 9101112131415161718192021222324
Hour
0 Hourly Traffic Volume +10th Percentile Daily Traffic Flow

Note: The 10th percentile daily traffic flow is the estimated daily routine traffic for each
day in the month. For each data point in an hour-column, the difference between the
point and the corresponding 10th percentile point is the estimated hourly tourism traffic
volume. If a data point is lower than the 10th percentile point, its tourism traffic volume
is assumed to be zero.

Figure 3-12. Example of Using the 10th Percentile Traffic Flow to Estimate
Routine Traffic (4049 North, Weekdays in October 1998).

69

3.4 Estimating Tourism Traffic

The following demonstrates the use of hour-column distributions and their 10th
percentile positions in the Removal of Routine Traffic Method. This method is designed
to achieve the main objective of the study, that is to estimate tourism traffic. It is
designed to remove hourly NRV traffic, estimate and remove hourly routine traffic, and
leave the residual as the estimate of hourly tourism traffic. This section presents the
detailed procedures for implementing the method. The cleaned hourly CPTR data were
used as major data sets to be fed into the removal of routine traffic procedure. For
illustration, the author used October 1998 CPTR data flom 4049 North (I-75 at
Vanderbilt) to demonstrate the removal of routine traffic operation in this section. To
help readers understand the procedures, matrix terminology is used in the following

description.

3.4.] Step One—Data Filtering and Transposing

1. Data Filtering.
In matrix terms, the CPTR data flom each day's recording comprise a 24 by 13
matrix (24 hours and 13 vehicle types)—24 is the number of rows and 13 is
the number of columns. Among the 13 vehicle types, only Classes 1 to 6 are
considered recreation-related vehicle types; therefore, data points generated by
non-recreational type vehicles need to be removed. Table 3-8 displays the
separation of NRV traffic data points flom RV traffic data points on October

1,1998.

70

Table 3-8. Separating NRV Traffic from RV Traffic (4049 North, 10/1/1998).

 

 

 

 

 

 

 

S e “i
> 0 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13
'< S

10 1 l o 31 24 o o 1 o 1 6 o o 0 2
10 1 2 0 15 7 0 o o o o 5 o o 1 1
10 1 3 o 13 10 0 0 o o o 1 o 1 o 1
10 l 4 o 12 10 0 3 o o 2 4 o 1 o o
10 1 5 o 13 9 1 3 0 o o 1 1 o 0 o
10 1 6 0 16 7 0 1 o o 6 10 4 1 0 3
10 1 7 0 42 23 o 3 1 o 5 15 2 0 2 4
10 1 s o 82 43 o 5 4 2 5 16 s 1 0 3
10 1 9 o 106 65 2 1 2 o 4 10 7 o o 5
1o 1 10 0 151 84 o 7 1 o 6 20 s 1 0 5
10 I ll 0 161 100 l 2 l 0 Remove C7 to C13, 4
10 l 12 o 236 102 5 10 2 o . 6
10 i 13 1 297 143 1 5 l 0 traffic of non-recreational 5
10 1 14 0 270 135 2 5 l 0 vehicle types 7
10 1 15 o 324 141 3 4 l o 10 13 13 o o 7
10 1 16 o 343 131 1 4 4 o 19 s 3 o o 7
10 l 17 o 356 153 0 4 2 o 15 17 7 0 o 7
10 1 18 o 343 152 2 4 2 o 10 s s l o 6
10 l 19 o 227 112 1 3 2 1 10 18 1 l 1 3
10 1 20 o 251 122 0 3 3 o 6 7 2 o o 2
10 l 21 o 200 101 1 0 o o 9 7 2 o o 6
10 1 22 o 214 86 l 3 1 o 5 7 o o o 3
10 1 23 o |24 83 o o 1 0 7 s o o 2 o
10 l 24 o 93 45 1 2 0 r 5 9 3 o o 3
10 7 l O '16 31 0 l (1 0 7 '1 0 O O l
10 2 2 o 19 21 o 1 0 o o 6 o o 0 1

Notations: NRV -- Traffic of non-recreational vehicle type (C1 to C6)

RV -- Traffic of recreational vehicle type (C7 to C13).

Vehicle Classes: C1. Motorcycle, C2. Car, C3. Pickup, C4. Bus,

C5. SU2AX, C6. SU3AX, C7. SU4AX, C8. ST4AX, C9. STSAX,

C10. ST6AX, C11. DT5AX, C12. DT6AX, C13. DT7AX

(Please see Section 3.1.2 for detailed definitions of FHWA vehicle types.)

71

 

2. Data Transposing.

After removing NRV traffic data points, a matrix of 24 by 6 remains for each
day's recording. Add the 6 columns horizontally, the result is a 24 by 1
column vector. Transpose the column vector (24 by 1) into a row vector (1 by
24) which results in a set of 24 hourly traffic data points generated by

recreational type vehicles (Table 3-9).

Table 3-9. Transposing RV Traffic Data Points (4049 North, 10/1/1998).

 

 

MONTH DAY HOUR C1 C2 C3 C4 C5 C6 Sum of C1 to C6 (RV Traffic)
10 1 1 0 31 24 0 0 l 56
10 1 2 0 15 7 O O 0 22
10 1 3 0 13 10 0 0 0 23
10 l 4 0 12 10 0 3 0 25
10 1 5 0 l3 9 l 3 0 26
10 l 6 0 16 7 0 1 0 24
10 l 7 0 42 28 O 3 1 74
10 1 8 O 82 43 0 5 4 134
10 1 9 0 106 65 2 1 2 176
10 l 10 0 151 84 O 7 1 243
10 1 11 0 161 100 1 2 1 265
10 l 12 0 236 102 5 10 2 355
10 1 13 l 297 143 1 5 1 448
10 1 14 O 270 135 2 5 1 413
10 1 15 O 324 141 3 4 l 473
10 1 16 0 343 131 1 4 4 483
10 l 17 0 356 153 0 4 2 515
10 1 18 O 343 152 2 4 2 503
10 1 19 O 227 112 1 3 2 345
10 l 20 0 251 122 0 3 3 379
10 1 21 O 200 101 1 0 0 302
10 l 22 0 214 86 l 3 1 305
10 l 23 0 124 83 0 0 l 208
10 1 24 0 93 45 1 2 0 141

 

 

 

+

Transposing the column vector to a row vector

Hourly traffic volumes on October 1, 1998

1 2 3 4 5 6 7 8 9 1O 11 12 1 2 3 4 5 6 7 8 9 10 11 12
AM AM AM AM AM AM AM AM AM AM AM AM PM PM PM PM PM PM PM PM PM PM PM PM

56 22 23 25 26 24 74 134 176 243 265 355 448 413 473 483 515 503 345 379 302 305 208 141'

 

 

72

3.4.2 Step Two—Data Grouping

3. QatafSorting b1 Direction.
Afier the Step-One operation, the data format is the same as for PTR data with
24 data points on each record. For each counter station, sort the data records
by direction of traffic flow.

4. Data Sorting by Month.
For each directional record set, separate the data records into monthly groups.

5. Dat_a Sorting by Dav of the Week.
For each monthly group, separate the traffic data records into weekend and
weekday groups (see examples in Table 3-10a and 3-10b). Weekend days are
Saturdays and Sundays and weekdays are Monday through Friday. Holidays
are grouped into weekend day group. An example of data grouping is
displayed in Table 3-10. The selection of holidays is in accordance with
observed national holidays. Specifically, the selected holidays in 1998 were
New Year’s Day (January lst), January 2nd, Martin Luther King, Jr. Day
(January 19th), President’s Day (February 16th), Memorial Day (May 25th),
July 3rd, Independence Day (July 4th), Labor Day (September 7th),
Thanksgiving (November 26), November 27, and Christmas (December 25th).
A copy of a 1998 calendar is provided in Appendix C. While each of these
days is not a paid day off for all people employed, a substantial percentage of
all households' normal travel routines are altered on these days because, for

example, schools, post offices, and some businesses may be closed.

73

Table 3-10a. Example of Data Grouping—Weekend Groups (4049 North, 10/ 1998).

 

512345 89 3456789101112
{,1 AMAMAMAMAMAMAMAMAMAMAMAMPMPMPMPMPMPMPMPMPMPMPMPM

I ‘33M
10 ha

 

10/4 33 26 20 14 17 10 27 70 104 215 287 388 502 548 561 499 477 438 370 336 255 159 88 61
10/11 51 31 20 21 12 12 43 55 127 190 344 467 537 619 606 613 311 471 429 320 247 165 96 60
10/18 49 26 21 17 18 42 42 102 170 249 334 426 457 417 439 429 363 273 231 203 105 77 59
10/25 40 24 13 10 23 26 74 95 145 268 328 367 390 404 409 397 332 303 272 217 145 64 53
10/3 167 94 62 94 145 248 423 560 407 708 381 532 507 424 339 309 236 155 139 117 71
10/10 157 74 55 41 48 79 160 291 424 613 813 818 707 695 310 501 436 292 283 208 180 138 95
10/17 141 64 66 38 42 54 91 146 266 439 384 719 418 597 566 282 232 358 334 238 171 173 96 64
Ill/24 121 69 43 23 39 59 83 110 190 293 349 429 477 406 420 371 315 284 241 196 184 174 112 68
10/31 102 39 28 32 28 44 68 98 163 205 277 266 254 210 195 236 189 169 113 113 124 75 48 36

b) u—I

6.3666

b

N

\1

oo
\lstlQ‘l-HH—Id

 

Table 3-10b. Example of Data Grouping—Weekday Groups (4049 North, 10/1998).

 

123456789101112]23456789101112
AMAMAMAMAMAMAMAMAMAMAMAMPMPMPMPMPMPMPMPMPMPMPMPM

u ””M
10 ha

 

10/5 36 17 17 15 13 35 93 137 159 218 301 345 393 383 383 375 341 289 195 134 99 91 56 33
10/12 31 22 20 11 20 26 107 144 141 201 280 385 367 395 408 410 293 303 241 175 132 98 60 55
10/19 21 9 20 14 19 24 81 145 139 178 247 256 301 264 298 291 268 265 202 121 87 77 78 38
10/26 34 16 10 10 21 34 99 135 155 157 207 251 240 230 243 251 261 239 233 144 109 91 39 43
10/6 21 7 9 8 14 11 69 108 173 175 277 302 325 305 342 310 308 281 196 125 98 87 69 51
10/13 37 21 11 8 10 18 73 128 142 201 242 286 232 285 318 282 269 249 200 132 120 103 65 35
10/20 35 17 13 7 9 24 64 104 132 140 191 231 199 227 238 271 259 226 156 124 90 82 66 45
10/27 36 14 14 10 14 24 61 130 136 147 200 204 235 214 215 240 240 244 185 125 91 76 45 42
10/7 39 17 13 16 27 26 69 131 151 192 252 269 302 312 303 284 311 277 209 185 143 127 94 54
10/14 38 19 13 12 17 27 77 137 159 173 245 326 318 311 371 324 297 249 223 168 132 108 83 49
10/21 32 12 18 8 8 31 67 124 146 162 193 184 227 193 237 272 271 256 195 158 138 107 72 52
10/28 32 7 9 7 13 29 67 135 147 170 195 204 258 222 240 212 272 269 193 144 104 105 71 41
10/1 56 22 23 25 26 24 74 134 176 243 265 355 448 413 473 483 515 503 345 379 302 305 208 141
10/8 37 24 20 26 26 41 87 131 144 257 253 310 408 460 455 484 490 442 415 348 312 328 240 147
10/15 39 21 13 15 23 29 80 160 147 217 275 320 354 365 384 402 395 411 393 336 275 250 180 113
10/22 45 20 13 13 14 23 71 123 159 223 208 234 226 274 247 326 343 351 304 270 219 194 121 99
10/29 33 15 7 11 10 24 80 142 146 167 177 218 239 236 245 263 287 279 246 187 166 149 91 68
10/2 68 41 26 27 30 30 91 161 213 338 464 682 683 491 780 925 905 931 906 915 565 707 558 293
10/9 96 6O 21 31 32 36 93 152 220 304 435 604 674 530 895 910 960 975 939 885 899 741 567 332
10/16 75 45 25 35 32 44 102 164 230 289 450 588 661 586 910 919 819 912 889 626 835 683 564 317
10/23 68 19 26 15 25 28 108 124 184 215 324 383 441 493 581 689 649 633 419 600 592 491 397 213
10/30 41 25 21 10 26 26 66 154 150 207 254 319 355 350 392 454 475 511 484 437 479 301 202 112

O‘O‘OO‘O‘MMMMMbh-fiAWWWWNNNN

 

* Day of the week: 1 = Sunday, 2 = Monday, 3 = Tuesday, 4 = Wednesday,
5 =Thursday, 6 = Friday, 7 = Saturday.

74

3.4.3 Step Three-Estimating and Removing Routine Traffic

6. Estimating Routine Traffic.
After performing the prescribed data grouping procedures (according to travel
direction, month, and day of the week), each traffic data block is a matrix of n
rows by 24 columns, CW4, where n is the available number of records in each
month (either a weekday or a weekend group). The value in each cell
represents traffic volume of a given hour of the day. From the n x 24 matrix,
Cnx24, the 10th percentile value (Pj) of each hour-column is calculated. This
results in a 1 X 24 row vector, P1x24, which is an estimate of the routine traffic

flow for the traffic data group (Cum).

 

 

 

 

 

 

 

 

 

 

 

 

 

Cnx24 = C11 C12 C13 C124
C21 C22 C23 C224
C31 C32 C33 C324
1 | | |
Cnl Cn2 Cn3 Cn24
1 1 Jr 1 1

p =

“‘24 [P1 [P2 [P3 1 IP24 j

 

CU- represents traffic volume of the ith day at the jth hour, and Pj represents the
10th percentile value of the jth hour-column. Table 3-11 and Figure 3-13
display an example of estimating hourly routine traffic volume fi'om the hour-

column distributions of weekend days’ traffic data in October 1998.

75

Table 3-11.
Weekends in October 1998.

Example of Estimating Hourly Routine Traffic for 4049 North,

 

 

g12345678910111212345678910111253

Q AMAMAMAMAMAMAMAMAMAMAMAMPMPMPMPMPMI’MI’MI’MI’MPMI’MI’M’:9q
10/4 33 26 20 14 17 10 27 70 104 215 287 388 502 548 561 499 477 438 370 336 255 159 88 61 1
10/11 51 31 20 21 12 12 43 55 127 190 344 467 537 619 606 613 311 471 429 320 247 165 96 60 1
10/18 49 26 21 18 17 18 42 42 102 170 249 334 426 457 417 439 429 363 273 231 203 105 77 59 1
10/25 40 24 13 9 10 23 26 74 95 145 268 328 367 390 404 409 397 332 303 272 217 145 64 53 1
10/3 167 94 62 60 42 78 94 145 248 423 560 407 708 381 532 507 424 339 309 236 155 139 117 71 7
10/10 157 74 55 34 41 48 79 160 291 424 613 813 818 707 695 310 501 436 292 283 208 180 138 95 7
10/17 141 64 66 38 42 S4 91 146 266 439 384 719 418 597 566 282 232 358 334 238 171 173 96 64 7
10/24 121 69 43 23 39 59 83 110 190 293 349 429 477 406 420 371 315 284 241 196 184 174 112 68 7
10/31 102 39 28 32 28 44 68 98 163 205 277 266 254 210 195 236 189 169 113 113 124 75 48 36 7
5:11.31); 39 26 l9 13 12 12 27 52 101 165 264 316 344 347 362 273 223 261 215 179 149 99 61 50
Traffic

 

Note: The calculation of each hourly routine traffic volume was based on the 10th
percentile position in each hour column. This table uses the same data as Table 3-
10a. * Day of the week: 1 = Sunday, 7 = Saturday.

§§§§

Hourly Traffic Volume
(.0
O
O

7

8

0 Hourly Traffic Volume

910

11

12 13 14 15 16 17 18

Hour

+Rout1neTraffic on Weekends Oct. 1998

Figure 3-13. Routine Traffic for 4049 North, Weekends in October 1998.

76

 

19 20 2122 23 24

7. Removing Routine m.
The Removal of Routine Traffic is completed by subtracting the 10th
percentile value calculated for each hour column from the value in each cell of
the column. Let Dij represent the difference of the value in a cell from the
10th percentile value for the column. If Dij is less than zero, then zero should

be assigned to this value. Therefore, a matrix of differences, Dnm, is derived

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

as follows:
an24 =
C11-P1 C12-P2 C13-P3 C124-P24
C21-P1 C22-P2 C23-P3 C224-P24
C31—P1 C32-P2 C33-P3 C324-P24
I I l I
Cn1-P1 an-Pz Cn3-P3 an4-P24
AISO, an24 =
D11 D12 D13 D124
D21 D22 1)23 D2 24
D31 D32 D33 D3 24
l l I |
Dnl Dn2 Dn3 Dn 24

 

 

 

 

 

 

 

 

Djj 2 0 for all i = 1 to n, and j = 1 to 24. An example of this operation is
displayed in Table 3-12. Table 3-12 is derived from Table 3-11 by subtracting

the corresponding hourly routine traffic volume from cells in each column.

