Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Vancouver personal travel market : a multivariate statistical analysis Nichols, Peter Royall 1968

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1968_A4_5 N53.pdf [ 10.33MB ]
Metadata
JSON: 831-1.0102414.json
JSON-LD: 831-1.0102414-ld.json
RDF/XML (Pretty): 831-1.0102414-rdf.xml
RDF/JSON: 831-1.0102414-rdf.json
Turtle: 831-1.0102414-turtle.txt
N-Triples: 831-1.0102414-rdf-ntriples.txt
Original Record: 831-1.0102414-source.json
Full Text
831-1.0102414-fulltext.txt
Citation
831-1.0102414.ris

Full Text

THE VANCOUVER PERSONAL TRAVEL MARKET: A MULTIVARIATE STATISTICAL ANALYSIS by PETER ROYALL NICHOLS B.Comm., University of Briti s h Columbia, 1961 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF BUSINESS ADMINISTRATION i n the Faculty of Graduate Studies We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA Apr i l , 196^ i i In presenting t h i s thes is i n p a r t i a l fu l f i lment of the requirements fo r an advanced degree at the Univers i ty of B r i t i s h Columbia, I agree that the L ib ra ry s h a l l make i t f r ee l y ava i lab le fo r reference and study. I fur ther agree that permission fo r extensive copying of t h i s thes is for scho lar ly purposes may be granted by the Head of my Department or by h i s representat ives. I t i s understood that copying or pub l ica t ion of t h i s t hes i s fo r f i n a n c i a l gain s h a l l not be allowed without my wr i t ten permission. Facul ty of Commerce and Business Administrat ion The Un ivers i t y of B r i t i s h Columbia, Vancouver 8, Canada. A p r i l , 1968. ABSTRACT This thesis has two objectives. The f i r s t i s to determine the c r i t i c a l variables i n the Vancouver personal travel market. The second objective i s to suggest additional surveys which might be undertaken. To determine the c r i t i c a l variables i n the Vancouver personal travel market, an analysis i s made of the data collected during the 1966 Air Canada Vancouver Survey. This analysis involves two separate but mutually reinforcing approaches. The f i r s t i s concerned with the pro-pensity to travel by a i r and by a l l modes combined for any given adult with known characteristics. The second approach i s concerned with the evaluation of particular factors affecting the choice between modes of travel. Trips by a i r are compared to tr i p s by auto, r a i l and bus i n terms of the characteristics of the tr i p s and i n terms of the characteristics of the persons taking the t r i p s . In suggesting additional surveys, this study focuses not only on the Air Canada Vancouver Survey but also on other recent travel surveys and analyses by transportation companies, government departments, travel associations, and academic institutions. This part of the analysis i n -volves two main areas of consideration. F i r s t , suggestions are made as to additional types of data which might prove valuable i n the transport firm's marketing decisions. Second, suggestions are made as to more use-f u l and revealing methods of analysing the data collected. The study finds past travel market research to be characterized by two main shortcomings. The f i r s t i s the failure of many studies to consider and analyse the interrelationships between t r i p variables and market variables. It i s suggested that only by matching t r i p character-V i s t i c s to market cha rac te r i s t i cs can the f i rm get some idea of the type of se rv ice , adver t is ing appeal and p r i c i ng po l i cy that are most e f fec t i ve fo r pa r t i cu la r segments of the market. The second major shortcoming found i n most t r a v e l market research i s the f a i l u r e to separate the d i rec t re la t ionsh ips between two var iab les from t h e i r i nd i rec t re la t ionsh ips -with other va r iab les . While i t i s recognized that var iab les such as income and occupation are c lose ly i n t e r -re l a ted , very few studies have made any attempt to determine the separate importance of occupation beyond that which i s i n d i r e c t l y a t t r ibu tab le to income. In the l i g h t of the shortcomings found i n past t r a v e l market research, the ana lys is of the Vancouver personal t r a v e l market involves the use of mul t i var ia te s t a t i s t i c a l methods. The basic method employed i s mul t ip le l eas t squares regress ion. However, the inherent quant i ta t ive nature of t h i s s t a t i s t i c a l procedure necessi tates that severa l mod i f i -cat ions be made fo r i t s app l ica t ion to the predominantly qua l i t a t i ve market and t r i p var iab les i n the survey. The primary modi f icat ion i s concerned .with the determination of numerical scales fo r the dependent and independent va r i ab les . A second major modi f icat ion i s concerned with the presentat ion of the mul t ip le regression r e s u l t s . This modi f icat ion i n -volves taking the mean of the dependent var iab le as a f i r s t approximation of the magnitude to be estimated and adjust ing t h i s value according to whether the pa r t i cu la r observation being considered belongs to a p a r t i c -u l a r group or possesses a pa r t i cu la r cha rac te r i s t i c . The thes is concludes that cer ta in market var iab les and cer ta in t r i p vi" variables have particularly strong and reliable relationships with personal travel. Income i s found to be the single most important market variable, with the domicile of the adult's children and parents being second in significance and reliability. However, trip variables are found to have generally more significant and reliable associations with travel than market variables. The trip variables showing greatest impor-tance are trip distance and trip purpose. The study further concludes that additional surveys of the Vancouver personal travel market should involve deeper investigation of those variables found in this study to be of particular significance and that some attempt should be made not only to f i l t e r out the indirect in-fluences between the variables but also to measure the magnitude of these indirect influences. TABLE OF CONTENTS CHAPTER PAGE I. INTRODUCTION 1 Objectives of the Study 1 Importance of the Study 2 Methodology 4 Limitations of the Study 6 D e f i n i t i o n of Terms 6 Plan of the Study _ 7 I I . DESIRABLE MARKET INFORMATION 8 Product Development 9 Price 12 S e l l i n g and Advertising . . . . . . 15 Summary 17 I I I . REVIEW OF PAST STUDIES 19 IV. THE AIR CANADA VANCOUVER SURVEY: BACKGROUND 44 The Objectives of the Survey 44 C e l l 45 The Sample 46 Method of Data Collection 47 The Questionnaire 47 The Analysis . 48 V. RESEARCH PROCEDURES 49 Multiple Least Squares Regression 49 Scaling the Dependent Variable 55 v i i i CHAPTER PAGE Multiple Classification Analysis . . . . 59 Interaction Effects 65 Weighting 67 Stratification 68 St a t i s t i c a l R e l i a b i l i t y . . 74 VI. PERSONAL TRAVEL FREQUENCY—ALL MODES 76 Introduction > 76 Influence and R e l i a b i l i t y of Variables . . 80 General Effectiveness of the Regression Analysis 104 Conclusions . . . . . . . . . . 108 VII. AIR TRAVEL FREQUENCY 112 Introduction 112 Influence and R e l i a b i l i t y of Variables 114 General Effectiveness of the Regression Analysis 133 Conclusions 134 VIII. CHOICE OF MODE 137 Introduction 137 -i Influence and Re l i a b i l i t y of Variables 139 General Effectiveness of the Regression Analysis 155 Conclusions 157 IX. CONCLUSIONS 160 Advantages of the Methodology . . . 160 C r i t i c a l Variables 162 General Effectiveness of the Regression Analysis 168 i x CHAPTER PAGE Shortcomings i n the Analys is . • • • 170 Suggestions fo r Add i t iona l Surveys 171 APPENDICES A. DEFINITION OF INDEPENDENT VARIABLES IN MULTIPLE REGRESSION ANALYSIS 175 B. ESTIMATES OF REGRESSION EQUATIONS BY ALL MODES, STRATIFIED BY FAMILY LIFE CYCLE:. 179 C. ESTIMATES OF REGRESSION EQUATIONS FOR AIR TRAVEL FREQUENCY 181 D. ESTIMATES OF REGRESSION EQUATIONS FOR MODE CHOICE 183 E. MULTIPLE DISCRIMINANT ANALYSIS IN SCALING THE DEPENDENT VARIABLE . . . 185 BIBLIOGRAPHY LIST OF TABLES TABLE PAGE I. A Comparison of Regression Equations for Business Automobile Travel for the Year 1962, for Married Adults with Children, using Discriminant Function Scales for 1955 and 1962 58 I I . Percentage of Adults Who Took any Air Trip by Age and by Family Life Cycle Stage . . 70 I I I . A Comparison of Regression Equations for Non-Business Auto Travel for Specified Stages i n Family Life Cycle for the Year 1955 . . 72 IV. Multiple Classification Analysis of Factors Affecting Total Travel Frequency by Stage i n Family Life Cycle . . . . 78 V. Residence Location and Total Travel Frequency 102 VI. Variance Analysis of Regressions for Total Travel Frequency Stratified by Family Life Cycle Stage 106 VII. Multiple Classification Analysis of Factors Affecting Air Travel Frequency 113 VIII. Multiple Classification Analysis of Factors Affecting Mode Choice 140 IX. Residence Location and Mode Choice 154 X. Variance Analysis of Regressions for Mode Choice . . . . . . . 156 LIST OF GRAPHS GRAPH PAGE 1 L i f e Cycle Stage and Tota l Travel Frequency 81 2 Income and Tota l Travel Frequency 82 3 Dwell ing Ownership and Tota l Travel Frequency 87 4 Dwell ing Type and Tota l Travel Frequency 88 5 F inanc ia l Interest and Tota l Travel Frequency 91 6 Occupation and Tota l Travel Frequency . . . . . . . . 92 7 Education and Tota l Travel Frequency 94 8 Domici le of Parents and Chi ldren and Tota l Travel Frequency . . 96 9' Place of Schooling and Tota l Travel Frequency 98 10 Place of B i r t h and Tota l Travel Frequency 100 11 L i f e Cycle Stage and A i r Travel Frequency 115 12 Income and A i r Travel Frequency 117 13 Dwell ing Type and A i r Travel Frequency 120 14 F inanc ia l Interest i n Major Assets 121 15 Occupation and A i r Travel Frequency 123 16 Education and A i r Travel Frequency . " 125 17 Domici le of Chi ldren and Parents and A i r Travel Frequency . . . 126 18 Place of Schooling and A i r Travel Frequency 128 19 Place of B i r t h and A i r Travel Frequency 129 20 Residence Location and A i r Travel Frequency 132 21 Income and Mode Choice 142 22 Dwell ing Type and Mode Choice . . . 145 23 L i f e Cycle Stage and Mode Choice 146 24 Tr ip Duration and Mode Choice 147 x i i GRAPH " PAGE 25 Inverse of Number Who Went on Trip and Choice Between A i r and Auto 148 26 Distance and Mode Choice 150 27 Travel Frequency and Mode Choice . 151 28 Trip Purpose and Mode Choice 152 ACKNOWLEDGEMENTS This t hes i s could not have been poss ib le without the whole-hearted encouragement, assistance and guidance of Professor Trevor D. Heaver, Chairman of the D i v i s i on of Transportat ion. Grati tude must a lso be expressed to A i r Canada fo r i t s f i n a n c i a l assistance and also fo r i t s permission to use the data from i t s Vancouver Travel Survey. The wr i te r reserves to himself the r e s p o n s i b i l i t y fo r any errors of commission and omission. CHAPTER I INTRODUCTION Objectives of The Study The purpose of this study i s two-fold; f i r s t , to determine the c r i t i c a l variables i n the Vancouver personal travel market, and second, to suggest additional surveys that could be undertaken to further personal travel market research. To determine the c r i t i c a l variables i n the Vancouver personal travel market, the study w i l l analyse the data collected during the 1966 Air Canada Vancouver Survey. The main part of this analysis w i l l involve two separate but mutually reinforcing approaches. The f i r s t w i l l be concerned with the propensity to travel by air and by a l l modes for any given adult with known characteristics. The second approach w i l l be concerned with the evaluation of particular factors affecting the choice between, modes of travel. Trips by air are compared to trips by r a i l , bus and auto i n terms of characteristics of the trips and i n terms of characteristics of the person taking the t r i p s . In suggesting additional surveys that could be undertaken to further personal travel market research, this study w i l l focus not only on the Air Canada Vancouver Survey but also on other recent travel surveys and analyses by transportation companies, governmental departments, travel associations, and academic institutions. This part of the analysis w i l l involve two main aspects of research. Fi r s t , suggestions w i l l be made as to additional typ.es of data-which might 2 prove valuable i n the transport f i r m ' s marketing dec is ions . Second, suggestions w i l l be made as to more usefu l and reveal ing methods of analysing the data co l l ec ted . As the foregoing statements have ind ica ted , t h i s study i s concerned*mainly with a i r t r a v e l . However, contrary to the usual prac t ice i n such a study, which i s to consider the one mode i n i s o l a t i o n from a l l o thers, i t i s an essen t ia l feature of t h i s study that i t involves constant comparison and i n t e r r e l a t i o n of a i r t r a v e l with t r a v e l by a l l other modes. Importance of The Study Market knowledge i s essen t ia l to any f i r m ' s s e l l i n g a c t i v i t i e s . Only wi th a thorough and accurate knowledge of i t s present and po ten t ia l markets can the f i rm form p o l i c i e s and i n i t i a t e serv ice development, adver t i s ing , and p r i c ing that w i l l enable i t to re ta in and expand i t s markets. Market research i s p a r t i c u l a r l y v i t a l to the t ransportat ion indust ry . Travel i s a nonstoreable commodity, making i t s under-production or over-production a matter of immediate l oss of revenue or reduction of p r o f i t s to the f i rm . Without inventory to act as a buffer between production and sa les , accurate knowledge of the shape and f l uc tua t ion of the demand curve i s c r i t i c a l . Another fac tor emphasizing the importance of market research i n the transport industry i s the fact that the t r a v e l "product" i s subject to a derived demand; i t s demand i s not fo r i t s own sake but rather for the dest inat ion to which the t r a v e l 3 i s e f fec ted. A i r l i n e s are looking on personal t r a v e l as one of t he i r br ightest horizons for expansion. Business t r a v e l , whi le provid ing a s o l i d base fo r many a i r l i n e s , i s recognized as being a r e l a t i v e l y i n e l a s t i c market. Except fo r r e l a t i v e l y short distance t r i p s , i t i s u n l i k e l y that even very substant ia l fare changes by any mode would appreciably a l t e r the a i r l i n e s ' share of the business t r a v e l market. This i s due to the fac t that fare d i f ferences between a i r t r a v e l and other modes tend to be considerably smaller than t he i r cost d i f f e r e n t i a l s i n terms of sa la r ied time consumed i n t r a v e l . In the case of personal t r a v e l , however, which i n most cases involves recreat ion where time tends not to be of as great importance, there are many s i tua t ions i n which the higher a i r fare o f fse ts the perceived benef i ts of the greater speed of a i r t ravel .^" This s i t ua t ion i s i l l u s t r a t e d i n a survey of 1964-1965 t r ave l behaviour by the Univers i ty of Michigan.^ This report ind ica tes that a i r t ransportat ion already has the vast major i ty (82 per cent) of business t r a v e l fo r distances of over 500 m i les . In the case of i vacat ion t r i p s , however, only 17 per cent of the people t r a v e l l i n g over 500 mi les use the a i r l i n e s . Thus, the personal t r a v e l market For fur ther d iscussion see K.W. Studn ick i -Gizber t , Economics of  the Canadian A i r Transport Industry (unpublished Ph.D. Thesis presented to the Facul ty of Research and Graduate Studies, M c G i l l Un ivers i ty , p. 227) and Richard E. Caves A i r Transport and I t s Regulators (Cambridge, Mass: Un ivers i ty Press, 1962), p. 35. John B. Lansing, The Travel Market: 1964-1965 (Ann Arbor, Michigan: Survey Research Center, October 1965), p. 40. 4 provides airlines with a wide potential for expansion and points out the need for applying proper research methods to this market to discern what changes the a i r l i n e might implement to grasp this potential profitably. Methodology The analysis of the Air Canada Vancouver Personal Travel Survey to be presented i n this study w i l l make use of quantitative multivariate s t a t i s t i c a l techniques, including multiple least squares regression, multiple classification analysis, and analysis of variance. . To the writer's knowledge, multivariate methods of s t a t i s t i c a l analysis such as multiple regression have been used i n the analysis of the travel market by only one organization, The Survey Research Center 3 of .the University of Michigan. The application of these techniques on a wider basis would seem to have been held back for two reasons. Fi r s t , multiple regression and other such techniques are basically quantitative and have been cursorilly labelled as unsuitable for qualitative data such as that gathered i n travel market surveys. Those who may have made some attempt to actually put theses methods to use on travel market data were probably soon turned away by the need for several modifications to the standard formulations. The second and probably the most powerful reason for the general avoidance of multi-J.B. Lansing and D.M. Blood, Mode Choice i n Intercity Travel: A  Multivariate S t a t i s t i c a l Analysis (Ann Arbor, Michigan: Survey Research Center, 1964). J.B. Lansing and D.M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1964). 5 variate s t a t i s t i c a l techniques in travel market research was expressed by Lansing and Blood of the University of Michigan*s Survey Research Center. In preparing a report l i k e this there i s always a dilemma about using formal, multivariate methods of s t a t i s t i c a l analysis . . . .most people interested i n the travel market are not trained as statisticians. A report built around a l o t of equations would be about as readily understood as a report written i n Sanskrit.4 Multiple regression offers several powerful advantages over the commonly used tabular methods of analysis. F i r s t , i t compresses a large amount of s t a t i s t i c a l data into a compact form. Second, and of major importance, i t can be used to substantially reduce the problem of multicollinearity.5 Multiple classification analysis provides a conversion of the multiple regression results for essentially qualitative variables into a more meaningful format. It involves taking the mean of the dependent variable as the f i r s t approximation of the magnitude which i s to be estimated (e.g. frequency of travel or probability of choosing one of two modes) and adjusting this value according to whether a particular observation belongs to a particular group or possesses a particular characteristic. This study also u t i l i z e s analysis of variance. This method i s used to measure the contribution to the explanation of the t o t a l "^op. c i t . , p. i i i . 5 Multicol.linearity i s the tendency for explanatory variables to be intercorrelated to the point where the effect of one variable i s "easily mistaken for the effect of another variable. 6 variance made by fami ly l i f e , cyc le stage, the s t r a t i f i c a t i o n var iab le used i n the t o t a l t r i p frequency regress ions. L imi ta t ions of The Study This study i s r es t r i c t ed to ana lys is of the data -which are ava i lab le from the A i r Canada Vancouver Survey. The survey concentrated on cer ta in aspects of t r ave l which were of pa r t i cu la r i n te res t to A i r Canada management and gave only b r i e f considerat ion, or no considerat ion at a l l , to other aspects which might a lso have borne i nves t i ga t i on . No d i rec t considerat ion i s given i n t h i s study to demand fo re -cas t ing , an area equal ly wide and needful of at tent ion as market research. Although of the same genre, i t i s important to d is t ingu ish between these two aspects of a i r t ransport research. Whereas market research re fers to the de f i n i t i on and study of the market as i t i s , demand forecast ing re fers to the pred ic t ion of how the market w i l l change i n the fu ture . The considerat ion of only the former does not reduce the importance of t h i s study, fo r a knowledge of how the market might change necessi tates a knowledge of what the market i s . This study i s fur ther l im i ted i n that i t involves the considerat ion and evaluat ion of only a small proport ion of the t r a v e l market studies undertaken by a i r l i n e companies. Due to t h e i r competit ive value, the a i r l i n e s hold a l l but t he i r most general studies con f i den t i a l . De f i n i t i on of Terms T r ip . Transportat ion by an adult to a dest inat ion more than 250 miles from Vancouver and re turn. 7 Business Travel. Travel which i s undertaken for the sole purpose of conducting or engaging i n some activity of business. Personal Travel. Travel which i s undertaken for any of the following purposes: i ) vacation and pleasure; i i ) personal affairs; i i i ) only partly business; and iv) business for only some of the party. Plan of The Study 'This chapter has indicated the objectives, importance, methodology, and limitations' of this study. Chapter II suggests the kind of information about the personal travel market that would be desirable for the transport firm i n determining i t s marketing strategy. Chapter III presents a review and evaluation of past personal travel market studies. The Air Canada Vancouver Survey i s introduced i n Chapter IV, with a brief background indication of i t s purpose, the sample, the survey methodology, and the type of data collected. Chapter V provides a detailed explanation of the analytical techniques applied to the Vancouver Survey. The results of this analysis are presented i n Chapters VI, VII and VIII. Chapter IX offers the conclusions to the study; a summary of the c r i t i c a l variables revealed by the.analysis of the Vancouver Survey, and some suggestions as to further surveys. CHAPTER II DESIRABLE MARKET INFORMATION The purpose of a market research program i s to provide the company wi th deta i led and accurate information that w i l l ass i s t i t i n shaping a marketing strategy that i s appropriate to the pa r t i cu la r nature of i t s product and environment. Thus, before any market study i s ca r r ied out, i t i s important to determine what sort of information about the market i s des i rab le . The purpose of t h i s chapter i s to discuss the type of information that would be desi rable fo r the a i r l i n e i n i t s e f fo r ts to serve the personal t r a v e l market. The d iscussion w i l l be of a general nature as the de ta i l s w i l l depend on the pa r t i cu la r a i r l i n e and i t s circumstances. Before get t ing in to the main d iscuss ion , one important d i s t i n c t i o n should be made; the essen t ia l d i f ference between passenger surveys and market research. Passenger surveys r e s t r i c t themselves to the examination of those people who are already using the serv ice . Market research, the subject of t h i s study, covers a much broader area. I t includes an examination of those people who are not customers, i n an e f for t to f i nd out why they are not using the service and what modi f icat ions would have to be made to a t t rac t them. The fo l lowing d iscuss ion i s d iv ided under three main areas of marketing a c t i v i t y : product development, p r i c i n g , and se l l i ng , and adver t i s ing . 9 l ) Product Development; The economic concept of consumer sovereignty w e l l describes the essence of modern marketing theory and is the f o c a l point upon which a i r transport marketing strategy i s based. This p r i n c i p l e embodies the idea that i t i s the consumer rather than the producer who determines what i s produced. Instead of deciding on his own what to produce and then attempting to force i t upon the buying public, the manufacturer f i r s t undertakes consumer or market research and then decides what to produce. Thus, along with cost and c a p a b i l i t y considerations, the firm's product development i s shaped by i t s knowledge of consumer wants and preferences. Very b a s i c a l l y t h i s involves finding out from the consumer about h i s l i k e s and d i s l i k e s concerning current products and his ideas about what he would l i k e to see on the market. In seeking out market opinions i t i s important that the research t r y to establish the nature of the firm's competitive surroundings and i t s place i n those surroundings. This implies much more than simply an estimate of market share. I t involves t r y i n g to determine why other firms are successful or unsuccessful i n the market and what the public l i k e s and d i s l i k e s about t h e i r services. With t h i s knowledge the f i r m can adjust i t s own competitive a c t i v i t i e s . To consider i n a more s p e c i f i c manner the type of information an a i r l i n e would want from the market i n developing i t s product, the discussion w i l l be formulated around the three main aspects of the a i r transport product: a) frequency and scheduling, b) cabin service and airport handling, and c) p r i c i n g . The f i r s t two of these w i l l be d i s c u s s e d here and the t h i r d w i l l be l e f t t o the next s e c t i o n . Frequency and Scheduling Frequency and scheduling are o f t e n the most d i f f e r e n t i a t e d c h a r a c t e r i s t i c s of the s e r v i c e s provided over the same route by d i f f e r e n t a i r l i n e s and may loom hig h i n the consumers' c o n s i d e r a t i o n s , not o n l y as to a i r l i n e choice but a l s o as t o mode choi c e . They are a l s o of major importance t o the a i r l i n e s themselves as a change i n e i t h e r of these may n e c e s s i t a t e a l a r g e c a p i t a l expenditure on a d d i t i o n a l a i r c r a f t . F l i g h t frequency c o n s i d e r a t i o n s i n v o l v e a t r a d e - o f f between opt i m a l a i r c r a f t u t i l i z a t i o n and optimal seat-mile c o s t s . The technology i n the a i r c r a f t i n d u s t r y i s such t h a t a i r c r a f t can now be b u i l t t o c a r r y over f i v e hundred passengers at a sea t - m i l e cost c o n s i d e r a b l y l e s s than the 140 seat a i r c r a f t p r e s e n t l y i n common use. However, t o use these a i r c r a f t on most routes and at the same time maintain a d e s i r a b l e frequency of s e r v i c e would mean t h a t the a i r c r a f t would probably be f l y i n g w i t h a quarter of t h e i r designed passenger paylqad, p u t t i n g a s i d e the a t t r a c t i v e s e at-mile cost f o r a f o u r - f o l d higher passenger-mile cost."*" What t h i s amounts t o i s a choice between lower f a r e s and gr e a t e r frequency. I n t h e i r i n t e r e s t s of maximizing p r o f i t s i t i s very important f o r t h e a i r l i n e s t o d i s c e r n which of these two s i t u a t i o n s , or what t r a d e -"^One p o s s i b i l i t y which t he a i r l i n e s are c o n s i d e r i n g t o reduce t h i s problem i s t h e ope r a t i o n o f mixed cargo and passenger f l i g h t s . However there i s some question as t o whether enough revenue can be taken from the cargo t r a f f i c on these f l i g h t s t o maintain passenger f a r e s s i g n i f i c a n t l y lower than those, o f f e r e d on the present a l l - p a s s e n g e r f l i g h t s o f smaller „(140 seat) a i r c r a f t . 