77

All negative values were replaced by zeros during the operation, because
traffic volume cannot be less than zero. For example, in Table 3-11, the
traffic volume at 1AM on 10/4/1998 is 33, but the routine traffic volume is 39.
A negative value would occur in the subtraction; therefore, a zero value is
used in Table 3-12 to represent the tourism traffic volume at this hour on

10/4/1998 (at Station 4049 North).

Table 3-12. Examples of Estimating Hourly Tourism Traffic for 4049 North,
Weekend Days in October 1998.

 

 

E123456789101112123456789101112§§
g AMAMAMAMAMAMAMAMAMAMAMAMPMPMPMPMPMPMPMPMPMPMPMPMazg
1014 0 0 1 1 5 0 0 18 3 50 23 72158 201199226254177155157106 60 27 11 1
10111 12 5 1 8 o o 16 3 26 25 80151193 272244340 88 210214141 98 66 35 10 1
10118 10 0 2 5 5 6 15 0 1 5 0 18 82110 55166206102 58 52 54 6 16 1
10125 1000011022004122343421361747188936846331
1013 128 68 43 47 30 66 67 93147 258296 91364 34170234201 78 94 57 6 40 56 21 7
10110 118 48 36 21 29 36 52 108 190 259 349 497 474 360 333 37 278175 77104 59 81 77 45 7
10117 102 38 47 25 3o 42 64 94165 274120403 74 250204 9 9 97119 59 22 74 35 14 7
10124 82 43 24 1o 27 47 56 58 89128 85113133 59 58 98 92 23 26 17 35 75 51 18 7
1013163139191632414662401300000000000007

 

Note: The values in this table is based the data in Table 3-11.
"' Day of the week: 1 = Sunday, 2 = Monday, 3 = Tuesday, 4 = Wednesday,
5 =Thursday, 6 = Friday, 7 = Saturday.

78

3.4.4 Estimating Tourism Traffic

By summing values in the difference matrix (e.g., Table 3-13) vertically

(Dy. = 2D,]. ) and then averaging the sum ( 5.,- = D" n ), an average of hourly tourism

i=1

traffic volume is derived. Further, the average daily tourism traffic (ADTT) is calculated

24 __
by summing the hourly averages horizontally ( ADTT 2 2D.)- ). Table 3-13 displays

j=1
this operation using the weekend data for October 1998 at 4049 North. The sum of the
column average (the last row) is the average daily tourism traffic (ADTT) which
represents the average tourism traffic volume on each weekend day in October 1998.

The above procedures are applied to both weekend and weekday groups in every
month to estimate weekend day (ADTIL) and weekday (ADTTd) tourism traffic. An
estimation of a week's tourism traffic (Tl‘w) thus is:

TTw =ADT1‘d x5 +ADTTex2
The above details the three-step operation for removing routine traffic from total

traffic with the residual traffic being the desired estimate of tourism traffic.

79

Table 3-13.

Example of Difference Matrix Derived from Removal of Routine
Traffic Operation (4049 North Weekend Days in October 1998).

 

 

 

gl23456789101112123456789101112
: AMAMAMAMAMAMAMAMAMAMAMAMPM PM PM PM PM PM PMPM PM PM PM PM
10/4 0 0 1 1 5 O 0 18 3 50 23 72 158 201 199 226 254 177 155 157 106 60 27 11
10/11 12 5 1 8 O O 16 3 26 25 80 151 193 272 244 340 88 210 214 141 98 66 35 10
10l18 10 0 2 5 5 6 15 0 1 5 0 18 82 110 55166 206102 58 52 54 6 l6 9
10/25 1 O 0 0 0 11 O 22 0 0 4 12 23 43 42 136 174 71 88 93 68 46 3 3
10/3 128 68 43 47 3O 66 67 93 147 258 296 91 364 34 170 234 201 78 94 57 6 40 56 21
10/10 118 48 36 21 29 36 52 108 190 259 349 497 474 360 333 37 278 175 77 104 59 81 77 45
10/17 102 38 47 25 30 42 64 94 165 274 120 403 74 250 204 9 9 97 119 59 22 74 35 14
10/24 82 43 24 10 27 47 56 58 89 128 85 113 133 59 58 98 92 23 26 17 35 75 51 18
10I31 63 13 9 19 16 32 41 46 62 40 13 O 0 0 0 0 0 0 O 0 0 0 0 0

Sum

£3; 58 24 18 15 16 27 35 49 76 115 108 151 166 148 145 139 144 104 92 75 50 50 34 15 1854

 

Note: The horizontal sum of column averages is the average daily tourism traffic
(ADTT). This table uses the same data as Table 3-12.

The desired estimate of tourism traffic for each day can be calculated by summing

24
the cells in each difference matrix horizontally (D,, = 2 Di]. ). The derived estimates are
j=l

the estimated tourism traffic of corresponding dates (see Table 3-14). Let us call these

values the "daily tourism estimates (DTEs)" and distinguish them from average daily

tourism traffic (ADTT). ADTT is used as a general estimate of daily tourism traffic for a

period of time (e.g., a month), which is useful in a statistical summary and comparison to

other ADTTs from different periods of time. A DTE is simply the estimated tourism

traffic on a specific day (e. g., in Table 3-14, tourism traffic volume is 1,905 on October 4,

80

1998) and is useful for creating further statistical analyses. Note that, in the same time
frame, the average of the DTEs is equal to its corresponding ADTT. For example, the

last number, 1854, in Table 3-14 is equal to the last number, 1854, in Table 3-13.

Table 3-14. Daily Tourism Estimates (DTEs) for 4049 North, Weekend Days in
October 1998.

 

1234 67891011121234567891011

and

12

5
AMAMAMAMAMAMAMAMAMAMAMAMPMPMPM PMPMPMPM PMPMPMPM PMDTE

 

0 0 18 3 50 23 72 158 201 199 226 254 177 155 157 106 60 27
0 16 3 26 25 80 151 193 272 244 340 88 210 214 141 98 66 35
1 5 0 18 82110 55 166 206102 58 52 54 6 16
11 0 22 0 0 4 12 23 43 42136174 71 88 93 68 46 3
10/3 128 68 43 47 30 66 67 93147 258 296 91364 34170 234 201 78 94 57 6 40 56
10/10 118 48 36 21 29 36 52 108 190 259 349 497 474 360 333 37 278 175 77 104 59 81 77
10l17102 38 47 25 3O 42 64 94165 274120 403 74 250 204 9 9 97119 59 22 74 35
10/24 82 43 24 10 27 47 56 58 89128 85113133 59 58 98 92 23 26 17 35 75 51
Ill/31631391916324146624013000000000000
1Avg

CN—nu—I
OM00—
001001

05

yd

U!

C

0
5
10/18 10 0
0

11 1905
10 2240
9 985
3 840
21 2691
45 3845
14 2372
13 1449
0 356

1854

Note: A Daily Tourism Estimate (DTE) is the horizontal sum of hourly tourism traffic of

a day. This table uses the same data as Table 3-12.

81

3.5 Validation of the Removal of Routine Traffic Method

3.5.1 Validation Designs

In addition to developing the Removal of Routine Traffic Method, this section
presents study designs for assessing the validity of the method in estimating tourism
traffic. From the above discussion, it is clear that the Removal of Routine Traffic Method
is designed to remove non-tourism traffic and leave residual traffic serving as an estimate
of tourism traffic. Thus, this method meets face validity (or content validity)
requirements for a valid procedure to estimate tourism traffic volume. However, the
method's construct validity must also be confirmed. Construct validation involves
determining whether or not the method measures what it is intended to measure (i.e.,
tourism traffic). Assessing construct validity involves accmnulating evidence that
measurement results are consistent with what relevant or established theory would
suggest.

Five hypotheses were developed to assess the construct validity of the proposed
method for estimating tourism traffic. The first hypothesis is to test whether there exists
any difference in the percentages of tourism traffic on weekend day and weekday traffic.
A paired two-sample T test will be used to test this hypothesis. The other hypotheses
(i.e., Hypotheses H to V) test the theoretical expectations of variables, such as region,
long weekend, direction, and percentage of recreational vehicle type traffic, on either
weekend day or weekday tourism traffic. Rather than using several T tests to examine
these hypotheses individually, these four hypotheses are combined and tested in two
regression models. Note that Regression Analysis is simply a variation of T tests. Using

Regression Analyses is more efficient to handle Null hypotheses H to V in this study.

82

The two regression models are: (1) a weekend regression model using percentage
of weekend day tourism traffic as dependent variable, and (2) a weekday regression
model using percentage of weekday tourism traffic as dependent variable. Region, long
weekend, direction, and percentage of recreational vehicle type traffic are used as
independent variables in both regression analyses to examine whether each variable
provides theoretically expected explanatory power for the dependent variables (i.e.,
percentages of weekend and weekday tourism traffic). A stepwise method is used in both
the weekend and weekday regression analyses. Independent variables are entered into or
removed fiom the regression models depending on the significance of the F values.
Criteria for entering and removing are the following: to enter a variable if its F value is
less than or equal to 0.05; to remove a variable if its F value is greater than or equal to
0.10. In order to be entered into the regression models, all the above variables must pass
the collinearity tolerance criterion of 0.0001. A variable is not entered if it would cause
the tolerance of another variable already in the model to drop below the collinearity

tolerance criterion.

3.5.2 Hypotheses

Null Hypothesis 1: There is no significant difference in the "percentages" of
tourism traffic on weekend days and weekdays.

Intuitively, weekend day tourism traffic should be proportionally greater than
weekday tourism traffic. Therefore, statistical results should lead to rejecting this
hypothesis. A paired two-sample T test will be used to test this hypothesis. The paired

samples are the percentages of weekend day and weekday tourism traffic.

83

Null Hypothesis 11: Region has no effect on the percentages of tourism traffic on
either weekend days or weekdays.

Region (REG) indicates the location of the traffic counter. A counter's region is
coded according to its location in one of the three major regions in Michigan. The
variable is coded 1 for the Upper Peninsula (UP), 2 for the Northern Lower Peninsula
(NLP), and 3 for the Southern Lower Peninsula (SLP). However, counters with 9000-
level ID numbers (i.e., those located in urban areas) are coded 4, because populations are
more concentrated in urban areas. Thus, a smaller value of this variable indicates a less
populated (rural) area, and a higher value indicates a more populated area. The
population estimates of Michigan counties in 1998 are exhibited in Appendix D.

This hypothesis is based on previous tourism infrastructure research findings that
highway development has different influences in rural and urban areas (Stephanedes and
Eagle, 1986; Humphrey and Sell, 1975; Kuehn and West, 1971). The theoretical
expectation is that there is less routine, especially commuter, traffic in rural areas; thus,
tourism traffic should be proportionally higher in less populated areas than in more
populated areas. Therefore, statistical results should lead to rejecting Null Hypothesis H
by indicating that the regression coefficient on this variable is negative and significantly
different from zero in both weekend and weekday regression models.

Null Hypothesis III: Percentages of recreational vehicle type traffic have no
effect on the percentages of tourism traffic on either weekend days or weekdays.

Percentage of recreational vehicle type (PRV) traffic is derived by dividing
recreational vehicle type traffic volume by total traffic volume. Recreational vehicle type

(RV) traffic is the base (or a superset) of tourism traffic. This variable should have

84

positive explanatory power on the dependent variables. In another words, if a route has
more non-recreational vehicle type traffic (i.e., truck traffic), then there will be less
tourism traffic. Therefore, statistical results should lead to rejecting Null Hypothesis III
by indicating that the regression coefficient on this variable is positive and significantly
different from zero in both weekend and weekday regression models.

Null Hypothesis IV: Long weekends have no significant effect on the percentages
of tourism traffic on either weekend days or weekdays.

A long weekend (LW) is a weekend with at least three non-work days. This
variable indicates whether a month contains any long weekend. If a month contains a
long weekend, the month is coded 1; otherwise the month is coded 0 thereby creating a
dummy variable for insertion in the regression models. In 1998, January, February, May,
July, September, November, and December each contained a long weekend. The
theoretical perception is that a long weekend should have a significant positive effect on
the percentage of weekend tourism traffic, but it should not have an effect on weekday
tourism traffic of that week. Therefore, statistical results should lead to rejecting the Null
Hypothesis IV on weekend regression model by indicating that the regression coefficient
on this variable is positive and significantly different from zero. However, statistical
results should lead to accepting the Null Hypothesis IV on weekday regression model by
indicating that the regression coefficient on this variable is not significantly different from
zero. Therefore, the variable, long weekend (LW), can be removed from the weekday
regression equation.

Null Hypothesis V: Direction of traffic flow has no significant effect on the

percentages of tourism traffic on either weekend days or weekdays.

85

Direction (DIR) indicates recorded traffic direction. North-and southbound are
generally perceived as the major tourism traffic flow directions in Michigan. In this
study, counters providing east and westbound traffic data are coded 0; otherwise, they are
coded 1. The theoretical expectation is that, during weekends, counters on north-
southbound highways should record higher percentages of tourism traffic than do
counters on east-westbound highways. Therefore, statistical results should lead to
rejecting Null Hypothesis V by indicating that the regression coefficient for this variable
is positive and significantly different from zero in the weekend regression model.
However, the effect of DIR on weekday tourism traffic is not as intuitive as its effect on
weekend tourism traffic. Basically, there is no clear theoretical expectation about the
direction of weekday tourism traffic. Although north-southbound rural-type tourism
traffic can take place during weekdays, it is also possible that a large amount of east-
westbound urban-type tourism traffic take place during weekdays. Thus, it is likely that

this variable is not significant in the weekday regression model.

3.6 Measurement of Data Improvement

This section provides a discussion on how to measure the degree to which the
Removal of Routine Traffic Method improves the tourism relevancy of original traffic
data. A data improvement measurement was derived to evaluate the performance of the
Removal of Routine Traffic Method.

Let’s assume that there is x% tourism traffic in total traffic and x is between 0 and
100. The tourism relevancy of traffic data can be indicated as 31%. However, the true

percentage of tourism traffic in total traffic is usually unknown; therefore, x' % is used as

86

an estimates of the percentage of tourism traffic. If we remove y% of non-tourism traffic

x )x100%. Inthis
lOO—y

 

from the total, the adjusted data relevancy to tourism becomes (

study, (100 — y)% is the proposed estimate of tourism traffic (x' %) after removing routine
and non-recreational vehicle type traffic from the original data. Thus, a measurement of
the degree to which the Removal of Routine Traffic Method improves the tourism

relevancy of traffic data can be derived as follows:

 

 

 

 

x _ x x x
Percentage of Data Improvement = 100 _: 100 = _x'_x100
106 W
= 100 - x'
x.
___ y
100 — y ’

where x is the true percentage of tourism traffic, y is the estimated non-tourism traffic,
and x' is the estimated percentage of tourism traffic and x' is equal to 100 — y. Note that
both x' and y are percentage values, their possible values are between 0 and 100.

The possible range of data improvement afier the Removal of Routine Traffic
operation is from zero percent (i.e., when all traffic is tourism and x' = 100) to infinity
(i.e., when tourism traffic is very small and x' is close to zero). In normal situations,
these extreme cases are unlikely to happen. Nonetheless, the formula suggests that when
a substantial amount of non-tourism traffic exists in untreated data, the Removal of
Routine Traffic Method can greatly improve the data’s relevancy to tourism. For

example, when the percentage of non-tourism traffic, y, is greater than 5 0%, the method

87

will improve the tourism relevancy of traffic data by more than 100% (Table 3-15).
Table 3-11 also displays that the amount of non-tourism traffic removed from the total
has an exponential relationship with the percent change of data improvement. This
exponential characteristic in data improvement strengthens the utility of the Removal of

Routine Traffic Method in tourism studies.

Table 3-15. A Chart of Data Improvement at Hypothesized Percentage Points.

 

 

Percentage of Non-Tourism Traffic Improved Data Relevancy to
Removed Tourism
10% 1 1%
20% 25%
30% 43%
40% 67%
50% 100%
60% 150%
70% 233%
80% 400%
90% 900%

 

88

Chapter 4

Results

This chapter presents statistical results derived from the highway traffic data refinement
procedures introduced in Chapter 3. It consists of four parts: (1) a data processing phase--
a set of descriptive and inferential statistics derived from the initial analyses and the
Removal of Routine Traffic operation from Michigan's highway traffic data, (2) a method
validation phase--the results of hypotheses tests using paired samples T test and
regression analyses for the verification of the Removal of Routine Traffic Method in
estimating tourism traffic, (3) an evaluation phase-«the results of measured data

improvement, and (4) some example results from daily tourism estimates (DTEs).

4.1 Results from the Data Processing Phase

4.1.1 Results from Data Preparation

Prior to removing of routine traffic from hourly CPTR data, a couple of initial
analyses were performed. The results of these statistical analyses are presented in this
section. The overall distributions of highway traffic counts generated by the 13 FHWA
vehicle types, on average, are:

' Class 1: motorcycle traffic accounts for 1.92% of total traffic,

' Class 2: passenger cars, 67.84%,

' Class 3: other two-axle, four-tire single unit vehicles (pickups), 16.11%,

I Class 4: buses, 0.23%,

' Class 5: two-axle, six-tire single unit trucks (SU2AX), 2.64%,

89

I Class 6: three-axle single unit trucks (SU3AX), 0.69%,

I Class 7: four or more axle single unit trucks (SU4AX), 0.13%,

I Class 8: four or less axle single trailer trucks (ST4AX), 1.18%,

I Class 9: five-axle single trailer trucks (STSAX), 5.19%,

I Class 10: six or more axle single trailer trucks (ST6AX), 1.09%,

I Class 11: five or less axle multi-trailer trucks (DT5AX), 0.77%,

I Class 12: six-axle multi—trailer trucks (DT6AX) 0.69%, and

I Class 13: seven or more axle multi-trailer trucks (DT7AX), 1.53%.

The distribution of traffic according to vehicle classification is also visually

displayed in Figure 4-1.

      

  
  
  
 
  

 

    
   
     
   

ST4AX ST6AX DT5AX DT6AX DT7AX mMOTOR
“8% 0.69% 1.53% CYCLE
su4Ax MOTOREYCLE ICAR
913% 1'9“ IPICKUP
SU3AX
_ , I 1 1 1 1 111111.11111‘1111111111111111!-
0.69% ' i“1,:‘1V-i11111111111111111111111111111111'111“ DBUS
111121111111111111'1111111111111111111111111111111.
SUGZAo/X . 1111111111111111111111111111111111111111111111111111111“ E SUZAX
2_ ‘ 1111111111111111111111111111111111'111111111111111111111111111111111111
':11‘11'11‘1u'11|111.11lt1‘11 11111111111111111111111111111111111111111 '5U3Ax
11111111111111111111111'1‘1111111111111111111111 11 l
BU? ~ 1111111111111111111 ““““""1111111 “111111111. 111111111 DSU4AX
0.234. 111 1 1 11111111111111111
11 11111111111111111111111111111111111111111111111111111111 1 11 1 ‘
1111111111 M1"WW"““1“"111111111111111111-11111111111111111 “11111111111111 “$1.4“
1.1111111111111113.‘111111111111111111111211111":1111““‘ 11111111111111111111111111111.1111“:1| H
1 111111111 11
11111111111111111111111111110111" 1WI‘I‘1”:|:“\“:.:|H 11111111011111“ - STSAX
111111111|1|Mm ‘33 1111111111‘111‘1“
“‘111111111111111111111é111m l1 a?'11111111111111MH .STSAX
‘ 1” 11111111111I -.
11111111111111.l 1'11" 1 111 El DT5Ax
I DT6AX
DT7AX

 

Distribution of Traffic by Vehicle Types.

90

Traffic generated by recreational type vehicles (that is, Classes 1-6) account for
89% of total traffic generated on Michigan’s highways. Passenger cars alone account for
about two-thirds of highway traffic. Motorcycles, passenger cars, and pickups account
for more than 85% of traffic. Among the 13 vehicle types, buses usually carry much
more (probably, 10 or 20 times more) passengers than the other types of vehicles.
Although buses generate only 0.23% of the total traffic, this small percentage of traffic is
highly tourism related and should not be neglected. Overall traffic generated by each
non-recreational type vehicle is relatively small. On average, Classes 7 to 13 only
account for 11% of the total traffic.

The distribution of traffic generated by recreational and non-recreational vehicle
types on each monitored route are presented in Table 4-1. Among the counters, 3069
North, 5249 North, 5299 West, 6369 East, 7029 East, 5159 West and East, 7179 West
and East, 8249 North and South, 8689 North and South, 8729 North and South, and 9369
West and East bear above average non-recreational vehicle (NRV) traffic as can be seen
in Table 4-1. It appears that major highways in the relatively populated Southern Lower
Peninsula bear much higher percentages of NRV traffic especially 7029, 7159, 7179, and
7369 all located on I-94 which links Detroit and Chicago. Therefore, one might generally
assume that the NRV traffic required to serve all areas would be proportional to their
population bases. Areas with less NRV traffic relative to total traffic are likely to be
more tourism dominant areas. This is not because they have relatively less NRV traffic
but because they have more recreational vehicle type (RV) traffic. This suggests that
highways in more populated areas are potentially less tourism dominant routes as

compared to highways in less populated areas.

91

 

 

 

 

 

Table 4-1. Traffic Distributions by Vehicle Types on Different Routes (%).
Motor— Car Pickup Buses Su2ax Su3ax Class Class
Station Location (511:1?) (Class 2) (Class 3) (Class 4) (Class 5) (Class 6) 1-6 7-13
3069N US-l3l, M-66 Kalkaska 0.42 55.40 21.21 0.05 0.58 8.55 86.2 13.8
30698 0.58 57.76 23.78 0.08 0.87 6.17 89.2 10.8
4049N 1-75 Vanderbilt 0.09 64.36 26.29 0.13 0.94 0.37 92.2 7.8
40498 0.15 68.82 22.82 0.05 0.76 0.41 93.0 7.0!
4129N US-27 Houghton Lake 0.52 65.46 25.27 0.10 0.77 0.31 92.4 7.6
41298 0.31 65.55 24.95 0.12 1.08 0.33 92.3 7.7
5029N US-27 St. Johns 0.06 73.30 20.92 0.05 0.67 0.29 95.3 4.7
50298 0.08 72.26 21.88 0.07 0.74 0.29 95.3 4.7
50398 US-27 By-Pass, St. Johns 0.15 66.70 22.79 0.19 3.67 0.31 93.8 6.2
5059NE I-l96 Hudsonville 0.00 64.94 20.29 0.25 3.04 1.44 90.0 10.0I
5059SW 0.00 71.51 14.50 0.20 2.54 1.10 89.8 10.2
5249N US-l3l Morley 0.46 55.84 19.89 0.10 8.06 0.38 84.7 15.3
52498 0.04 63.26 22.48 0.09 4.88 0.44 91.2 8.8
5299W 1-96 Ionia 0.03 68.96 14.68 0.16 1.78 0.30 85.9 14.1
5309N U8-131 Big Rapids 0.02 59.03 26.19 0.10 3.89 0.49 89.7 10.3
53098 0.07 63.73 22.92 0.13 3.50 0.39 90.7 9.3
6369B I—69 Capac 0.42 51.79 16.09 0.22 8.55 0.58 77.6 22.4
7029B I-94 Grass Lake 0.19 53.59 12.35 0.24 2.47 0.64 69.5 30.5
7109N U8-l3l Schoolcraft 3.62 71.73 14.95 0.06 0.85 0.97 92.2 7.8
71098 2.86 69.75 16.97 0.06 1.13 0.84 91.6 8.4
7159W I-94 Battle Creek 0.44 60.84 15.27 0.24 0.96 2.79 80.5 19.5
715913 0.12 59.48 13.32 0.29 1.31 0.60 75.1 24.9
7179W I-94 Coloma 0.59 61.79 13.36 0.24 0.93 0.55 77.5 22.5
717913 0.14 61.75 13.21 0.26 0.99 0.53 76.9 23.1
8219W 1-96 Howell 0.02 70.80 20.12 0.15 1.11 0.30 92.5 7.5
821913 0.02 72.58 18.91 0.12 1.04 0.23 92.9 7.1
8229N US-23 Brighton 0.17 70.37 19.39 0.11 1.07 0.46 91.6 8.4
82298 0.07 65.78 24.52 0.32 1.19 0.38 92.3 7.7
8249N I-75 Luna Pier 1.38 50.68 14.37 0.00 1.77 2.39 70.6 29.4
82498 0.80 52.79 11.62 0.47 1.76 2.18 69.6 30.4
8689N US-23 Dundee 0.41 68.60 14.67 0.10 1.34 2.19 87.3 12.7
86898 0.29 66.59 15.27 0.05 0.75 1.98 84.9 15.1
8729N US-23 Lambertville 0.83 55.14 24.85 0.92 2.11 0.42 84.3 15.7
87298 0.61 68.09 13 .42 0.07 0.77 1.62 84.6 15.4
9049N U8-127 Lansing 0.09 70.23 16.07 0.09 5.00 0.25 91.7 8.3
90498 0.04 69.46 16.94 0.14 6.13 0.44 93.1 6.9
9369W I-94 Kalamazoo 0.03 63.10 13.80 0.15 2.71 0.40 80.2 19.8
9369B 0.00 60.45 12.73 0.16 5.55 0.39 79.3 20.7
9829W 1-696, E of Southfield Rd. 0.29 87.33 6.41 0.15 0.89 0.24 95.3 4.7
9829B 0.08 80.95 13.85 0.07 1.09 0.25 96.3 3.7
9959N I-75, at Mack Ave 0.09 67.87 20.22 0.31 5.1 l 0.45 94.0 6.0I
99598 0.08 75.69 15.06 0.26 2.85 0.54 94.5 5.5
9979N I-75, at Wattles Rd. 0.02 71.64 17.44 0.12 4.90 0.33 94.4 5.6
99798 0.01 72.79 19.36 0.09 2.07 0.34 94.7 5.3
Overall Average 1.92 67.84 16.1 1 0.23 2.64 0.69 89.43 10.57

 

 

 

 

 

 

92

After coding the direction variable north-southbound highways as 1 and east-west
bound as 0, one can observe the relationship between traffic flow and vehicle type. A
Pearson correlation analysis demonstrates that the correlation coefficient between
direction (DIR) and percentage of recreational vehicle type traffic (PRV) is 0.403 (p-
value = 0.007) (Table 4-2). This correlation analysis indicates that north-southbound
highways are more likely to be associated with a higher percentage of RV traffic than
east-westbound highways. This goes along with the general perception that that north-

south bound highways are likely to be more tourism dominant routes in Michigan.

Table 4-2. Correlations between Direction (N8 vs. EW) and Percentage of RV

 

 

 

Traffic.
% of RV
Dire tion
c Traffic

Direction Pearson 1 .000 .403 *"'
Correlation

Sig. (2-tailed) . .007

N 44 44

% of RV Pearson .403" 1,000
Traffic Correlation

Sig. (2-tai1ed) .007
N 44 44

 

 

 

 

 

"Correlation coefficient is significant at the 0.01 levels.

93

After performing data filtering and transposing, the original CPTR data sets
(containing 214,319 records) were transformed into a set of data with 8,645 records (in a
format similar to PTR data sets). In Step Two of the data preparation operation (i.e., Data
Grouping), the data were grouped according to traffic flow directions and months. Due to
missing values, the data grouping operation resulted only in 315 monthly groups. A
complete data set should result in 576 monthly groups (i.e., 24 CPTR stations x 12
months x 2 directions).

After the monthly groups were further broken down into weekend and weekday
groups, 315 weekend and 315 weekday groups result. In both weekday groups and
weekend groups, there are 102 north-, 130 south-, 37 east-, 24 west-, 11 northeast-, and
11 southwest-bound traffic record sets. Data for the month of June 1998 were not
available from any of the CPRT traffic counters. Thus, there are 630 groups (i.e., 315
weekend groups and 315 weekday groups) to feed into the Removal of Routine Traffic
operation. A summary of the number of records and groups in the Data Preparation Phase

is presented in Table 4-3.

94

 

 

 

 

Table 4-3. Summary of Data Preparation.
Group Description Number of Groups

Original Counter Stations 24
Hourly CPTR data 214,319
(records)
Step One Filtered and Transposed CPTR 8,645
(records)
Step Two Monthly groups“ 315
Weekday groups" 315

Northbound 102

Southbound 130

Eastbound 37

Westbound 24

Northeast-bound 1 l

Southwest-bound 1 1
Weekend groupsM 315

Northbound 102

Southbound 130

Eastbound 37

Westbound 24

Northeast-bound 1 1

Southwest-bound 1 1
Total week groups (i.e., weekend and 630

weekday)

 

* June 1998's data are not available from all CPTR.
** No CPTR is Northwestern and Southeastern oriented.

95

A statistical summary indicates that, on average, the percentage of traffic
generated by recreational type vehicles is 89.43%. After being broken down to weekend
and weekday groups, the percentage of RV traffic is 93.37% during weekend days and
85.49% during weekdays. After being broken down to directional groups, the results
indicate that the percentages of north-southbound RV traffic are greater than the
percentages of east-westbound RV traffic during the weekend days as well as during the

weekdays (Table 4-4).

 

 

 

Table 4—4. Percentages of RV Traffic in Total Traffic.
Day Mean of Average of
of Direction N Average Daily RV traffic
Week Traffic Percentages (%)

Weekend day North 102 13714.92 93.49
South 130 15142.84 94.49
Weekday North 102 15968.09 86.31
South 130 16837.85 86.96
Weekend day East 37 27526.43 89.62
West 24 21187.50 90.39
Weekday East 37 35322.35 79.69
West 24 24466.25 80.82
Weekend day All directions 315 19392.92 93.37
Weekday All directions 3 15 23 148.63 85 .49
All Days All Direction 630 21270.7776 89.43

 

96

The statistics derived from comparing weekday RV traffic to weekend day RV
traffic indicate that, in all directions, weekday RV traffic volumes are significantly higher
than weekend day RV traffic volumes. However, the opposite results are derived when
comparing percentages of weekday RV traffic to weekend day RV traffic. The
percentages of RV traffic during the weekend day are significantly higher than during the
weekday. The statistics from a paired two-sample T test are presented in Table 4-5. For
visual convenience, the opposite results are graphically presented in Figure 4-2 and
Figure 4—3. Figure 4-2 displays that the RV traffic volumes are higher during the
weekday, and Figure 4-3 displays that the percentages of RV traffic are higher during the
weekend day.

Results presented in Table 4-5 reveal that most t statistics are significant at the
0.01 levels of significance. The only exception is the westbound pair (p-value = 0.03),
where the difference in RV traffic volumes between weekend day and weekday is only
significant at 0.05 levels. Note that the negative signs on the t statistics (i.e., the sixth
column in Table 4-5) indicate stations where the weekend days have a lower traffic
volume (or percentage) than do weekdays, and the positive t statistics indicate stations
where weekend days have a higher traffic volume (or percentage) than do weekdays.
Although the weekday RV traffic volume appears to be higher than that of the weekend
day, it is possible that the higher volume of weekday RV traffic is mainly generated by
commuters.

In Table 4-5, the statistics also indicate that the percentage of north-southbound
weekend day RV traffic is higher than that of east-westbound weekend day RV traffic.

Also, the percentage of north-southbound weekday RV traffic is higher than that of east-

97

westbound weekday RV traffic. This re-confirms that north-southbound highways are

potentially more tourism dominant than east-westbound highways.

 

 

 

 

 

 

Table 4-5. Paired Two-Sample T-Test: Weekend Day vs. Weekday.
Paired Degree 1’
Direction Paired Variables Differences Deisritaiifion $131212." t of (Sig. 2-
Mean Freedom tailed)

North RV Traffic Volume -2253.1690 5505.5034 545.1261 4.133” 101 .000
Toul'l‘rlfflc Volume -3525.1815 6119.5993 605.9306 -5.818" 101 .000

RV Traffic Percentile 7.176E-02 4.357E-02 4.3 1413-03 16.634" 101 .000

South RV Trimc Volume -l695.0085 6896.2619 604.8422 -2.802" 129 .006
TotalTnfflc Volume -3001.1816 7477.3236 655.8047 4.576” 129 .000

RV Traffic Percentage 7.530E-02 4.953E-02 4.344E-03 17.334” 129 .000

East RV Traffic Volume -7795.9189 13003.8560 2137.8208 -3.647** 36 .001
ToulTrIfflc Volume -10524.1622 13243.3145 2177.1875 -4.834"”" 36 .000

RV Truffle Perccnuse 993513-02 594113-02 9.766E-03 10.173" 36 .000

West RV Tfllmc Volume -3278.7500 6990.3542 1426.9001 -2.298"' 23 .031
Toni Traffic Volume -6088.1250 7337,4116 1497,7429 4.065" 23 .000

RV Traffic Percent.“ 9.572E-02 3.867E-02 7.893E-03 12.128" 23 .000

All RV Trams V°|Ilm¢ -2823.4709 7474.8135 421.1580 -6.704" 314 .000

Directions

Total'l'rnmc Volume -4416.8702 8051.2059 453.6341 -9.737** 314 .000

RV Trtmc Percentlse 7 .877E-02 471913-02 265913-03 29.621" 314 .000

 

"‘ Statistic is significant at 0.05 levels.
** Statistic is significant at 0.01 levels.