11 of f point i n between the market f inds most des i rab le . The scheduling dec is ion involves another t rade-of f s i t ua t i on . I t i i s essen t i a l l y a peaking problem; a t rade-of f between serving the peaks whi le having large amounts of unused capaci ty i n the off-peak periods or maintaining an a i r c ra f t f l ee t at off-peak capaci ty and ignor ing the peak surpluses i n demand. The f i r s t of these s i tua t ions involves the buildup of a high c a p i t a l investment i n a i r c r a f t which w i l l not be earning any revenue during a large proport ion of the t ime. This leads to higher fa res , probably both i n the peak and off-peak per iods. The second s i tua t ion involves the turning away of customers at peak periods and the i nev i t ab le advent of competit ion, whether through the processes of i government l i cenc ing or unrestrained competit ive mob i l i t y , that w i l l not only take the unserved peak t r a f f i c but probably a good proport ion of the off-peak t r a f f i c as w e l l . Thus, i t becomes important fo r the a i r l i n e to have knowledge of the importance the market puts on optimum scheduling r e l a t i v e to lower fa res . Cabin Service and Ai rpor t Handling Cabin serv ice i s looked on by some a i r l i n e s as t he i r only remaining area of competition on domestic routes. This i s ind icated by a quick perusal of a i r l i n e advertisements i n the United States, with t h e i r common emphasis on the meals, i n - f l i g h t movies, and stewardess fash ions. L i t t l e i s known, however, as to jus t how e f fec t i ve t h i s type of competit ion i s . An area often overlooked, but needful of improvement and having 12 considerable p o t e n t i a l i t y fo r competit ive advantage, i s a i rpor t handl ing. I t i s f a i r to say that no other phase of a i r l i n e operations i s as much af fected by serv ice or qua l i t y competit ion as t r a f f i c handl ing. The ef fec ts of service competition are also transmitted to monopoly routes, par t l y because a mono-poly c a r r i e r i s p a r t i c u l a r l y sens i t i ve to publ ic c r i t i c i s m and pa r t l y because i t i s d i f f i c u l t w i th in the same a i r l i n e organizat ion to provide d i f fe ren t serv ice standards on com-pe t i t i ve and non-competitive routes.2 Airpor t handling i s one of the greatest sources of passenger f rus t ra t i on and annoyance and may have, a greater e f fect upon an i n d i v i d u a l ' s use of a i r t ransportat ion i n s i tua t ions where there i s a c lose choice between a i r and another mode, espec ia l l y auto, than a i' considerable margin of f i n a n c i a l expense. This i s another area where research of passenger and non-passenger at t i tudes may be of considerable va lue. 2) P r i c e : Of the three aspects of the a i r transport product mentioned at the beginning of the previous sec t ion , p r i c i ng i s usua l ly the most important to the consumer. As i s the case i n a l l i ndus t r i es , a i r l i n e s would l i k e to know the p r ice e l a s t i c i t y of the demand fo r t h e i r serv ices . I f i t i s e l a s t i c they may consider lowering t h e i r p r i ce i n hopes that the resu l t i ng increase i n patronage w i l l provide a greater than o f fse t t ing increase i n p r o f i t s . On the other hand, i f the f i rm i s faced with an K.W. Studn ick i -Gizber t , Economics of the Canadian A i r Transport  Industry (unpublished Ph.D. Thesis presented to the Facul ty of Research and Graduate Studies, McG i l l .Un i ve rs i t y , 1964), p. 34. 13 i n e l a s t i c demand curve, i t may f i nd i t s e l f i n the pos i t i on of being able to ra i se i t s fares without any appreciable decrease i n patronage. Thus, i n order to get some idea of what i t can safe ly and p ro f i t ab l y do as fa r as changing i t s fares i s concerned, i t i s important to the a i r l i n e to discover the nature of the demand funct ion for i t s serv ices over i t s var ious routes. This might involve enquiring of t r a v e l l e r s the reasons fo r t h e i r use of a pa r t i cu la r mode or even a d i rec t question as to the ef fect of p r i ce upon t h e i r dec is ion . Further to providing information to ass i s t the a i r l i n e i n deciding p r ice changes, knowledge about the e l a s t i c i t y of i t s demand curves may be he lp fu l i n i t s determination of the f e a s i b i l i t y of i n s t i t u t i n g cer ta in types of d i f f e r e n t i a l p r i c i n g . Two of the most common forms of d i f -f e r e n t i a l p r i c i ng pract iced by a i r l i n e s are excursion fares and family p lan fa res . These reduced fares become advantageous to the a i r l i n e when of fered to sectors of the market which a|re character ized by a reasonable degree of p r i ce e l a s t i c i t y . I t i s a w e l l recognized fact that a i r l i n e s are faced with a high proport ion of f i xed costs to. t o t a l costs and that the out-of-pocket cost incurred by the carr iage of add i t iona l passengers i n a l e s s than f u l l y loaded a i r c r a f t i s marginal . Thus, the carr iage of t r a f f i c at considerably reduced rates s t i l l provides the a i r l i n e with some contr ibut ion to f i xed costs which i t would not otherwise rece ive . This added contr ibut ion to f i xed costs also allows the a i r l i n e to lower i t s rates on the otherwise higher rated t r a f f i c . The net resu l t i s a s i tua t ion advantageous to a l l concerned; acceptable rates to those who 14 would not otherwise be able to af ford to t r a v e l , reduced rates to those who would be w i l l i n g to pay higher ra tes , and higher p r o f i t s to the a i r -l i n e . As mentioned above, the one proviso fo r the i n s t i t u t i o n of such reduced rates i s that the market sector i n which the o r i g i n a l fare reduct ion i s made must have a p r i ce e l as t i c demand for the se rv i ce . This gives fur ther emphasis to the d e s i r a b i l i t y of applying market research to determine the nature and e l a s t i c i t y of the a i r l i n e ' s demand funct ion. Market research may also be of value i n determining the f e a s i b i l i t y of i n s t i t u t i n g several other types of fa res , namely package tours and f l y -dr ive f a res . Although not as d i r e c t l y connected to pr ice e l a s t i c i t y of demand as the fares mentioned above, the success of e i ther of these "combination" fares great ly depends upon ithe existence of su f f i c i en t demand and might involve considerable wasted adver t is ing and admini-s t ra t ion expense i f i n s t i t u t ed and then withdrawn due to lack of demand. In in te rna t iona l commercial a i r t ranspor tat ion there i s very l i t t l e l a t i t u d e fo r p r ice competit ion due to agreements and r e s t r i c t i o n s i n i t i a t e d by I .A.T.A. Any s ign i f i can t degree of qua l i t y d i f ference i n cabin serv ice i s also proh ib i ted . This leaves s e l l i n g and adver t is ing as perhaps the most important aspects of marketing-strategy i n i n t e r -na t iona l a i r l i n e competit ion. These w i l l be discussed i n the fo l lowing sec t ion . 15 3) S e l l i n g and Adver t i s ing : The s e l l i n g of any commodity requires not only a knowledge of the product to be sold but a lso a knowledge of the market.towards which the s e l l i n g e f fo r ts are to be d i rec ted . Certa in parts of the market may be more predisposed to purchasing the product than others and d i f fe rent parts may be suscept ib le to d i f fe ren t types of sales and advert is ing appeal. The basic aim i n s e l l i n g i s to a f fec t the la rges t volume of revenue sales per do l l a r of s e l l i n g and adver t is ing cos ts . Theoret ica l ly and i d e a l l y the f i rm w i l l continue to expand i t s s e l l i n g e f fo r ts u n t i l the cost of such add i t iona l e f fo r t surpasses the net revenue gained from the add i t iona l sales made. I t has long been rea l i zed that properly conducted market research, p r i o r to and during the actua l s e l l i n g of the product, w i l l g reat ly reduce the amount of misdirected and ine f fec t i ve s e l l i n g e f fo r t , and resu l t i n lower average t o t a l costs per sale and, u l t imate ly , higher t o t a l sales and net p r o f i t s . B a s i c a l l y , t r ave l market research for "sel l ing and advert is ing involves t r y ing to match market cha rac te r i s t i cs to t r a v e l charac te r is -t i c s . On the market charac te r i s t i c s ide , information on l i f e cycle stage, education, occupation, monetary and d iscre t ionary income, ..and domici le of parents or ch i ldren might be sought. On the t rave l charac-t e r i s t i c s s ide , research might seek information on d is tance, durat ion, dest inat ion, purpose, frequency, and seasonal i ty . With t h i s information the a i r l i n e can be aware of what type of person buys i t s serv ice and some d e t a i l as to the type of serv ice he buys. This w i l l ass i s t the _ a i r l i n e i n deciding the d i rec t ion and nature of i t s s e l l i n g approaches 16 and adver t is ing appeals. In formulating i t s s e l l i n g and adver t is ing a c t i v i t i e s , i t i s also important fo r the a i r l i n e to discover the causual l i n k between the customer and the pa r t i cu la r serv ice he des i res ; the reasons causing him to buy or not to buy a pa r t i cu la r type of serv ice . I t i s the customer's motivat ion and reasoning, whether r a t i ona l or i r r a t i o n a l , that leads him to purchase any pa r t i cu la r se rv ice . With t h i s type of information the a i r l i n e w i l l not only know to whom i t should d i rec t i t s s e l l i n g e f for ts but w i l l a lso have some idea of what type of reasoning and motivation i t should appeal to i n i t s adver t is ing and s e l l i n g approaches. A fac tor c lose ly re la ted to those just mentioned i s the determin-a t ion of the most desirable channels of d i s t r i bu t i on and most e f fec t ive adver t is ing media. At present most a i r l i n e s r e l y on two main channels of d i s t r i b u t i o n , t he i r own o f f i ces and independent t r a v e l agencies. More s p e c i f i c a l l y , the a i r l i n e ' s own channel consis ts of three l i n e s of communication or sub-channels; telephone, terminal v i s i t , and t i c k e t -o f f i ce v i s i t . Likewise the t rave l agency channel encompasses the choice between telephone and o f f i c e v i s i t . The strength and low cost of the a i r l i n e ' s s e l l i n g e f fo r ts w i l l be great ly dependent upon i t s use of the appropriate channel. As a resu l t of determining the market's preference for a cer ta in channel, the a i r l i n e w i l l know whether to increase i t s telephone reservat ions personnel or es tab l i sh more or bigger t i c ke t o f f i c e s . Beyond t h i s , research of populat ion preferences may reveal a new type of channel, or even a simple modi f icat ion to those channels present ly used, which might open up h i ther to unserved markets. In the case of adver t i s ing , the a i r l i n e has to determine which media to use. A i r t r a v e l l e r s and po ten t ia l a i r t r a v e l l e r s may be more exposed to ce r ta in media than others. A lso , persons l i k e l y to t r a v e l to ce r ta in p laces, such as Europe, may be more exposed to or inf luenced by a pa r t i cu la r magazine, or newspaper, or b i l l boards i n a pa r t i cu la r sect ion of downtown, or pa r t i cu la r radio and t e l e v i s i o n programs. With-out knowledge; of t h i s type the a i r l i n e might we l l undertake large and expensive adver t is ing programs which do not reach the market fo r which they are intended. Although the adver t is ing mater ia l i t s e l f might be d i rec ted towards the r igh t people and might appeal to the r igh t motives and reasoning, i t i s of l i t t l e value i f inappropriate media are employed. Summary A i r l i n e marketing management, l i k e marketing management i n most commercial enterpr ises, i s p r imar i l y concerned with those var iab les that i t can con t ro l . McCarthy has l abe l l ed these var iab les "the four P * s " — product, p r i c e , place and promotion.^ In developing i t s "product" , a i r -l i n e management requires information as to the preferences of i t s customers, and po ten t ia l customers, concerning schedul ing, frequency, cabin serv ice , a i rpor t handl ing, and even a i r c r a f t type. In p r i c i ng i t s product, the a i r l i n e i s in teres ted i n the d i f ferences i n fare e l a s t i c i t y i n the d i f fe ren t segments of i t s market so that i t may i n s t i t u t e reduced ^ E . Jerome McCarthy, Basic Marketing - A Managerial Approach (Homewood, I l l i n o i s : R.D. I rw in , I960), p. 45. 18 rates and other special fares which are not only advantageous to the public but also to itself. Airline management is also concerned with determining the best "place" to sell i t s product. This requires informa-tion as to the type of person that buys air service, some detail as to the type of service he buys, and information as to those persons who don't presently buy the service but who. may be a potential market. In promoting i t s product, the airline i s interested in information as to the advertising media which are most effective in reaching its market and the advertising appeals which have the greatest influence upon i t s potential customers. Establishing the specific types of information which are important to i t s marketing strategy i s only the first step in the airline's market research. The next steps are to collect the appropriate data and to undertake an analysis which will indicate the important market variables and the ways in which these variables are related to the controllable factors in the firm's marketing process. These aspects of marketing research will be discussed throughout the balance of this study. CHAPTER I I I REVIEW OF PAST STUDIES I t i s the purpose of t h i s chapter to examine the "s ta te of the a r t " of t r a v e l market research i n terms of the types of data co l lec ted and the methods of ana lys is used i n past t r a v e l market s tud ies . Past t r a v e l market studies were sought from numerous a i r l i n e s , a i r c r a f t manufacturers, academic i n s t i t u t i o n s , governmental departments, t r a v e l assoc ia t ions , and some un l i ke l y sources such as women's magazines. In a l l , twenty-seven studies were obtained. Unfortunately the a i r l i n e s are poorly represented. Due to t h e i r competit ive va lue, most a i r l i n e s hold a l l but t h e i r most general studies con f i den t i a l . This chapter w i l l not review a l l the studies gathered. A large proport ion of those received are el iminated from considerat ion because they do not r e a l l y involve market research. Some are simply o r i g in and dest inat ion s tud ies, being concerned only wi th estimating the number of passengers t r a v e l l i n g between cer ta in points."'" Others are concerned only with re l a t i ng changes i n the volume and types of t r a v e l to changes i n the nat iona l economy. An example of t h i s l a t t e r type of study i s "The Ef fec t of Selected Demographic Charac te r is t i cs on U.S. C i t i z e n Travel Abroad" by the Douglas A i r c ra f t Company. This study involves a general descr ip t ion of the s i ze and importance of U.S. C i t i zen t r a v e l to overseas 1 See The New Yorker A F ive Year Study of T ra f f i c to the P a c i f i c (New York: The New Yorker, 1967) and a lso A Five Year Study of Travel  Between the U.S.A. and Mexico, Centra l America, Panama, and South America (New York: The New Yorker, 1966). dest inat ions over the l a s t decade with some emphasis on a i r t r a v e l , and also goes in to some general economic cha rac te r i s t i cs of the United States such as populat ion, consumer consumption pat terns, and trends i n export a c t i v i t i e s . However, t h i s study does not delve in to any spec i f i c cha rac te r i s t i cs of the people i n the t r a v e l market or of the t r i p s they take. This i s not to say that t h i s report and others l i k e i t do not provide usefu l informat ion, but rather that there i s a l o t l ess true market research than a simple count of the t o t a l reports on the t r a v e l market would suggest. This chapter w i l l review fourteen s tud ies ; those which are considered c losest to true market research as discussed i n the l a s t chapter. Each study w i l l be considered separately, except where several studies by the same organizat ion are c lose ly s i m i l a r . The review w i l l be concerned wi th the purpose of the study, the sampling method, the data, and the methods of ana lys i s . l ) Better Homes and Gardens Travel Questionnaire Study . . . A Report  on Family Travel Bet ter Homes and Gardens undertook t h i s study i n 1966 to determine the t r a v e l i n te res ts and a c t i v i t i e s of i t s subscr ibers. I t defends i t s i n te res t i n t h i s area by s ta t ing : Our very v i t a l i t y centers on the home and the fami ly . But from there i t spreads to a l l a c t i v i t i e s that make up a family or iented l i f e . And one of the most important—certain ly one of the most exc i t ing—of these a c t i v i t i e s i s t r a v e l . 2 Better Homes and Gardens, Travel Questionnaire Study . . . . A  Report on Family Travel (New York: Better Homes and Gardens, 1966), p. 1. An eight page questionnaire was attached to a l l subscriber copies of the February and March 1966 issues of the Better Homes and Gardens magazine. The February enquiry contained a set of questions different from those in March. As an incentive td response, $500.00 in Travellers Cheques were awarded to each of f i f t y randomly selected respondents. From, the 210,000 questionnaires returned, a statistically random sample of 10,000 questionnaires was selected for the analysis. The data collected in the questionnaire were extremely varied. The topics covered were: characteristics of the respondents, vacation planning, most memorable vacation, auto travel, accommodations, travel services, domestic travel, Western hemisphere travel, foreign travel, family camping, o i l company credit cards, rented cars, and travel photo-graphy. However, data were collected on only five personal character-istics: age, education, residence, number of children and auto ownership. Conspicuously absent are income and occupation. Much of the information requested was; of a hypothetical nature; the respondents* plans for future trips and the characteristics of the trip they would "most like to take". The analysis of the data i s superficial, consisting entirely of trivariate tabulations with two of the variables, circulation region and percentage of total respondents, maintained throughout. Nowhere does the analysis relate one travel characteristic to another, or, with the exception of circulation region, one market characteristic to another. Even more important, there i s no consideration of the interrelation of market characteristics with travel characteristics. The analysis is essentially a description of the travel behaviour and attitudes of the 22 magazine's audience as a,whole, with nothing to indicate that these people are any different from the rest of the population or that any particular type of person i s l i k e l y to travel i n any different manner than any other type of person. It can only be concluded that this study provides l i t t l e , i f any, information that would be useful to the transport firm i n making decisions as to i t s product development and marketing strategy. 2) Alberta Department of Industry and Development Survey and Analysis  of Non-Resident Travel. This study was conducted by the Transportation Division of Kates, Peat, Marwick and Co., consultants, during 1966 and was presented i n April, 1967. The objective of the study was to obtain factual information regarding travel motivations, travel mode, and spending char-acteristics of non-resident visitors to the Province of Alber-ta . . . . The survey results and'subsequent analyses were to provide guidelines for non-resident tourist promotion, guidelines to recreational development, and an assessment of the direct dollar value of !tourism to the province.3 The survey was made during the peak tourism months, from May 28th to September 5th. The sample was taken as tourists l e f t the province. In the case of travel by automobile, a highway cordon was established along the provincial borders with interview stations at points where major highway f a c i l i t i e s crossed the cordon. In the case of travel by common carriers, interview stations were set up at major air , r a i l and bus terminals within the province. The sampling i t s e l f was a randomized -"Province of Alberta, Department of Industry and Development, Survey and Analysis of Non-Resident Travel, April 1967, p. 1. 23 intermittent procedure. The particular days of interviewing at each station were chosen randomly, then hour periods, and f i n a l l y non-resident automobiles. The questionnaire was divided into two sections. The f i r s t section was completed i n person at the interview stations. This part of the questionnaire sought information relating to travel character-i s t i c s , t r i p purpose, and eaqoenditure. The second section of the questionnaire was handed to the v i s i t o r with a request to complete i t at a later date and return i t by mail. This questionnaire pursued in greater depth the inducement-factors influencing the t r i p making decision and the v i s i t o r ' s reaction to the province's recreational environment. A t o t a l of 8,999 direct interviews were completed and 3,700 of the mail questionnaires were returned. The analysis consists of bivariate and trivariate tabulations, histograms, and origin and destination maps. These methods of analysis and presentation are put to somewhat better use than' was the case i n the Better Homes and Gardens Study, with consideration of the relationships between t r i p characteristics and also between t r i p characteristics and income. The most severe limitation to this study i s that i t considers only one market characteristic. Income level can hardly constitute a tourist p r o f i l e . Further, to better pin-point the nature of the d i f -ferent sectors of the market, i t would have been desirable to relate more than only two t r i p characteristics or one t r i p characteristic and one market characteristic at a time. 3 ) Air Canada i ) Southern Traveller Survey This study was completed i n May 1966 by the Market Research Division of Air Canada. Its purpose was "to determine the motivations, expectations and experience, and descriptive image of the Southern Traveller."^ The study i s based on data collected from two random samples. One sample was drawn from people i n Air Canada's reservations f i l e s across Canada who had. made a reservation to travel to Southern points, i n February and March of 1965. This sample consisted of 194 persons, a l l of whom were interviewed i n person and 70 percent of whom were re-interviewed after returning from their t r i p . The second sample was drawn from the general public across Canada who had taken a t r i p South to the Caribbean or Florida by any mode of transportation any. time within the period December 1964 to May 1965. This sample involved 256 personal interviews. Data were collected on three main aspects of Southern travel; the decision making process, the journey i t s e l f , and the southern holiday. In examining the decision making process, enquiries were made about the amount of planning, the use of a travel agency, and the influence of advertising. The journey i t s e l f was questioned i n terms of the length of stay, day of departure, mode choice and air l i n e choice. The investi-gation of the Southern holiday sought information about the attractiveness of the places visited, the purpose of the v i s i t , the amount spent, and the jftlr Canada, Market Research Division, "Southern Traveller Survey", May, 1966, p. 1. . 25 type of accommodation. The analysis i n this study involves primarily bivariate l i s t i n g s with the variations i n each t r i p characteristic evaluated by the percentage of respondents giving that answer to the appropriate question. With only two exceptions the rest of the analysis i s trivariate. In these cases one of the variables i s percentage of respondents and another i s one of three dichotomous variables; destination (Caribbean versus Florida), mode (air passenger versus general public), or planning (planned versus actual). The two exceptions simply involve stratification of several of these same dichotomous variables. A l l of these tabulations consider only t r i p characteristics. Market characteristics (age, sex, income, occupation, and language spoken) are brought into the analysis only at one point-'and are related only to whether the_respondent went to Florida or to the Caribbean and whether he travelled by a i r or by any of the other modes. The analysis could have been made much more revealing and useful i f these market characteristics had been related to further t r i p variables such as purpose, length of t r i p , expenditure, etc., especially i f these variables were given more than just a dichotomous breakdown. i i ) Air Canada Passengers Departing from Eight  Canadian Cities, Aug.-Sept., 1966 This study v/as undertaken for Air Canada by O.R.C. International Ltd. The objectives of the study were to obtain detailed information on the characteristics of Air Canada passengers and their t r i p s with special emphasis on three groups of passengers: those who increased their air 26 t r a v e l r e l a t i v e to three years e a r l i e r , frequent t r a v e l l e r s , and passengers who were f l y i n g only because of the r a i l s t r i ke during the time of the . survey. The study was based on 17,238 personal interviews with persons departing from eight major Canadian c i t i e s on regu la r ly scheduled A i r Canada f l i g h t s during the per iod August 26 to September 3, 1966. Quotas were set to r e f l e c t the r e l a t i ve volume of f l i g h t s departing from each c i t y . Interviews were conducted i n the departure lounges of the a i rpor ts jus t p r i o r to each f l i g h t . Data were co l lec ted on f i v e d i f fe ren t types of t r a v e l l e r s ; regular t r a v e l l e r s , heavy t r a v e l l e r s , t r a v e l l e r s on t he i r f i r s t f l i g h t , t r a v e l -l e r s who had increased t h e i r a i r t r a v e l , and t r a v e l l e r s who were going by a i r only because of the ra i lway s t r i k e . Each of these types of t r a v e l were examined on the bas is of eight c h a r a c t e r i s t i c s : des t ina t ion , purpose, s i ze of t r a v e l party, age, sex, whether business, and what k ind of bus-i n e s s . The an lays is i n t h i s study i s tabular and most of the tables invo lve the re la t ionsh ip between two or three t r i p cha rac te r i s t i c s . No considerat ion i s given to the i n t e r r e l a t i o n of market cha rac te r i s t i c s . In f a c t , only three market charac te r i s t i cs are considered at a l l : sex, age, and c i t y of residence. Although the ana lys is re la tes these market cha rac te r i s t i cs to t r i p cha rac te r i s t i c s , they are too few to provide a usable p r o f i l e of the d i f fe ren t sectors of the market. Even without co l l ec t i ng data on any add i t iona l market va r iab les , the ana lys is could have been improved by s t r a t i f y i n g in to the same tables the three var iab les that were co l l ec ted . The sample s i z e , which i s usua l ly a l i m i t i n g fac tor i n the number of var iab les i n te r re la ted at one t ime, was ce r ta in l y large enough to warrant considerable s t r a t i f i c a t i o n . i i i ) The Canadian A t l an t i c Trave l le r This was a study sponsored by A i r Canada i n conjunction with B r i t i s h Overseas Airways Corporation and the B r i t i s h Travel Assoc ia t ion . The object ive of the study was to provide information on: 1 ) The demographic cha rac te r i s t i cs of t ransa t lan t i c t r a v e l l e r s resident i n Canada. 2 ) The reasons for t r a v e l w i th in the framework of non-business t r a v e l . 3) The planning, t iming and length of t r i p ; past and fu ture. 4) Preferences fo r countr ies to v i s i t with accompanying reasons. 5) Mode of t ranspor tat ion used and cost of t r i p . 6 ) Timing of the booking and dec is ion to t rave l . - ' The study was based on information co l lec ted i n 1 , 2 0 0 personal interv iews conducted i n ten major metropol i tan centers i n Canada. Respondents were selected on the basis of having t r ave l l ed i n the past f i v e years, or going i n the next three years, on a non-business t r i p across the A t l a n t i c . The t o t a l sample was d iv ided equal ly among these two types of respondents. In examining those respondents who had been on a t ransa t lan t i c t r i p , the survey sought information about t h e i r economic and demographic A i r Canada, Market Reserarch D i v i s i o n , "The Canadian A t l an t i c T rave l le r - a Preview of F ind ings" , June, 1 9 6 6 , pp. 1 - 2 . 