98

40000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
   

 

 

 

 

 

 

 

35000
30000
25000
E
3 20000
‘>’
15000
10000
5000
0
North South East West All Directions
Figure 4-2. Volumes of RV Traffic: Weekend Day vs. Weekday.
100%
95% T
90% -~—- __
5 ”‘7
g 85% ~-— . I “ll“"lllm
n- M I ‘ ‘ . V
.1 , “(WW
o s__ — M " __ , .
80 A .. . . 1.. lmmljfl j’ Wu... lithium“
‘ V ‘ JF'W “‘h‘mm
75% r— 1‘ ‘ — W— 1.35"“ W
0‘. ..w ‘\:.“‘l“:““ I h ,l Infllllllu‘
70% l . I z.‘ I H. w.‘ I w I (10 .
North South East West A I' actor
Direction
Figure 4-3. Percentages of RV Traffic: Weekend Day vs. Weekday.

99

4.1.2 Results from the Removal of Routine Traffic Operation

The 10th percentile traffic flow, as noted earlier, was used to estimate routine
traffic on each route. After using the Removal of Routine Traffic Method on each
monthly weekend and weekday group, the estimated tourism traffic volumes were
derived. Examples of using actual data in the processing procedures have already been
provided in the fourth section of Chapter 3. The estimated average daily tourism traffic
(ADTT) volumes are summarized in Table 4-6. The estimates indicate that the volumes
of tourism traffic are generally higher in more populated areas than in less populated
areas. This may be due partially to the higher volumes of total traffic in populated areas
as well as the definition of tourism (including both pleasure and business day trips and
ovemight trips) in this study. As ADTT is a summarized statistic, its major use is for the
derivation of percentage of daily tourism traffic (PDTT). As mentioned in Chapter 3,
another useful gauge for tourism traffic is the daily tourism estimate (DTE). Note that a
DTE, which is an estimate of tourism traffic for a specific day, is different from ADTT.
The number of DTEs created by the Removal of Routine Traffic Method are as many as
the available records which are fed into this data refinement procedure. For example,
there were 8,645 records (after the Data Filtering and Transposing Phase) fed into the
removal process; thus, 8,645 DTEs were created. Due to the large amount of DTEs
created, the author will demonstrate their use only on a few selected stations at the end of

this chapter.

100

 

 

Table 4-6. Estimated Average Daily Tourism Traffic Volumes (ADTT).
Day of
Station Location Week Jan Feb Mar Apr May Jul Aug Sep Oct Nov Dec
3069N US-131, M-66Ka1kaska weekend 863 620 566 635 792 673 812 795 848 664 646
weekday 648 622 756 485 727 693 954 715 807 676 742
30698 weekend 975 7 1 6 847 1002 1083 1 100 1349 742 707
weekday 1036 526 464 463 728 857 678 524 698
4049N [-75 Vanderbilt weekend 2131 2259 3314 3061 1854 1496 1693
weekday 2361 3826 2917 2858 2168 1602 1336
40498 weekend 3706 5060 3692 5664 4695 3 172 1 610
weekday 1340 2639 2343 1704 1931 1718 1168
4129N US-27 Houghton Lake weekend 1302 1605 1156
weekday 1700 1429 1057
41298 weekend 3085 2087 1 1 14
weekday 822 981 801
5029N US-27 St. Johns weekend 3175 6636 2961 4788 2463
weekday 3906 5547 3333 4488 3347
50298 weekend 5873 6776 5908 3573 2418
weekday 1 843 1 905 2045 2369 277 3
5039s US-27 By-Pass, St. Johns weekend 2138 2925 2490 2677 3574 4784 4967 4269 3570 2307
weekday 970 1070 1218 1 163 1277 1674 1971 2092 2285 2929
5059NE 1—196 Hudsonville weekend 3063 3078 2499 2867 3142 3289 2897 2535 2727 2947 2508
weekday 2695 31 17 3783 2758 3451 3461 3994 3626 2893 2568 4089
5059SW weekend 3308 2644 2621 2904 3400 3282 3623 2847 31 17 3158 3076
weekday 2999 2690 4103 2993 2824 2544 2468 241 1 2982 3241 4301
5249N US-131 Morley weekend 1752 1850 1474 2416 2655 3183 2945 2634 3605 2348
weekday 1866 1872 1701 2801 2387 3436 31 13 3254 2426 1802
52498 weekend 1752 1850 1474 2416 2655 3183 2945 2634 3605 2348
weekday 1866 1872 1701 2801 23 87 3436 31 13 3254 2426 1802
5299W 1-96 Ionia weekend 3040 4075 3567 3922 4112 4420 3399
weekday 3401 3824 4894 3642 3148 3250 3431
5309N U8-131 Big Rapids weekend 3162 2203 2366 1240
weekday 2933 2118 1713 1184
53098 weekend 2405 2291 2179 2746 4267 4578 3664 4853 4217 3251 1741
weekday 931 849 950 905 1 129 2260 2153 1 130 1066 1256 1088
6369!: 1-69 Capac weekend 982 928 1105 1198 1089 1532 2333
weekday 855 658 829 765 872 1008 1415
7029!: 1-94 Grass Lake weekend 5151 5126 3233
weekday 3538 4208 3420
7109N US-131 Schoolcraft weekend 1606 1598 1663 2241 1909 1647 2000
weekday 1424 1363 1943 1792 2020 1582 2785
71098 weekend 1567 1653 2244 1583 1898 21 16 1692 2052
weekday 1300 1897 2035 1798 1656 1464 1 180 1902
7159W 1-94 Battle Creek weekend 4589 2629
weekday 4806 3 1 93
715915 weekend 33 18 4922 4203 5562 3509
weekday 4286 3509 3874 3904 2587
7179“! 1-94 Coloma weekend 2085 3141 4078 2981
weekend 2364 3174 3491 3305
7179B weekend 1956 3379 4649 3537
weekday 2410 3 167 2720 2788

 

101

 

 

Table 4-6. (cont’d)
. Day of
Station Location Week Jan Feb Mar Apr May Jul Aug Sep Oct Nov Dec
8219W 1-96 Howell weekend 7774 4717
weekday 6980 743 1
82198 weekend 5915 4482
weekday 6100 5884
8229N US-23 Brighton weekend 3796 6706 5301 5899 3928
weekday 4726 5591 5392 5451 4482
82298 weekend 5138 7706 7249 7259 3163
weekday 4212 3496 4531 3679 2915
8249N 1-75 Luna Pier. weekend 3223 3916 3291 3304 2521 4403
weekday 2460 3149 3436 2549 3330 3012
82498 weekend 2536 2858 2517 2980 3149 2499 2713 2526 3178 3511 3886
weekday 1626 2156 1831 2413 2114 3160 3327 1873 1667 2146 2841
8689N U8-23 Dundee weekend 2495 2353 2587 2612 3458 4579 4803 3884 3615 4052 2583
weekday 2099 2395 2441 2463 2758 3475 3856 2897 2773 2609 2514
86898 weekend 2371 1932 2565 2462 2309 4578 4534 3810 3170 3167 2884
weekday 1978 2015 2541 2825 2795 3167 2964 2389 2934 2849 2916
8729N US-23 Lambertville weekend 3626 2371 3603 3732 3821 3954 4644 3455
weekday 2254 2381 3379 2624 3055 4399 5202 3159
87298 weekend 3817 3347 2871 4819 4312 3579 3221 3935 3568
weekday 2613 3344 2678 3247 2607 3029 2674 3052 3172
9049N U8-127 Lansing weekend 4431 3080 2943 3198 4197 4395 3962 4450 4839 4795 4644
weekday 3566 3431 3591 4831 4247 3681 4815 4819 5236 5532 6205
90498 weekend 4332 3002 2770 2853 3132 4569 4302 6177 5901 5655 5096
weekday 3682 3887 4382 3640 4256 3836 3756 4072 3505 3944 7008
9369W [-94 Kalamazoo weekend 6538 4673 4872 6494 5221 5097 6090
weekday 3872 4810 5489 7899 4641 5476 5925
93691-3 weekend 6825 5156 4670 6183 4430 6057
weekday 5134 4929 5895 7530 4086 4574
9829W 1-696, E of Southfield weekend 18807 13962
weekday 16961 1 521 3
9829!: weekend 14662 14860 14895 15100 17065 1 1979 14810 14761 13845 14124
weekday 15887 14274 15973 13926 13255 16758 16334 14692 15429 20380
9959N 1-75, at Mack Ave. weekend 13626 7921 8656 9114 9586 9447 8339 9628
weekday 12492 7352 7875 8520 7687 8312 8557 10995
99598 weekend 13488 9652 9028 9138 12073 1 1559 6975 9347 8519
weekday 9872 9879 12189 10685 9368 10329 7764 9799 12062
9979N 1.75, at Wattles Rd. weekend 1 1802 8687 7879 8620 10147 10687
weekday 12169 1 1804 8239 1 1382 10559 17208
99798 weekend 1 1583 9839 7366 8089 7680 9216 7106 7123 7464 10130 1 1 145
weekday 10274 9799 10160 13640 10669 9671 8216 9871 9295 10306 16750

 

102

As tourism dominance is defined by the percentage of tourism traffic on a route,
the percentages of daily tourism traffic (PDTT) are summarized and displayed in Table 4-
7. The percentage of daily tourism traffic is calculated by dividing the average daily

tourism traffic (ADTT) by the average daily traffic (ADT) of a counter: PDDT = Ari—11:.

For instance, the average daily traffic (ADT) on a January weekend day around
Station 3069 North was 2,616 vehicles (derived along with the Removal of Routine
Traffic Method) and the estimated average daily tourism traffic (ADTT) is 863 vehicles
(i.e., the first traffic volume data point displayed in Table 4-6). Therefore, the percentage
of daily tourism traffic (PDDT) is 33% (i.e., the first percentage data point displayed in
Table 4-7). Estimates in Table 4-7 indicate that the percentages of tourism traffic were
higher during weekend days than during weekdays for most routes and months in 1998.

In September 1998, Station 5029 (on US-27) northbound weekend day tourism
traffic accounted for more than 5 0% of the total traffic; therefore, this highway segment
can be considered to be "absolutely" tourism dominant. Generally, the observations of
tourism dominance in Table 4-7 are consistent with the fact that north-southbound
highways are the major corridors accessing most visited destinations and natural
recreation areas in northern Michigan. For example, Stations 4049 (on 1-75), 4129, 5029,
5039 (on US-27), 5309 (on US-131), and 8729 (on U8-23) frequently registered higher
percentages of tourism traffic and displayed "relative" tourism dominance as compared to
other routes. Also, highways in more populated areas (e. g., stations with II) number
greater than 7000) during the weekday are generally less tourism dominant. Applications

of tourism dominance will be discussed in Chapter 5.

103

 

 

Table 4-7. Percentages of Tourism Traffic.
8 . . Day of
tatron Location Week Jan Feb Mar Apr May Jul Aug Sep Oct Nov Dec
3069N US-131, M-66 Kalkaska weekend 33% 21% 21% 21% 24% 17% 20% 21% 23% 23% 23%
weekday 20% 18% 24% 13% 18% 15% 21% 17% 19% 18% 21 %
30698 weekend 35% 23% 23% 22% 24% 25% 33% 24% 25%
weekday 35% 15% 12% l 1% 17% 22% 17% 14% 20%
4049N 1-75 Vanderbilt weekend 38% 23% 33% 39% 33% 40% 47%
weekday 36% 37% 32% 39% 38% 40% 39%
40498 weekend 43% 35% 24% 46% 48% 45% 39%
weekday 23% 28% 24% 24% 31% 32% 28%
4129N US-27 Houghton Lake weekend 27% 36% 28%
weekday 37% 35% 30%
41298 weekend 45% 39% 30%
weekday 21% 27% 25%
5029N US-27 St. Johns weekend 25% 57% 20% 44% 23%
weekday 31% 46% 27% 43% 32%
50298 weekend 35% 39% 36% 28% 23%
weekday 1 5% 16% 1 8% 22% 28%
50398 US-27 By—Pass, St. Johns weekend 33% 40% 35% 35% 38% 48% 35% 31% 33% 25%
weekday 19% 19% 21% 18% 19% 20% 19% 20% 24% 34%
5059NE 1-196 Hudsonville weekend 31% 28% 23% 24% 24% 23% 21% 20% 22% 24% 22%
weekday 1 7% 1 9% 24% 1 6% 19% 2 1% 23% 20% 16% 15% 25%
5059SW weekend 32% 22% 24% 24% 27% 23% 26% 22% 25% 26% 26%
weekday 19% 16% 26% 17% l 5% 14% 14% 14% 17% 18% 25%
5249N US-131 Morley weekend 29% 29% 22% 29% 28% 24% 27% 26% 39% 31%
weekday 28% 27% 25% 35% 25% 30% 30% 35% 28% 23%
52498 weekend 29% 29% 22% 29% 28% 24% 27% 26% 39% 31%
weekday 28% 27% 25% 35% 25% 30% 30% 35% 28% 23%
5299W 1-96 Ionia weekend 25% 30% 24% 22% 27% 27% 19%
weekday 22% 23% 29% 1 8% 16% 16% 1 7%
5309N 03-131 Big Rapids weekend 39% 31% 36% 22%
weekday 34% 29% 27% 21%
53098 weekend 36% 32% 30% 33% 42% 35% 29% 44% 45% 41% 30%
weekday 18% 15% 15% 12% 14% 23% 23% 17% 17% 21% 21%
6369!: 1-69 Capac weekend 22% 18% 22% 21% 17% 20% 32%
weekday 15% 1 1% 14% 11% 12% 13% 20%
7029!: 1-94 Grass Lake weekend 27% 26% 24%
weekday 1 8% 20% 20%
7109N U8-13l Schoolcraft weekend 20% 20% 23% 30% 25% 24% 30%
weekday 15% 14% 21% 19% 22% 19% 36%
71098 weekend 20% 2 1 % 29% 20% 24% 26% 23% 30%
weekday 14% 20% 22% 19% 17% 15% 1 3% 22%
7159W 1-94 Battle Creek weekend 24% 17%
weekday 21% 18%
7159!: weekend 17% 25% 23% 29% 29%
weekday 1 8% 1 8% 20% 19% 17%
7179“! 1-94 Coloma weekend 14% 22% 29% 24%
weekend 1 5% 20% 23% 24%
7179!: weekend 12% 22% 29% 27%
weekday 15% 19% 17% 18%

 

104

 

 

Table 4-7 (cont’d)
Day of
Station Location Week Jan Feb Mar Apr May Jul Aug Sep Oct Nov Dec
8219W 1-96 Howell weekend 27% 18%
weekday 22% 23%
8219!: weekend 20% 17%
weekday 18% 18%
8229N US-23 Brighton weekend 27% 30% 25% 31% 24%
weekday 25% 22% 22% 24% 21%
82298 weekend 29% 3 1% 30% 33% 1 7%
weekday 23% 14% 18% 15% 13%
8249N 1-75 Luna Pier. weekend 26% 32% 24% 25% 17% 29%
weekday 16% 20% 22% 1 8% 20% 1 8%
82498 weekend 21% 21% 19% 21% 22% 15% 16% 16% 22% 24% 26%
weekday 10% 13% 1 1% 14% 12% 18% 18% 1 1% 10% 12% 16%
8689N US-23 Dundee weekend 29% 28% 26% 22% 26% 27% 28% 25% 29% 32% 24%
weekday 23% 28% 25% 23% 21% 21% 25% 21% 23% 22% 23%
86898 weekend 27% 20% 27% 27% 20% 29% 27% 28% 25% 28% 27%
weekday 20% 19% 23% 25% 21% 20% 19% 17% 23% 24% 27%
8729N US-23 Lambertville weekend 45% 22% 34% 25% 25% 24% 25% 26%
weekday 23% 22% 29% 18% 18% 25% 28% 23%
87298 weekend 54% 32% 23% 27% 23% 21% 21 % 28% 27%
weekday 32% 26% 19% 19% 1 6% 18% 16% 1 9% 2 1%
9049N US-127 Lansing weekend 30% 19% 19% 19% 25% 27% 22% 23% 24% 25% 27%
weekday 17% 16% 17% 21% 17% 16% 20% 18% 19% 21% 25%
90498 weekend 28% 18% 17% 17% 18% 27% 23% 29% 26% 29% 29%
weekday 17% 18% 21% 17% 18% 17% 16% 15% 13% 15% 29%
9369W 1-94 Kalamazoo weekend 26% 18% 19% 22% 19% 18% 23%
weekday 12% 15% 17% 22% 13% 15% 18%
9369!: weekend 30% 20% 1 9% 22% 17% 22%
weekday 17% 16% 19% 21% 12% 13%
9828W 1-696, E of Southfield weekend 31% 26%
weekday 19% 18%
982915 weekend 23% 24% 23% 25% 27% 18% 23% 22% 23% 23%
weekday 17% 15% 17% 14% 14% 17% 17% 15% 16% 23%
9959N 1-75, at Mack Ave. weekend 33% 19% 21% 22% 24% 23% 22% 25%
weekday 23% 14% 14% 16% 14% 16% 1 8% 20%
99598 weekend 32% 22% 21% 20% 28% 27% 19% 24% 23%
weekday 17% 17% 21% 18% 16% 19% 16% 18% 22%
9979N 1-75, at Wattles Rd. weekend 36% 23% 20% 26% 24% 27%
weekday 22% 20% 14% 20% 1 7% 28%
99798 weekend 33% 26% 22% 22% 20% 25% 17% 17% 19% 26% 30%
weekday 18% 17% 18% 24% 17% 15% 13% 16% 15% 17% 29%

 

105

A summary of estimated tourism traffic volumes (i.e., ADTTs in Table 4-6) and
percentages (i.e., PDTTs in Table 4-7) is displayed in Table 4-8. On average, the daily
tourism traffic volume was 4,350 during the weekend day and 4,108 during the weekday.
Among the monthly groups, the percentages of estimated weekend day tourism traffic
range from 12.12% to 56.86%. On average, weekend day tourism traffic is 26.50% of the
day's traffic. The percentages of estimated weekday tourism traffic range from 9.92% to
46.47%. On average, weekday tourism traffic is 20.56% of the day's traffic (Table 4-8).
A paired T -test of the weekend day versus weekday comparisons indicates that the
difference in percentages of tourism traffic between weekend days and weekdays is
significant at 0.01 levels. The difference in daily tourism traffic volumes is also
significant at 0.01 levels (Table 4-8). These results confirm the theoretical expectation

that weekend day has more tourism traffic (in volume and percent) than weekday.