28 cha rac te r i s t i c s , t h e i r t r a v e l experience, and a descr ip t ion of t he i r l a s t t r i p across the A t l a n t i c . The most extensive questioning involved the l a s t of these aspects, -with" enquir ies about the manner i n which the booking was made, the purpose of the t r i p , the a i r l i n e used, the places v i s i t e d , and the durat ion of the t r i p . In the case of the respondents who were planning a t ransa t lan t i c t r i p i n the next three years, the survey concentrated, on t h e i r preferences as to which country they would l i k e to v i s i t and t h e i r reasons fo r these preferences. L ike the other studies reviewed so f a r , the analys is i n t h i s study i s tabu lar . However, un l ike these other analyses, many of the tabu la -t i ons consider the re la t ionsh ips not only among t r i p charac te r i s t i cs and among market cha rac te r i s t i cs but a lso the i n te r re l a t i ons between these two basic areas of in format ion. As discussed i n the l a s t chapter, t h i s k ind of ana lys is can provide very usefu l information to the t ransportat ion f i rm i n deciding i t s product,development and marketing strategy. By matching the charac te r i s t i cs of the t r a v e l l e r s to the charac te r i s t i cs of the t r i p s they take, the f i rm w i l l have a bet ter idea of areas fo r improvement and areas with po ten t ia l fo r expansion. For example, by r e l a t i ng data about peoples' reasons fo r going on t r i p s to the places they go, the f i rm can adjust i t s adver t is ing appeals fo r pa r t i cu la r f l i g h t s and dest inat ions so that they are i n l i n e wi th the market's mot ivat ions. By re l a t i ng the income of the t r a v e l l e r to the length of time he stays away on a t r i p , the f i rm may have some ind ica t ion of how i t might change .the time period of i t s excursion fares i n order to increase the demand fo r t h i s type of se rv ice . The analys is considers only t r i v a r i a t e i n t e r r e l a t i o n s . Idea l ly i t would have been more reveal ing i f more var iab les had been added, wi th several s t r a t i f i c a t i o n s of the market charac te r i s t i cs i n te r re la ted with several s t r a t i f i c a t i o n s of t r i p cha rac te r i s t i c s . However, i n t h i s study, wi th only a 1,200 uni t sample and some var iab les wi th large numbers of c lasses , such an expansion would probably have reduced the s t a t i s t i c a l r e l i a b i l i t y of the resu l t s to an unacceptable l e v e l . i v ) The C a l i f o r n i a Study This study was undertaken i n 1966 by A i r Canada's Marketing Research D i v i s i o n . There were two purposes . . . one to provide A i r Canada'a.management wi th an i n ten -s ive ana lys is of the C a l i f o r n i a , SouthernCal i fornia and Los Angeles markets and the other to evaluate and quant i fy , where poss ib le , the cha rac te r i s t i cs of t h i s market with the end to providing the bas is fo r determining A i r Canada's short and long range market object ives fo r the Toronto-Los Angeles rou te . ° In i t s en t i re ty , t h i s study covers more than just market research. I t involves eight sect ions, s ta r t i ng with a general socio-economic p r o f i l e of C a l i f o r n i a , and the Los Angeles area i n pa r t i cu la r , and fol lowing-on to forecasts of passenger and f re igh t t r a f f i c between Los Angeles and Eastern Canada. Our concern here i s wi th two intermediary sec t ions , "The Los Angeles Travel Market" and "The Eastern Canada Travel Market", which involve analyses of surveys on the t r a v e l behaviour and t r a v e l l e r cha rac te r i s t i cs i n these areas. A i r Canada, Market Research D i v i s i on , "The C a l i f o r n i a Study", November, 1966, Foreword, p. 1. 30 a) The Los Angeles Market The purpose of t h i s sect ion of the study was stated as : Although our prime in te res t i n t h i s consumer research was d i rec ted towards increas ing our marketing knowledge, hope-f u l l y to i den t i f y opportunity areas which A i r Canada might c a p i t a l i z e upon, we also intended that t h i s survey would provide an important "bench-mark" against which A i r Canada might evaluate i t s future promotional a c t i v i t i e s . 7 The survey was made by a Los Angeles based research f i rm . The sample was taken from only those census t rac ts with median family incomes of $7,500.00 or more. The reason stated for t h i s was that an i n i t i a l ana lys is had shown that there was a decided increase i n the percentage of the populat ion which had1 ' t rave l led outside the state i n income l e v e l s at $7,500.00 and above.^ The ac tua l method of sampling employed i s known as unrest r ic ted proport ional mul t i -s tage sampling. Within the designated census t r a c t s , s t reets and then houses were selected at random. The number of un i t s taken from each census t rac t was proport ional to i t s populat ion i n terms of number of fami l iesJ The data was co l lec ted by means of telephone in terv iews. The interv iews co l lec ted data concerning seven fac to rs ; the general t r a v e l market out of C a l i f o r n i a , long distance t r a v e l , the long distance t r a v e l l e r , the European t r a v e l l e r , the t o t a l vacat ion t r a v e l market, vacat ions i n Canada, and the importance of f r iends or re la t i ves i n Canada on t r a v e l to Canada. The general t r a v e l market out of C a l i f o r n i a , long distance t r a v e l , ^ A i r Canada, Market Research D i v i s i o n , "The C a l i f o r n i a Study", November, 1966, p. 66. 8 I b i d . p. 67. 31 the long distance t r a v e l l e r , and the European t r a v e l l e r were considered i n a s u p e r f i c i a l way, wi th the ana lys is consis t ing en t i re l y of b ivar ia te or t r i v a r i a t e tabulat ions i n t e r re l a t i ng several t r i p cha rac te r i s t i c s . Market cha rac te r i s t i cs were not considered at a l l i n these parts of the. ana l ys i s . The ana lys is of vacations i n Canada was considerably more de ta i l ed , re la t i ng market cha rac te r i s t i cs ( t rave l experience, education, income, e tc . ) to whether or not a t r i p to Canada was planned. Aside from t h i s one part of the ana lys is , however, i t would appear that the study was pr imar i l y in teres ted i n forecast ing the t o t a l amount of t r a v e l between C a l i f o r n i a and Canada with very l i t t l e concern fo r determining the important market va r i ab les . This lack of concern for l i n k i n g t r a v e l -l e r cha rac te r i s t i cs to t r i p cha rac te r i s t i cs i s i l l u s t r a t e d by the analy-s i s of the ef fect of having f r iends or r e l a t i ves i n Canada on t r a v e l to Canada. Three items are considered: ' the number of long distance t r a v e l -l e r s who have f r iends or r e l a t i ves i n Canada, the proport ion of f r iends or r e l a t i ves which have been v i s i t e d i n the past three years, and the fami ly s ize of people having f r iends or re la t i ves i n Canada. As sepa-rate re la t ionsh ips i n t h i s form, no r e a l knowledge i s gained about the nature of the market fo r t r a v e l to Canada. Although we know that a proport ion of the people who have f r iends or r e l a t i ves i n Canada have v i s i t e d Canada and that the people who have f r iends or re la t i ves i n Canada tend to have cer ta in fami ly s i z e s , we do not know anything about the fami ly s ize of the people who ac tua l l y v i s i t t he i r f r iends or r e l a -t i ves i n Canada. This information would be valuable i n formulating the a i r l i n e ' s adver t is ing and s e l l i n g programs. 32 b) The Eastern Canadian Travel Market The purpose of t h i s study was to invest igate the extent to which t r a v e l i s made by residents of Eastern Canada to C a l i f o r n i a . A survey was conducted by telephone interv iews on a random sample of the adult populat ion of Montreal and Toronto. Some questions concerning C a l i f o r n i a t r a v e l were a lso asked i n a survey on Expo '67 made i n Ontario, Quebec and the Marit ime provinces. Data were co l lec ted on personal cha rac te r i s t i c s , reasons fo r t r a v e l l i n g , vacat ion cha rac te r i s t i cs and the ef fects of having f r iends or r e l a t i ves i n C a l i f o r n i a . In comparison to the Los Angeles Market Study, the analys is of t h i s study i s or iented more towards providing information that would be usefu l to decis ions on marketing strategy, rather than just to fo re -cast ing volume of t r a f f i c . Considerable emphasis i s placed on the re la t ionsh ips between t r ave l to C a l i f o r n i a and the market charac te r i s t i cs of income, education, occupation, and language. 4) Port of New York Author i ty i ) New York 's Domestic A i r Passenger Market: 1963-1964  i i ) New York 's Overseas A i r Passenger Market: 1963-1964 These studies are repeats of s im i la r studies conducted i n 1956-1957• They embody two general ob jec t ives : l ) to describe the market cha rac te r i s t i cs of a i r passengers and t he i r t r i p s , and 2) to determine the changes i n the composition of the market s ince 1956-1957. For the survey, a sample was taken from a l l passengers twelve years of age and older on a l l f l i g h t s , except those destined fo r Puerto ' 33 Rico , which departed from any of the three major Port Author i ty a i rpor ts during the f u l l year A p r i l 1, 1963 through March 29, 1964. F l i gh t s were chosen at random and a l l passengers over the age of twelve on the chosen f l i g h t s were given quest ionnaires. In the domestic survey 63,168 persons completed the questionnaire of which 22,263 were considered i n the ana l ys i s . In the overseas survey 43,641 persons completed the quest ion-na i re of which 27,923 were considered. In both studies data were co l lec ted on an extensive number of t r i p and market va r i ab les . The t r i p var iab les inc luded: purpose, c lass of t i c k e t , des t ina t ion , length of hau l , durat ion, time and day of departure, frequency, l o c a l o r i g i n of t r i p , and mode of ground t ranspor ta t ion. The market var iab les included education, residence, occupation, industry , age, sex, year of f i r s t f l i g h t , and income. Data were also co l lec ted on t rans fer passengers; o r i g i n , des t ina t ion , a i rpo r ts of a r r i v a l and depar-ture , and mode of i n te r -a i r po r t t ranspor ta t ion. L ike a l l the other studies considered so f a r , the analyses i n the Port Author i ty studies are tabu la r . However, w i th in the l im i t a t i ons imposed by tabu la t ion , these studies are considerably more de ta i led and reveal ing than those discussed before. This i s p a r t i c u l a r l y t rue of- the ana lys is of the domestic survey. In t h i s analys is considerat ion i s given to the re la t ionsh ip between each t r i p charac te r i s t i c and, with few exceptions, each and every other t r i p charac te r i s t i c and every market cha rac te r i s t i c . The ana lys is i s made deta i led by the s t r a t i f i c a t i o n of as many as s i x var iab les in to one tabu la t ion . I t i s the considerat ion of the i n te r re la t i onsh ips between the t r i p and market charac te r i s t i cs •which i s most important and which may provide very valuable information to the transport f i rm i n deciding i t s marketing strategy. The overseas study i s somewhat l e s s reveal ing and deta i led than the domestic study. I t considers market charac te r i s t i cs only i n t he i r re la t i onsh ip to three t r i p cha rac te r i s t i c s ; des t ina t ion , purpose, and c lass of t r a v e l . A valuable facet of both the domestic and overseas studies i s t h e i r comparison of market and t r i p i n te r re la t i onsh ips between 1956 and 1963. This comparison enables the assessment of changes over time i n the r e l a t i v e impact of the var ious t r i p and market, va r i ab les . 5) Survey Research Center, Un ivers i ty of Michigan. i ) The Travel Market: 1958. 1959-1960, 1961-1962, 1964-1965. Since 1955 the Survey Research Center of the Ins t i tu te for Soc ia l Research at the Un ivers i ty of Michigan has conducted a ser ies of seven Nat iona l Travel Market Surveys. These surveys have had numerous sponsors; large na t iona l and in te rna t iona l a i r l i n e s , a i r c r a f t manufac-tu re rs , governmental regulatory bodies, ra i lways, and magazines. Each of these surveys was s im i la r to the others i n many ways but each also sought unique informat ion. Although the stated object ives for each study were somewhat more s p e c i f i c , the general object ives were to determine and describe the ef fec ts of important market var iab les on the i n d i v i d u a l ' s frequency and type of t r a v e l , to analyse the choice between modes of t ranspor ta t ion, and to examine trends i n both of these over t ime. 35 The universe studied i n these surveys was defined as a l l fami l ies i n the cont inenta l United States excluding the i n s t i t u t i o n a l populat ion, t rans ien ts , and most m i l i t a r y personnel. In a l l the surveys the sampling technique was mult i -stage area sampling. In the f i r s t stage 66 sampling uni ts were se lected, inc lud ing 12 of the la rges t metropoli tan areas and 54 other counties or groups of counties selected at random from a l l the non-metropolitan counties i n the country. Within these primary sampling un i t s c i t i e s , towns, and open country areas were se lected, then c i t y b locks, and f i n a l l y i nd i v idua l dwel l ing u n i t s , always i n a random choice process. Thus, the sample i n each survey gave a representat ion of a cross-sect ion of pr ivate dwell ing un i t s i n the cont inenta l United Sta tes . In the surveys up to and inc lud ing that made i n 1958^  information was obtained fo r only the respondent, one to each dwel l ing un i t . However, i n the three most recent surveys, information' was co l lec ted fo r each adult i n the fami ly . In the f i r s t surveys, up to and inc lud ing that made i n 1959-1960, the sample data was co l lec ted en t i r e l y by means of personal in terv iew. However, i n the 1961-1962 survey the o r i g i n a l phase of personal interviews was fol lowed by telephone re interv iews of a random sub-sample of those 1 fami l ies which had reported ten or more t r i p s by any mode. The most recent survey, "The Travel Market: 1964-1965", was even more thorough > i n i t s data c o l l e c t i o n procedure. In addi t ion to personal interviews and telephone re interv iews, there were also reinterviews by mai l of those respondents who had no phone. The sample s i ze fo r each survey was as fo l lows : 36 Year Sample Size 1959-1960 1961-1962 1964-1965 1958 1,456 8,329 2,651 3,101 A l l the surveys co l lec ted data concerning f i v e main aspects of t r a v e l ; a t t i t udes , experience, frequency, reg ional d i f ferences, and mode choice. These aspects were considered by t o t a l t r a v e l and by mode i n terms of a large number of market and t r i p va r iab les . Some emphasis was placed on vacat ion t r a v e l and overseas t r a v e l . In add i t ion , data was co l lec ted i n one study or .another concerning time to reach the a i rpo r t , a t t i tudes towards the in t roduct ion of j e t a i r c r a f t , use of rented cars on a i r t r i p s , react ions to changes i n the p r ice of a i r t r a v e l , importance of speed, and superhighways. 1 One of the most s i gn i f i can t charac te r i s t i cs of these studies i s that they co l lec ted data not only on people who d id take t r i p s but also on people who d id not take any t r i p s at a l l , by any mode. As discussed e a r l i e r i n Chapter I I , t h i s i s one of the f l a g charac te r i s t i cs of t rue market research. Information about t r a v e l l e r s i s of much greater value when i t can be compared to information on the same var iab les about non-t r a v e l l e r s . The analyses of these surveys are comprehensive, p a r t i c u l a r l y i n the most recent s tud ies . Although the Port of New York Author i ty Studies undertook a much greater volume of tabular cross-comparisons of va r iab les , the "Michigan" Studies tend to be more purposeful i n t h e i r analyses. These studies are general ly d i rec ted towards inves t iga t ing spec i f i c 37 questions and, i n doing t h i s , tend to give more thorough treatment to these areas. Considered together, the ser ies of seven studies provides ins igh t i n to a large number of important aspects of t r a v e l behaviour. The analyses make optimum use of tabu la t ion , both i n terms of the type of re la t ionsh ips and the number of var iab les considered. Market var iab les are re la ted to t r i p var iab les and i n several cases s t r a t i f i -cat ions of several of each type of var iab le are i n te r re l a ted . The number of var iab les i n te r re la ted i n any one tab le i s su f f i c i en t to revea l usefu l d e t a i l about t r a v e l a c t i v i t y , but the number of c e l l s i s not so large as to cast any serious doubt on the s t a t i s t i c a l r e l i a b i l i t y of the r e s u l t s . In some instances graphs and histograms are employed for p i c t o r i a l presentat ion of the analys is but by nature these are r es t r i c t ed to b i va r ia te and t r i v a r i a t e i n te rac t ions . As mentioned e a r l i e r , many of the same questions were asked i n several or a l l of the surveys. This enabled considerat ion, i n l a t e r s tud ies , of changes and trends i n the r e l a t i v e importance of explanatory va r i ab les . • An in te res t ing and usefu l aspect of the analys is undertaken i n several of the studies i s the comparison of cohort age groups. As opposed to the ef fects of the aging process i t s e l f , cohort e f fects are those e f fec ts a t t r ibu tab le to the unique circumstances i n which a person obtains a given age l e v e l . The actual cohort analys is involves a compar-ison of t r a v e l behaviour and the importance of d i f fe rent explanatory var iab les at d i f fe rent periods i n time for the same age group. For example, the 25-35 age group i n the 1964 Survey would be compared to the 38 25-35 age.group i n the 1955 Survey. In these studies the only var iab le to which t h i s method of ana lys is was. ac tua l ly appl ied was a i r t r a v e l experience, but i n other works by the same authors t h i s type of analys is has a lso been appl ied to residence, income, occupation, education, and mar i ta l s ta tus .^ i i ) a) The Changing Travel Market b) Mode Choice i n I n te rc i t y T rave l : A  Mu l t i va r ia te S t a t i s t i c a l Ana lys is . These two studies, .have been l e f t to separate d iscuss ion, not because they involve separate surveys or d i f fe ren t types of data than those just reviewed, but rather because of t h e i r i r a d i c a l l y d i f fe rent a n a l y t i c a l approaches. To the knowledge of the author these studies are the only instances to date i n which mul t ivar ia te s t a t i s t i c a l techniques have been appl ied to the ana lys is of the t r a v e l market. "The Changing Travel Market" i s ac tua l l y a two-part ana lys i s . The f i r s t part of the report involves a summary of the whole body of research connected wi th the Nat iona l Travel Market Surveys discussed prev ious ly . -The methods of ana lys is are s im i la r to those used i n the o r i g i n a l s tud ies . However, i t i s the second part of the report which i s of concern here; the mul t i var ia te s t a t i s t i c a l ana lys is of t r a v e l frequency and mode choice. Mode choice i s fur ther analysed, i n some-what more depth but using the same techniques, i n "Mode Choice i n In ter -- c i t y T rave l : A Mu l t i va r ia te S t a t i s t i c a l Ana l ys i s " . Dwight M. Blood, "A Cross-Sect ion Analys is of the Domestic I n te rc i t y Travel Market"., (unpublished doctora l d i sser ta t ion i n Economics, Univers i ty , of Michigan, Ann Arbor, 1963), pp. 68-69. The data fo r these analyses was drawn from the Nat ional Travel Market Surveys. "The Changing Travel Market" made use of both the 1955 and 1961-1962 Surveys i n order to assess the extent of changes over time i n the r e l a t i ve importance of the explanatory va r iab les . "Mode Choice i n In te rc i t y Travel . . . . " r e l i e d pr imar i l y on the 1961-1962 Survey wi th some supplementary use of the 1959-1960 Survey. The spec i f i c methods of mul t i var ia te analys is used were mul t ip le regression and mul t ip le discr iminant ana lys is . Mu l t i p le regression was used to estimate the s t ruc tu ra l parameters of the factors a f fect ing demand fo r t r a v e l . The dependent var iab les were scaled by the use of mul t ip le discr iminant ana l ys i s . Mu l t ip le regression has two important advantages over tabular ana lys i s . F i r s t , i t enables the compression of a large amount of s ta t -i s t i c a l data in to a compact and meaningful form. Second, i t provides a means of reducing to a substant ia l degree the problem of m u l t i c o l l i n -ea r i t y . This i s the tendency fo r explanatory var iab les to be corre lated with one another to the point where the ef fect of one var iab le can eas i l y become confused with the ef fect of another va r i ab le . In reference to "The Changing Travel Market" the authors point out: Throughout (the f i r s t part of) t h i s volume many factors taken separately have been shown to have an ef fect on t rave l d e c i -s ions . These fac tors r e a l l y operate simultaneously. Whether a person decides to take a t r i p depends on a wide var ie ty of personal cha rac te r i s t i cs of the t r a v e l l e r at a given moment i n t ime. Each of these charac te r i s t i cs (income, education, and the l i k e ) can be shown to have a discernable impact on t r a v e l . Yet , i f t r a v e l were predicted by adding up the resu l t s obtained by re la t i ng each separate var iab le to t r a v e l , the t o t a l would l i k e l y be overstated, or ce r ta in ly misstated. What' we attempt to do ( in the second part of the study) i s to show whether each characteristic (or independent variable) that i s used to explain travel has any remaining impact on travel (which i n this case i s the dependent variable) after the effects of a l l the other independent variables simultaneously have been taken into account.10 Due to the qualitative nature of the data, several modifications had to be made to the normal application of the multiple regression technique; the use of dichotomous dummy variables, multiple discriminant analysis, and multiple classification analysis. These s t a t i s t i c a l tools w i l l be described i n some detail i n Chapter V and the Appendices. The variables considered i n the regression equations were those which had been indicated.by the tabular analyses of the original surveys to be the most important. In the case of the travel frequency regres-sions these were: experience, sex, residence, income, occupation, education, and auto ownership, with l i f e cycle stage as a stratification variable. In the mode choice regressions four general areas were considered; financial considerations, mode avai l a b i l i t y , attitudes and preferences, and t r i p distance, each containing several specific variables. Summary and Conclusions It has been the purpose of this chapter to examine the "state of the art" of travel market research so as to provide a point of departure for an analysis of the Air Canada Vancouver Survey. In reviewing past travel market studies, i t has become apparent John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1964), p. 161. that t h e i r number i s over - ind ica t ive of the true market research which has been car r ied out. A large number of studies involve merely a numerical count of the volume of passengers t r a v e l l i n g between cer ta in centers, with no considerat ion of the nature of the passengers or of the t r i p s . These are proper ly termed o r i g i n and dest inat ion s tud ies. These studies contr ibute very l i t t l e to the f i r m ' s understanding of the market beyond an i nd i ca t i on of i t s s ize and dispersement. Other studies involve a macro-approach, re la t i ng the changes i n the general s o c i a l , economic and demographic cha rac te r i s t i cs of the country to changes i n the t o t a l volume of t r a v e l . Although t h i s kind of ana lys is i s basic to the f i r m ' s demand forecasts , p a r t i c u l a r l y i t s long run fo recas ts , i t does not provide information that could be useful i n deciding such important parts of marketing strategy as adver t is ing appeals, s e l l i n g approaches, p r i c i n g , and "product" development. Even studies which do examine t r i p charac te r i s t i cs and market cha rac te r i s t i cs often provide l i t t l e usefu l informat ion. Their shor t -comings a r i se not from the nature of the data on which they are based or the r e l i a b i l i t y of the survey sample, but rather from the l im i ted extent of t h e i r analyses. Many studies do not go beyond simple b ivar ia te l i s t i n g s of each separate t r i p and market va r i ab le , with number or percentage of respondents as the quant i fy ing va r iab le . This type of ana lys is provides only a loose general character sketch of the ent i re survey sample. The market i s described i n terms of the proport ion of the t o t a l respondents exh ib i t ing each cha rac te r i s t i c , with no i nd i ca t i on of the degree of concurrence between any of the cha rac te r i s t i c s . Other 42 studies do consider the re la t ionsh ips between two or more qua l i t a t i ve va r iab les , but r e s t r i c t t he i r a t tent ion to the in te r re la t ionsh ips within, the two basic spheres of informat ion; market cha rac te r i s t i cs and t r i p cha rac te r i s t i c s . However, t h i s type of analys is does provide some improvement over the type jus t mentioned. To the extent of the number of var iab les i n te r re la ted i n one tabu la t ion , there i s an ind ica t ion of the concurrence of one charac te r i s t i c with another. This type of analy-s i s , however, does not give any i nd i ca t i on of the di f ferences i n the charac te r i s t i cs of the t r i p s taken by the d i f fe rent segments of the market. Only a few studies have ventured to consider the in te r re la t ionsh ips between t r i p var iab les and market va r i ab les . This i s the primary ingredient i n t rue market research. By matching t r i p charac te r i s t i cs to market cha rac te r i s t i cs the f i rm can get some idea of the type of serv ice , adver t is ing appeal, p r i c ing po l i c y , et cetera that are most e f fec t ive for pa r t i cu la r segments of the market, as defined by such charac te r i s t i cs as l i f e cyc le , education, income, e tc . Another important shortcoming common to much of- the current t r a v e l research i s i t s confinement to the considerat ion of only those persons who are already t r a v e l l e r s . Firms undertaking studies such as these would appear to be making the assumption that t he i r best l i n e of expansion i s i n get t ing present customers to t r a v e l more. In ac tua l fact i t may be that those who do not t r a v e l at present provide the greatest po ten t ia l fo r expansion. At any rate i t would be desi rable fo r the f i rm to use some sort of empir ica l research to determine the degree to which the non- t rave l le r provides a po ten t ia l market and the type of marketing strategy to which he might be sens i t i ve . The most advanced studies of the t r a v e l market have been under-taken by the Survey Research Center of the Un ivers i ty of Michigan. Of pa r t i cu la r value are the two studies invo lv ing the use of mul t ip le regression and mul t ip le discr iminant ana lys i s . These techniques have two advantages over the commonly used tabular methods of ana lys is ; the compression of a large volume of data in to a compact and useable form, and the reduct ion of the d i s to r t i ng ef fects of multico.11 i nea r i t y . To conclude, current market research i s character ized by weak ana lys i s . Considerable at tent ion has been, paid to the nature of the data co l lec ted and the randomness and adequacy of the survey samples, but these strengths tend, wi th some noteable exceptions, to be over-r idden by analyses which are s u p e r f i c i a l and undirected. To provide resu l t s which w i l l be of use to the f i rm i n determining i t s marketing strategy, t r a v e l market research must be concerned with the purposeful i n t e r r e l a t i o n of market var iab les with t r i p va r iab les . CHAPTER IV THE AIR CANADA VANCOUVER SURVEY: BACKGROUND The Objectives of The Survey The A i r Canada Vancouver Survey was undertaken i n the F a l l of 1 9 6 6 . Personnel of the Market Research Department of A i r Canada were ass is ted by a group of students from the Univers i ty of B r i t i s h Columbia and a pr iva te consul t ing f i rm based i n Vancouver. In i t s en t i re ty , the survey sought information on both personal t r a v e l and business t r a v e l . However, as mentioned e a r l i e r , t h i s present study i s concerned only with the analys is of the personal t r ave l port ion of the survey. The basic object ive of the personal t r a v e l survey was to determine the s o c i a l and economic charac te r i s t i cs of people l i v i n g i n Vancouver i n r e l a t i o n to t he i r t r a v e l . Essen t i a l l y i t was to provide a survey p r o f i l e inventory of t r a v e l cha rac te r i s t i c s ; where people t r a v e l and why they t r a v e l , or why they don't t r a v e l . An important fac tor was that the survey would consider not only the t r a v e l l e r but also the non- t rave l le r , an area commonly neglected i n past s tud ies . Within the general object ive of obtaining a p r o f i l e of t r a v e l l e r and non- t rave l le r cha rac te r i s t i c s , i t was the purpose of the survey to fu rn ish data that could be analysed to provide a complete range of information usefu l to A i r Canada i n deciding i t s marketing strategy and expanding i t s markets. A major aim was to invest igate the factors in f luenc ing mode choice. Other areas queried were: the impact of area of residence on t r a v e l , the in f luence of Expo on European t r a v e l , and fac tors a f fec t ing the choice between t r a v e l agency and t ransportat ion company fo r t i c k e t purchase. A i r Canada's in te res t i n t h i s survey derived from two sources; the corporate l e v e l and the l o c a l sa les force l e v e l . At the corporate l e v e l , i t was important that the survey ass i s t i n the long run develop-ment of the Vancouver personal t r a v e l market and that i t provide information useful- to the company's long run or ien ta t ion of . i ts market research a c t i v i t i e s . At the l o c a l sales force l e v e l , i t was important that the survey i d e n t i f y the immediate Vancouver personal t r a v e l market and that i t i den t i f y 1 t he immediate aspects of a i r l i n e serv ice that i n f l u -ence the market. Although concerned wi th a l l modes of t r a v e l , some emphasis was to be placed on a i r t r a v e l . C e l l The d i v i s i o n of the- survey area in to geographic c e l l s was a basic feature of the survey. This was done i n order to determine whether va r ia t i ons ex is t i n the t r a v e l behavior of people who have s im i la r s o c i a l and economic charac te r i s t i cs but who choose to l i v e i n d i f fe rent geogra-phic areas of the c i t y . The c e l l s are essen t i a l l y combinations of census t r a c t s . The composition of each c e l l was establ ished on the bas is of s i x c r i t e r i a : 1) Average fami ly income on an annual bas i s . 2) Percentage of the male labour force engaged i n managerial or pro fess iona l occupations. 3) Percentage of the adult population not present ly attending school who completed elementary school standing only, i . e . who d id not graduate from secondary school . 4) Percentage of the adult population which immigrated to Canada i n the period 1946-1961. 5) Percentage of t o t a l population 14 years or younger. 6) Percentage of one person dwell ings i n the t o t a l population of dwel l ings. The o r i g i n a l breakdown involved 26 c e l l s , but, due to budget l im i t a t i ons and a spec i f i ed requirement that each c e l l contain a minimum of 240 sample observat ions, t h i s was reduced to 17. The boundaries of the c e l l s are shown i n Figure 1. I t w i l l be noted that , with the excep-t i o n of c e l l s 7» 11 and 13, a l l c e l l s are made up of geographical ly contiguous census t r a c t s . The Sample The universe to be studied was metropoli tan Vancouver. Within t h i s , the target was a l l adu l ts , regardless of whether they had ever t r ave l l ed and regardless of any other of t h e i r cha rac te r i s t i c s . At an intermediary stage i n the survey i t was found necessary to el iminate considerat ion of the skid-row d i s t r i c t of the downtown area due to d i f f i c u l t i e s i n in terv iewing. The sample un i ts were selected on a random basis from the Vancou-ver c i t y d i rec to ry . A pre-set budget d ic ta ted the sample s i z e . The f i n a l sample was comprised of a primary sample and a supplementary sample. The primary sample involved 2,981 observations chosen at random from the c i t y as a whole. A supplementary sample was necessi tated by M A R K E T C E L L S MUNICIPALITY OF NORTH VANCOUVER CELL BOUNDARIES 47 the fact that the primary sample resu l ted i n several c e l l s having l e s s than the 240 observations spec i f ied as being the minimum adequate fo r a c e l l ana lys i s . This supplementary sample involved a fur ther random se lec t ion from the de f ic ien t c e l l s . The t o t a l sample fo r the ent i re personal t r a v e l survey comprised 4,158 observations. Method of Data Co l lec t i on The data was co l lec ted by means of personal interviews at the homes of the respondents. Only one adult was questioned i n each sample household. The pa r t i cu la r adult questioned was determined on the bas is of random se lec t ion from a l l the persons over 18 i n the fami ly . To minimize the chances of selected respondents not being home, a l l i n i t i a l c a l l s were made i n the evenings or on Saturdays. The Questionnaire The questionnaire s o l i c i t e d 117 items of informat ion; 73 on t r a v e l cha rac te r i s t i cs and 44 on market c h a r a c t e r i s t i c s . The, t r ave l questions were aimed at the respondents' frequency of t r a v e l by the d i f fe rent modes and at spec i f i c cha rac te r i s t i cs of h i s most recent and next to l a s t personal t r i p s . In addi t ion to information on the t im ing, durat ion, expenditure, mode, dest inat ion and purpose of the l a s t two t r i p s , questions were also d i rected towards the spec i f i c a i r l i n e used on a i r t r i p s , the exact rout ing of t r i p s to Europe ( inc luding the name of each ca r r i e r invo lved) , the mot ivat ional factors in f luenc ing the choice of mode, passenger assessment of good and bad points of the c a r r i e r ' s se rv ice , the s i ze and composition of the t r ave l par ty, and the t r a v e l l e r ' s behavior i n making h i s a i r l i n e reservat ions. The market cha rac te r i s t i cs questioned i n the survey include monetary income; ownership of major assets such as dwel l ing, auto, cottage, and boat; occupation; education; fami ly l i f e cyc le stage; and pred ispos i t ion to t r a v e l as ind ica ted by being born, having ch i ldren or parents, or having gone to school i n places outside B r i t i s h Columbia. The Analys is The next chapter w i l l ou t l ine the methods of analys is appl ied to the survey data. The fo l lowing three chapters w i l l present the resu l t s obtained from these analyses. CHAPTER V RESEARCH PROCEDURES The main research methods employed i n the analys is of the A i r Canada Vancouver Survey are mul t ip le least squares regression, mul t ip le c l a s s i f i c a t i o n ana lys is , and ana lys is of var iance. Mu l t ip le regression estimates the s t ruc tu ra l parameters for factors a f fec t ing demand fo r t r a v e l . Mu l t i p le c l a s s i f i c a t i o n converts these parameters in to more meaningful form. Variance ana lys is ind icates the contr ibut ion to the explanation of the t o t a l var iance made by the var iab les used as the basis fo r s t r a t i f i c a t i o n . In t he i r basic mathematical form these methods are not d i r e c t l y appl icable to the data co l lec ted i n the survey. This i s due pr imar i l y to the fact that much of the data are qua l i t a t i ve rather than quant i ta t i ve . Thus, i t i s necessary to make several modi f icat ions to the basic procedures. These w i l l be described i n t h i s chapter. Mu l t ip le Least Squares Regression The advantages of mul t ip le regression ana lys is were explained i n the l a s t chapter. I t w i l l be su f f i ce to say here that the a b i l i t y of t h i s method of analys is to compress a large amount of data in to a compact and useable form and to great ly reduce the d i s to r t i ng ef fects of mu l t i -c o l l i n e a r i t y make i t s use very des i rab le . I t s sca rc i t y of usage i n the past would seem to be a t t r ibu tab le to the complexity of i t s app l icat ion and consequent ial ly, i t s cos t , and also the d i f f i c u l t y of i t s i n te rp re -ta t i on by those, without some s t a t i s t i c a l t r a i n i n g . 50 The decision as to the factors to include as independent variables i n the regression equations was based on several c r i t e r i a . The choice of variables was naturally constrained by the type of data collected i n the survey. Within this constraint the f i r s t step involved the selection of those variables which demonstrated s t a t i s t i c a l significance or explana-tory importance in,previous market studies. To this basis were added any variables which intuition suggested might offer some contribution to the explanation of the t o t a l variance. Further, several variables were included whose effect was not even suggested by intuition, but about which i t was f e l t that indication of even a neutral effect would be valuable information. Care was taken, through an examination of partial correlation coefficients to exclude any variable which obviously did not have an explanatory effect of i t s own but which rather just reflected the effect of another variable with which i t was highly intercorrelated. Inclusion of such a variable would have a negligeably small or even negative effect on the value of the coefficient of multiple correlation. Although the direct effect of any independent variable on the multiple correlation coefficient may at worst be zero, i t s indirect effect may be positive or negative depending on whether the variable acts i n conjunction with each of the other independent variables to increase or decrease the overall relationship.^ - ' Care was also taken to keep the number of independent variables Ttobert Ferber, S t a t i s t i c a l Techniques i n Market Research (New York: McGraw-Hill, 1949), p. 363. 51 i n the regression equations within reasonable limits. There are two important reasons for this. F i r s t , the number of subdivisions which can be made i n any numerical analysis while maintaining an acceptable degree of s t a t i s t i c a l significance depends upon the number of observations i n the sample. It i s not the sample size i t s e l f which i s of greatest impor-tance but rather the number of observations i n each data c e l l which i s to p be attributed with s t a t i s t i c a l meaning or s t a t i s t i c a l l y manipulated. The second factor limiting the number of independent variables i s the loss of mathematical accuracy from rounding-off by the computer during the regression matrix inversions. Although the amount of accuracy lost during any one matrix inversion w i l l be negligeable, the compounded effect of a large number of inversions may be quite significant. The expansion of the equation to include a few marginally important independent v a r i -ables i s l i k e l y to reduce the mathematical accuracy of a l l the regression coefficients with l i t t l e compensating benefit. A complete l i s t of the independent variables considered i n any or a l l of the regression equations i s set out In Appendix A. It w i l l be noted that each classification of each t r i p or market characteristic i s considered as a separate variable and assigned a numerical range of value. These peculiarities are the result of several modifications which had to be made to the data to put i t i n a form suitable for the use of multiple least squares regression. J. Johnston, Econometric Methods (New York: McGraw-Hill, I960), p. 227. Bas i ca l l y the modi f icat ion involved the set t ing of numerical scales fo r qua l i t a t i ve data. While t h i s numerical transformation was necessary fo r the app l ica t ion of the regression method i t was also important that the scale detract as l i t t l e as possib le from the bas i ca l l y non-numerical nature of the va r i ab les . The method of numerical transformation involved the use of dummy va r i ab les . Each t r i p and market var iab le was div ided in to mutually exclusive c lasses . Then each c lass was given a dichotomous scale such that a respondent was assigned a value of one i f he possessed the p a r t i c -u l a r cha rac te r i s t i c or took a t r i p with the pa r t i cu la r a t t r ibu te and a value of"zero otherwise. These dummy var iab les were used even i n the case of var iab les already convent ional ly sca led. This has the advantage that i t necessi tates no r e s t r i c t i v e assumptions about the l i n e a r i t y of the e f f ec t . I t w i l l be noted from Appendix A that there i s one exception to the use of dichotomous dummy var iab les fo r the measurement of the inde-pendent va r i ab les . T h i s . i s i n the case of the var iab le speci fy ing the number of people who went on a t r i p , which i s considered i n the mode choice equation invo lv ing auto and a i r . This var iab le i s considered as an ind ica to r of r e l a t i v e p r i c e , based on ,the assumption that the t o t a l cost of t r a v e l by common ca r r ie r , increases i n proport ion to the number who t r a v e l i n a party whereas the marginal cost fo r add i t iona l passengers i n an auto i s negl igeable up to a party of f i ve or s i x . Under t h i s assumption the r e l a t i v e cost of common ca r r i e r and auto -.changes wi th the s i z e of the party i n proport ion to the rec ip roca l of 53 3 the number i n the party. A further modification to the use of linear multiple regression i s necessary. Where the population i s divided up into several groups on the basis of a given characteristic i t i s necessary to leave one of the dummy variables out of the computations i n order to achieve a determinate solu-tion. The reasons for this modification can be shown by reference to the normal regression model: Y = aX+b 1 1I 1+b 1 2I 2 + b 1 3 I 3 + C1 + u (l) where 1]_, I 2 and I3 are three exhaustive income groups; under $5,000, $5,000-10,000, and over $10,000. It i s immediately clear that the least squares estimates of Cj_ and b ^ are indeterminate. This i s indicated by the fact that for each unit i n the sample only one of the I i w i l l have the value of 1 with the others being equal to 0. Therefore, the addition of any arbitrary number to each of the b^ simultaneously subtracted from Ci w i l l not affect the value of Y. This indeterminacy may also be con-sidered i n another way.. There i s a perfect linear multiple correlation among the I i and any attempt to estimate the regression coefficients w i l l f a i l because of singularity i n the moments matrix. Thus, to make the regression coefficients determinate an additional constraint must be imposed upon the equation. There are two basic ways of doing t h i s . ^ John B. Lansing and Dwight M. Blood, Mode Choice i n Intercity  Travel: A Multivariate S t a t i s t i c a l Analysis, report to the Under Secretary for Transportation, U.S. Department of Commerce (Ann Arbor, Michigan: Survey Research Center, 1964), p. 7. ^Daniel B. Suits, "Use of Dummy Variables i n Regression Equations", Journal of the American S t a t i s t i c a l Association, Vol. 52, December, 1958, p. 549. 54 One of the ways of making the regression coe f f i c ien ts determinate i s to set C]_=Q which w i l l convert the normal regression equation in to the homogeneous form: Y = aX + bgjI-L + b 2 2 I 2 + b 2 3 I 3 + u (2) This method has two advantages.1 F i r s t , the fac t that the normal equations used i n the estimation of regression coe f f i c i en ts of a homogeneous form involve moments around zero rather than around means, saves a step i n computation. Second, since 1^1^ = 0 fo r a l l the i ^ j , there i s a diagonal matr ix of moments between the dummy va r iab les . This diagonal matr ix can then be set at the top of the D o o l i t t l e format, fur ther reducing ca l cu la t i on . An a l te rnat ive way of making the regression coe f f i c ien ts determi-nate i s to set any one of the b]j_ = 0. This w i l l y i e l d an ordinary non-homogeneous regression equation i n which one of the Ij_ i s not included as an independent va r i ab le . Y = aX + b 3 1 I 1 +.-b32I2 + c3 + u (3) The removal of I3 does not reduce the amount of information considered i n the analysis'"as i t s value i s determinable from I]_ and I 2. Both of these methods of making the regression equation determinate y i e l d the same estimates of Y . While the d i rec t in te rpre ta t ion of the . two methods d i f f e r , the regression estimates from one are eas i l y derived from those obtained i n the other. The b ^ of equation (2) measure income inf luences as deviat ions from zero and t h i s equation can be considered as a l i nea r regression of Y on X, the in tercept of which ( b ^ ) var ies between each income c l a s s . By adding -b 23 to each b ^ of equation (2) and 4- b23 55 to the zero constant term of equation (2) equation (3) i s obtained. The second of the above methods i s used i n t h i s study. I t w i l l be noted that i n Appendix A one category of e a c h ' c l a s s i f i e d var iab le i s not shown. Scal ing the Dependent Var iab le The dependent var iab le i n mul t ip le regressions considering cross sect ion data i s usual ly e i ther spec i f i ed i n terms of the o r i g i n a l data or scaled dichotomously. However, both of these, methods involve serious shortcomings. Spec i f i ca t i on of a dummy va r i ab le , such as t r i p frequency, i n terms of the o r i g i n a l data has two shortcomings. F i r s t , i t al lows extremes i n the data, such as someone tak ing 100 t r i p s a year, to d i s to r t the resu l t s of the ana l ys i s . Second, i t does not face the problem of response error which usual ly takes the form of over- repor t ing. Dichotomous sca l ing of the dependent var iab le provides computed p robab i l i t i e s for i nd i v i dua l cases which are usefu l f o r c l ass i f y i ng these cases in to one of two c lasses . However, t h i s approach provides only an est imation of whether a pa r t i cu la r observation belongs i n a pa r t i cu la r c lass wi th no i nd i ca t i on whatsoever as to the extent of pa r t i c i pa t i on i n that c l a s s . This method f a i l s to d i s t i ngu ish between what have been l a b e l l e d the "Threshold" and the "Satura t ion" ba r r i e r s . ^ The Threshold ba r r i e r i s the ba r r i e r between tak ing one t r i p and taking no t r i p s at a l l . -^Dwight M. Blood; "A Cross Sect ion Analys is of The Domestic I n te rc i t y Travel Market";(unpublished doctoral d i sser ta t ion i n economics, Un ivers i ty of Michigan, Ann Arbor, 1 9 6 3 i p. 3 0 . 56 The Saturation barrier i s the barrier involved i n taking an additional t r i p after already having taken many t r i p s . A dichotomous dependent variable would consider only the threshold effect, treating the person who takes only one t r i p the same as the. person who takes many tr i p s . The method of scaling the t r i p frequency dependent variable used very successfully i n the Michigan Studies was multiple discriminant analysis.^* The use of this multivariate s t a t i s t i c a l tool enabled the formulation of a scaling for the dependent variable which simultaneously included elements of the Threshold and Saturation effects and yet dampened down the distorting effects of extremes i n the data. . The multiple discriminant method and i t s application to travel market research are discussed i n more detail i n Appendix E. It was the original intention of this study to undertake i t s own multiple discriminant analysis to formulate scales for i t s dependent variable which would directly reflect the data collected i n the Vancouver Survey. However, the unavailability of either a systematically written computer program for multiple discriminant analysis, or the time and expenditure required to write such a program, necessitated reliance on the results obtained by the Michigan Study. For several reasons this would not appear l i k e l y to distort the results of this present study. Fi r s t , the data upon which this Michigan Study i s based are comparable i n nature to that considered i n this study. Many of the survey questions asked i n the Air Canada Vancouver Survey were the same as those i n the John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1964), p. 163. 57 Michigan Study questionnaire. Second, i t has been shown that the regression coefficients calcu-lated under different scalings of the dependent variable, but based on the same' data, are different i n magnitude but not i n relative value. This i s illustrated i n Table I, which compares the regression results obtained by the University of Michigan i n i t s 1962 study using the scales developed from the 1962 data with the results for the same study using the scales developed from i t s 1955 data. It w i l l be noted that the regression coefficients for the equation based on the 1955 scale are a l l slightly higher than those for the equation based on the 1962 scale and that the relative importance of the different variables i s the same in both cases. As the purpose of using multiple discriminant analysis i s to remove from the data the distorting effects of extreme values and over-reporting, the scale values developed for the Michigan Study w i l l be adequate. Third, there i s a positive value i n using the scales employed by the Michigan Study. Although not a direct purpose of this present study, the use of the same scale values w i l l contribute to the comparability of the results. In the mode choice regressions the dependent variables are zero-one dichotomies. This means that only two modes are considered i n the same regression equation, suggesting that the persons considered i n these equations only had a two-way choice. In reality, and probably i n the majority of cases, people may have a choice between more than two modes of travel. However, a test made by the Survey Research Center of the 58 TABLE I A COMPARISON OF REGRESSION EQUATIONS FOR BUSINESS AUTOMOBILE TRAVEL FOR THE YEAR 1962, FOR MARRIED ADULTS WITH CHILDREN, USING DISCRIMINANT FUNCTION SCALES FOR 1955 AND 1962 Equations Based on Equations Based on 1955 Scale 1962 Scale Variable Coefficient Standard Coefficient Standard Error Error C .1152 .0901 EXP .0901 .0256 .0649 .0188 SEX -.1602 .0248 -.1170 .0182 METRO -.2131 .0255 -.1551 .0187 Y l .0707 .0399 .0511 .0293 Y 2 .0753 .04L2 .0452 .0302 Y3 .1507 .0466 .1008 .0342 occ .4108 .0313 .3200 .0230 EDUC .0423 .0256 .0346 .0188 AUTO .0265 .0422 .0229 .0310 R2 .1855 .1948 N 2451 2451 Source: Dwight M. Blood, A Cross-rSection Analysis of the Domestic  Intercity Travel Market.(unpublished doctoral dissertation i n Economics, University of Michigan, Ann Arbor, 1963), p. 36. Univers i ty of Michigan showed that the simultaneous considerat ion of choice between a l l modes ac tua l l y makes "no substantive contr ibut ion to the ana l ys i s " over a paired comparison of auto and common c a r r i e r . ^ In t h i s tes t case mul t ip le discr iminant ana lys is was appl ied to four modes and f i f t e e n independent va r i ab les . This resu l ted i n the f o l -lowing scale fo r the dependent va r i ab le . Mode Scale Pos i t i on Auto 0.00 R a i l 1.00 Bus 1.15 A i r 1.15 With the scale values fo r r a i l , bus and a i r so c lose together, there i s no e f fec t i ve d i f ference between these resu l t s and a zero-one dichotomous comparison of auto and common c a r r i e r . A very advantageous feature of def in ing the dependent var iab les of mode choice regression i n terms of zero-one dichotomies i s the fact that t h i s est imation procedure i s equivalent to that of assigning l i nea r p robab i l i t i e s to observations by means of a l i n e a r discr iminant funct ion. This w i l l be deal t wi th i n d e t a i l during the fo l lowing d iscussion of mul t ip le c l a s s i f i c a t i o n ana l ys i s . Mu l t i p le C l a s s i f i c a t i o n Analys is Mu l t i p le c l a s s i f i c a t i o n ana lys is provides an a l te rna t ive format fo r the presentat ion of r esu l t s from mul t ip le regressions which are based 7 John B. Lansing and Dwight M. Blood, Mode Choice of I n te rc i t y  Trave l : A Mu l t i va r ia te S t a t i s t i c a l Ana lys is , report to the Under Secretary f o r Transportat ion, U.S. Department of Commerce (Ann Arbor, Michigan: Survey Research Center, 1964), pp. 39-40. 60 en t i r e l y on dummy va r i ab les . I t involves taking the mean of the depen-dent var iab le as the f i r s t approximation of the magnitude which i s to be estimated (e .g . frequency of t r a v e l or p robab i l i t y of choosing one of two modes) and adjust ing t h i s value according to whether the pa r t i cu la r observation being considered belongs to a pa r t i cu la r group or possesses a pa r t i cu l a r cha rac te r i s t i c . The advantage of using mul t ip le c l a s s i f i c a t i o n ana lys is i s p a r t i c -u l a r l y evident i n the case of a t t r ibu tes which are d iv ided in to a number of dummy va r iab les ; e .g . income. In such cases the ef fect of being assigned to pa r t i cu la r income groups upon the frequency of t r a v e l or choice between two modes i s represented d i r e c t l y by pos i t i ve and negative deviat ions from the mean of the dependent va r i ab le . A fur ther advantage to the use of mul t ip le c l a s s i f i c a t i o n analys is revolves around the fact that the constant term i n a regression equation based en t i re l y on dummy var iab les can no longer be considered as an in te rcep t . The constant term i n such an equation r e f l e c t s the trends i n the e f fec ts of a l l the dummy var iab les for which the regression c o e f f i -c ients have been constrained to zero as we l l as the trends i n the mean. Thus, i t i s of greater relevancy to consider changes i n the mean of the dependent var iab le and the ef fect on t h i s value that resu l t s from being c l a s s i f i e d in to var ious populat ion sub-classes than i t i s to consider changes i n the constant term.** To i l l u s t r a t e how mul t ip le c l a s s i f i c a t i o n ana lys is converts 8 j o n n Lansing and Dwight M. Blood, The Changing Travel Market. op. c i t . , p. 165-166. regression coefficients into deviations about the mean of the dependent variable, consider the following example.^ The equation resulting from the application of the standard method of multiple least squares regression i s as follows: Y = .03 + .041! + . 1 2 I 2 + .30I 3 + ,15E + .270! + .llOg + .020^.(1)' Y (mean of the dependent variable) = .25 where: Yj_ = 1 i f .jth adult has travel experience. = 0 i f j_th adult has no travel experience. I i i = 1 i f j t h adult has income class 1 ($3,000-$5,999). = 0 i f otherwise. ±2i = 1 i f i t h adult has income class 2 ($5,000-$9,999). = 0 i f otherwise. I 3 i = 1 i f i t h adult has income class 3 ($10,000 or more). = 0 i f otherwise. E i = 1 i f j_th adult i s high school graduate. = 0 i f otherwise, and so on for occupation class 0 i . It w i l l be noted that there i s no variable for one income class (under $3,000) or for one education class (non-high school graduate). As explained earlier this i s because one categorical classification of each market or. t r i p variable must be constrained to zero to achieve a 9 J.B. Lansing & W. Ladd; "An Example of the Conversion of Regres-sion Coefficients into Deviations about the Grand Mean", (unpublished paper to the Seminar i n Quantitative Economics, University of Michigan, Ann Arbor, October, 1962). determinate solution to the regression equation. To convert the regression equation into a form i n which the con-stant w i l l be replaced by the mean of the dependent variable and coef-f i c i e n t s w i l l be specified f o r the variable classes constrained to zero, two properties w i l l be considered. F i r s t , f o r any t r i p or market varir-able such as income, the sum of the deviations about the mean of the dependent variable weighted by the number of observations must equal zero. Second, the transformation must not change the difference between the regression c o e f f i c i e n t s f o r the different income classes. For example, i f the regression c o e f f i c i e n t f o r income class 2 i s .08 higher than that f o r income class 1 before the transformation, the difference must s t i l l be .08 after the transformation. Now, l e t t i n g Ijt represent the deviation from the mean of the dependent variable for the missing income class and considering the following income d i s t r i b u t i o n : Income Class Dummy Variables (which equal 1 for those i n t h i s class) Observations i n Each Class Under $3,000 341 $3,000-5,999 966 $6,000-9,999 407 $10,000 and over *3 120 1,834 the two properties may be expressed symbollically as: * + Nplp* + Nolo* + N, I . * =0 (2) and !Lj = I J + .04 (3a) 63 I* = I* + .12 (3b) 2 .4 I*. = I* + .30 (3c) .3 4 Then substituting i n (2) 966 (I* + .04) + 407 (I* + .12) + 120 (I* + .30) + 341 I* = 0 4 4 4 4 Simplifying 1834 I-x- + 966 (.04) + 467 (.2) + 120 (.3) = 0 4 I* = -123.4 4 1834 I* = -.07 4 Since I-x- = I-x- + .04, an estimate of Ix- can be obtained. Similarly for 4 . 1 I-x- and I-H-. Applying the same procedure to the education and occupation groups y ie lds E g = -.07 and 0* = - .09. Calculating a l l the adjusted deviation values, the desired mult i -ple c lass i f icat ion equation i s : Y = .26 - .07Ij - .031! + , 0 5 I 2 + . 2 3 ± 3 + .OSE^ - .07Eg + .180 1 + ,020 2 - .970 3 - .090-x- (4) In summary th i s equation has three characteristics which d i s t i n -guish i t from the normal regression equations containing dummy variables: 1) The constant term i s the mean of the dependent variable. 2) Each regression coefficient i s i t s e l f an adjusted deviation from the mean of the dependent variable. 3) There i s an adjusted deviation value for each class of each t r i p and market variable . The interpretation of the multiple c lass i f icat ion equation i s straight forward. Further consideration of the above example w i l l serve to i l l u s t r a t e . To f a c i l i t a t e this consideration, a complete l i s t i n g of 6 4 the independent variables and their corresponding deviation values i s set out below. Independent Variable Income: Under $3,000 $3,000-5,999 $6,000-9,999 $10,000 and over Education: High School Graduate Non-high School Graduate Occupation: Professional, managerial Clerical, sales, blue collar Farm A l l Others No. of cases 3 4 1 966 4 0 7 1 2 0 Deviations from mean of . 2 6 - . 0 7 -.03 .05 .23 1,834 884 950 1,834 258 838 67 671 .08 - . 0 7 .18 .02 -.07 -.09 1,834 The parameter which i s to be estimated i s the probability that a given individual i s an experienced traveller. If nothing at a l l was known about an adult, consideration would be given to only the mean o f the dependent variable, suggesting a . 