Table 4-8. A Paired Two-Sample T-Test on Tourism Traffic: Weekend Day vs.

 

 

Weekday.
Tourism Day of . . . Mean
Traffic Week N Minimum Maxrmum Mean Difference

Volume

Weekend day 315 4349.7462

Weekday 315 4108.0989 241.6474“
Percent

Weekend 315 12.12% 56.86% 26.50%

Weekdayday 315 9.93% 46.47% 20.56% 5.93%"

 

** Statistic is significant at 0.01 levels.

106

 

In the previous section, statistical results (Table 4-4) show that there is 93.37%
RV traffic in weekend day traffic and 85.49% RV traffic in weekday traffic. Therefore, a
simple calculation reveals that the Removal of Routine Traffic Method has removed
66.87% of routine traffic from weekend day traffic and 64.93% of routine traffic from
weekday traffic (see Table 4-9). To date, the Federal Highway Administration has
conducted two Nationwide Personal Transportation Surveys (telephone surveys). Survey
results indicate that 24% and 18% of highway traffic was tourism traffic in 1990 and
1995, respectively (FHWA, 1997). Therefore, the estimates derived by the Removal of
Routine Traffic Method are comparable to the Federal Highway Administration’s

estimates.

Table 4-9. Summary of Traffic Components after the Removal of Routine

 

 

 

Traffic Operation.
RV Traffic
NRV Traffic Iii-23in: T3332" Total
Weekend Day 6.63% 66.87% 26.50% 100%
Weekday 14.51% 64.93% 20.56% 100%

 

Notations: NRV—non-recreational vehicle type, RV--recreati0na1 vehicle type.

The percentages of weekend day and weekday tourism traffic by direction are
summarized in Table 4-10. On average, the percentage of northbound tourism traffic was
27.12% of the total traffic on weekend days and 23.73% on weekdays. The percentage of

southbound tourism traffic was 28.09% of the total traffic on weekend days and 19.78%

107

on weekdays. On average, the percentage of eastbound tourism traffic was 22.67% of the
total traffic on weekend days and 16.66% on weekdays. The percentage of westbound

tourism traffic was 22.93% of the total traffic on weekend days and 19.00% on weekdays.

Table 4-10. Summary of Tourism Traffic Percentages by Weekend Day,

 

 

Weekday, and Direction
Day of Week Direction N Mean (%) Std. Deviation
Weekend day North 102 27.12 6.62E-03
South 130 28.09 6.85E-03
Weekday North 102 23.73 7.22E-03
South 130 19.78 4.69E-03
Weekend day East 37 22.67 4.38E-02
West 24 22.93 4.59E-02
Weekday East 37 16.66 2.89E-02
West 24 19.00 4.02E—02

 

The results from a correlation analysis indicate that weekend day and weekday
tourism traffic volumes are significantly correlated with each other with a correlation
coefficient equal to 0.981 and p-value < 0.001 (Table 4-11). Another correlation analysis
indicates that the percentages of weekend day and weekday tourism traffic are
significantly correlated with a correlation coefficient equal to 0.478 and p-value < 0.001
(Table 4-12). The significant correlation between weekend day and weekday tourism

traffic may imply that the variations in these two variables are simultaneously affected by

108

a third variable, such as the same seasonality or geographical (spatial) factors. Therefore,
when a month is in a tourism season, then both weekend day and weekday traffic appear
to have higher tourism traffic volumes and percentages. Also, when a route is an access
to tourist destinations, then both weekend day and weekday traffic appear to have higher
tourism traffic volumes and percentages. In other words, the statistical results from these
correlation analyses indicate that the seasonal and spatial variation natures of tourism are

preserved in the estimated tourism traffic with the Removal Routine Traffic Method.

Table 4-1 1. Correlation of Weekend Day and Weekday Tourism Traffic Volumes.

 

 

 

 

 

 

 

Tourism
Traffic Weekday Weekend day
Volume
Weekday 1.00 .918""'I
(.000)
Weekend day .918“ 1.00
(.000)

 

Table 4-12. Correlation of Weekend Day and Weekday Tourism Traffic

 

 

 

Percentages.
Percentage of
Tourism Traffic Weekday Weekend day
Weekday 1 .00 .478"
(.000)
Weekend day .478M 1.00
(.000)

 

 

 

 

 

** Correlation is significant at the 0.01 levels (2-tailed). Numbers in the parentheses are
the p-value.

109

4.2 Results from the Method Validation Phase

This section presents test statistics from hypotheses testing which utilized a paired
two-sample T test and two regression analyses. These tests were conducted to assess
construct validity of the Removal of Routine Traffic Method in estimating tourism traffic.

The results of the test statistics suggest that the Removal of the Routine Traffic Method

provides construct validity in estimating tourism traffic.

4.2.1 Results from Paired Samples T Test on Null Hypothesis 1

The general perception is that weekend day tourism traffic should be
proportionally greater than weekday tourism traffic. Ifthe Removal of Routine Traffic
Method is effective in removing routine traffic and leaving the residual traffic as the
closest estimate of tourism traffic, then the statistical results from estimated tourism
traffic should lead to the rejection of the Null Hypothesis I that there is no significant
difference in percentages of tourism traffic between weekend days and weekdays.

In Table 4-13, the test statistics fi'om a paired two-sample T test indicate that
significant differences exist between the percentages of weekend day and weekday
tourism traffic. The difference is so significant that the p-value of the t statistic is less
than 0.001. Thus, the test statistic leads to rejecting Null Hypothesis 1. The positive sign
on the 1 statistic also indicates that percentages of tourism traffic on weekend days are
higher than that on weekdays (Table 4-13). Statistics in Table 4-8 indicate that the
Percentages of weekend day tourism traffic on the average are 6 percentage points higher
than the percentages of weekday tourism traffic. These results support the validity of the

Rev'nloval of Routine Traffic Method by showing that the estimated results from the

110

method do not deviate from the general perception of most people (criterion-related

validity).

Table 4-13. Paired Samples T Test.

 

 

 

 

. . P
Paired Differences t D.F (Sig. 2-tail ed)
Mean Std. Std. Error 95% Confidence
Deviation Mean Interval of the
Difference
Lower Upper
5.93E-02 6.75E-02 3.80E-03 5.18E-02 6.68E-02 15.606” 314 .000

 

 

 

 

 

 

 

 

 

 

** Statistic is significant at 0.01 levels.
D.F. = Degree of Freedom

4.2.2 Results from Regression Analyses on Hypothesis II to V

Hypotheses H to V were collectively tested in two regression analyses. In other
words, two regression analyses were used to test the theoretical expectations associated
with Hypotheses II to V. In the regression models, percentages of weekend day and
weekday tourism traffic (PWETT and PWDTT) were used as the dependent variables
and were regressed on region (REG), percentage of traffic generated by recreational type
vehicles (PRV), long weekend (LW), and direction (DIR). The author would like to
indicate that these regression analyses were not used to identify the best prediction model
for tourism traffic, rather, they were designed to assess the construct validity of the

Removal of Routine Traffic Method.

111

Regression Anglvsis I--WLd<end Tourism Traffic Stepwise Regression Models

Using stepwise method (i.e., variables are entered into the model one by one with
the most significant variable entered first), the analysis of weekend tourism traffic
suggests the following four models:

Dependent Variable: Percentage of weekend day tourism traffic (PWETT)

Model l-- PWETT: (Constant), REG

Model 2-- PWETT: (Constant), REG, PRV

Model 3-- PWETT: (Constant), REG, PRV, LW

Model 4-- PWETT: (Constant), REG, PRV, LW, and DIR

All of the models are significant at 0.01 levels (Table 4-l4b). The final stepwise
regression model (with R square equal to 0.194) explains 19.4% of the variations in
PWETT (Table 4-l4a). The statistical results indicate that region (REG), percentage of
traffic generated by recreational type vehicles (PRV), long weekend (LW), and direction
(DIR) have significant explanatory power for the percentage of weekend day tourism
traffic (Table 4-14b and c). All the p-values of these variables are smaller than 0.01
expect the value on (DIR) (Table 4-14c). However, DIR is significant at the 0.05 levels.
Note that REG (i.e., location) was the most significant and the first variable entered into
the model. PRV was the second most significant variable with the largest unstandardized
coefficient (0.257 in Model 4 Table 4-14c). LW was the third most significant, and DIR
was the least significant. After standardization, REG has the largest standardized

coefficient (-5.529 in Model 4 Table 4-140).

112

Table 4-14a. Summary of Weekend Tourism Traffic Regression Models.

 

 

 

 

 

Model R R Square Adjusted R Std. Error of the
Square Estimate
1 .307 .094 .092 6.596E-02
2 .399 .159 .154 6.366E-02
3 .424 .180 .172 6.297E-02
4 .440 .194 .183 6.254E-02
Table 4-14b. AN OVA of Weekend Regression Models.
Sum of Mean p
Model Squares df Square (Sig.)
1 Regression .142 1 .142 32.664" .000
Residual 1.362 313 4.350E-03
Total 1.504 314
2 Regression .239 2 .120 29.529M .000
Residual 1 .264 312 4.053E-03
Total 1.504 3 14
3 Regression .270 3 9.015E-02 22.732“ .000
Residual l .233 3 l l 3.966E-03
Total 1.504 314
4 Regression .291 4 7 .285E-02 18.627" .000
Residual 1.212 310 3.991 E-03
Total 1.504 314

 

** Statistic is significant at 0.01 levels.

113

Table 4-14c. Coefficients of Weekend Tourism Traffic Regression Models.

 

 

 

 

 

. Unstandardized Std. Standardized p
Model Variables Coefficients Error Coefficients (Sig.)
(B) (Beta)
1 (Constant) .363 .018 20.713“ .000
REG -3.190E-02 .006 -.307 -5.715** .000
2 (Constant) 7 .794E-02 .061 1 .287 .199
REG -3.512E-02 .005 -.338 -6.471** .000
PRV .316 .064 .256 4.898M .000
3 (Constant) 6.770E-02 .060 1.128 .260
REG -3.494E-02 .005 -.337 -6.506** .000
PRV .311 .064 .252 4.878" .000
LW 2.104E-02 .008 .144 2.801 ** .005
4 (Constant) 8.727E-02 .060 1.449 .148
REG -3.097E-02 .006 -.298 -5 .529" .000
PRV .257 .068 .209 3.806" .000
LW 2.174E-02 .007 .149 2.912" .004
DIR 2.274E-02 .010 .130 2314* .021

 

"' Statistic is significant at 0.05 levels.
** Statistic is significant at 0.01 levels.

114

The results fiom weekend regression analysis agree with theoretical expectations
about these variables. That is, the effects of REG, PRV, LW, and DIR on PWETT are
significantly negative, positive, positive, and positive, respectively, which lead to the
rejection of the Null Hypotheses II to V on weekend tourism traffic model. Therefore,
this study confirms that:

(1) Tourism traffic is proportionally higher in less populated areas than in more
populated areas during weekend.

(2) Routes with smaller percentages of non-recreational vehicle type traffic have
higher percentage of tourism traffic during weekend.

(3) A long weekend has significant positive effects on the percentage of tourism
traffic on that weekend.

(4) North-south oriented highways in Michigan have higher percentages of

tourism traffic during weekend than do east-west oriented highways.

Regession Analfiis II--Weekday Tourism Traffic Stepwise Regression Models

Using a stepwise method, the analysis on weekday tourism traffic suggests the
following two models:

Dependent Vania—He: Percentage of weekday tourism traffic (PWDTT)

Model 1-- PWDTT: (Constant), REG

Model 2-- PWDTT: (Constant), REG and PRV

The weekday regression models examined provide similar results to those

reported for weekend models, except that the regression terminated on the second step.

115

 

The final stepwise regression model has R-square equal to 0.258 (Table 4-15a). As
predicted in Chapter 3, the indicator of the long weekend (LW) was dropped out of the
weekday's regression model, because it had no explanatory power on weekday tourism
traffic (Table 4-15a, b and c). DIR was also dropped out of the models. Again REG was
the most significant and the first variable entered into the model, and PRV was entered

next though it has the largest unstandardized coefficient (in Model 2 Table 4-15c).

Table 4-15a. Summary of Weekday Tourism Traffic Regression Models.

 

Adjusted R Std. Error of the

 

Model R R Square Square Estimate
1 .434 .188 .186 5.638E-02
2 .508 .258 .253 5.401E-02

 

Table 4-15b. AN OVA of Weekday Regression Models.

 

 

Sum of p
Model Squares df Mean Square F (Sig.)
1 Regression .231 l .231 72.689 .000
Residual .995 3 13 3. 179E-03
Total 1.226 314
2 Regression .316 2 1.58 54.144 .000
Residual .910 312 2.917E-03
Total 1 .226 3 14

 

116

Table 4-15c. Coefficients of Weekday Tourism Traffic Regression Models.

 

 

 

. Unstandardized Std. Standardized p
Model Variables Coefficients Error Coefficients (Sig.)
(B) (Beta)
1 (Constant) .330 .012 22.064 .000
REG -4.068E-02 .002 -.434 -8.526 .000
2 (Constant) .181 .015 5.777 .000
REG -4.472E-03 .002 -.477 -9.654 .000
PVR .190 .008 .367 5.392 .000

 

The results from weekday regression analysis also agree with theoretical
expectations about these variables. That is, the effects of REG and PRV on PWDTT are
significantly negative and positive, respectively, but the effects of LW and DR on
PWDTT are not significant (or show no effect). The statistical results lead to the
rejection the Null Hypotheses H and III and to the acceptance of the Null hypotheses IV
and V on weekday tourism traffic regression analysis. Therefore, this study confirms the
following:

(1) Tourism traffic is also proportionally higher in less populated areas than in
more populated areas during weekdays.

(2) It is also true that routes with smaller percentages of non-recreational vehicle
type traffic have higher percentages of tourism traffic during weekdays.

(3) A long weekend has no effect on the percentages of tourism traffic on

weekdays.

117

 

(4) There is no clear pattern of tourism traffic direction (i.e., north-south versus.
eats-west) during weekdays.

Overall, the results fi'om hypotheses tests correspond to the opinions of tourism
experts as well as to theoretical expectations. The inferential statistical results
demonstrate that the tourism traffic estimates derived fi'om the Removal of Routine

Traffic Method provide construct validity.

4.3 Results from Data Improvement Measurement

The evaluation of to what degree the Removal of Routine Traffic Method
improves the relevancy of highway traffic data to tourism is based on the estimated
percentages of daily tourism traffic (PDTT) for each of the counter stations in Table 4-7.

Using the formula derived in Chapter 3, percentage of data improvement = 100 .— x ,
x

 

where x' is the estimated percentage of tourism traffic.

As previously described, the amount of non-tourism traffic removed from total
traffic has an exponential relationship with the percentage change of data relevancy to
tourism. Based on non-recreational vehicle type traffic information in Table 4-1, even

simply removing NRV traffic would improve overall traffic data relevancy to tourism by
12% (i.e., 10.57 + (100 - 10.57) = 0.12). The measurement results indicate that, on
average, the application of the Removal of Routine Traffic Method on Michigan highway
traffic data improved the overall data relevancy to tourism by 364% (Table 4-16). The

improvements on weekend and weekday data relevancy to tourism are 301% and 427%,

respectively. The reason that weekday traffic estimates exhibit more improvement is due

118

 

to the fact that there is more non-tourism traffic in weekday data. Estimated

improvements for each counter station are presented in Appendix E.