2 6 probability of travel experience. If there was more information about the particular adult, the mean of the dependent variable would be adjusted by the deviation value representing each of his known characteristics. For example, i f i t was known that the adult earned income of $4,000 per year, was a high school graduate and was employed as a sales clerk, the equation would adjust the mean estimate of . 2 6 by +.07 (-.03 + .08 + .02). It would then be suggested that the adult had a .33 chance of being an experienced traveller. Another way of putting this would be to say that, of a l l adults with the same or s im i l a r cha rac te r i s t i c s , one t h i r d have t r a v e l experience. The only d i f ference between the above example and in terpre ta t ion of the mode choice and t r a v e l frequency equations i n t h i s study i s i n the meaning of the estimates of the dependent va r i ab les . In the mode choice equations the value of the dependent var iab le ind ica tes the p robab i l i t y that the i nd i v i dua l under considerat ion w i l l choose to t r a v e l by the mode designated with the scale pos i t ion 1 rather than the mode wi th the scale pos i t ion 0. In the t r a v e l frequency equations, the dependent var iab les are not i n terms of p robab i l i t i e s at a l l , but rather i n terms of numerical estimates of the number of t r i p s taken per year . In teract ion E f fec ts The most important l i m i t a t i o n to mul t ip le regression and c l a s s i -f i c a t i o n ana lys is i s the assumption that each explanatory var iab le a f fec ts the dependent var iab le i n an independent manner, regardless of the values of the other explanatory va r iab les . I t i s assumed that a cer ta in l e v e l of education w i l l have the same ef fect on an i n d i v i d u a l ' s t r a v e l frequency no matter what h i s income l e v e l , occupation, sex, e tc . Referr ing once again to the example considered prev ious ly , the in te rp re -ta t i on of the equation i s to add to the p robab i l i t y of t r ave l experience an adjusted deviat ion f igure of .08 i f a person i s a high school graduate, no matter what other adjustments have been made because of h i s other cha rac te r i s t i c s . This assumption of a d d i t i v i t y i s only an approximation of r e a l i t y . I t ignores the p o s s i b i l i t y of i n te rac t ion ef fects between 66 va r i ab les . I t may be that the combination of pa r t i cu la r values of par-t i c u l a r independent var iab les have an ef fect on the dependent var iab le that i s considerably more (or l ess ) than the simple addi t ion of the e f fec ts that these fac tors exhib i t when they occur alone. This w i l l be i l l u s t r a t e d i n the next chapter by the fac t that i t i s poss ib le to combine scores on d i f fe rent independent var iab les i n such a way as to lead to a predicted frequency of l ess than zero t r i p s per year or a predicted p robab i l i t y of t r a v e l by a pa r t i cu la r mode of la rger than 1 or l e s s than 0. The basic method of encorporating in te rac t ion e f fec ts i n regres-s ion equations involves the inc lus ion of an add i t iona l va r iab le for each poss ib le combination of the separate independent va r i ab les . However, i n the case of t h i s study, where the regression equations involve as many as f i f t y separate va r iab les , the app l i ca t ion of t h i s method would be untenable. The number of var iab les i n such an expanded equation would fa r surpass the number of observations i n the sample. By way of example, reference can be made to the f i r s t regression equation considered i n the next chapter. I t involves 49 separate dummy var iab les which, i f f u l l account were taken fo r i n te rac t i on , would be expanded to 1,566,720. With only 313 observations fo r t h i s regress ion, i t can eas i l y be seen that the vast major i ty of the data c e l l s would have no observat ions. Although beyond the time and cost l im i t a t i ons of t h i s present study, Orcutt et a l have suggested a method of adjust ing the data so that the pred ic t ions can be used i n an i t e r a t i v e simulat ion process . -^ lOGuy H. Orcutt et a l . , Micro Analys is of Socioeconomic Systems; A  Simulat ion Study (New York: Harper, 1961), pp. 229-231. B a s i c a l l y the method cons is ts of accumulating by i n te rva l s the expected values of a l l the poss ib le combinations of the independent va r iab les , as ca lcu la ted from the basic regressions without in te rac t ion terms, and arraying these against a corresponding ordering of accumulated observed values fo r the dependent va r i ab le . Each expected value and each observed value i s then weighted by i t s frequency of occurrence, summed wi th in the i n t e r v a l s , and the i n t e r v a l t o t a l s d iv ided by the sum of the i n te r va l weights. This resu l t s i n a ser ies of mean expected values and a cor-responding ser ies of mean observed va lues. Thus, there i s establ ished a transformation from expected values, 'as ca lcu lated from a regression equation considering only the separate ef fects of each va r iab le , to t rue values tempered by in te rac t ion e f fec ts . Weighting The mode choice regressions consider only data about the respon-dents ' most recent and next to l a s t t r i p s . Thus, there i s inadequate representat ion of the frequent t r a v e l l e r and the means of the mode choice dependent var iab les do not provide use fu l estimates of the r e l a t i ve f r e -quency of t r a v e l by the d i f fe rent modes. This problem i s p a r t i a l l y overcome by the reca lcu la t ion ' of the means of the dependent var iab les on the bas is of the t o t a l number of t r i p s taken by each mode during the survey year . However, t h i s does not give representat ion to the cha rac te r i s t i cs of the e a r l i e r t r i p s (those p r i o r to the most recent and next to l a s t t r i p s but s t i l l i n the survey year) i n the determination of the regression coe f f i c i en ts fo r the inde-pendent va r i ab les . Any attempt to weight each pa r t i cu la r respondent's 68 most recent t r i p and next to l a s t t r i p cha rac te r i s t i cs by h i s t r a v e l f r e -quency would be open to j u s t i f i a b l e skept ic ism concerning the impl ica t ion that these t r i p s were t y p i c a l of h i s other t r i p s . I t was considered un-des i rab le to have the survey s o l i c i t data on more than most recent and next t o , l a s t t r i p s because the questionnaire was already long and also because-of the problem of accurate r e c a l l by the respondent. S t r a t i f i c a t i o n The regressions for t o t a l t r a v e l frequency are s t r a t i f i e d by stage i n fami ly l i f e cyc le . This embodies a composite va r iab le invo lv ing the j o in t considerat ion of age, mar i ta l s tatus, and fami ly s ta tus. On the bas is of a survey ana lys is Lansing and Kish have shown the importance of l i f e cyc le as an independent var iab le i n the inves t iga t ion of va r ia t i ons i n s i x economic cha rac te r i s t i c s . For each of these cha rac te r i s t i cs i t ( fami ly l i f e cycle) proved i t s e l f superior i n 'explanatory* power to age c lasses . This resu l t i s consistent with s o c i a l theory since F . L . C . ( fami ly l i f e cyc le) should be a bet ter r e f l e c t i o n than age of the i n d i v i d u a l ' s s o c i a l ro le .12 This conclusion would seem reasonable from r e a l i z a t i o n of the fac t that the in f luences of age are ef fected i n large part through changes i n l i f e cyc le . Since age i s not the only fac to r determining stage i n l i f e cyc le , stage i n l i f e cyc le w i l l l o g i c a l l y be the more "^"John B. Lansing and L e s l i e K i s h / Family L i f e Cycle as an Inde-pendent V a r i a b l e / American Soc io log i ca l Review, V o l . 22, October, 1957, pp. 512-519. 1 2 i b i d . , p. 518. 69 powerful independent va r i ab le . An i l l u s t r a t i o n of the explanatory power of l i f e cyc le stage i s shown i n Tables I I (a and b ) . These Tables are taken from "The Changing Travel Market" study reviewed i n Chapter I I I . ^ Bimodali ty i s shown by fami ly l i f e cyc le whi le hidden by the considerat ion of age alone. In the case of age (Table 11(a)), the percentage who took a t r i p increases up to the 35-44 age group and then s tead i l y dec l ines . In the case of fami ly l i f e cyc le stage, the percentage who took an a i r t r i p s ta r ts out high i n the f i r s t stage, drops i n the second, r i s e s again i n the t h i r d , and then f a l l s o f f fo r a second time i n the l a s t two stages. An important point of considerat ion was the number of l i f e cyc le groups to use i n the s t r a t i f i c a t i o n . Many past studies have confined t h e i r l i f e cyc le breakdown to three groups; s ing le people and married people without ch i ld ren under the age of 45, s ing le people and married people without ch i ld ren over the age of 45, and married people with ch i l d ren . However, research car r ied out by Blood ind ica tes that there i s a considerable d i f ference i n the t r a v e l behaviour of s ing le people and married people without ch i ldren i n both age d i v i s i o n s . " ^ His research consisted of running regressions of seven independent var iab les on non-business auto t r a v e l s t r a t i f i e d f i r s t wi th the under-and-over 45 groups pooled and then with the s ing le adul ts i n each of these l i f e cyc le groups 13John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan-: Survey Research Center, March 1964), p. 97 and p. 102. ^^Dwight M. Blood, "A Cross Sect ion Analys is of the Domestic In ter -c i t y Travel Market" (unpublished doctora l d i sser ta t ion i n economics at the Un ivers i ty of Michigan, Ann Arbor, 1963), pp. 42-43. 70 TABLE 11(a) PERCENTAGE OF ADULTS WHO TOOK ANY AIR TRIP BY AGE Age of Adult Percentage Who Took Any A i r Tr ip 18-24 11 25-34 12 35-44 13 45-54 10 55-64 9 62 and over 5 Source: John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1964), p. 97. TABLE 11(b) PERCENTAGE OF ADULTS WHO TOOK ANY AIR TRIP BY FAMILY LIFE CYCLE STAGE Family L i f e Cycle Stage Percentage Who Took Any A i r Tr ip (1) Young, Single (2) Under 45, Married No Chi ldren under 18 (3) Marr ied, Chi ldren under 18 (4) Over 45, Married No Chi ldren under 18 (5) Older, S ingle 14 8 12 8 Source: John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1964), • p. 102. 71 separated from the married people without ch i l d ren . The resu l t s of these regressions are shown i n Table I I I . Table I I I ind icates that a i r t r a v e l experience (E X P) i s a more important explanatory fac tor for young s ing le people than for young mar-r i e d people without ch i ld ren , the regression coe f f i c i en ts being .18 for the former group and .008 for the l a t t e r group. Residence i n one of the 12 largest metropoli tan areas (MET) has a more powerful negative i n f l u -ence on non-business auto t r a v e l by young married people without ch i ldren than on young s ing le adu l ts ; - .18 fo r the former and -.036 for the l a t t e r . The reverse re la t ionsh ip i s exhib i ted at the other end of the l i f e cyc le . Family income appears to be of ..somewhat greater importance to married adults without ch i ld ren , both young and o l d , than-to s ingle adults i n the same age d i v i s i o n . I t i s apparent that education and.ai r experience tend to be subst i tu tes fo r income i n explanatory importance for s ing le adults i n comparison to married adul ts without ch i l d ren . On the bas is of h i s ana lys i s , Blood reached the conclusion that the l i f e cyc le var iab le should be broken down in to f i v e groups as fo l l ows : 1) Single adults under the age of 45. 2) Married adults under the age of 45 without ch i ld ren . 3) A l l married adults wi th ch i ld ren . 4) Marr ied adults over the age of 45 without ch i ld ren . 5) Single adults over the age of 45. In order to determine the contr ibut ion to the explanation of the t o t a l variance made by the l i f e cyc le stage s t r a t i f i c a t i o n , a modif ied variance ana lys is i s undertaken. Essen t i a l l y t h i s involves pa r t i t i on ing TABLE I I I A COMPARISON OF REGRESSION EQUATIONS FOR NON-BUSINESS AUTO TRAVEL FOR SPECIFIED STAGES IN FAMILY LIFE CYCLE FOR THE YEAR 1955 Var iab le Under 45 Single Under 45 Married No Chi ldren Under 45, S ingle Under 45, Married No Chi ldren Over 45 Single Over 45 Married No Chi ldren Over 45, Single Over 45, Married No Chi ldren C .4078 .4434 .4133 .1875 -.0235 .1759 EXP .1800 .0083 .0945 .3068 .2128 .1995 SEX -.0462 -.0143 -.0051 .0930 .0261 .0495 METRO -.0361 -.1852 -.0967 -.1821 .0325 -.0986 Y l .0115 .3711 .1594 .0238 -.0040 .1735 Y 2 .0944 .4841 .2441 .0477 .0492 .1455 Y3 .1342 .5836 .2758 -.0214 .1264 .3156 OCC .1431 .1756 .1436 .2001 -.0573 .0894 EDUC .3654 .0260 .2343 .2301 .0063 .2376 AUTO .3137 .3197 .3243 .2327 .0260 .2999 R2 .1925 .1584 .1689 .1496 .1521 .1843 N 425 348 782 439 970 1419 Source: Dwight M. Blood, "A Cross-Sect ion Analys is of the Domestic I n te rc i t y Travel.Market". (un-published doctora l d isser ta t ion i n economics, Univers i ty of Michigan, Ann Arbor, 1963} p. 42. 73 the t o t a l sum of squares i n t o the percentage explained by the separate r e g r e s s i o n s and the percentage explained by the s t r a t i f i c a t i o n v a r i a b l e . The procedure i s as follows:''"-' 1) The p r o p o r t i o n of the t o t a l sum of squares explained by the subgroup means i s estimated i n the f o l l o w i n g equation. n _ 2 = 2 T N i x i - N x • i = l where r e f e r s t o the group frequencies, X-j_ r e f e r s t o the subgroup means, N r e f e r s t o the t o t a l number of observations, X i s the grand mean of a l l combined subgroups, and the s u b s c r i p t ^ r e f e r s t o the number of subgroups. 2) The p o r t i o n o f the t o t a l sum of squares explained by the separate r e g r e s s i o n s i s estimated as f o l l o w s : (R? N ± <r±2) ? 2 where the R^'s are those obtained from the separate r e g r e s s i o n s and ^ r e f e r s t o the v a r i a n c e of the dependent v a r i a b l e i n the subgroup equa-t i o n s . 3) The t o t a l explained sum of squares i s the sum of the va r i a n c e explained by the means obtained i n step ( l ) and the v a r i a n c e explained by the separate r e g r e s s i o n s obtained i n step (2). 4) The unexplained sum of squares i s de f i n e d as: N i - o - i 2 - [ ( R f N i o - i 2 ) ] 5) The t o t a l sum of squares i s obtained by adding the expl a i n e d sum of squares obtained i n (3) t o the unexplained sum of squares obtained ^ i b i d . , p. 45-46. 74 i n (4). S t r a t i f i c a t i o n by l i f e cyc le was done only for the t o t a l t r i p f r e -quency regress ion. The a i r t r a v e l frequency and mode choice regressions involved i n s u f f i c i e n t observations to permit such subdiv is ion. S t a t i s t i c a l R e l i a b i l i t y The resu l t s of the regression ana lys is , i n terms of regression co-e f f i c i e n t s fo r each independent va r iab le , have l i t t l e meaning i f cons id-ered i n i s o l a t i o n of t h e i r s t a t i s t i c a l r e l i a b i l i t y . In regression ana lys is , s t a t i s t i c a l r e l i a b i l i t y i s usual ly measured i n terms of the standard errors of the regression coe f f i c i en t s . Given the assumption of a normal d i s t r i b u t i o n , 68% of a l l the i nd i v i dua l obser-v a t i o n s i n the sample w i l l f a l l w i th in one standard error of the regres-s ion coe f f i c i en t , 95% w i l l f a l l w i th in two standard er rors , and 99% w i l l f a l l w i th in three standard errors of the regression coe f f i c i en t . Thus, the s i ze of the standard error of the regression coe f f i c ien t gives an i nd i ca t i on of how we l l the regression coef f i c ien t : represents the^data upon which i t i s based; the la rger the standard er ror , the greater i s the scat ter i n the data and the poorer i s the regression coe f f i c ien t as a representat ive of that data. In t h i s study i t i s des i rab le to compare the r e l i a b i l i t y of the d i f fe ren t var iab les i n the regressions. However, t h i s cannot be done by considerat ion of the standard error terms alone. This i s because each var iab le i s measured i n terms of c l a s s i f i c a t i o n s which are en t i re l y d i f fe ren t from the c l a s s i f i c a t i o n s of each other va r iab le . For example, there i s no unequivocal way of making a quant i ta t ive comparison of the 75 Over $10,000 income c l a s s i f i c a t i o n and the College Graduate education c l a s s i f i c a t i o n . However,: the d i f f i c u l t y of making comparisons of s t a t i s -t i c a l r e l i a b i l i t y s t r i c t l y i n terms of standard errors may be overcome by looking at the standard error for each var iab le i n r e l a t i on to the regression coe f f i c ien t for that va r i ab le . The larger the regression coe f f i c ien t fo r a var iab le i n comparison to i t s standard er ror , the more sure we can be that the re la t ionsh ip ind icated by that regression coe f f i c ien t ac tua l l y does ex is t and i s not just appearing as the resu l t of random sampling va r ia t i on i n the data. For example, i f the regression coe f f i c ien t fo r the Over $10,000 income c l a s s i f i c a t i o n i s .150 and i t s standard error i s .075, there i s about a 95% p robab i l i t y that the r e l a -t ionsh ip between having income over $10,000 and t r a v e l i s more than the resu l t of random sampling v a r i a t i o n . However, a standard error of .100 instead of .075 would not necessar i l y ind ica te a poorer degree of r e l i a b i l i t y i n the re la t i onsh ip . I f the regression coe f f i c ien t were .300, the .100 standard error would, i n f ac t , ind ica te an even greater l i k e l i -hood {99%) of a t rue re la t i onsh ip . On the other hand, a very smal l standard error of .025 does not necessar i l y mean a high degree of r e l i a b i l i t y i n the re la t i onsh ip . I f the regression coe f f i c ien t i s .020, the .025 standard error would ind ica te l ess than a 68% p robab i l i t y that the apparent re la t ionsh ip i s more than just the resu l t of sampling v a r i a t i o n . CHAPTER VI PERSONAL TRAVEL FREQUENCY—ALL MODES Introduction The personal travel frequency regressions were stratified by five stages in the family l i f e cycle; single and under 45 years of age; mar-ried, under 45 with no children; married with children; married, over 45 with no children; and single and over 45- The independent variables were the same in each equation and the regressions took the following form: Index of personal travel by a l l modes = a + b]_ (income under $5,000) + b 2 (income $5,000-$9,999) + b 3 (income $10,000 and over) + b^ (owns auto) + b^ (owns boat) + b^ (owns cottage) + by (own home, no mortgage) + bg (own home, mortgage) + b Q (single detached dwelling) + b-j_0 (apt./flat) + b^ (manager) + bj_2 (blue collar) + b^j (retired) + b ^ (grade school) + b ^ (high school) + b ^ (college graduate) + bj_y (child/parents—West Can. or U.S.) + bj_Q (child/parents—East Can. or rest U.S.) + b-^ (child/parents—outside Can. & U.S.) + b2Q (school—West Can. or U.S.) + b ^ (school—East Can. or rest U.S.) + b 2 2 (school—outside Can. & U.S.) + b 2 3 (birth— 77 West Can. or U.S.) + b 2 ^ (bir th—East Can. or rest U.S.) + b 2 5 (bir th—outside Can. & U.S.) + b 2 £ (male) + b 2 y ( C e l l j) + b 2 g ( C e l l 2 ) + b 2 9 ( C e l l 4 ) + b ^ ( C e l l 1 6 ) + u The dependent var iab le i n these equations was defined according to the sca le values set out i n Chapter V. These scale values are 0, 1.0, 1.6, 2.4 and 3 .2 , depending on whether an adult took no t r i p s , one t r i p , 2-4 t r i p s , 5-15 t r i p s or 16 or more t r i p s . The independent var iab les i n the equation are defined i n Appendix A. As explained i n Chapter V, one c lass of each var iab le has been excluded from the equation to permit determinacy i n the r e s u l t s . For example, the white c o l l a r c lass of the occupation var iab le i s not included i n the equation. The b*s i n the equation represent estimated regression coe f f i c i en ts for the var iab les i nd i ca ted ; a represents the constant term; and u represents the error term. The resu l t s of the regressions for personal t rave l frequency as s t r a t i f i e d by fami ly l i f e cyc le are set fo r th i n Appendix B, accompanied by the standard error of each regression coe f f i c i en t . The mul t ip le c l a s -s i f i c a t i o n conversion of these r e s u l t s , i n terms of an adjusted deviat ion f igure fo r each regression coe f f i c i en t , i s set out i n Table IV. I t w i l l be noted that adjusted deviat ion f igures are also spec i f ied fo r the c lass of each var iab le which had to be excluded from the o r i g i n a l regression equation (e .g . the white c o l l a r category i n the occupation va r i ab le ) . For t h i s and the other reasons explained f u l l y i n Chapter V, the d iscus-s ion which fo l lows w i l l re fe r mainly to these mul t ip le c l a s s i f i c a t i o n TABLE IV MULTIPLE CLASSIFICATION ANALTSXS OF FACTORS APFECTIN3 TOTAL TRAVEL FREQUENCY BY STAGE IN FAMILY LIFE CYCLE Under 45; Single Under 45 i Married Married; Children Over 45; Married Over 45 ; Single No Children No Children Adjusted Adjusted Adjusted Adjusted Adjusted Deviation Deviation Deviation Deviation Devlatior No. of From Mean No. of From Mean No. of From Mean No. of From Mean No. of From Mear Variable Cases Of .84 Casss Of .90 Cases Of .71 Casee Of .72 Cases Of .56 Income Under »5,000 131 .00 40 +.01 183 -.16 213 .00 211 -.08 »5,O0O-*9,999 91 +.03 91 +.04 671 -.02 187 +.05 . 40 +.17 110,000 «nd Over 24 +.22 24 +.31 203 +.28 60 +.18 4 +.24 Answer Refused 115 -.08 43 -.09 309 -.04 216 -.09 113 +.07 Auto Ownership Own 175 .•.06 175 +.03 1,260 +.02 495 +.07 100 +.10 Not Own 136 -.06 23 -.a 106 -.19 181 -.20 268 -.04 Boat Ownsrshlp -. Own 20 •.04 . 17 •.11 95 +.04 46 -.01 6 -.11' Not Own 341 .00 181 -.01 1,271 .00 630 .00 . 362 .00 Cottage Ownership Own 27 •.11 6 +.47 74 +.08 30 -.18 10 +.26 Not Own 334 -.01 192 -.01 1,292 .00 646 +.01 358 -.01 Dwelling Ownership -Own—No Mortgage 12 •.14 29 .00 258 -.04 380 +.02 128 +.10 Own—Mortgage 22 -.12 69 -.17 784 -.02 176 -.08 40 +.09 Rent 327 .00 100 +.12 324 +.07 120 +.04 200 -.08 Dwelling Type Single-Detached 207 -.02 119 +.07 1,169 .00 541 .00 201 -.10 Single-Attached 28 •.03 16 -.19 127 +.05 47 +.06 27 -.13 Apt./Flat 126 •.03 63 -.08 70 -.03 88 -.03 140 +.17 Occupation ' Managerial; Profes- 97 +.15 57 +.13 490 +.07 131 +.04 36 -.02 sional; Proprietary Whits Collar 144 -.07 62 -.01 333 -.05 97 .00 45 +.10 Blue Collar 95 +.02 76 -.08 • 523 -.03 201 -.04 38 +.05 Retired 25 -.26 3 -.43 20 -.08 247 +.01 249 -.02 Education Grade School; Some High 93 -.17 • 68 -.06 593 -.03 . 401 -.04 220 -.04 School or Less High School Grad. or 168 +.04 98 +.03 627 +.03 239 •.07 127 +.08 Some University University Student 70 +.17 6 +.10 16 +.19 0 +.36 0 +.76 Univereity Graduate 25 -.06 26 +.02 130 -.05 36 -.01 21 -.04 Child or Parents Within 250 mi. of 255 -.10 118 -.04 817 -.03 430 -.08 255 -.05 Vancouver West Can. or West U.S .A. 43 ^ +.22 25 -.12 231 +.16 141 +.19 67 +.13 East Can. or Rest of 20 +.13 17. .00 98 -.03 70 +.10 34 +.06 U.S. A. Outside Can. & U.S.A. 43 +.31 38 +.19 220 -.03 35 •.01 12 +.17 Schooling 176 Within 250 mi. of 197 .00 87 +.02 541 +.03 -.04 74 . -.03 Vancouver West Can. or West U.S.A. 74 +.14 38 +.27 349 +.10 189 .00 83 +.11 East Can. or Rest of 26 +.13 20 +.32 118 .00 79 -.04 62 .00 U.S.A. Outside Can. & U.S.A. 64 -.22 53 -.35 358 -.14 232 +.04 , 149 -.05 Birth Inside B.C. 170 -.15 66 -.13 392 -.06 107 +.12 40 +.08 West Can. or West U.S.A. 73 +.17 45 +.02 413 -.06 161 +.07 55 -.09 East Can. or Rest of 38 +.03 28 -.17 139 +.01 106 +.01 94 -.05 U.S.A. Outside Can. & U.S.A. 80 +.15 59 +.20 ^22 +.11 302 -.08 • 179 +.04 Sex / Male 194 . . -.03 101 +.07 566 +.02 342 .00 87 -.06 Female 167 +.04 97 .00 800 -.02 325 .00 281 +.02 Cell 1 15 -.12 " 14 +.05 142 -.06 58 +.08 25 -.10 2 6 -.23 7 -.41 30 +.16 13 -.40 6 +.42 4 12 +.10 11 •.19 87 +.10 20 +.20 6 -.41 5 57 -.18 11 +.12 10 +.01 25 -.06 45 -.20 6 23 +;o2 11 -.33 » 60 +.07 59 •.17 28 ' +.01 7 34 +.13 10 +.15 97 -.14 55 +.08 39 +.13 8 57 -.02 32 +.01 110 -.05 67 +.03 55 +.07 9 7 -.05 7 -.06 91 .00 10 -.03 6 +.02 10 21 +.17 14 +.23 173 -.01 105 +.10 34 +.07 11 25 +.13 4 -.16 42 -.07 46 -.06 19 -.21 12 31 ' -.22 23 -.28 120 -.11 75 -.15 45 -.01 13 . 13 +.19 11 +.18 62 -.09 23 -.23 5 +.06 14 8 +.04 7 -.32 58 +.05 17 -.31 4 +.10 15 24 +.06 23 +.12 162 •.13 39 +.06 17 .00 16 9 +.21 3 +.14 28 -.07 9 -.12 4 +.10 ' 17 19 +.14 10 +.18 94 +.13 55 -.09 30 +.10 79 results rather than the original regression results. The interpretation of the multiple classification format was de-scribed in Chapter V. As an example of the meaning of the data in Table IV, consider the results for the stage in the family l i f e cycle composed of single adults under the age of 45. The mean travel index for a l l adults in this group i s .84.