Table 4-16. Improvement Measures Using Methods Developed in This Study.

 

Percentage Change in Data

 

Operational Phase Relevancy to Tourism
Only NRV Traffic Removed 12%
Weekend day--with NRV & routine traffic removed 301%
Weekday-- with NRV & routine traffic removed 427%
Overall 364%

 

4.4 Example Results from Using Daily Tourism Estimates (DTEs)

The previous statistical analyses were based on the estimated average daily
tourism traffic (ADTT) and percentage of daily tourism traffic (PDTT). This section
provides statistical results based on the derived daily tourism estimates (DTEs) for some
selected counter stations.

As previously mentioned, differences in daily traffic flows should be studied,
because they provide useful information about tourism traffic flow. For example, using
DTEs the author estimated that, during the Memorial Day weekend (May 22"d to 25th,
1998) around Station 4049 on I-75 at Vanderbilt, there were 23,155 counts of tourism
traffic going north and most took place on Friday May 22nd, and there were 26,352 counts
of tourism traffic going south and most took place on Monday May 25th (Table 4-17).

Using the smaller of the two numbers, one may say that there were about 23,000 counts

119

of tourism traffic passing through Vanderbilt on I-75 during the Memorial weekend in
1998.

Table 4-17. Example of Using the Removal of Routine Traffic Method to Estimate
Tourism Traffic on Memorial Day Weekend.

 

 

 

 

4049 North 4049 South
Day Of the Week Tourism Traffic Tourism Traffic
May 22"“, 1998 (Friday) 12,128 2,745
May 23’“, 1998 (Saturday) 8,235 2,642
May 24‘“, 1998 (Sunday) 2,072 6,313
May 25‘“, 1998 (Monday) 720 14,651
Total 23,155 26,352

 

Also, two statistical summaries of average DTEs for each day of the week from
Station 4049 and 9369 are presented in Table 4-18 and Table 4-19, respectively. Note
that these examples are utilized to demonstrate possible practical use of the Removal of
Routine Traffic Method. The results relate to Station 4049 on highway I-75 (north-south)
and Station 9369 on I-94 (east-west) only. Three time frames are specified: (1) all
weeks—including all available days of any weeks, (2) regular weeks—including any days
that are not in the vicinity of any long weekends (i.e., 2 days before and 2 days after a
long weekend, totally 7 days), (3) weeks with a long weekend—including any days that
are in a long weekend or its neighboring days.

Around Station 4049 the results from average weekly tourism traffic flow indicate
that, in all of the three time frames, most rural-type tourism trips started on Friday and

returned on Sunday (Table 4-18). Readers may notice that a significant portion of trips

120

 

 

started even earlier (on Thursday) and ended later (on Monday), also notice that a high

 

portion (19%) of tourism trips flowed fi'om north to south on Friday in a regular week.
Figures 4—4a, 4-4b, and 4-4c display substantial differences between northbound and

southbound weekly tourism traffic flows.

Table 4-18. Averaged Weekly Tourism Traffic Flows around Station 4049.

 

 

Day of the

 

 

 

 

 

 

 

 

 

Time Frame week North South % North % South
A11 weeks
Sunday 1551 5903 9% 34% , 1
Monday 1091 3403 7% 20%
Tuesday 671 1390 4% 8%
Wednesday 1228 1316 7% 8%
Thursday 2654 1233 16% 7%
Friday 6237 2683 ' 38% 16%
Saturday 2950 1364 18% 8%
Total 16383 17292 100% 100%
Regular weeks
Sunday 1327 5493 9% 36%
Monday 942 1994 7% 13%
Tuesday 525 882 4% 6%
Wednesday 965 1234 7% 8%
Thursday 2073 1307 15% 9%
Friday 5951 2869 42% 19%
Saturday 2484 1301 17% 9%
Total 14267 15080 100% 100%
Weeks with a Long
Weekend
Sunday 2630 7871 10% _ _ _ 29%
Monday 1776 9885 7% 36%
Tuesday 1368 3832 5% 14%
Wednesday 2800 1810 10% 7%
Thursday 4885 948 18% 3%
Friday 7957 1566, 30% 6%
Saturday 5283 1677 20% 6%
Total 26699 27589 100% 100%

 

121

All Weeks

-\ /\ +Norrh

\ // A \. +South
W i/ \-

Sun. Mon. Tue. Wed. Thu. Fri. Sat.

 

 

 

 

 

 

 

 

 

 

.EEEE

 

 

Figure 4-4a. Weekly Tourism Traffic Flow around Station 4049, All Weeks.

 

 

 

Regular Weeks
0
:E‘ __
E +North
% +80uth
..-

 

Sun. Mon. Tue. Wed. Thu. Fri. Sat.

Figure 4-4b. Weekly Tourism Traffic Flow around Station 4049, Regular Weeks.

Weeks with a Long Weekend

 

+ North
—I-- South

 

 

 

 

Traffic Volume

Sun. Mon. Tue. Wed. Thu. Fri. Sat.

Figure 4-4c. Weekly Tourism Traffic Flow around Station 4049, Weeks with a
Long Weekend.

122

 

 

The weekly tourism traffic flows around Station 9369 indicate that in all of the
three time frames most urban-type tourism trips took place on Friday (Table 4-19). The
weekly flow patterns displayed in Figure 4-5a, 4-5b, and 4-5c suggest that these urban-
type tourism trips are likely to be day trips. This is because there are no substantial
differences between westbound and eastbound tourism flows except that, during a long

weekend, a greater portion of tourism trips started on Friday and ended on Monday.

Table 4-19. Averaged Weekly Tourism Traffic Flows around Station 9369.

 

 

 

 

 

 

 

 

 

 

Time Frame Da'yvgkthe West East % West % East
A11 weeks

Sunday 3993 4392 1 1% 12%
Monday 4539 5071 12% 13%
Tuesday 4164 4338 11% 11%
Wednesday 451 1 4220 12% 1 1%
Thursday 5433 5089 14% 13%
Friday 8544 8644 23% 23%
Saturday 6749 61 16 18% 16%

Total 37932 37870 100% 100%

Regular weeks

Sunday 3940 4104 10% 1 1%
Monday 4083 4614 l 1% 12%
Tuesday 4136 4158 11% 11%
Wednesday 4869 4348 13% 12%
Thursday 5827 5397 15% 14%
Friday 8661 8511. _, , 23% 23%
Saturday 6808 6216 18% 17%

Total 38324 37346 100% 100%

Weeks with a Long
Weekend

Sunday 4194 5486 12% 14%
Monday 6180 6718 17% 17%
Tuesday 4247 4877 12% 12%
Wednesday 3078 3710 8% 9%
Thursday 4094 4043 1 1% 10%
Friday 8100 9152 22% 23%
Saturday 6466 5643 18% 14%

Total 36359 39628 100% 100%

 

123

All Weeks

 

+ West
—I- East

 

 

 

Traffic Volume

 

Sun. Mon. Tue. Wed. Thu. Fri. Sat.

Figure 4-5a. Weekly Tourism Traffic Flow around Station 9369, All Weeks.

 

Regular Weeks

10000

 

+ West
+ East

6000

 

 

 

Traffic Volume

2000

 

Sun. Mon. Tue. Wed. Thu. Fri. Sat.

Figure 4-5b. Weekly Tourism Traffic Flow around Station 9369, Regular Weeks.

 

 

 

Weeks with a Long Weekend
8
(>5 +West
0 +15»:
E __._.
S

 

Sun. Mon. Tue. Wed. Thu. Fri. Sat.

Figure 4-5c. Weekly Tourism Traffic Flow around Station 9369, Weeks with a
Long Weekend.

124

CHAPTER 5

SUMMARY AND CONCLUSIONS

A number of topics were discussed in previous chapters including: the inherent
problems in highway traffic data, related literature, traffic data characteristics, a
conceptual model linking traffic data to tourism studies, and statistical results derived
from applying the Removal of Routine Traffic Method to highway traffic data in
Michigan. This chapter presents the major findings and their implications, and concludes

with discussion of applications and recommendations for firrther research.

5.1 Summary of the Study

Ahnost all tourism trips involve using ground vehicles and highways either
exclusively or in concert with other means of transportation. Although the highway
traffic data collected by the Department of Transportation is a potentially useful set of
comprehensive recording of highway travel activities, inherent problems in the data have
deterred tourism researchers from using the data to their firllest potential. Neither the
government agencies that collect these data nor tourism researchers who look for useful
sources of information have exerted much effort toward creating scientifically sound
tourism information out of highway traffic data that are both readily available and of high
quality. When researches do use highway traffic data, they often utilize untreated traffic
data that contain both tourism and non-tourism traffic information. The derived results

are often difficult to interpret and can be misleading.

125

The purposes of this study were to improve the relevance of remotely sensed
highway traffic to tourism studies, to enhance the connections between highway traffic
data and tourism studies by developing sound procedures for removing non-tourism
elements from highway traffic data sets, and to demonstrate that valuable tourism
information can be derived from underutilized highway traffic data. These goals have
been achieved by the development of a new data processing method called the Removal
of Routine Traffic Method. The method is designed to mitigate the problem of separating
tourism traffic from non-tourism traffic thereby facilitating the use of highway traffic by
tourism researchers, planners, and business operators.

Prior to the development of the method, highway traffic time series were studied.
Useful but less known characteristics of traffic data were identified for tourism studies.
Characteristics of the highway traffic data indicated that the behavior of travelers was
highly patterned. Based on the patterned behavior of highway travelers, a conceptual
model was constructed to link highway traffic data to tourism studies. The behavior
theory in the model was used to guide the development of the Removal of Routine Traffic
Method that mitigates the inherent problems in the traffic data.

To demonstrate its validity, the method was applied to a set of Classified
Permanent Traffic Recorder (CPT R) data acquired from the Michigan Department of
Transportation (MDOT). The estimates derived by the Removal of Routine Traffic
Method were demonstrated to provide face validity for measuring tourism traffic. In
summary, the Removal of Routine Traffic Method functions as a filter to screen out non-

tourism traffic from total traffic and leaves the residual as the closest estimate of tourism

126

traffic. Further, statistical results from hypotheses tests indicate that the estimates from
the method satisfy theoretical expectations. The study has thus verified that:

(1) Weekend day tourism traffic is proportionally greater than weekday tourism
traffic.

(2) Tourism traffic is proportionally higher in less populated areas than in more
populated areas during either weekend days or weekdays.

(3) Routes with smaller percentages of non-recreational vehicle type traffic have

”'21

higher percentages of tourism traffic during either weekend days or weekdays.

(4) A long weekend has significant positive effects on the percentage of tourism
traffic on that weekend but has no effect on tourism traffic on weekdays.

(4) North-south oriented highways in Michigan have higher percentages of
tourism traffic during weekend days than do east-west oriented highways. This pattern is
not significant (or may not exist) on weekday tourism traffic.

Previously, tourism researchers avoided using highway traffic data in tourism
studies, because there was no appropriate and cost effective method to estimate tourism
traffic on highways. Now, with the Removal of Routine Traffic Method, researchers not
only can better understand the behavior of highway travelers but also utilize the
extensively collected highway traffic data with confidence that resulting estimates are
closer to the true values. An estimate of the overall improvement in traffic data relevancy
to tourism is 364%. Even simply removing traffic generated by non-recreational type
vehicles would improve overall data relevancy to tourism by 12%. Hopefully, the method
will encourage more researchers to use highway traffic data for regional tourism studies

and planning.

127

5.2 Major Findings and Implications

The results of this study demonstrate that mitigating the problem of separating
tourism traffic and non-tourism traffic is possible and that more valuable tourism
information can be derived from highway traffic data. This section summarizes the major

findings and associated implications from the study.

Vehicle Types and Tourism TraLfic

On Michigan highways, traffic generated by recreational type vehicles (RV traffic)
accounted for 89% of total traffic. Since RV traffic is the base of tourism traffic, in the
regression analyses, the percentage of RV traffic (PRV) always had the largest regression
coefficients (Table 4-14c, and 4-15c). Passenger cars and pickups (including SUVs,
vans, and mini vans) are the major vehicle types used on highways (Figure 4-1). Before
the removal of routine traffic process, weekday RV traffic volume was significantly
higher than on weekend day. However, after the removal of routine traffic process,
weekend day tourism traffic volume was significantly higher than on weekday (Table 4-
8). This indicates that the higher volume of RV traffic is most likely generated by
commuters during weekdays.

On average, weekend day tourism traffic is 26.50% of total traffic and weekday
tourism traffic is 20.56% of total traffic, also the percentages and volumes of tourism
traffic during weekend days are significantly higher than during weekdays (Table 4-8).
Therefore, this study confirms the general perception that weekend day traffic is more

tourism dominant than weekday traffic. The author believes that the differences between

128

 

weekend day and weekday statistics would be even more significant if Friday evening is
included as a part of weekend. However, the purpose of defining weekday and weekend
day as they are in this study is to facilitate the comparison of different traffic patterns
across days. That is, the purpose is to include 24 hours in each "day" and not to regard its
social meaning fully (especially for Friday). To decide the starting time of a weekend is
beyond the scope of this study.

The significant correlation between weekend day and weekday tourism traffic
(Table 4-11 and 4-12) may imply that the variations in these two variables are
simultaneously affected by a third variable, such as the same seasonality or geographical
(spatial) factors. For example, when a month is in a prime tourism season, both weekend
day and weekday traffic appear to have higher tourism traffic volumes and percentages.
Also, when a route is along a major access route to tourist destinations, both weekend day
and weekday traffic appear to have higher tourism traffic volumes and percentages. In
other words, the statistical results from these correlation analyses indicate the seasonal
and spatial variation natures of tourism are preserved in the estimates of tourism traffic

with the Removal Routine Traffic Method.

Regions

The major highways on the Southern Lower Peninsula of Michigan carry much
higher percentages of traffic generated by non-recreational type vehicles (NRV traffic)
(Table 4-1). This observation agrees with the author's postulation that highways on more
populated areas are "potentially" less tourism dominant as compared to highways in rural

areas. Further, in the regression analyses, region (REG) was the most significant variable

129

among the independent variables and entered the regression models before other variables
(Table 4-140 and Table 4-15c). This verifies the proposition that the location of a traffic

counter is the dominant factor that affects the recording of tourism traffic data.

Directions

North-southbound highways carry higher percentages of RV traffic than east-

westbound highways (Table 4-4). As a matter of fact, the most beautiful scenery and
natural outdoor recreation areas of Michigan are in the Northern Lower and Upper
Peninsulas. The statistical results indicate that north-southbound highways in Michigan
are more tourism dominant than east-westbound highways during weekends (Table 4-10).
Another interesting observation is that travelers tended to start their northbound
tourism trips before the weekend (e.g., on Friday or even earlier) and return south at the
end of the weekend or even later on Monday (Table 4-18). There are substantial
differences between northbound and southbound weekly tourism traffic patterns, which
indicate that a lot of north-southbound rural-type trips are overnight trips (Figure 4-4a,
4-4b, and 4-4c). On the other hand, most urban-type tourism trips are likely to be day
trips. This is so because there are no substantial differences between westbound and
eastbound tourism flow patterns, except, during a long weekend, when a greater portion
of tourism trips started on Friday and ended on Monday (Table 4-19 and Figure 4-5a,

4-5b, and 4-5c).

130

5.3 Applications

Application of the Removal of Routine Traffic Method on highway traffic data
can be straightforward, because the final results derived from the method are estimated
tourism traffic volumes and percentages. The method and its results can easily be applied
by tourism researchers, planners, and business operators.

The Removal of Routine Traffic Method can be effectively used to estimate
tourism traffic for a specific day or period of time. Ifone is interested in knowing the
amount of tourism traffic going through a site during certain holidays, the Removal of
Routine Traffic Method can promptly provide that information. A couple of such kinds
of examples have been presented at the fourth section of Chapter 4.

While the estimated tourism traffic going through a site may be valuable to
business operators, long-term tourism tracking may be the focus of tourism researchers
and planners. In a long-term study, the estimated tourism traffic can be used in
comparing tourism traffic for the same location over different periods of time, for
example a year-to-year or month-to—month comparisons, to observe its trend line.

Traffic counter data can be used individually or in a group. In the conceptual
model, two major applications were suggested: site—specific and regional tourism

applications.

5.3.] Site-Specific Tourism Application
In a site-specific tourism monitoring system, the ideal traffic counter location
should be right at the entrance of a recreational site. For a multiple-entrance site, multiple

counters will be needed. For instance, at the University of California at Los Angeles

131

 

(UCLA) an automated traffic monitoring system is installed at all entrances to the
campus. This system records entering and exiting traffic information for historical trend
analysis and long-range development strategies (UCLA, 2001).

The benefits of site-specific traffic tracking accrue only to that site. However, its
advantage is that it provides more accurate tourism traffic estimates for that site. Usually,
a site-specific tourism monitoring system is developed and maintained by the
administrators of the site who can quickly adopt and use the Removal of Routine Traffic
Method. Thus, the method need not be limited to traffic data collected by a state's
Department of Transportation. Usually, the officially installed permanent traffic
recorders are always some distance away from most recreational sites. As the distance
between a traffic counter and monitored site increases, the availability of the Removal of
Routine Traffic Method should serve to encourage installation of more site-specific traffic

monitoring system by individual communities or larger businesses.

5.3.2 Regional Tourism Application

In a regional tourism monitoring system, a logical configuration is to set traffic
counters on a region’s borders to monitor highway traffic flowing into and out of the
region. A regional tourism monitoring system is an enlarged version of a multiple
entrance site-specific tourism monitoring system. Consider an island as a tourist
destination which can only be accessed by ferries or by a bridge, tracking its tourism
traffic can be accomplished simply by counting how many people use the ferries or drive
over the bridge during a period of time. Analogously, traffic counters can be installed at

all of the major highway entrances to a region to monitor in-flow and out-flow tourism

132

traffic. Since international air traffic data collection fits with the analogy of regional
tourism monitoring, this may be the reason why international arrival data are more often
utilized in monitoring international travel and tourism activity (Witt and Martin, 1987;

Martin and Witt, 1989).

5.3.3 Applications of Tourism Dominance

Tourism dominance is defined by the percentage of tourism traffic on a route. In
application, tourism dominance is used to compare tourism—relatedness of different
routes or locations. Also it can be used to infer the tourism-dependence of a location by a
set of near-by counters when no traffic data are available specific to that location.

One may also use tourism dominance in the following ways.

1) For surveys that need to target tourists, researchers may select strong
tourism dominant routes to improve sampling efficiency. Or,
researchers may balance the sampling proportionally according to the
tourism dominance on different routes or days of the week.

2) A region’s tourism planning and marketing strategies may rely on
tourism traffic flowing through the region. After highway traffic data
have been collected, tourism researchers and planners can determine the
tourism dominance of each route according to the Removal of Routine
Traffic Method, and then design marketing and promotion strategies.
For example, installing highway display boards along more tourism

dominant routes to promote the attractions in the region.

133

 

 

 

3) Researchers can select the most tourism dominant locations to
incorporate in regional and/or statewide travel activity monitoring
systems such as that used by Michigan State University's Tourism

Center.

5.4 Projected Cost for Using the Removal of Routine Traffic Method

The above discussion demonstrates selected applications of the Removal of
Routine Traffic Method. Readers may be interested in knowing the cost and time
required to apply the method. Projected costs and time involved in using the Removal of
Routine Traffic Method are provided in Table 5-1. Compared to most tourism research
methods, the Removal of Routine Traffic is efficient and economical, because operation
of the method can be highly automated by computers and source data can be obtained at

no or minimum cost fi‘om state Departments of Transportation.

134

 

Table 5-1. Projected Cost and Time Investments Required to Implement the
Removal of Routine Traffic Method.

 

 

 

Phase Cost Description Needed Time
Data acquisition Minor Government traffic data Relatively short
are readily available to the
general public.
Data storage $3,000 The cost of a database Relatively short--
server. Data transferring
Data cleaning $1,200 Two-week wages of a data 2 weeks--(semi-
(recursive) analyst automation is
($15 x 80 hours) possible)
Data Processing
Database $4,800 Three-week wages of a 3 weeks--program
programming database programmer ($40 for automation of
x 120 hours) the operation
Creating computer $100 A few hour wages of a A few hours
repots (recursive) computer operator
First Time Cost $9,100
Recursive Cost $1,300

 

135

5.5 Limitations

It was previously mentioned that it is not possible to perfectly separate tourism
traffic fi'om non-tourism traffic. The major limitation of the Removal of Routine Traffic
Method is that it can only miti gate the problem of separating tourism from non-tourism
traffic. However, until a perfect separation method is invented, the Removal of Routine
Traffic is a theoretically suitable method for estimating highway tourism traffic. To
perfectly separate tourism traffic from non-tourism traffic, one would need to overcome
inherent invasion of privacy, traveler inconvenience, and traffic disruption problems.
Using the Removal of Routine Traffic Method to estimate tourism traffic avoids all of
these problems and yields much improved, but not perfect, tourism traffic estimates at a
very little cost in comparison to what would be required to make further improvements in
estimates. Nonetheless, the accuracy of tourism traffic estimates is affected by traffic
counters’ ability to accurately count vehicles and to classify them correctly. The accuracy
would also be affected by the selected percentile point used to estimate routine traffic
when applying the method.

The statistical results for this study were limited by the availability of classified
Permanent Traffic Recorder data (i.e., the number of CPTR stations and the regions
where data were available). The author believes that more complete CPTR data from
more counters and from different regions especially the Upper Peninsula would have
enhanced the results of this study and the test statistics developed would have proven to
be even more significant. The application of the method in a regional tourism study is
also limited by the locations of traffic counters. Although, the method can be used to

estimate highway tourism traffic, the estimates can only be used to infer tourism traffic on

136

major highways. There is always a small portion of tourists who enjoy driving on
meandering country roads and estimating their numbers falls outside the scope of this

study.

5.6 Recommendations
This study has demonstrated that highway traffic data are a useful source of
information for tourism studies. Throughout the study process, some useful

characteristics of highway traffic data were identified and a conceptual model and traveler

 

behavior theory were developed to guide the use of highway traffic data in tourism
studies. There are many questions and issues that this study failed to explore. The
following ideas are the author’s recommendations to the Department of Transportation

and tourism researchers to begin to resolve these questions and issues.

5.6.1 Recommendations to the Department of Transportation

As discussed in the study, location is an import factor affecting the potential uses
of collected traffic data. For regional tourism studies, the locations of traffic counters are
even more important. The ideal locations for traffic counters should be at the major
highways entering the region. However, when installing traffic counters, a state's
Department of Transportation usually does not consider the information needs of the
tourism industry. Thus, to improve the usefulness of the traffic data in tourism studies,
the author recommends that the Department of Transportation install new traffic counters
in the future at the major entrances to major tourism regions. Regional tourism planners,

such as local Conventions and Visitors Bureaus (CVBs), should also request that the

137

Department of Transportation install traffic counters at entry points to major tourism
regions.

Another way of improving the usefulness of government collected traffic data is to
process and report the data by vehicle types so that more relevant and meaningful
information can be derived from the data. Ifthe Department of Transportation fails to
process the data in more creative ways, the uses of data will be continuously limited to a
narrow range of public sector applications.

While the in-flow and out-flow of traffic are accurately recorded in bridge
crossing data, these data are currently only available on a daily basis and not by the hour.
Yet, bridge crossing data are very important in a regional tourism monitoring system,
because bridges are usually the only connection between two regions. For example, for
this study Mackinaw Bridge crossings were not available on an hourly basis; therefore,
one could only use the near—by counter (e.g., 4049) to estimate how much tourism traffic
visited the Upper Peninsula. Thus, the author recommends that the Department of
Transportation provide hourly bridge crossing counts along with vehicle classification
information in the future. By providing more useful traffic data, the Department of

Transportation can help tourism researchers to improve highway tourism studies.

5.6.2 Recommendations for Future Research

The Removal of the Routine Traffic Method is in a preliminary stage, and there is
room for improvements. As mentioned in Chapter 3, routine traffic may vary across
seasons and locations. The appropriate cut-off point may vary with studied areas and

time periods. It may be that a higher percentile cut-off point should be applied in urban

138

 

 

areas, because there is a higher percentage of commuter traffic in urban areas. A lower
percentile cut-off point should probably be applied in rural areas, because there is less
commuter traffic in rural areas. Also, a higher percentile cut-off point may have to be
applied on weekdays and a lower percentile cut-off point, on weekends, because
commuter traffic is higher on weekdays than on weekends. Further research could be
directed to fine tuning the method, to discover the best cut-off point for each traffic
counter location for example.

More creative ways of using traffic data and traffic data characteristics could also
be developed. When vehicle speed data are widely available, vehicle speed would
become valuable information on traveler behavior for example. Based on the consumer
behavior theory in economics, researchers can study travelers' preferences about time
spent on highways. Some travelers may enjoy highway scenery; therefore, they drive at
lower speeds for example. Some travelers may dislike the boredom of traveling;
therefore, they drive at higher speeds (Yang, 1996).

As more complete CPTR data are collected, regression analyses can be applied to
examine the relationship between regional tourism traffic and tourist spending.
Researchers may also study the relationship between tourism traffic and weather
conditions. In this study, only Michigan data were used. The Removal of Routine Traffic
Method could also be applied to highway traffic data in different states and countries to

test the generalizability of the method.

139

APPENDICES

140

 

APPENDIX A

Format of Permanent Traffic Recorder Data

141

Format of Permanent Traffic Recorder Data

 

 

a:
S g E 5 g “>3 E E E can. a. 3' 3 £1 E.
*3 8 >' >1 E ‘I :1 :1 ("I 2' 2. E —' 3 °1
a a E a a' a ’3 ’6 ‘3 a a 8 3 8 5’
9969 7 0 1995 12 1 884 607 38 2454 1823 74712 FALSE 6
9969 7 o 1995 12 2 1221 866 448 2468 1916 59752 FALSE 7
9969 7 o 1995 12 3 1348 950 483 1772 1305 47490 FALSE 1
9969 7 0 1995 12 4 729 552 263 1820 1168 67439 FALSE 2
9969 7 0 1995 12 5 807 514 48 1935 1465 67749 FALSE 3
9969 7 0 1995 12 6 790 517 463 1807 1453 67887 FALSE 4
9969 7 0 1995 12 7 766 523 265 2063 1394 69905 FALSE 5
9969 7 0 1995 12 8 775 673 465 1868 1311 70618 FALSE 6
9969 7 0 1995 12 9 916 593 536 2058 1643 50337 FALSE 7
9969 7 0 1995 12 10 1164 885 443 1514 1104 41706 FALSE 1
9969 7 0 1995 12 11 660 450 240 1911 1220 60485 FALSE 2
9969 7 0 1995 12 12 653 444 251 1948 1307 65886 FALSE 3
9969 7 0 1995 12 13 726 480 297 1437 965 63603 FALSE 4
9969 7 0 1995 12 14 538 391 258 2137 1420 63588 FALSE 5
9969 7 0 1995 12 15 869 693 333 2634 2102 74469 FALSE 6
9969 7 0 1995 12 16 1329 872 737 2586 2274 59908 FALSE 7
9969 7 o 1995 12 17 1512 1119 529 1854 1282 48701 FALSE 1
9969 7 0 1995 12 18 745 589 305 1946 1346 67547 FALSE 2
9969 7 0 1995 12 19 756 562 279 2048 1411 66951 FALSE 3
9969 7 o 1995 12 20 733 547 287 2046 1446 68254 FALSE 4
9969 7 0 1995 12 21 820 540 352 2423 1686 70382 FALSE 5
9969 7 0 1995 12 22 1020 863 652 2785 2024 73563 FALSE 6
9969 7 o 1995 12 23 1431 984 614 2417 2030 55873 FALSE 7
9969 7 0 1995 12 24 1419 917 793 2525 2193 48726 FALSE 1
9969 7 0 1995 12 25 1217 955 324 2555 1812 45194 FALSE 2
9969 7 0 1995 12 26 1168 696 297 1747 1349 55600 FALSE 3
9969 7 0 1995 12 27 777 561 295 1896 1447 60018 FALSE 4
9969 7 0 1995 12 28 871 565 289 2005 1570 62234 FALSE 5
9969 7 o 1995 12 29 1087 860 320 2274 1965 64433 FALSE 6
9969 7 0 1995 12 30 1407 905 457 2125 1811 52427 FALSE 7
9969 7 0 1995 12 31 1196 839 461 2117 1257 44141 FALSE 1

 

142

APPENDIX B

Format of Classified Permanent Traffic Recorder Data

143

Format of Classified Permanent Traffic Recorder Data

 

 

III
a: 8 E E E
8 8 F D E a 5 8 a
8=§E§§§g>§ihgfimggggggggi
eaSeSESEaEBEezaaaaeasEEE
C 26 58 8689 5 0 98 10 19 l 153 0 88 22 0 l 2 0 3 0 24 l 6 6
C 26 58 8689 5 0 98 10 19 2 105 0 46 16 0 l 4 0 l 0 23 2 6 6
C 26 58 8689 5 0 98 10 19 3 85 0 23 19 0 l 2 0 1 l 23 2 5 8
C 26 58 8689 5 0 98 10 19 4 78 0 21 8 0 4 5 0 2 2 29 0 2 5
C 26 58 8689 5 0 98 10 19 5 114 0 45 20 0 1 7 0 2 6 19 l 4 9
C 26 58 8689 5 0 98 10 19 6 179 0 60 39 0 5 5 2 5 13 29 3 5 13
C 26 58 8689 5 O 98 10 19 7 414 0 229 85 0 9 5 l 4 18 41 5 8 9
C 26 58 8689 5 0 98 10 19 8 493 0 241 120 l 8 12 0 ll 29 44 8 2 17
C 26 58 8689 5 0 98 10 19 9 719 0 464 125 1 5 3 0 ll 7 58 7 12 26
C 26 58 8689 5 0 98 10 19 10 747 l 485 135 0 l 3 0 l6 0 53 21 9 23
C 26 58 8689 5 0 98 10 19 11 418 2 157 121 0 0 11 0 23 2 40 24 29
C 26 58 8689 5 0 98 10 19 12 475 0 197 116 0 3 7 l 13 2 73 35 5 23
C 26 58 8689 5 O 98 10 19 13 469 0 213 109 0 7 35 0 35 4 35 10 4 17
C 26 58 8689 5 0 98 10 l9 14 553 0 254 103 0 l3 l3 0 28 0 69 10 12 51
C 26 58 8689 5 0 98 10 19 15 810 0 512 110 l 9 7 1 l7 4 66 26 6 51
C 26 58 8689 5 0 98 10 l9 16 965 0 643 120 0 l4 0 38 10 57 29 13 36
C 26 58 8689 5 O 98 10 l9 17 101 2 704 143 l 3 25 0 37 4 46 14 7 26
C 26 58 8689 5 0 98 10 19 18 102 0 755 115 0 5 l3 1 10 2 85 11 11 16
C 26 58 8689 5 0 98 10 19 19 779 0 535 110 0 l 5 0 20 0 76 ll 8 13
C 26 58 8689 5 0 98 10 19 20 519 0 348' 58 0 2 l4 0 21 1 45 5 8 17
C 26 58 8689 5 0 98 10 19 21 397 0 213 59 0 l 3 0 6 3 81 4 12 15
C 26 58 8689 5 0 98 10 19 22 354 0 203 46 0 2 2 0 6 l 51 3 20 20
C 26 58 8689 5 0 98 10 19 23 281 0 127 32 0 2 2 0 3 l 63 6 22 23
C 26 58 8689 5 0 98 10 19 24 229 0 102 14 0 4 2 0 5 l 51 8 19 23

 

144

 

APPENDIX C

1998 Calendar and Major Holidays

145

 

 

January

SMTWTFS

 

4
11
18
25

1 2
5 6 7 8 9
12 l3 14 15 16
19 20 21 22 23
26 27 28 29 30

 

 

 

 

CD

February
M T W T F

 

.—
v.“—

2 3 4 5 6
9 10 11 12 13
16 17 18 19 20
23 24 25 26 27

14
21
28

 

 

 

 

March
MTWTF

 

2 3 4 5 6
9 10 11 12 13
16 17 18 19 20
23 24 25 26 27
30 31

 

 

 

 

April
M T W T F

 

12

26

l 2 3
6 7 8 9 10
13 14 15 16 17
20 21 22 23 24
27 28 29 30

25

 

 

 

 

(I)

May
MTWTF

U)

 

tau——
—&~rou

l
5 6 7 8
12 13 14 15
19 20 21 22
26 27 28 29

~——
uu—&

un—
oonON

 

 

 

 

 

14

28

15 16 17 18 19
22 23 24 25 26
29 30

 

 

1998

Calendar and Major Holidays

January
1 NewYear's Day

2 January 2nd

19 Martin Luther King, Jr.

February

16 Presiden t's Day

May

25 Memorial Day

July
3 July 3rd

4 lnckpendence Day

Septerrber

7 Labor Day

October

31 Halloween

Novenber
26 Thanksgiving

27 After Thank sgivin g

Decenber
25 Ch rist ma 5

146

 

 

July
MTWTF

S

 

12

26

l 2 3
6 7 8 9 10
13 14 15 16 17
20 21 22 23 24
27 28 29 30 31

4
ll
18
25

 