^ * If nothing else were known about an individual adult, this value might be taken as a crude approximation to an average expected travel frequency index for an adult taken at random from this l i f e cycle stage. If, on the other hand, the income status, occupation and education of an adult i s known, a closer approx-imation to the estimated travel frequency index for that particular adult can be obtained. For example, a positive value of +.22 would be added to the mean of .84 i f the adult earned annual income over $10,000; a value of -.07 would be subtracted from the cumulated value of 1.06 i f the adult was employed in a white collar occupation, resulting in a cumulated value of .99. Finally, i f the adult had attained high school education, a positive value of +.04 would be added to .99, resulting in a final cumu-lative estimated trip index for an adult with the indicated character-istics of 1.03. The discussion which follows will involve a detailed consideration of each variable in terms of: the magnitude and direction of its influence upon personal travel frequency in each family l i f e cycle stage, the "'"The "mean travel index" term used here should not_be interpreted as a probability as i t may easily exceed 1.0 or conceivably f a l l below 0.0. It is rather an index of relative travel frequency and i t s exact numerical value has l i t t l e importance in itself, being based only on the scales defined on page 76. manner i n which i t s in f luence changes over the fami ly l i f e cyc le , and the s t a t i s t i c a l r e l a i b i l i t y of i t s in f luence . This w i l l be fol lowed by an i nd i ca t i on of the contr ibut ion of each regression equation, and the separate contr ibut ion of s t r a t i f i c a t i o n by l i f e cycle stage, to the explanation of the t o t a l variance i n the data. The chapter w i l l be concluded wi th an i nd i ca t i on of those var iab les which have shown them-selves to be most important i n the determination of the personal t r a v e l market. Influence & R e l i a b i l i t y of Var iab les  L i f e Cycle Stage Graph I sets out the mean t r a v e l index for each l i f e cyc le stage. This graph suggests that , ce te r i s par ibus, adul ts who are under 451 and married without ch i ld ren t r ave l more f requent ly than other adul ts , whi le those leas t l i k e l y to t r a v e l are adults who are over 45 and s ing le . S ingle adults under .45,travel s l i g h t l y l ess than married adults under 45 without ch i l d ren . Married adults who have ch i ld ren or who are over 45 without ch i ld ren t r a v e l l e s s f requent ly than adults i n e i ther of the f i r s t two l i f e cyc le stages but considerably more than s ing le adults over 45. Of course, these resu l t s re fe r only to the present day t r a v e l habi ts and could not be taken to suggest that those adults present ly i n the Under 45, Married without Chi ldren l i f e cyc le stage w i l l necessar i l y show the same t r a v e l frequency as those adul ts present ly i n the Married wi th Chi ldren stage when they pass in to that stage. GRAPH I LIFE CYCLE STAGE AND TOTAL TRAVEL FREQUENCY2 Mean Travel 1.00 -.90 .80h .70 .60 Frequency ^ Q .40 .30 .20 Single:'. Married Married Married Single Under 45 Under 45 C h i l d . Over 45 Over 45 No C h i l d . No C h i l d . ' Income The resu l t s of the mul t ip le c l a s s i f i c a t i o n ana lys is set out i n Table IV suggest that adults wi th higher incomes tend to t r a v e l more f requent ly than other adu l ts . Graph I I i l l u s t r a t e s . Adults earning annual income over $10,000 t r a v e l more f requent ly than adults earning income between $5,000 and $10,000, and both of these groups tend to t r a v e l more than adul ts earning l ess than $5,000 a year. Graph I I fur ther i l l u s t r a t e s that the d i f ference i n t r a v e l frequency between adul ts earning over $10,000 and adul ts earning $5,000-$9,999 i s p a r t i c u l a r l y pronounced i n the f i r s t four l i f e cycle Although t r u l y app l icab le only to continuous phenomena, l i n e graphs are used throughout t h i s chapter because they enable a number of c lasses to be superimposed upon the same graph and, thereby, ease comparison and require l e s s space than other v i s u a l methods. GRAPH I I INCOME AND TOTAL TRAVEL FREQUENCY +.40 +.30 Pos i t i ve +.20 Adjustment" To Mean' Travel Frequency +.10 .0 -.10 Negative -.20 -.30 -.40' Over $10,000 — $5,000-$9,999 • ' • Under $5,000 Single Married Married Married Single Under 45 Under 45 C h i l d . Over 45 Over 45 No C h i l d . No C h i l d . stages. In these same stages, the d i f ference i n the s ign i f i cance of i n -come between $5,000 and $9,999 and income under $5,000 i s much l e s s . The l a s t l i f e cyc le stage, s ing le adul ts over 45, shows the reverse s i t ua t i on , wi th the d i f ference i n t r a v e l frequency between those with income over $10,000 and those wi th income between $5,000 and $9,999 being much l ess pronounced than the d i f ference i n t r a v e l frequencies between the l a t t e r income group and those earning l e s s than $5,000 annual ly . A s i gn i f i can t proport ion of the respondents i n the survey refused to ind ica te t h e i r income group. This proport ion measured between one-quarter and one-th i rd of the t o t a l respondents i n each l i f e cyc le stage. The mul t ip le c l a s s i f i c a t i o n resu l t s fo r these respondents, as set out i n Table IV, ind ica te a t r a v e l frequency e i ther l ess than that for adul ts ad-mi t t ing income of l e s s than $5,000 or between t h i s income group and 83 the $5,000-$9,999 income group. This leads to the suggestion that people i n the low income groups tend to want to hide their income more than do people i n the medium and high income groups. Graph II also points out the changes i n the effect of each income group upon personal travel frequency through the various stages i n the family l i f e cycle. High income ($10,000 and over) has i t s greatest positive significance i n the case of married adults with children and married adults under 45 without children, and i t s least significance i n the case of married adults over 45 without children. Middle income ($5,000-$9,999) has i t s most significant positive association with travel by single adults over 45. This income level i s of only negligible significance i n the other four l i f e cycle stages. Low income (under $5,000) has i t s greatest negative association with the travel of married adults with children and a less, though s t i l l significant, negative association with the travel of single adults over 45. Examination of the standard errors of the regression coefficients set out i n Appendix B indicates that the results for the highest income classification ($10,000 and over) are very reliable, whereas the results for the lower income classifications are characterized by considerably poorer r e l i a b i l i t y . Auto Ownership Families which own at least one automobile undertake more personal travel (including by a i r , r a i l and bus) than families which do not own even one automobile. This positive relationship i s apparent i n a l l l i f e 84 cycle stages, with particular prominence i n the case of married adults without children. Appendix B indicates that the relationship i s highly reliable i n the case of married adults with children and married adults over 45 without children, but only f a i r l y reliable i n the case of the other l i f e cycle stages. However, despite the consistency and r e l i a b i l i t y of the results for this variable, the small absolute size of these results indicate that the significance of auto ownership i n to t a l travel frequency i s relatively slight when the relationships of a l l the other variables i n the study are simultaneously taken into account. Boat Ownership The relationship between personal travel frequency and owning a boat over 14 feet i n length appears insignificant, inconsistent over the family l i f e cycle, and unreliable. If any consideration at a l l can be given to the results for this variable, i t would be suggested that those owning a boat i n the f i r s t three l i f e cycle stages have a slightly greater frequency of travel ithan other adults, while those owning a boat i n the last two stages tend to travel slightly less than other adults. The peak positive relationship i s i n the case of married adults under 45 without children and the peak negative relationship i s i n the case of single adults over 45. Cottage Ownership Although equally inconsistent over,the family l i f e cycle, cottage ownership exhibits a somewhat more prominent and more reliable relation-ship with personal travel frequency than boat ownership. Cottage owner-ship i s p o s i t i v e l y associated wi th t o t a l t r a v e l frequency i n a l l l i f e cyc le stages except the Over 45 Married Without Chi ldren stage, i n which case there i s a very s ign i f i can t negative re l a t i onsh ip . A p a r t i c u l a r l y s i gn i f i can t pos i t i ve re la t ionsh ip i s apparent i n the case of married adul ts under 45 without ch i l d ren . Dwell ing Ownership Whether an adult owns h i s home without a mortgage, owns i t wi th a mortgage, or simply ren ts , appears to have only a weak assoc iat ion with h i s personal t r a v e l frequency. This i s shown i n Graph I I I . I t i s apparent that the d i f fe ren t c l a s s i f i c a t i o n s of the dwel l ing ownership va r iab le do not exh ib i t a pattern of re la t ionsh ip wi th personal t r a v e l frequency which i s consistent over the var ious l i f e cyc le stages. A loose suggestion might be that the t r a v e l frequency of married adu l ts , regardless of number of ch i l d ren , i s p o s i t i v e l y re la ted to dwel l ing i ren ta l and negat ive ly re la ted to dwel l ing ownership wi th mortgage; where-as t r a v e l by s ing le adul ts , both under and over 45, i s negat ively a s s o c i -ated with dwel l ing ren ta l and p o s i t i v e l y re la ted to dwel l ing ownership without mortgage. However, t h i s general statement ignores several exceptions. Thus, i t may be more usefu l to consider each l i f e cyc le stage separately. Single adul ts under 45 are most l i k e l y to t r a v e l i f they own a home without mortgage and are l eas t l i k e l y to t r a v e l i f they own a home wi th mortgage. Married adul ts under 45 without ch i ldren have a greater tendency to t r a v e l i f they rent whereas they tend to t r a v e l l eas t i f 86 they own wi th mortgage. In the Married wi th Chi ldren l i f e cyc le stage, dwel l ing ren ta l i s p o s i t i v e l y associated wi th t r a v e l frequency whi le dwel l ing ownership, both with and without mortgage, shows a s l i gh t neg-a t ive assoc ia t ion . Married adul ts who are over 45, and have no ch i ldren under 18, tend to t r ave l more i f they rent or own without mortgage than i f they own wi th mortgage. S ing le adul ts over 45 who own the i r home, e i ther wi th or without mortgage, have a greater tendency to t r ave l than those who ren t . Graph I I I a lso I l l u s t r a t e s the va r ia t ions i n the re la t ionsh ip between each dwel l ing ownership c l a s s i f i c a t i o n and t r a v e l frequency over the fami ly l i f e cyc le . I t w i l l be noted that dwel l ing ownership without mortgage has i t s greatest pos i t i ve assoc ia t ion wi th the personal t r a v e l frequency of those adults i n the f i r s t and l a s t l i f e cyc le stages while i t i s negl igeably associated with the middle three l i f e cyc le stages. Dwell ing ren ta l exh ib i ts the opposite re la t i onsh ips , having a pos i t i ve assoc ia t ion wi th t r a v e l by the middle three l i f e cyc le stages, no assoc ia t ion wi th t r a v e l by s ing le adul ts under 45, and a s i g n i f i c a n t l y negative assoc ia t ion with the t r a v e l frequency of s ing le adults over 45. The t h i r d c l a s s i f i c a t i o n , dwel l ing ownership wi th mortgage, gener-a l l y has a negative associat ion wi th t r a v e l . I t s greatest negative assoc ia t ion i s wi th adul ts under 45, both s ing le and married without ch i l d ren . However, i t a lso has a s i gn i f i can t pos i t i ve assoc iat ion with the t r ave l frequency of s ing le adul ts over 45. Examination of Appendix B ind ica tes that the resu l t s fo r the re la t ionsh ips between dwel l ing ownership and t r a v e l frequency have only 87 GRAPH III DWELLING OWNERSHIP AND TOTAL TRAVEL FREQUENCY + . 4 0 -+ . 3 0 -Positive + . 2 0 • Adjustment+.IO " To Mean > 0 _ Travel Frequency - . 1 0 • - . 2 0 -Negative - . 3 0 -- . 4 0 -a f a i r degree of r e l i a b i l i t y . Dwelling Type Dwelling type, whether single-detached, single-attached, or apart-ment/flat, shows significant relationships with the personal travel frequency of married adults under 45 without children and single adults over 4 5 . However, as shown i n Graph IV, i t s relationships with the travel frequencies of the remaining three l i f e cycle stages appear negligeable. Married adults under 45 without children tend to travel more frequently i f they l i v e i n a single-detached dwelling. Living i n a single-attached dwelling or apartment has a negative association with the personal travel frequency of adults i n this l i f e cycle stage, with — Owns No Mortgage — — Owns Mortgage ••«» Rents Single Married Married Married Single Under 45 Under 45 Child. Over 45 Over 45 No Child. No Child. 88 GRAPH IV DWELLING TYPE AND TOTAL TRAVEL FREQUENCY Positive + . 4 0 + . 3 0 + . 2 0 -Adjustment4"* 10 r To Mean #Q Travel Frequency - « 1 0 r Negative - . 2 0 • - . 3 0 • - . 4 0 • Single-Detached — - Single-Attached • Apt./Flat Single Married Married Married Single Under 45 Under 45 Child. Over 45 Over 45 No Child. No Child. the former being particularly significant. Single adults over 45 exhibit relationships between personal travel frequency and dwelling type which are almost the complete opposite to those just described for married adults under 45 without children. In the case of single adults over 45, l i v i n g i n an apartment or f l a t has the greatest positive relationship with travel. Living i n either a single-attached or a single-detached dwelling has a significant negative re l a -tionship with the travel frequency of adults i n this group, with the former showing a slightly greater association. As mentioned above and indicated i n Table IV, the influence of dwelling ownership on the personal travel frequency of adults i n the remaining three l i f e cycle stages i s very small. Graph IV also sets out the changes i n the effect of each c l a s s i -89 fication of dwelling type during the family l i f e cycle. j. It w i l l be noted that l i v i n g i n a single-detached dwelling has i i t s greatest positive association with, the personal travel frequency of married adults under' 45 without children, while i t has about an equal negative association with single adults over 45. Living i n an apartment or f l a t shows the opposite relationships; negative i n the case of mar-ried adults under 45 without children and positive i n the case of single adults over 45. Living i n a single-attached dwelling has a significant negative association with the travel of adults i n both of these l i f e cycle stages. With the general lack of significance i n the results for the i Dwelling Type variable, i t i s not surprising to find from perusal of Appendix B that these results are also characterized by a poor degree of r e l i a b i l i t y . Financial Interest i n Major Assets One of the main purposes behind the inclusion i n the survey of questions on dwelling type and ownership of dwelling, auto, boat and cottage, was the hope that these factors might provide a reasonable i n d i -cation of the significance of financial interest i n certain major assets. It has been suggested that ownership of an auto, boat, cottage or dwelling, would each represent a drain upon the adult's income, i . e . the portion of his income which would otherwise be free to be spent on personal travel. Thus, the hypothesis was that adults who do not own an auto, boat, cottage or single-detached dwelling travel more frequently than other adults. 90 The results of the regression and multiple classification anal-yses do not substantiate this'lfinancial interest" hypothesis. In fact, perusal of Graph V suggests that the hypothesis i s opposite to the true situation. This graph i s formed from the combined multiple c l a s s i f i -cation results for non-ownership of auto, boat, cottage and single-detached dwelling for each l i f e cycle stage. It w i l l be noted that the combined results are negative for the f i r s t four l i f e cycle stages and only slightly positive for single adults over 45. The most obvious conclusion i s that those individuals who have the money to spend on a home, boat, etc., are s t i l l l i k e l y to have money for travel. Occupation The respondents to the survey were cl a s s i f i e d into four types of occupation; managerial/professional/proprietary, white collar, blue collar and retired. The multiple classification results for the rel a -tionship of these different classifications to personal travel frequency i n the various l i f e cycle stages are illu s t r a t e d i n Graph VI. In the f i r s t two l i f e cycle stages, the occupation variable exhibits a f a i r l y clear, three-level pattern of relationship with personal travel frequency. Managerial, professional and proprietary occupations are associated with high personal travel frequency; retirement i s associated with low personal travel frequency; and white collar and blue collar occupations are associated with medium frequencies of personal travel. The same pattern of relationship i s maintained i n the third l i f e cycle stage but the relative difference between the occupations i s consid-91 GRAPH V FINANCIAL INTEREST AND TOTAL TRAVEL FREQUENCY +.30 P o s i t i v e +.20 Adj ustment+.10 To Mean Travel * U Frequency -.10 Negative -.20 -.30 | 1 J2Z2L Single Married Married Married Single Under 45 Under 45 Child. Over 45 Over 45 No Child. No Child. erably reduced. In the fourth l i f e cycle stage there i s no p r a c t i c a l difference between the significance of the different occupations. The l a s t l i f e cycle stage exhibits a pattern of relationship which, despite only s l i g h t d i f f e r e n t i a l s between the occupation groups, i s a reversal of the pattern during the f i r s t three l i f e cycle stages; the white and blue c o l l a r occupations are associated with higher frequencies of personal t r a v e l than the managerial, professional or proprietary occupations, or retirement. I t may be useful to consider each occupation class separately and i n terms of the changes i n i t s relationships over the family l i f e cycle. Occupation i n the managerial/professional/proprietary class shows a f a i r l y s i g n i f i c a n t positive association with the personal t r a v e l frequency of single adults under 45 and married adults under 45 without children but t h i s declines steadily over the l a s t three l i f e cycle stages, ending i n a s l i g h t negative association with single adults over 45. The relationship 92 GRAPH VI OCCUPATION AND TOTAL TRAVEL FREQUENCY +.40 • Positive +.20 Single Married Married Married Single Under 45 Under 45 Child Over 45 Over 45 No Child. No Child. between the white collar occupation and travel frequency i s slightly neg-ative i n the f i r s t four L i f e cycle stages and then becomes significantly positive for single adults over 45. Blue collar occupation has only a slight relationship with personal travel frequency over the entire family l i f e cycle but i s positive for a l l single -adults and negative for a l l married adults. Retirement has a significant negative association with the f i r s t two l i f e cycle stages, particularly married adults under 45 without children, but this decreases to a moderate negative association with marriedradults with children and a negligeable association with the last two l i f e cycle stages. In general, i t i s apparent that Type of Occupation has f a i r l y significant relationships with personal travel frequency i n the f i r s t three l i f e cycle stages but only slight" r e l a -tionship with the last two stages. 93 Perusal of Appendix B suggests a very mixed degree of r e l i a b i l i t y for the occupation variable. The results for blue collar occupation are characterized by poor r e l i a b i l i t y , the results for the retired c l a s s i f i -cation are somewhat better, and the managerial/professional/proprietary results are quite good, particularly i n the case of single adults under 45 and married adults with children. Education The results for the education variable are surprising. In con-tradiction to what would be expected, adults i n the highest education classification (university graduate) do not travel the most. In fact, these adults tend to travel less than adults who have only high school graduation or just some university, and quite significantly less than university students. As illustrated i n Graph VII, this phenomenon appears consistently over a l l the l i f e cycle stages. Only those adults with grade school, some high school, or less education travel less fre-quently than university graduates. In fact, i n the Single Over 45 l i f e cycle stage, the least educated adults exhibit the same propensity to travel as the university graduates. It w i l l also be noted that there i s only a slight difference between the travel propensity for these two levels of education i n the Married Over 45 Without Children l i f e cycle stage. Graph VII also ill u s t r a t e s how the relationships between the d i f -ferent education levels and travel frequency vary over the family l i f e cycle. The positive relationship between travel frequency and being a 94 GRAPH VII EDUCATION AND TOTAL TRAVEL FREQUENCY +.40 Pos i t i ve +.20 Adjustment To Travel + * 1 0 Mean ,0 Frequency -.20 -.30 -.40 Negative — Univ. Grad. ***** Univ. Student - - High School Grad. • ' • Grade School Single Marr ied Married Married Single Under 45 Under 45 C h i l d . Over 45 Over 45 No C h i l d . No C h i l d . un i ve rs i t y student i s very high i n the Single Under 45 l i f e cyc le stage, becomes considerably reduced although s t i l l s i gn i f i can t i n the case of married adul ts under 45 without ch i l d ren , and returns to i t s o r i g i n a l l e v e l i n the Married wi th Chi ldren stage. High school graduation a lso maintains a pos i t i ve assoc iat ion wi th t r a v e l frequency throughout the fami ly l i f e c y c l e . However, t h i s re la t i onsh ip i s smal l i n magnitude over the f i r s t three l i f e cyc le stages and r i s e s only s l i g h t l y i n the l a s t two stages. The re la t ionsh ip of un i ve rs i t y graduation to t r a v e l frequency f luc tuates between s l i g h t l y pos i t i ve and s l i g h t l y negat ive. I t s ta r ts by being negative for s ing le adul ts under 45, becomes pos i t i ve for married adul ts under 45 without ch i ld ren and married adul ts with ch i l d ren , and returns to negative fo r married adul ts over 45 without ch i ldren and s ing le adul ts over 45. The lowest education category, grade school/some 95 high school/or less, has a very marked negative association with the travel frequency of single adults under 45 but rises and maintains only a small negative association i n the remaining l i f e cycle stages. The r e l i a b i l i t y of the relationships between education and per-sonal travel frequency i s revealed i n Appendix B. The estimates of the relationships for the Single Under 45 and the Married With Children l i f e cycle stages have a f a i r degree of r e l i a b i l i t y but the r e l i a b i l i t y of the results for the other three l i f e cycle stages i s very poor. Domicile of Children and Parents Respondents i n the survey were asked whether they have children or parents l i v i n g farther than 250 miles from Vancouver. A l l those answering i n the affirmative were further requested to indicate where their children or parents l i v e . In the analysis this variable was given four classifications; those having children or parents within 250 miles of Vancouver; those having children or parents l i v i n g i n Western Canada (Manitoba and westward) or the Western United States (Washington, Oregon, Idaho, California); those having children or parents l i v i n g i n Eastern Canada or the rest of the United States; and those having children or parents l i v i n g outside Canada_.and the United .States. The results, as il l u s t r a t e d i n Graph VIII, suggest that there i s a significant relation-ship between this variable and personal travel frequency. . Graph VIII shows the variations i n the relationships between travel frequency and each classification of the parents/children domicile v a r i -able over the family l i f e cycle. In the case of single adults and married 96 GRAPH VIII DOMICILE OF PARENTS AND CHILDREN AND TOTAL TRAVEL FREQUENCY +.40 • +.30 • Positive • +.20 Ad j ustment+> -J_Q L To Mean Travel * 0 Frequency _.io|-Negative - . 2 0 - . 3 0 - . 4 0 Within 250 Mi. West Can. & US • • • East Can. & US Outside Can. & US Single Married Married Married Single Under 45 Under 45 Child. Over 45 Over 45 No Child. No Child. adults under 45 without children, those travelling most frequently tend to have children or parents l i v i n g outside Canada or the United States. In the remaining two l i f e cycle stages, married adults with children and married adults over 45 without children, the greatest travel frequen-cies are shown by those with children or parents l i v i n g i n Western Canada or the Western United States. Adults with the lowest frequencies of travel tend to be those who do not have any children or parents l i v i n g more than 250 miles from Vancouver. The only exception to this are married adults under 45 without children, i n which case those with children (over 18 years of age) or parents l i v i n g i n Western Canada or the Western United States show a slight tendency to travel less than adults with a l l their children and their parents l i v i n g i n the Vancouver area. 97 The r e l i a b i l i t y c f the re la t ionsh ips exhib i ted by the parents/ ch i ld ren domici le va r iab le f luctuates widely over the "f i f teen c ross -c l a s s i f i c a t i o n s of domici le and l i f e cyc le stage. This i s shown i n Appendix B. I t might be suggested that the resu l t s fo r having ch i ldren or parents i n Western Canada or the Western United States are reasonably r e l i a b l e , whereas the r e l i a b i l i t y of the resu l t s fo r the other c l a s s i -f i c a t i o n s of the var iab le; i s general ly poor. Place of Schooling Respondents i n the survey were a lso asked i f they went to school more than 250 mi les away from Vancouver. I f the answer was i n the a f f i r -mative, the respondent was fur ther requested, to ind ica te where he went to school . This var iab le was scaled in to the same four c l a s s i f i c a t i o n s as the Domici le of Parents and Chi ldren var iab le discussed prev ious ly . The resu l t s fo r t h i s var iab le are i l l u s t r a t e d i n Graph IX. Graph IX ind ica tes that the most s i gn i f i can t re la t ionsh ips between place of schooling and t r a v e l frequency are i n the f i r s t two l i f e cyc le stages. In both of these stages, Single Under 45 and Married Under 45 Without Ch i ld ren, adul ts t r a v e l l i n g most f requent ly are those who have gone to school somewhere i n Canada or the United States more than 250 miles from Vancouver. Adults showing the lowest frequencies of t r a v e l are those who have gone to school outside Canada and the United States. The re la t ionsh ips are much l ess s i gn i f i can t i n the l a s t three l i f e cycle stages. However, there i s some s i m i l a r i t y between married adults with ch i ld ren and s ing le adul ts over 4 5 . In both of these stages adults 98 GRAPH IX PLACE OF SCHOOLING AND TOTAL TRAVEL FREQUENCY Pos i t i ve +.40-+.30-+.20-Adju stment+.10 To Mean Travel »° Frequency _.io Negative -.20 -.30 -.40 Within 250 M i . West Can. & US • • East Can. & US Outside Can. & US Sing le Marr ied, ..Married Married Single Under 45 Under 45 C h i l d . Over 45 Over 45 No C h i l d . No C h i l d . showing the greatest frequencies of t r a v e l are those having gone to school i n Western Canada or the Western United Sta tes, whi le those t r a v e l l i n g the leas t f requent ly are those having gone to school outside Canada and the United Sta tes . In the case of married adul ts over 45 without ch i ld ren , the general pattern exhib i ted i n the other four l i f e cyc le stages i s reversed. Those adul ts i n t h i s stage who t r a v e l the most f requent ly are those having gone to school outside Canada or the United States, whereas those t r a v e l -l i n g leas t are adul ts who went to school i n Eastern Canada or the Eastern United Sta tes . In genera l , the resu l t s fo r the school ing var iab le have a poor degree of r e l i a b i l i t y . The only exceptions are the resu l t s for adul ts i n the f i r s t three l i f e cyc le stages who went to school i n Western 99 Canada or the Western United States, or outside Canada and the United Sta tes . However, even i n these cases the r e l i a b i l i t y i s only f a i r . Place of B i r t h Respondents i n the survey were asked where they were born. For the mul t ip le regression and c l a s s i f i c a t i o n analys is t h i s va r iab le was scaled in to four c l a s s i f i c a t i o n s . These c l a s s i f i c a t i o n s were the same as those used for the Domicile of Parents and Chi ldren and the Place of Schooling va r iab les , wi th the exception that the f i r s t c l a s s i f i c a t i o n was widened to inc lude a l l of B r i t i s h Columbia rather than just a 250 mile radius around Vancouver:. Graph X ind ica tes that no two l i f e cyc le stages exhib i t patterns of assoc ia t ion between place of b i r t h and t o t a l t r a v e l frequency which are even remotely s im i l a r . B i r t h outside Canada and the United States shows s ign i f i can t pos i t i ve assoc iat ions with t r a v e l frequency i n the f i r s t three l i f e cyc le stages and the greatest negative assoc iat ion of any b i r t h c l a s s i f i c a t i o n i n the case of married adul ts over 45 without ch i l d ren . B i r t h i ns ide B r i t i s h Columbia shows subs tan t ia l l y the opposite pattern of r e la t i onsh ip ; s ign i f i can t negative associat ions i n the f i r s t three l i f e cyc le stages and the greatest pos i t i ve assoc ia t ion of any b i r t h c l a s s i f i c a t i o n i n the l a s t two stages. The re la t ionsh ip between b i r t h i n Western Canada or the Western United States and t r a v e l frequency f luc tuates widely over the l i f e cyc le , but i s s i g n i f i c a n t l y pos i t i ve i n the case of s ing le adul ts under 45 and negative i n the case of married adults with ch i ld ren and s ing le adults 100 GRAPH X PLACE OF BIRTH AND TOTAL TRAVEL FREQUENCY +.40-+.30-Posxt ive +.20K Adjustment+.IO -To Mean Travel * u Frequency -,10k -.20--.30 -.40 Negative Inside B.C. West Can. & US • - • East Can. & US Outside Can. & US i Single Married Married Marr ied Single Under 45 Under 45 C h i l d . Over 45 Over 45 No C h i l d . No C h i l d . over 45* B i r t h i n Eastern Canada or the res t of the United States shows a s i gn i f i can t negative assoc ia t ion wi th t r a v e l frequency i n the case of married adul ts under 45 without ch i l d ren , but i s only negl igeably asso-c ia ted wi th t r a v e l frequency i n the other l i f e cyc le stages. The r e l i a b i l i t y of the resu l t s fo r the Place of Schooling var iab le i s general ly poor. Perusal of Appendix B ind ica tes that the resu l t s fo r the c r o s s - c l a s s i f i c a t i o n s fo r b i r t h outside Canada and the United States have the only acceptable degree of r e l i a b i l i t y . C e l l As explained i n the previous chapter, the Vancouver survey area was div ided i n to seventeen c e l l s . Each of these c e l l s embodied a rough degree of homogeneity i n terms of the average income, occupation, edu-101 ca t ion , percentage immigration, age and type of dwel l ing of the fami l ies i n i t s const i tuent census t r a c t s . As a l l but one of these c r i t e r i a were the subject of other questions asked i n the survey and were included as separate var iab les i n the same regression equations as the c e l l va r iab le ; i t was the w r i t e r ' s expectation that the mul t ip le regression analys is ' would a t t r ibu te very l i t t l e va r ia t i on i n personal t r a v e l frequency to.- t h i s c e l l va r i ab le . However, t h i s would appear not to be the case. The r e -gression resu l t s set out i n Appendix B and" the .mult iple c l a s s i f i c a t i o n resu l t s set out i n Table IV suggest that i n cer ta in instances there are s i gn i f i can t d i f ferences i n the average personal t r a v e l frequency of the adul ts i n d i f fe rent c e l l s , which might be re la ted to the choice of r e s i d e n t i a l l oca t i on—in the sense that r e s i d e n t i a l l oca t ion may be associated wi th some socio-economic cha rac te r i s t i cs not otherwise measured. Table V presents the mul t ip le c l a s s i f i c a t i o n resu l t s fo r the c e l l va r iab le i n a graded form. P codes ind ica te a pos i t i ve re la t ionsh ip between the corresponding c e l l and l i f e cyc le stage; N codes ind ica te a negative re la t i onsh ip . The subscr ipts to these codes ind ica te the strength of the re la t i onsh ip : 1 fo r negl igeable, 2 fo r weak, 3 f o r mod-erate and 4 fo r st rong. The ac tua l mul t ip le c l a s s i f i c a t i o n values which correspond to these codes are set out at the bottom of the Table. The tabu la t ion points out that the c e l l va r iab le has i t s greatest assoc ia t ion wi th t r a v e l i n the second l i f e cyc le stage, married adults under 45 without ch i l d ren ; f i v e of the c e l l s show strong re la t ionsh ips wi th t r a v e l frequency and an add i t iona l eight c e l l s show moderate r e l -102 TABLE V RESIDENCE LOCATION AND TOTAL TRAVEL FREQUENCY T " Married Married C e l l S ingle , Under 45 Under 45 , No C h i l d . Married Chi ldren Over 45 : No C h i l d . Single lOver 45 1 N3 P l N 2 P 2 Nl 2 N4 N4 P3 N4 P4 4 P2 p 3 P2 P 3 N4 5 N3 p 3 P l N 2 N 3 6 P l N 4 P 2 P 3 P l 7 P 3 P 3 N 3 P 2 P 2 8 N l P l N l P l P2 9 Nl N2 0 Nl P l 10 P 3 P4 Nl P 2 P2 11 P 3 ' N 3 N 2 N 2 N4 12 N 4 N 3 N 3 N i 13 P 3 P 3 N 2 N 4 P 2 14 P l N4 P l N4 P 2 15 P2 P 3 P 3 P 2 0 16 P4 P3 N 2 N 3 P 2 17 P 3 P 3 P 3 N 2 P 2 Codes Pos i t i ve Deviat ions Negative Deviat ions Deviat ion Code Code Strength of Code Code Deviat ion Value Le t te r No. Relat ionship Let ter No. Value 0 to +.05 P 1 Negligeable N 1 0 to - .05 +.06 to +.10 P 2 Weak N 2 -.06 to -.10 +.11 to +. 20 P 3 Moderate N 3 -.11 to - .20 +.21 and Over P 4 Strong N 4 -.21 and Over 103 a t ionsh ips . C e l l exh ib i ts it 's, poorest assoc ia t ion wi th t r ave l i n the case of married adults with ch i l d ren ; none of the c e l l s show strong re la t ionsh ips and only f i v e may be considered even moderate. Eaeh of the other three l i f e cyc le stages exh ib i t strong re la t ionsh ips i n the case of three c e l l s . The c e l l showing the most s i gn i f i can t re la t ionsh ips with t r a v e l frequency i s C e l l 2. This c e l l shows strong assoc ia t ions with t r a v e l i n four l i f e cyc le stages and a moderate assoc ia t ion i n the other stage. Other c e l l s exh ib i t ing s ign i f i can t re la t ionsh ips are C e l l s 4 and 12. The c e l l s having the leas t s i gn i f i can t associat ions wi th t r ave l frequency are C e l l s 8 and 9; both have neg l igeab le , re la t ionsh ips with a l l but one l i f e cyc le stage, and i n each case the exception i s , cha rac te r i zed by only a weak re la t i onsh ip . With only one exception, no c e l l shows a re la t ionsh ip wi th t r a v e l frequency which i s cons is tent ly pos i t i ve or cons is ten t ly negative i n a l l l i f e cyc le stages. The exception i s C e l l 12 which was rated above as having a r e l a t i v e l y s ign i f i can t assoc ia t ion wi th t r a v e l frequency. This c e l l shows a negative re la t ionsh ip i n a l l l i f e cyc le stages. The C e l l var iab le exh ib i ts a very mixed and general ly poor degree of r e l i a b i l i t y . The resu l t s fo r some c r o s s - c l a s s i f i c a t i o n s of c e l l and l i f e cyc le stage are reasonably r e l i a b l e whi le the resu l t s for others are extremely un re l i ab le . As ind icated by perusal of Appendix B, the r e l a t i o n -ships which are most r e l i a b l e are general ly those which show the greatest magnitude of assoc ia t ion wi th a i r t r a v e l . This i s p a r t i c u l a r l y t rue i n the case of C e l l 2 which shows the most r e l i a b l e as w e l l as the strongest 104 associations with travel frequency. One exception, however, i s Cell 1 for which the results are characterized by a relatively high degree of s t a t i s t i c a l r e l i a b i l i t y while being generally weak i n magnitude. The interpretation of the c e l l variable should be considered with some care. It would not seem correct to claim that the differences i n c e l l which accompany differences i n travel frequency are necessarily a cause of those differences i n travel frequency. Both factors may be influenced by a third factor which i s not considered elsewhere i n the regression, such that the same factor that leads an adult to buy a home i n a certain part of the city also leads him to undertake a high frequency of travel. However, i f there i s such a third factor (or many such'factors) with an indirect effect upon both the Cell variable and travel frequency, we can be f a i r l y sure that i t i s not one of the other variables being considered i n the same travel frequency regression equation^ Although there i s l i t t l e doubt that there i s a f a i r l y high degree of correlation between several of the variables i n the equation (e.g. income and occu-pation), one of the valuable facets of multiple regression i s that the results which i t produces for each variable have been f i l t e r e d free from the effects of variation i n a l l the other independent variables i n the same equation. This has been discussed further i n Chapter III and i n Chapter V. 3 General Effectiveness of The Regression Analysis This chapter, i n addition to discussing the magnitude and direction of the influence exerted by each variable upon personal travel frequency, 3p. 39 and pp. 65-66 . 105 has a lso considered the s t a t i s t i c a l r e l i a b i l i t y of the regression r e s u l t s . However, each var iab le has been discussed separately from a l l the other va r i ab les . Consideration w i l l now be given to ( i ) the c o n t r i -but ion of each ent i re regression equation to the explanation of the var iance i n the data fo r the l i f e cyc le stage i t represents, and ( i i ) the contr ibut ion of s t r a t i f i c a t i o n by l i f e cyc le stage to the explanation of the t o t a l var iance. The method used to determine the amount of the t o t a l variance explained by the separate regression equations and by the s t r a t i f i c a t i o n va r iab le was variance ana lys i s . The app l ica t ion of t h i s basic s t a t i s -t i c a l method was described i n Chapter V.4 Table VI sets out the proport ion of the var iance fo r each l i f e cyc le stage which i s explained by the regression equation for that stage. I t w i l l be noted that the explained variance ranges from 28.6% fo r the Married Under 45 Without Chi ldren regression to 12.7$ f o r the Married With Chi ldren regress ion. At f i r s t glance these percentages are sur -p r i s i n g l y low. However, they appear qui te acceptable when several fac tors are considered. F i r s t l y , i t must be remembered that these regression equations are being f i t t e d to very inexact , s o c i a l and emotional phenomena rather than phys i ca l , chemical or inherent ly math-ematical phenomena. This f i r s t fac tor i s compounded by the c lose ly re la ted fac tor that the var iab les considered i n the equations are only those which were considered l i k e l y to be most important among an almost i n f i n i t e number of var iab les having some ef fect upon the i n d i v i d u a l ' s ^ p . 71-74. TABLE VI VARIANCE ANALYSIS GF REGRESSIONS FOR TOTAL TRAVEL FREQUENCY STRATIFIED BY FAMILY LIFE CYCLE STAGE Married Under 45 Married Under 45 Single Under 45 No Children Married Children No Children Single Over 45 Sum of Squares 42.11 23.1 25.87 28.6 68.86 12.7 49.94 17.7 -22.81 16.5 Explained By 1 Regression Sum of Squares 140.38 76.9 64.59 71.4 473.48 87.3 232.33 82.3 115.27 83.5 Unexplained By Regression Total Sum 182.49 100.0 90.46 100.0 542.34 100.0 282.27 100.0 138.08 100.0 Of Squares o ON 107 t r a v e l behaviour. Th i rd ly , both of these fac to rs are compounded by the large number of the observat ions, each add i t iona l observation i n j ec t i ng in to the scat ter of data some add i t iona l va r i a t i on to be explained by the equation. The resu l t s of the var iance ana lys is fo r the contr ibut ion of s t r a t i f i c a t i o n by fami ly l i f e cyc le stage to the explanation of the t o t a l var iance i n the data are set out below: Sum of Squares Explained 21.39 1.7$ By S t r a t i f i c a t i o n Sum of Squares Not Explained 1,235.64 98.3$ By S t r a t i f i c a t i o n Tota l Sum of Squares 1,257.03 100.0$ These resu l t s i nd ica te that 1.7$ of the t o t a l variance was expla-ined by l i f e cyc le stage. Considering that t h i s f igure i s only fo r one var iab le associated with t o t a l t r a v e l frequency, i t would seem to be i n I l i n e wi th the resu l t s discussed above fo r the combined explanatory con-t r i b u t i o n of a l l the other var iab les i n the ana lys i s . Both of these resu l t s are c lose ly comparable to those found by Blood i n h i s analys is of t o t a l t r a v e l frequency. This previous study reported I7.7$-19.4$ of the t o t a l variance explained by the separate regressions and .2%-2.5% explained by the l i f e cyc le stage s t r a t i f i -ca t i on . ^ Dwight M. Blood; "A Cross Sect ion Analys is of the Domestic In ter -c i t y Travel Market";(unpublished doctoral d i sse r ta t i on i n Economics, Un ivers i ty of Michigan, Ann Arbor, 1963i p. 162. 108 Conclusion The foregoing ana lys is has revealed several var iab les which, by v i r t ue of the magnitude and r e l i a b i l i t y of t he i r re la t ionsh ips . with personal t r a v e l frequency, might be termed c r i t i c a l i n the determination of the market fo r personal t r a v e l . Income Income would appear the most important s ing le va r i ab le , both i n the magnitude and the r e l i a b i l i t y of i t s re la t i onsh ips . The higher an a d u l t ' s income the more frequent ly he tends to undertake personal t r a v e l . Domici le of Parents and Chi ldren The domici le of an adu l t ' s parents and ch i ld ren has shown i t s e l f to be the second most s i gn i f i can t and r e l i a b l e va r i ab le . Two c l a s s i f i -cat ions of t h i s var iab le are p a r t i c u l a r l y important. The f i r s t i s having parents or ch i ld ren l i v i n g i n Western Canada or the Western United Sta tes . This has a very s ign i f i can t pos i t i ve re la t ionsh ip wi th the t r a v e l frequency of adul ts i n a l l l i f e cyc le stages except those married and under 45 without ch i l d ren , i n which case i t i s st rongly negat ive. The second important c l a s s i f i c a t i o n of t h i s var iab le i s having ch i ldren or parents l i v i n g outside Canada and the United Sta tes . This has a p a r t i c -u l a r l y powerful pos i t i ve re la t ionsh ip wi th the t r a v e l of s ing le adults and married adul ts under 45 without ch i l d ren . C e l l In terms of the magnitude of i t s re la t i onsh ips , C e l l might be considered the second most important va r i ab le . However, i t s general ly 1 0 9 lower degree of reliability must jrelegate i t to third place. Cell 2 shows the most significant relationships, being positive in.the case of -married adults with children and single adults over 4 5 , and negative in the case of adults in the other l i f e cycle stages. Of somewhat lesser magnitude and reliability are: Cell 1 2 , with a negative association with travel in a l l l i f e cycle stagesj and Cell 4 , with a positive relationship with travel in the first four l i f e cycle stages and a significant negative relationship in the last stage. The most likely explanation for this cell relationship i s not that residential location per se i s related to travel but rather that there are factors associated with choice of residential location which are also associated with travel. Auto Ownership Throughout the family l i f e cycle, adults who do not have an automobile travel less frequently than those who do. This relationship i s particularly significant in the case of married adults. Cottage Ownership Cottage ownership, although characteristic of only about 5 . 5 $ of the Vancouver population, exhibits a significant association with personal travel frequency. This association i s positive for a l l but married adults over 4 5 without children and i s especially strong in the case of married adults under 4 5 without children. Education Although an adult's achieved level of education (university grad-110 uate, high school graduate, grade school, some high school, or less) has only a weak association with his personal travel frequency beyond that . which i s actually an indirect association with his income and the other variables in the regression equation, those adults who are presently undertaking university education (university students) show a sig-nificantly greater tendency to travel than other adults. Although the above variables have shown themselves to be critical in the determination of the personal travel market, this Is not to say that the other variables are unimportant. Most of the remaining variables exhibit significant relationships with personal travel frequency in at least some of their cross-classifications with the family l i f e cycle, and some show, a fairly high degree of reliability. However, their asso-ciations are generally of smaller magnitude and a lesser degree of reliability than those highlighted above. Stage in the family l i f e cycle, while not worthy of classification as a critical variable, shows a significant association with personal travel frequency. An adult's personal travel frequency tends to be greatest when he i s married under 45 with no children and least when he is over 45 and single. The next two chapters will present the results of the multiple regression and classification analyses of air travel frequency and mode choice. Besides examining different aspects of personal travel behaviour and bringing into consideration additional variables, these analyses will also serve as further investigation of the variables considered during this, chapter. ' ... < CHAPTER VII AIR TRAVEL FREQUENCY Introduct ion The previous chapter was concerned wi th t o t a l personal t r a v e l f r e -quency by a l l modes combined. At tent ion w i l l now be focused s p e c i f i c a l l y upon personal t r a v e l frequency by a i r . The a i r t r a v e l frequency regression was not s t r a t i f i e d by stage i n the fami ly l i f e cyc le because of budget l i m i t a t i o n s . However, t h i s market charac te r i s t i c was included i n the equation as an independent va r i ab le . In order that expansion of the equation would be kept to a minimum t h i s var iab le was c l a s s i f i e d in to three rather than f i v e stages: Under 4 5 , s ing le or married without ch i l d ren ; Married with ch i l d ren ; and Over 4 5 , s ing le or married without ch i l d ren . The dependent var iab le i n the equation was defined according to the sca le values set out i n Chapter V. These scale values are 0 , 1 . 0 and 1 . 4 , depending on whether an adult took no personal a i r t r i p s , one t r i p or two or more t r i p s . With the exception of the l i f e cyc le stage va r i ab le , the independent var iab les i n the equation were the same as those considered i n the regressions fo r t r a v e l by a l l modes combined. As ex-pla ined i n Chapter V, one of each var iab le was excluded from the equation to permit determinacy i n the r e s u l t s . For example, the regression d id not inc lude the Univers i ty Student category of the education va r iab le . The resu l t s of the regression fo r personal a i r t r ave l frequency are set out i n Appendix C, accompanied by the standard error of each 112 regression coe f f i c i en t . The mul t ip le c l a s s i f i c a t i o n conversions of these r e s u l t s , i n terms of an adjusted deviat ion f igure fo r each regres-s ion coe f f i c i en t , are set out i n Table V I I . I t w i l l be noted that adjusted deviat ion f igures are also spec i f i ed fo r the category of each var iab le which had to be excluded from the o r i g i n a l regression equation. Because of t h i s , and fo r other reasons explained f u l l y i n Chapter V, the d iscuss ion which fo l lows w i l l re fe r mainly to these mul t ip le c l a s s i f i -cat ion conversions rather than the o r i g i n a l regression r e s u l t s . The mul t ip le c l a s s i f i c a t i o n resu l t s fo r a i r t r a v e l frequency are spec i f i ed to three decimal places rather than the two which were con-sidered i n the regressions fo r t o t a l t r a v e l frequency by a l l modes combined. This was done because the adjusted deviat ion f igures for the a i r t r a v e l regression are general ly much smaller than those fo r t o t a l t r a v e l frequency by a l l modes. This i s the resu l t of two fac to rs . F i r s t l y , an adu l t ' s a i r t r i p s must l o g i c a l l y be l ess than, or at most equal t o , h i s t o t a l t r a v e l frequency by a l l modes combined. Secondly, the sca le value fo r the frequent a i r t r a v e l l e r i s considerably l ess than that fo r the frequent t r a v e l l e r by a l l modes combined (1.4 as compared to 3.2), thus applying a,greater dampening ef fect upon the weight of the frequent a i r t r a v e l l e r . Both of these fac tors are re f lec ted i n the mean of the dependent var iab le fo r the d i f fe rent equations. The mean a i r t r a v e l frequency has an index of .12, whereas the mean t o t a l t r a v e l f r e -quency has an index ranging from .56 to .90, depending on the l i f e cycle stage. Thus, an adjusted deviat ion of .05 associated wi th a i r t r a v e l frequency w i l l have about the same r e l a t i v e impact as an adjusted TABU T i l MULTIPLE CLASSIFICATION ANALISIS OF FACTOHS APPECTIH} AIR TRAVEL FREQUEBT Variable No. of Adjusted Deviation Cases From Mean of .12 Variable No. of Adjusted Deviation Cases Prom Mean of .12 Family Life Cycle Stage Domicile of Children 4t Parents -.026 Undar A5, Single or Married 1,091 -.0*2 Within 250 Miles of Vancouver 1,890 Without Children Western Canada or Western 536 +.026 Harried With Children 1,162 +.021 U.S.*. Over 45j Single or Married 728 •.029 Eastern Canada or Rest of U.S. A. 282 +.058 Without Children Out Bids Canada & U.S.A. 371 +.053 Monetary Income Place of Schooling Under 15,000 77* -.026 Within 250 Miles of Vancouver 1,073 - .01* »5,000-»9,999 elO.OOO and Over 1,087 -.003 Western Canada or Western 735 -.003 31* +.121 U.SJL. - Refused 802 -.018 Eastern Canada or Reet of U.S. A. 309 +.023 " - - Outaide Canada «. U.S.A. 86* +.012 Auto Ownership -Owns 2,212 +.006 Place of Birth Does Not Own 769 -.017 Inside B.C. 773 -.018 Western Canada or Western 750 +.005 Boat Ownership U.S.A. Owns 185 +.050 Eastern Canada or Seat of U.S. A. *08 - .00* Does Not Own 2,796 - . X 3 - Outside Canada & U.S.A. 1,0*9 +.011 Cottage Ownership Sex -.015 Owno 150 +.058 Male 1,29* Does Not Own 2,831 -.003 Female 1,686 +.011 Dwelling Ownerehip Cell Own—No Mortgage • 810 +.00* 1 25* +.013 Own—Mortgage 1,111 -.007 2 62 -.012 Rent 1,060 +.00* * 139 -.065 5 1*9 +.025 Dwelling Type 6 187 +.027 Single-Detached 2,2*2 -.009 7 2*1 +.013 Single-Attached 2** +.021 8 325 +.028 Apt./Flat *95 +.030 9 121 -.002 10 3*9 - .00* Occupation 11 1*1 -.047 Managerial 81* +.052 12 295 +.002 White Collar 681 -.003 13 117 -.0*9 . Blue Collar 938 -.032 1* 93 -.0*0 Retired 5*7 -.017 15 266 +.007 16 5* -.033 Education 17 +.013 . Some High School, Grade School 1,390 -.016 or Lese High School Graduate - 1,261 +.017 \ , University Student . 91 -.017 University Graduate 239 +.010 114 dev ia t ion of .25 to .40 associated wi th t o t a l t r a v e l frequency by a l l modes combined. The d iscussion which fo l lows w i l l involve a de ta i led examination of each var iab le i n terms of : the magnitude and d i rec t ion of i t s i n f l u -ence upon a i r t r a v e l frequency, the s t a t i s t i c a l r e l i a b i l i t y of i t s in f luence , and the di f ference i n i t s ef fect and r e l i a b i l i t y r e l a t i ve to the resu l t s fo r personal t r a v e l frequency by a l l modes combined. This w i l l be fol lowed by an ind ica t ion of the ove ra l l contr ibut ion of the regression to the explanation of the t o t a l variance i n the data. The chapter w i l l conclude with an i nd i ca t i on of those var iab les which have shown themselves to be most important i n the determination of the personal a i r t r a v e l market. t Inf luence and R e l i a b i l i t y of Var iables  Family L i f e Cycle Stage The mul t ip le c l a s s i f i c a t i o n resu l t s fo r the re la t ionsh ip between fami ly l i f e cyc le stage and a i r t r a v e l frequency are i l l u s t r a t e d i n Graph X I . These resu l t s ind ica te that adul ts who are over 45 and s ing le or married without ch i ld ren tend to t r a v e l more frequent ly by a i r than other adults whi le those leas t l i k e l y to t r a v e l by a i r are under 45 and s ing le or married without ch i l d ren . Married adults with ch i ldren t r ave l only s l i g h t l y l ess than adults i n the former stage but considerably more than those i n the l a t t e r stage. Discussion i n the previous chapter concerning the re la t ionsh ips between the l i f e cyc le var iab le and t o t a l t r a v e l frequency by a l l modes combined, ind icated a s i tua t ion opposite to that described here for a i r 115 GRAPH XI LIFE CYCLE STAGE AND APR TRAVEL FREQUENCY P o s i t i v e +.04h +.03 +.02 Adjustment m f To Mean + ' u l * m A i r Travel .00 Frequency _ -.02 -.03 -.04 Negative Under 45 No Ch i l d . Married C h i l d . Over 45 No Ch i l d . t r a v e l frequency. Adults wi th the greatest personal t r a v e l frequency by a l l modes combined were under 45 and married without ch i l d ren , whereas adul ts showing the leas t personal t r a v e l frequency were over 45 and s ing le or married without ch i l d ren . Appendix C suggests that the re la t ionsh ip between the Under 45 Single or Married Without Chi ldren l i f e cyc le stage and a i r t r a v e l f r e -quency has a high degree of r e l i a b i l i t y . However, t,he r e l i a b i l i t y of the resu l t s f o r the second l i f e cyc le stage, Married With Chi ldren, appears very poor. Income Income exh ib i ts a very strong pos i t i ve assoc iat ion with a i r t r ave l frequency. Adults with annual income over $10,000 tend to t r a v e l more f requent ly by a i r than other adu l ts . L ikewise, those adults earning 116 income in the $5,000-$9,999 bracket show a greater frequency of air travel than those with income less than $5,000 per year. However, the difference between the significance of the latter two groups i s much smaller than the difference between these groups and. the highest income group. Graph XII illustrates. The previous chapter indicated a pattern in the relationships between income and personal travel frequency by a l l modes combined similar to that found here for air travel frequency. However, one l i f e cycle stage showed an exception. In the Single Over 45 stage, the dif-ference in significance between income over $10,000 and income between $5,000-$9,999 was shown to be less than the difference in significance between the latter income group and income below $5,000. A further similarity between the relationship of income with air travel frequency and the relationship of income with total travel fre-quency i s in the results for those adults who refused to indicate their income level. In the case of air travel frequency, those adults who refused to specify their income level show a'tendency to travel signif-icantly less than adults with income in the $5,000-$9,999 range and only slightly greater than adults with income less than $5,000. In the case of total travel frequency by a l l modes combined, this same situation was revealed for two of the l i f e cycle stages, while those adults in the remaining three l i f e cycle stages who refused to state their income level showed a tendency to travel even less frequently by air than those admitting to being in the lowest income grouping. The mixed degree of reliability characterizing the income variable 117 GRAPH XII INCOME AND AIR TRAVEL FREQUENCY +.14 + . 1 2 K + . 1 0 P o s i t i v e +.08 +.06 Ad justment+. 0 4 To Mean b A i r Travel Frequency . 0 0 - . 0 2 - . 04 - .06 - .08 Negative $10,000 or Over $5,000-$9,999 $5,000 or Less Answer Refused i s a fur ther s i m i l a r i t y between the regression fo r a i r t r a v e l frequency and the regressions fo r t o t a l t r a v e l frequency by a l l modes combined. The resu l t s fo r the top income c l a s s i f i c a t i o n , $10,000 and Over, are h ighly r e l i a b l e . However, the resu l t s fo r the middle and lower income groups show poor r e l i a b i l i t y . Auto Ownership i Fami l ies which own at l eas t one automobile t r a v e l more frequent ly by a i r than fami l ies which do not own even one automobile. This f ind ing i s consistent wi th the resu l t s fo r the re la t ionsh ip between auto ownership and t o t a l t r a v e l frequency discussed i n the previous chapter. A fur ther consistency between these resu l t s i s t h e i r moderate degree of r e l i a b i l i t y . 118 Boat and Cottage Ownership Both boat ownership and co.ttage ownership exhibit positive asso-ciations with air travel frequency which are considerably greater than that described, above for auto ownership. These relationships are also considerably greater than were found for either of these variables in the analysis of total travel frequency. In fact, boat ownership exhibited a negative association with total travel frequency in the Single Over 45 l i f e cycle stage and both boat ownership and cottage ownership showed negative associations with the total travel frequency of married adults over 45 without children. The boat ownership and cottage ownership variables show further similarity in the relatively high degree of reliability in their relation-ships with air travel frequency. In their relationships with total travel frequency, both variables, and in particular boat ownership, exhibited a poor degree of reliability. Dwelling Ownership Whether an adult owns his home without a mortgage, owns with a mortgage, or rents, has l i t t l e association with his frequency of air travel. In fact, the multiple classification results indicate no dif-ference whatsoever between the air travel frequency of adults who own a home without a mortgage and those who just rent. In the total travel frequency regressions discussed in the previous chapter, the results for the dwelling ownership variable indicated some magnitude of association but there was no apparent consistency in the direction of the relationships or in the relative significance of the 119 d i f f e r e n t dwelling ownership c l a s s i f i c a t i o n s . The insi g n i f i c a n c e of the relationship between dwelling ownership and a i r t r a v e l frequency i s further emphasized by a very poor degree of s t a t i s t i c a l r e l i a b i l i t y . The s t a t i s t i c a l r e l i a b i l i t y f o r t h i s variable i n the t o t a l t r a v e l frequency regressions was somewhat better, but s t i l l poor. Dwelling Type The type of dwelling i n which an adult l i v e s appears to have a s i g n i f i c a n t association with his frequency of a i r t r a v e l . As shown i n Graph X I I I , adults l i v i n g i n an apartment or f l a t tend to have a greater a i r t r a v e l frequency than other adults, while those l i v i n g i n a single-detached dwelling are l i k e l y to t r a v e l l e s s frequently by a i r than other adults. Adults l i v i n g i n the t h i r d type of dwelling, single-attached, tend to undertake personal a i r t r a v e l more frequently than the l a t t e r , but l e s s frequently than the former. Graph XIII also i l l u s t r a t e s the d i r e c t i o n and r e l a t i v e magnitude of the relationship between each dwelling type and a i r t r a v e l frequency. I t w i l l be noted that l i v i n g i n a single-attached dwelling has a s i g n i f -icant p o s i t i v e association which i s s l i g h t l y less than that f o r l i v i n g i n an apartment or f l a t , but considerably greater than the negative association for l i v i n g i n a single-detached dwelling. There appears l i t t l e s i m i l a r i t y between the r e s u l t s f o r the relationships between dwelling type and a i r t r a v e l frequency and the re s u l t s disoussed i n the previous chapter f o r the relationships between t h i s variable and t o t a l t r a v e l frequency by a l l modes combined. In the 120 GRAPH XIII DWELLING TYPE AND AIR TRAVEL FREQUENCY Positive +.04 +.03 +.02 • Ad j ustment+. 01 p To Mean QQ Air Travel Frequency -.01 • -.02h Negative -.03 -.04 Single-Detached Single-Attached Apt./ Flat analysis of total travel frequency,.living in an apartment or flat was revealed to have the greatest positive association with travel in only one of the five l i f e cycle stages. In fact, this dwelling classification showed a negative association with three of the l i f e cycle stages and in two of these was associated with the lowest frequencies of total travel. The only similarity between the results for dwelling type in air travel frequency and the results for this variable in total travel fre-quency would seem to be in their poor reliability. Financial Interest in Ma.jor Assets Despite the fact that the "financial interest" hypothesis was not substantiated in terms of total travel frequency, i t would seem inter-esting to test i t further on air travel frequency. Graph XIV sets out the multiple classification values for dwelling rental, apartment/flat 121 GRAPH XIV FINANCIAL INTEREST IN MAJOR ASSETS Positive +.04 • +.03 • +.02 • Adjustment*. 