 

 

Augrst
M T W T

"11

U)

 

3 4 5 6
10 ll 12 13 l
17 18 19 20 2
24 25 26 27 2
31

NN—
OHM“—

 

 

 

September
M T W T F

 

1 2 3 4
7 8 9 10 ll
14 15 16 l7 18
21 22 23 24 25
28 29 30

 

 

 

October
MTWTF
2

 

1
5 6 7 8 9
12 13 14 15 16
19 20 21 22 23
26 27 28 29 30

 

 

 

November
M T W T F

 

15

29

2 3 4 5 6
9 10 11 12 13
16 17 18 19 20
23 24 25 26 27
30

14

28

 

 

 

December
MTWTF

 

1 2 3 4
7 8 9 10 ll
14 15 16 17 18
21 22 23 24 25
28 29 30 31

12
19

 

 

 

 

 

 

 

 

APPENDIX D

Population Estimates of Michigan Counties, 1998

147

Population Estimates of Michigan Counties, 1998

 

 

 

County Population UP County Population NLP County Population SLP
Baraga 8602 1 Antrim 21473 2 Clinton 63407 3
Dickinson 27062 1 Benzie 14743 2 Gratiot 40145 3
Gogebic 17243 1 Charlevoix 24496 2 Ionia 66710 3
Houghton 35617 1 Clare 29514 2 Kent 544781 3
Iron 12882 1 Emmet 28633 2 Montcalm 60602 3
Keweenaw 2099 1 Grand Traverse 74224 2 Muskegon 166849 3
Marquette 62585 1 Kalkaska 15554 2 Ottawa 225407 3
Menominee 24393 1 Lake 10424 2 Bay 109980 3
Ontonagon 7842 1 Leelanau 19142 2 Genesee 435691 3
Alger 9984 1 Manistee 23485 2 Huron 35273 3
Chippewa 37906 1 Mason 27896 2 Lapeer 88229 3
Delta 38936 1 Missaukee 13887 2 Midland 81562 3
Luce 6791 1 Osceola 22138 2 Saginaw 210032 3
Mackinac 11041 1 Wexford 29118 2 St. Clair 159465 3
Schoolcraft 8782 1 Alcona 11061 2 Sanilac 43051 3

Alpena 30475 2 Shiawassee 72489 3
Arenac 16405 2 Tuscola 57965 3
Cheboygan 23813 2 Allegan 101680 3
Crawford 14128 2 Barry 54465 3
Gladwin 25341 2 Berrien 159831 3
10500 25715 2 Branch 43702 3
Montrnorency 9999 2 Calhoun 140806 3
Ogemaw 21085 2 Cass 49975 3
Osooda 8890 2 Eaton 101022 3
Otsego 22232 2 Jackson 156130 3
Presque Isle 14535 2 Kalamazoo 229627 3
Roseommon 23355 2 St. Joseph 61 141 3
Isabella 58394 2 Van Buren 75637 3
Mecosta 40156 2 Hillsdale 46572 3
Newaygo 45769 2 Ingham 285874 3
Oeeana 24745 2 Lenawee 98609 3

Livingston 146317 3

Macomb 786866 3

Monroe 143365 3

Oakland 1 175057 3

Washtenaw 302787 3

Wayne 2116540 3

Sub-total 31 1,765 770,825 8,737,641
State Total 9,820,231

 

Source: US. Bureau of the Census, Population Estimates and Population Distribution

Branches.

Notations: UP--Upper Peninsula, NLP--Northem Lower Peninsula, and SLP--Southem

Lower Peninsula.

148

APPENDIX E

Detailed Percentages of Improvement in Highway Traffic Data Relevancy to Tourism

149

 

Detailed Percentages of Improvement in Highway Traffic Data Relevancy to

 

 

Tourism
Day of
Station Location Week Jan Feb Mar Apr May Jul Aug Sep Oct Nov Dec
3069N US-131, M-66 Kalkaska weekend 203% 377% 378% 377% 316% 495% 402% 371% 337% 340% 338%
weekday 402% 468% 318% 652% 465% 588% 387% 505% 431% 446% 373%
30698 weekend 184% 337% 338% 361 % 317% 292% 205% 317% 296%
weekday 182% 560% 716% 839% 504% 350% 484% 590% 388%
4049N 1-75 Vanderbilt weekend 1 66% 337% 204% 154% 201 % 153% 1 14%
weekday 178% 168% 209% 1 56% 161 % 150% 154%
40498 weekend 1 30% 184% 309% 1 17% 108% 121 % 156%
weekday 331% 257% 316% 311% 221% 214% 264%
4129N US-27 Houghton Lake weekend 267% 175% 254%
weekday 173% 184% 232%
41298 weekend 122% 159% 238%
weekday 367% 276% 298%
5029N US-27 St. Johns weekend 297% 76% 396% 130% 340%
weekday 227% 1 15% 266% 131% 208%
50298 weekend 182% 154% 181 % 254% 333%
weekday 570% 507% 471% 350% 253%
50398 US-27 By-Pass, St. Johns weekend 199% 150% 183% 189% 166% 110% 184% 228% 200% 299%
weekday 425% 421% 365% 448% 437% 388% 428% 395% 320% 196%
5059N E 1-196 Hudsonville weekend 222% 262% 334% 320% 310% 326% 380% 404% 352% 311% 354%
weekday 484% 417% 313% 527% 414% 380% 329% 393% 510% 585% 307%
5059SW weekend 21 1% 347% 315% 309% 276% 331% 282% 352% 296% 290% 280%
weekday 420% 510% 282% 479% 553% 602% 615% 608% 475% 455% 297%
5249N US-131 Morley weekend 239% 242% 364% 241% 260% 310% 267% 278% 159% 225%
weekday 252% 267% 296% 188% 297% 230% 230% 188% 258% 330%
52498 weekend 239% 242% 364% 241% 260% 310% 267% 278% 159% 225%
weekday 252% 267% 296% 188% 297% 230% 230% 188% 258% 330%
5299W 1-96 lonia weekend 296% 229% 310% 353% 274% 274% 428%
weekday 357% 329% 243% 462% 521% 538% 492%
5309N US-13l Big Rapids weekend 155% 228% 179% 347%
weekday 193% 249% 276% 369%
53098 weekend 176% 214% 232% 205% 141 % 186% 247% 126% 121 % 147% 236%
weekday 460% 587% 589% 715% 629% 327% 327% 502% 487% 387% 379%
6369E 1-69 Capac weekend 362% 456% 349% 380% 491% 41 1% 214%
weekday 553% 814% 599% 774% 743% 688% 402%
702913 [-94 Grass Lake weekend 272% 280% 310%
weekday 453% 400% 388%
7109N US-l 31 Schoolcraft weekend 402% 397% 338% 232% 306% 318% 229%
weekday 547% 627% 368% 423% 346% 433% 180%
71098 weekend 400% 383% 241% 400% 317% 287% 329% 235%
weekday 612% 411% 357% 428% 480% 546% 676% 363%
7159W 1-94 Battle Creek weekend 312% 493%
weekday 375% 452%
7159E weekend 492% 297% 344% 244% 241%
weekday 442% 451% 412% 418% 474%
7179W 1-94 Coloma weekend 632% 357% 249% 309%
weekend 580% 413% 336% 319%
71791-3 weekend 725% 360% 239% 274%
weekday 581% 423% 496% 463%

 

150

 

Detailed Percentages of Improvement in Highway Traffic Data Relevancy to
Tourism (cont’d)

 

 

Station Location 1:32;: Jan Feb Mar Apr May Jul Aug Sep Oct Nov Dec
8219W [-96 Howell weekend 269% 448%
weekday 357% 326%
8219i: weekend 389% 493%
weekday 447% 449%
8229N US-23 Brighton weekend 267% 231% 297% 222% 311%
weekday 293% 364% 358% 316% 366%
82298 weekend 247% 227% 236% 200% 491%
weekday 342% 610% 442% 556% 671%
8249N [-75 Luna Pier. weekend 283% 21 1% 310% 297% 475% 242%
weekday 542% 405% 361% 456% 394% 454%
82498 weekend 379% 375% 429% 375% 357% 555% 529% 528% 350% 316% 278%
weekday 876% 686% 810% 634% 743% 460% 447% 841% 907% 743% 526%
8689N US-23 Dundee weekend 248% 252% 278% 354% 278% 271% 259% 295% 248% 208% 311%
weekday 342% 262% 306% 342% 379% 366% 305% 371% 342% 345% 332%
86898 weekend 276% 406% 274% 274% 409% 244% 269% 256% 294% 259% 270%
weekday 412% 432% 338% 306% 381% 403% 439% 487% 336% 311% 276%
8729N US-23 Lambertville weekend 1 20% 348% 197% 305% 297% 309% 294% 283%
weekday 342% 357% 250% 467% 449% 307% 255% 338%
87298 weekend 87% 216% 337% 277% 344% 368% 370% 264% 276%
weekday 216% 292% 429% 417% 541% 455% 511% 413% 367%
9049N US-127 Lansing weekend 231% 422% 429% 422% 302% 274% 353% 332% 318% 301% 277%
weekday 474% 506% 475% 366% 476% 537% 409% 461% 417% 369% 294%
90498 weekend 262% 469% 479% 492% 465% 265% 329% 244% 283% 244% 244%
weekday 486% 447% 380% 505% 441% 472% 542% 578% 688% 583% 249%
9369W [-94 Kalamazoo weekend 281% 470% 436% 355% 429% 469% 340%
weekday 717% 573% 492% 359% 662% 554% 448%
936915 weekend 236% 398% 433% 362% 505% 352%
weekday 484% 538% 431% 371% 744% 667%
9828W [-696, E of Southfield weekend 225% 283%
weekday 414% 459%
9829i: weekend 341% 322% 331% 305% 265% 464% 329% 345% 338% 339%
weekday 480% 555% 487% 605% 640% 475% 494% 553% 509% 343%
9959N [-75, at Mack Ave. weekend 201% 419% 384% 351% 313% 340% 355% 302%
weekday 344% 640% 600% 541% 617% 539% 462% 407%
99598 weekend 209% 355% 376% 390% 261% 269% 430% 325% 344%
weekday 481% 489% 388% 444% 528% 421% 538% 450% 355%
9979N [-75, at Wattles Rd. weekend 180% 329% 392% 286% 312% 268%
weekday 349% 402% 607% 402% 504% 260%
99795 weekend 201% 280% 360% 350% 391% 301% 502% 506% 433% 282% 229%
weekday 453% 490% 465% 318% 488% 549% 681% 543% 577% 491% 250%

 

151

Bibliography

Babbie, E. (1994). The Practice of 80%] Research. (7th ed.). Behnont, CA: Wadsworth
Publishing Company.

Borg, W. R., & Gall, M. D. (1979). Educational_Research: An Introduction. (3rd ed.).
New York, NY: Longman, Inc.

Briggs, R. (1981). Interstate Highway System and Development in Nonmetropolitan
Areas. Mmonation Research Record, 812:9-12.

Brockwell, P. J ., & Davis, R. A. (1991). Time Series: Theory and Methods. (2nd ed.).
New York: Springer-Verlag.

Chadwick, R. A. (1994). Concepts, Definitions, and Measures Used in Travel And
Tourism Research. In J. R. B. Ritchie and C. R. Goeldner (Eds.), Travel
Tourismyand Hospitality Research: A Handbook for Managers and Resgarchers
(2nd ed.). New York: John Wiley & Sons, Inc.

 

Cuyahoga County Engineer's Office. (1999). Cuyahoga County Engineer's Office Traffic
Count Data. hm;://www.cuycflengineerscrgjtraffic.html (lO/22/2000)

Eggertsson, T. (1990). Economic Behavior and Inatitutions: Cambridge Surveys of
Economic Literature. New York: Cambridge University Press.

Federal Highway Administration. (1995). US. Department of Transportation. Traflg
Monitoring Guide. Washington, DC: Federal Highway Administration.

Federal Highway Administration. (1997). 1996 Highway Statistics.
http://www.fl1wa.dot.gov/ohim/ 1 996/index.html (7/25/2001)

Furubotn E. G., & Richter, R. (1998). hstitutions and Economic Theory: The
Contribution of the New Institutiongal Economics-Economics, Cognition aad
Socieg. Ann Arbor, MI: University of Michigan Press.

152

 

Greene, W. H. (1990). Econometric Analysis. New York, NY: Macmillan Publishing
Company.

 

Gunn, C. A. (1979). Tourism Planning. New York, NY: Taylor & Francis.

Holecek, D. F. (1991). Characteristics of Michigan’s Travel Market. In D. M. Spotts
(Ed.), lime] and Tourism in Michigan: A Statistical Profile. (2nd ed.). East
Lansing, MI: Travel, Tourism, and Recreation Resource Center, Michigan State
University.

Humphrey, C. R., and Sell, R. R. (1975). The Impact of Controlled Access Highways on
Population Growth in Pennsylvania Nonmetropolitan Communities, 1940-1970.
Rural Sociology, fl: 332-43.

Kelly, J. R. (1996). Leisure. (3rd ed.). Needham Heights, MA: Simon & Schuster
Company.

 

Kuehn, J. A., and West, J. G. (1971). Highway and Regional Development Growth and
Change, 2: 23-28.

Lichter, D. T., and Fuguitt G. V. (1980). Demographic response to Transportation
Innovation: The Case of the Interstate Highway. Social Forces, £2: 495-512.

Martin, C. A., and Witt S. F. (1989). Accuracy of Econometric Forecasts of Tourism.
Annals of Tourism Research, E (3): 407-30.

Travel Michigan. (1996). Michigan Travel Bureau Overview. Lansing, MI: Michigan
Jobs Commission, Travel Bureau.

Roehl, W. 8., Fesenmaier, J ., and Fesenmaier, D. R. (1993). Highway accessibility and
regional tourist expenditures. Journal of Travel Research. 11.

153

Singleton, R. A., Jr., Straits, B. C., and Straits, M. M. (1993). Approaches to Social
Research (2nd ed.). New York, NY: Oxford University Press

Spotts, D. M. (1991). Tourist Attractions. In D. M. Spotts (Ed.), [gavel and Tourism in
Michigan: A Statistical Profile. (2nd ed.). East Lansing, MI: Travel, Tourism, and
Recreation Resource Center, Michigan State University.

Spotts, D. M. (1997). Regional Analysis of Tourism Resources for Marketing Purposes.
Journal of Trayel Researgh, 3_5 (3): 3-15.

Statistica Canada. (19913). Air Passenger Origin and Destination, Domestic, Cat. No. 51-
204, and Air Passenger Origin and Destination, Canada-United States, Cat. No.
51-205, two annual reports, Ottawa: Statistics Canada.

Stephanedes, Y. J ., and Eagle, D. M. (1986). Time series Analysis of Interaction
between Transportation and Manufacturing and retail Employment.
Transportation Research Record, 1074: 16-24.

Stewart, W. P., Lue, C. C., Fesenmaier, D. R., and Anderson, S. (1993). A Comparison
between Welcome Center Visitors and General Highway Auto Travelers. Journal
of Travel Research, 3_1 (3): 40-46.

Transportation Research Board. (1982). Proceedings of a Conference on Urbg

Transportation Flaming in the 19808. Washington, DC: National Academy
Sciences.

UCLA. (2001). Campus Automated Traffic Monitoring System.
http://www.transportation.ucla.edu/CATMS.htm. (3/2/2001).

US. Bureau of the Census. (1990). St_atistica1 Abstract of the United gates 1990: the
National Data Book. (110th ed.). Washington, DC: Bureau of the Census.

US. Bureau of the Census. (1999).
http://www.census.gov/prod/www/abs/transpor.html. (10/8/1999).

154

US. Bureau of the Census. (1999). Population Estimates and Population Distribution
Branches. httl)://ww.Lu1te.mi.us/dmb/mic/censua/demo/pop est/9099micounty.htm
(6/23/2001).

US. Bureau of the Census. (2000). Statistical Abstract of the United States 2000: the
National Data Book. (120th ed.). Washington, DC: Bureau of the Census.

Witt, S. F., and Martin, C. A. (1987). Econometric Models for Forecasting Intemational
Tourism Demand. Journal of Tgvel Research, g (3): 23-30.

World Tourism Organization. (1991). Conference Resolutiona. International Conference
on Travel and Touriam Statistics. Ottawa, Madrid: World Tourism Organization.

Yang, S. (1996). Utility and Disutility of the Travel Time. (unpublished).

Yang, S., and Holecek, D. F. (1997). The effectiveness of using highway traffic data in
monitoring travel activity. In Proceedings of the Travel and Tourism Resea£h
Association 28th Annual Conference.

Yang, S., and Holecek, D. F. (1996-99). Michigan Travel Indicators: Monthly Reports,
No. 1-40. East Lansing, MI: Travel, Tourism and Recreation Resource Center,
Michigan State University.

155