01 f To Mean #QQ Air Travel Frequency -.01 • -.02 • Negative -.03 --.04 Net No Effect Auto No No Rent Apt./ Boat Cottage Flat dwelling type, and. non-ownership of auto, cottage and boat, indicating their relationships with a i r travel frequency. I t w i l l be noted that the net effect of a l l these factors together i s positive, showing support for the hypothesis. However, examination of the individual value for each variable indicates that non-ownership of an auto, non-ownership of a cottage, and non-ownership of a boat a l l violate the hypothesis. Despite the fact that the "financial interest" hypothesis holds up better i n the case of ai r travel frequency than i t did i n the case of tot a l travel frequency, i t would seem f a i r to echo the same conclusions suggested i n the previous chapter: f i r s t l y , that those individuals who have the money to spend on a home are s t i l l l i k e l y to have money for travel; and secondly, that the monetary income interaction with some of the "financial interest" variables was not very successfully eliminated i n the multiple regression analysis. 122 Occupation Occupation shows f a i r l y s i gn i f i can t re la t ionsh ips with a i r t r a v e l frequency. As shown i n Graph XV, adul ts i n managerial occupations tend to t r a v e l by a i r more frequent ly than other adu l ts , - whi le those i n blue c o l l a r occupations usua l ly t r a v e l l ess than other adu l ts . The remaining two occupation c l a s s i f i c a t i o n s have more moderate associat ions with a i r t r a v e l frequency, with r e t i r e d adul ts showing somewhat lower frequencies of a i r t r a v e l than adul ts i n white c o l l a r occupations. Graph XV ind icates that the managerial c l a s s i f i c a t i o n has a s i gn i f i can t pos i t i ve re la t ionsh ip wi th a i r t r a v e l frequency r e l a t i ve to the mean a i r t r a v e l frequency fo r a l l the respondents i n the survey. The other three occupational c l a s s i f i c a t i o n s show negative re la t ionsh ips r e l a t i v e to the mean a i r t r a v e l frequency. These negative re la t ionsh ips are i n s i g n i f i c a n t fo r white c o l l a r occupations, moderate fo r ret irement, and f a i r l y s i gn i f i can t fo r blue c o l l a r occupations. In the analys is of t o t a l t r a v e l frequency by a l l modes combined, the managerial occupation c l a s s i f i c a t i o n was found to have the greatest pos i t i ve re la t ionsh ip wi th a l l but the l a s t l i f e cyc le stage. I t s re la t ionsh ip wi th the l a s t stage (Single Over 45) was shown to be negative and associated wi th a lower frequency of a i r t r ave l r t han e i ther white or blue c o l l a r occupations. The greatest negative associat ions wi th t o t a l t r a v e l frequency were shown not by blue c o l l a r occupation, as was ind icated above fo r a i r t r a v e l frequency, but rather by ret irement. In f ac t , i n three l i f e cyc le stages blue c o l l a r occupation i s associated wi th higher frequencies of t o t a l t r a v e l than both white c o l l a r occupation 123 GRAPH XV OCCUPATION AND AIR TRAVEL FREQUENCY +.05 +.04 Positive +.03' +.02 Adjustment To Mean +.01 Air Travel Q^Q Frequency-Negative -.01: -.02 -.03 -.04 VUJl I Manage./ Prof./ Prop. White Blue Retired Collar Collar and retirement. The occupation variable exhibits a mixed degree of reliability in its relationships -with air travel frequency. As indicated by examination of Appendix C, the results for the managerial classification are highly reliable, -while those for the blue collar classification are moderate and those for retirement are very poor. The reliability of this variable in i t s relationships with total travel frequency by a l l modes combined was found to be equally mixed, but generally poorer than in the case of air travel frequency. Education When income, occupation, auto ownership, and a l l the other variables in the regression equation for air travel frequency are held constant, adults with only high school graduation appear to travel by air 124 more frequently than other adults. This shows some similarity to the results for the education variable in the regressions for total travel frequency by a l l modes combined. Although the university student clas-sification showed the greatest positive association with total travel frequency, high school graduation showed a greater positive association than university graduation. Those adults with only some high school, grade school or less education were found to travel least, both by air and by a l l modes combined. Graph XVI illustrates the magnitude and direction of the relation-ship between each education classification and air travel frequency. Both high school graduation and university graduation show moderate positive associations, with that for the former being almost twice that for the latter. The classifications for university student and some high school, grade school or less education both show moderate negative associations, with that for the university student being slightly greater. The regression results for the relationship between education and air travel frequency show the same poor degree of reliability as was found to characterize the relationships between education and total travel frequency by a l l modes combined. Domicile of Children and Parents The domicile of an adult»s children and parents was shown to be one of the most important variables associated with total travel fre-quency by a l l modes combined. This variable would also appear to have a very significant relationship with air travel frequency. As illustrated GRAPH XVI EDUCATION AND AIR TRAVEL FREQUENCY Pos i t i ve +.04-+.03 • +.02 • .00 Adjustment+.Ol • To Mean A i r Travel Frequency -.01 -.02 Negative e -.03 • -.04-Some High Sch . , High Sch. Un ivers i ty Un ivers i ty Grade Sch . , Grad. Student Grad. or Less i n Graph XVII , the fur ther an adu l t ' s parents or ch i ld ren l i v e from Vancouver, the more frequent ly he i s l i k e l y to t r a v e l by a i r . Having ch i ld ren or parents l i v i n g i n Eastern Canada or the Rest of the United States ( a l l states except Washington, Idaho, Oregon and C a l i f o r n i a ) , has a pos i t i ve assoc ia t ion jus t over twice the pos i t i ve assoc iat ion of having ch i ld ren or parents l i v i n g i n Western Canada or the Western United Sta tes . Having no ch i ld ren or parents L i v ing more than 250 mi les away from Vancouver has a negative assoc ia t ion equal to the pos i t i ve assoc ia -t i o n fo r l i v i n g i n Western Canada or the Western United Sta tes. However, t h i s general distance re la t ionsh ip does not hold fo r adults having ch i ld ren or parents l i v i n g outside Canada and the United States, who tend to t r a v e l s l i g h t l y l ess f requent ly by a i r than adul ts having ch i ld ren or parents l i v i n g i n Eastern Canada or the Rest of the United Sta tes . 126 GRAPH XVII DOMICILE OF CHILDREN AND PARENTS AND AIR TRAVEL FREQUENCY +.05 +.04 Positive +.03 Adjustment"*"*02 To Mean +.Q1\-Air Travel Frequency -.01k Negative -.02--.03; -.04. Inside 250 Mi. of Vane. West Can. & West USA East" Can. & Rest USA Outside Can. & USA In contrast to these relationships for a i r travel frequency, high t o t a l travel frequency was found to be associated either with having children or parents outside Canada and the United States or i n Western Canada or the Western United States. However, similar to the relation-ships for a i r travel frequency, low t o t a l travel frequency was found to be associated with having no children or parents more than 250 miles from Vancouver. One of the factors contributing to the importance of the Domicile of Children and Parents variable i n the regressions for t o t a l travel f r e -quency was i t s high degree of r e l i a b i l i t y . The results for this variable i n the regression for a i r travel frequency have even a higher degree of r e l i a b i l i t y . 127 Place of Schooling Place of Schooling shows considerably less association with a i r travel frequency than that discussed above for Domicile of Children and Parents. However, as shown i n Graph XVIII, Place of Schooling exhibits the same general distance relationship as Domicile of Children and Parents. A s t r i c t distance relationship i s broken only by schooling outside Canada and the United States which i s associated with a lower frequency of a i r trave l than schooling i n Eastern Canada or the Rest of the United States. Similar to these results for a i r trave l , the magnitude of the associations between Place of Schooling and to ta l trave l frequency were not found to be part icular ly s ignif icant, with the exception of married adults under 45 without children. However, unlike the results for a ir trave l frequency, not even a par t ia l distance relationship between Place of Schooling and t o t a l travel frequency was found. As indicated by examination of Appendix C, the relationship between Place of Schooling and a ir travel frequency i s characterized by a poor degree of r e l i a b i l i t y ; About the same degree of r e l i a b i l i t y was shown for th is variable i n i t s associations with to ta l travel frequency by a l l modes combined. Place of Bir th Place of Birth appears even less significant i n i t s relationships with a i r travel than Place of Schooling. In addition, the general distance relationship characterizing Place of Schooling and Domicile of Children and Parents i s not shown by Place of B ir th . This i s i l lus tra ted i n Graph XIX. 128 GRAPH XVIII PLACE OF SCHOOLING AND AIR TRAVEL FREQUENCY Positive +.04-+.03 +.02 Ad j ustment+. 01 To Mean Air Travel , 0 0 Frequency -.01 -.02 Negative & -.03 -.04 Within 250 Mi. of Vane. West Can. East Can. Outside & West USA & Rest USA Can. & USA Although birth within British Columbia has the greatest negative relation-ship, birth in Western Canada or the Western United States shows only a slight positive association, while birth in Eastern Canada or the Rest of the United States has a slight negative association. Birth outside Canada and the United States shows the greatest positive relationship with air travel. The results for the relationship between Place of Birth and total travel frequency were found to be equally as insignificant as those found here for air travel frequency. In addition, there was l i t t l e consistency over the different l i f e cycle stages. The relationships between Place of Birth and air travel frequency have only a slightly better degree of reliability than that described previously for the Place of Schooling variable. However, this i s some-129 GRAPH XIX PLAGE OF BIRTH AND AIR TRAVEL FREQUENCY Positive +.04 +.03 +.02 Adjustment+.Ol To Mean Air Travel » 0 0 Frequency -.Oik Negative -.02 • -.03 • -.04 B.C. West Can. East Can. Outside & West USA & Rest USA Can. & USA what poorer than the r e l i a b i l i t y exhibited by this variable i n i t s asso-ciation with total travel frequency by a l l modes combined. In that case - 1 the results for the f i r s t l i f e cycle stage and also those for birth outside Canada and the United States showed good r e l i a b i l i t y . Sex As indicated by the multiple classification results set out i n Table VII, female adults have a significantly greater tendency to under-take personal a i r travel than male adults. This suggestion i s supported by Appendix C which indicates.a f a i r l y high degree of r e l i a b i l i t y for these results. In contrast to this, the sex variable was found to have a negligeable and very unreliable association with t o t a l travel fre-quency by a l l modes combined. 131 Cell The particular area of Vancouver i n which an adult lives would appear to have a significant relationship with his a i r travel frequency. The map on the next page (Graph XX) il l u s t r a t e s the relative air travel frequency associated with each c e l l . In terms of deviation from the average air travel frequency for a l l respondents i n the survey, Cell 8 appears to have the greatest positive association with a i r travel, whereas Cell 4 has the greatest negative association. This suggests that adults l i v i n g i n Cell 8 tend to undertake air travel more frequently than other adults while adults l i v i n g i n Cell 4 are l i k e l y to travel by a i r less frequently than other adults. Other c e l l s having significant positive associations with a i r travel frequency are Cells 5 and 6. Other c e l l s with significant negative associations are Cells 11 and 13. The C e l l variable was found to have a similar significance i n i t s relationships with t o t a l travel frequency. Although only one c e l l showed consistency i n the direction of i t s associations over* a l l the l i f e cycle i stages (Cell 12), several other c e l l s were found to have a high magnitude of association regardless of the direction of the relationship. Cell 2 j showed the greatest association, with Cells 4, 7 1 3 and 17 also showing significance. The relationships between Cell and a i r travel frequency are c characterized by poor r e l i a b i l i t y , as was also found to be the case for total travel frequency. GRAPH X T RESIDENCE LOCATION and AIR TRAVEL FREQUENCY M A R K E T C E L L S H I VERY LOW -.045; and over 133 General Ef fect iveness Of The Regression Analys is This chapter, i n addi t ion to d iscuss ing the magnitude and d i r e c -t i o n of the re la t ionsh ips between each var iab le and personal a i r t r a v e l frequency, has a lso considered the s t a t i s t i c a l r e l i a b i l i t y of the regres-s ion r e s u l t s . However, each var iab le has been discussed separately,, from a l l the other variables,. Consideration w i l l now be given to the c o n t r i -but ion of the complete regression equation to the explanation of the t o t a l variance i n the data. As ind icated by the fo l lowing tabu la t ion , the a i r t r a v e l frequency regression equation explains only 8.8$ of the t o t a l var iance. Sum of Squares Explained 30.75 8.8% By Regression Sum of Squares Unexplained 317.04 91.2$ By Regression Tota l Sum of Squares 347.79 100.0$ This i s a considerably lower proport ion than the range of 28.6$ to 12.7$ ind icated i n the previous chapter fo r the t o t a l t r a v e l frequency regressions s t r a t i f i e d by l i f e cyc le stage. However, even the 8.8$ explained variance fo r the a i r t r a v e l frequency regression does not seem surpr is ing -when i t i s considered that many of the var iab les are essen t i a l l y qua l i t a t i ve , that r e l a t i v e l y few var iab les were considered out of an i n -f i n i t e number having some ef fect upon a i r t r a v e l , and that the equation i s based on a very large number of observat ions. These factors were discussed i n more d e t a i l i n the previous chapter. 134 Conclusions The foregoing analysis has revealed several variables which, by v i r t u e of the magnitude and r e l i a b i l i t y of t h e i r relationships with a i r t r a v e l frequency, might be termed c r i t i c a l i n the determination of the market for a i r t r a v e l . Income Income shows i t s e l f to be the single most important variable i n terms of the magnitude of i t s association with a i r t r a v e l . The higher an adult's income, the more frequently he tends to undertake personal a i r t r a v e l . Domicile of Children and Parents Although l e s s s i g n i f i c a n t than income i n terms of the magnitude of i t s relationship,,the domicile of an adult's children and parents shows the most r e l i a b l e association with a i r t r a v e l frequency. A general distance relationship i s suggested, with adults having children or parents i n Eastern Canada or the Eastern United States showing greater frequencies of a i r t r a v e l than adults with children or parents i n Western Canada or the Western United States, who i n turn tend to t r a v e l more frequently by a i r than adults having no children or parents more than 250 miles from Vancouver. However, a s t r i c t distance relationship i s violated by the tendency for adults with children l i v i n g outside Canada and the United States to t r a v e l somewhat less frequently by a i r than adults with children or parents l i v i n g i n Eastern.Canada or the Eastern United States. 135 Cell In terms of the magnitude of i t s relationships, C e l l might be considered the second most important variable, but i t s generally lower degree of r e l i a b i l i t y , relative to both Income and Domicile of Children and Parents, must put i t i n third place. The c e l l showing the greatest association i s Cell l+, with a very significant negative relationship with a i r travel frequency. Other c e l l s having significant negative associ-ations are Cells 11, 13 and 14. Cells exhibiting notable positive associations are Cells 5, 6 and 8. Occupation Contrary to what i s often expected, occupation appears to have a significant relationship with personal a i r travel frequency beyond that which i s actually an indirect association with other variables considered i n the equation (e.g. income and education). Managerial occupations are associated with the highest frequencies of ai r travel, followed by white collar, retirement and blue collar. Family L i f e Cycle Stage In Chapter VI married adults under 45 without children were revealed to have the greatest t o t a l travel frequencies by a l l modes combined. In this chapter, however, i t has been shown that the adults with the greatest tendency to travel by a i r are those over 45 and single or married without children. Cottage and Boat Ownership Adults who own summer cottages or boats over 14 feet i n length (or 136 both) show a significantly greater frequency of personal air travel than other adults. These variables are also characterized by a fairly high degree of statistical reliability. Although the above variables have been revealed to be of particular importance in the determination of the air travel market, this is not to suggest that the other variables considered in the analysis are unimpor-tant. However, the relationships of the remaining variables are generally of lower magnitude and a poorer degree of reliability than the variables pointed out here. The next chapter will present the results of the multiple regres-sion and classification analyses of mode choice: air vs auto, air vs r a i l and air vs bus. Besides examining a different aspect of personal travel behaviour and bringing into consideration additional variables, these analyses will also.serve as further investigation of the variables considered during these last two chapters. CHAPTER VIII CHOICE OF MODE Introduction The preceding chapter was concerned with the factors affecting a i r travel frequency. This chapter w i l l present the results of the multiple regression and classification analyses of factors influencing the choice of mode i n personal travel. Three paired choices w i l l be considered: the choice between a i r and auto, the choice between a i r and r a i l , and the choice between a i r and bus. The reasons for pairing the choices rather than considering a l l modes together were discussed i n Chapter V."*" The data for the analyses of mode choice were obtained from infor-mation provided by each respondent concerning his most-recent and next-last t r i p s . Thus, detailed data were available from a maximum of two trip s per respondent. Whereas the dependent variable for the ai r travel frequency regres-sion was measured on a scale having three positions, the dependent v a r i -ables i n the mode choice regressions are defined as zero-one dichotomies. This estimation procedure i s equivalent to that of assigning linear probabilities to observations by means of a linear discriminant function. For each paired choice, trips by ai r were assigned the scale value of 1.0 while the other mode i n the choice (auto, r a i l , bus) was assigned the scale value of zero. Thus, the regression results could be interpreted i n terms of the estimated probability of choosing a i r travel. The d i f -•pp. 57-59. 138 ference between the p robab i l i t y of choosing a i r t r a v e l and 100$ would then be the p robab i l i t y of choosing the other mode ' in the p a i r . As ind icated i n Chapter V, the mean of the dependent var iab le i n each mode choice equation i s formulated on the bas is of only most-recent and nex t - las t t r i p s and, thus, does not provide a true i nd i ca t i on of the r e l a t i v e frequency with which each of the paired.modes i s chosen.2 For t h i s reason the means of the dependent var iab les considered i n t h i s ana lys is are not those produced by the regression equations themselves, but rather those ca lcu la ted independently from the t o t a l number of t r i p s by each mode during the survey year . The regressions for t o t a l t r a v e l frequency and a i r t r a v e l frequency considered only market cha rac te r i s t i c s , i . e . cha rac te r i s t i cs of the persons taking the t r i p s , such as income and education. The mode choice regressions replace some of these market var iab les wi th var iab les which descr ibe the charac te r i s t i cs of the t r i p i t s e l f . The var iab les for occupation, education, domici le of ch i ld ren and parents, place of schooling and place of b i r t h are replaced by var iab les ind ica t ing the durat ion, distance and purpose of the pa r t i cu la r t r i p . The respondents' t o t a l t r a v e l frequency i s a lso included as an independent va r i ab le . In the case of the A i r vs Auto regress ion, one fur ther var iab le i s added which measures the inverse of the number of persons who went on the t r i p . Based on the assumption that the t o t a l cost of t r a v e l by common ca r r i e r increases i n proport ion to the number who t r a v e l i n a party whereas the marginal cost fo r add i t iona l passengers i n an auto i s negl igeable up to 2pp. 67-68. 139 a party of five or six, i t was expected that groups of five or six people would have a greater tendency to choose auto than persons travelling alone or in smaller groups. The discussion which follows will involve a detailed examination of each variable in terms of both the magnitude and direction of i t s relationships with choice between each mode pair, and the statistical reliability of i t s influence. This will be followed by an indication of the overall contribution of each regression equation to the explanation of the variance in the data. The chapter will conclude with an indication of those variables which have shown themselves to be most important in the deterinination of mode choice. Influence and Reliability of Variables The results of the multiple regression analysis for each mode choice pair are set out in Appendix D. The conversions of these results into the multiple classification format are presented in Table VIII. Table VIII also shows the mean of the dependent variable for each equation. These means represent the total number of air trips taken by the survey respondents during the survey year as a proportion of the i combined total of these air trips and the trips taken on the other mode in the particular paired choice. Thus, for example, the mean of .18 for the Air vs Auto regression indicates that air was the mode used in 18% of a l l the trips by either air or auto. This may be interpreted as the probability that an adult will take his next personal trip by air when nothing is known about the particular adult or the characteristics of TABLE VIII MULTIPLE CLASSIFICATION ANA1ISIS OF FACTORS AFFECTING MODS CHOICE Variable "  Income Under 15,000 *5,00O-$9,999 $10,000 and Over Answer Be fused Auto Ownership Own Not Own Boat Ownership Own Not Own Cottage Ownership 0^*m Not Own Dwelling Ownership Own—No Mortgage Own Mortgage Btont Dwelling Hype Single-Deta ched Single-Attached Apt./Plat Life Cycle Stage Under 45, Single or Married Without Children • Married With Children Over 45, Single or Married Without Children Time Away 6 Days or Leas 7-14 Days 15 Days or Over Inverse of No. Who Went Seal One Two Three, Four, Seven or More Five, Six Distance Leas Than 700 Miles 700-1,000 Miles 1,000-2,000 Miles 2,000-4,000 Miles 4,000 Miloe and Over Frequency of Travel 1 TripAaar 2 Trips/Tear 3 Trips or More/loar Purpose of Trip Vacation Personal: Emergency Personal: Non-Emergency Partly Business Sex Male Female Cell 1 . -2 U 5 6 7 . 8 9 10 11 12 13 U 15 16 17 ' No. of Cases Adjusted Deviation Prom Mean of .18 No. of Cases Adjusted Deviation Prom Mean of .71 No. of Cases Adjusted Deviation Prom Mean of .74 317 -.03 112 -.07 128 -.07 771 .00 153 +.02 131 +.08 310 +.04 94 +.03 83 +.04 448 .00 109 +.01 117 -.04 1,60.7 -.01 448 .00 435 +.01 229 •.08 20 -.04 24 -.13 142 . •.04 38 -.02 35 +.01 1,704 .00 430 .00 424 .00 120 •.04 43 +.02 36 +.08 1,726 .00 425 .00 423 -.01 464 •.01 118 +.02 124 .00 770 •.01 158 +.01 141 -.01 612 -.02 192 -.02 194 +.01 1,413 .00 312 -.03 295 -.02 280 +.01 37 +.11 44 -.05 153 +.01 119 +.05 120 •.06 389 +.08 202 -.03 184 +.04 1,193 -.03 130 +.04 151 .00 264 +.02 136 +.01 124 -.05 529 +.01 67 +.20 81 •.04 799 .00 130 •.06 140 -.03 518 -.01 271 -.08 238 .00 Position 10 198 +.33 5 661 +.04 3 727 -.08 2 260 -.13 845 -.14 27 -.05 63 -.35 530 -.06 122 -.17 106 -.13 88 -.06 49 -.42 25 -.03 185 +.11 103 -.10 101 -.02 198 +.68 167 +.31 164 +.23 985 .00 272 .00 273 -.03 466 +.01 132 -.04 120 +.04 395 -.01 64 +.10 66 +.05 1,134 -.04 159 -.09 184 -.11 33 +.31 26 +.23 27 +.U 629 +.06 269 +.03 225 +.08 50 -.01 14 +.04 23 -.07 857 -.01 165 +.01 156 +.03 989 +.01 303 .00 303 -.02 187 +.04 59 -•.08 55 •.09 39 -.05 6 +.14 7 -.06 109 -.04 • 12 -.20 9 -.03 74 +.06 39 / -.05 41. .00 148 .00 48 +.01 48 +.03 150 +.02 54 +.02 49 +.05 168- +.06 71 +.06 78 .00 87 +.01 H +.21 13 +.20 212 -.04 35 -.03 39 -.08 77 -.05 17 -.19 15 -.19 105 +.02 37 .00 40 -.06 75 +.01 10 -.02 9. •.05 48 * -.01 9 -.11 8 -.04 212 -.02 33 -.06 31 -.07 30 .00 8 -.10 6 -.05 125 -.04 16 -.13 11 +.02 141 his trip and when i t is assumed that he is restricted to a choice between air and auto. As further knowledge about the particular adult and his trips are assumed, this mean probability of choosing air over auto is adjusted by the appropriate adjusted deviation figures. These adjusted deviation values are set out in Table VIII and will be discussed below. Income Income shows very moderate relationships with a l l three mode choice pairs. As illustrated in Graph XXI, this i s particularly true in the case of Air vs Auto. Adults in the higher income levels are shown to have a slightly greater tendency to choose air than adults in the lower income levels. Those who refused to answer the income question exhibit the same propensity to travel by air as adults in the $5,000-$9,999 income clas-sification. In the Air vs Rail regression, income shows a slightly greater positive association with air travel than that indicated above for Air vs Rail. The difference between the Under $5,000 and $5,000-$9,999 income classification i s considerably more significant than the difference between the $5,000-$9,999 classification and the Over $10,000 classification. Adults refusing to specify their income classification show a tendency to travel by air just slightly less than those in the $5,000-$9,999 clas-sification. The regression for choice between air and bus assigns income a greater significance than in either of the above paired choices. However, the direct relationship between level of income and propensity to travel by air, shown in both situations above, i s not evident here. Adults in 142 GRAPH XXI INCOME AND MODE CHOICE Adjustment To Mean Choice Auto vs A i r R a i l vs A i r Bus vs A i r Under $5,000 $5,000-$9,999 $10,000 & Over Refused 1/5" = .10 Deviat ion From Mean Choice the $5,000-$9,999 income c l a s s i f i c a t i o n have a greater tendency to choose a i r than those i n the Over $10,000 c l a s s i f i c a t i o n . Although adults with income Under $5,000 are those most l i k e l y to choose bus, t he i r propensity towards t h i s a l te rna t i ve i s only s l i g h t l y greater than that shown by the survey respondents who refused to ind ica te t h e i r income l e v e l s . The s t a t i s t i c a l r e l i a b i l i t y of the re la t ionsh ips between income and mode choice i s general ly qui te poor. Auto Ownership Auto ownership exh ib i ts about the same magnitude of assoc iat ion wi th mode choice as that discussed above fo r income. The d i rec t ion of the re la t ionsh ip var ies between the d i f fe ren t choice p a i r s . When the ' a l t e r -143 nat ives are a i r and auto, adul ts who do not own an auto have a greater tendency to choose a i r than adul ts who do have an auto. However, i n the case of A i r vs R a i l and A i r vs Bus, non-ownership of an auto i s associated more wi th choosing r a i l or bus than a i r . The r e l i a b i l i t y of the re la t ionsh ip between auto ownership and mode choice i s general ly good, p a r t i c u l a r l y i n the case of A i r vs Auto and A i r vs Bus. Boat and Cottage Ownership Both boat ownership and cottage ownership are i n s i g n i f i c a n t i n t h e i r re la t ionsh ips with mode choice. I f any considerat ion can be given to the r e s u l t s , i t would be suggested that cottage ownership i s associated wi th the choice of a i r i n a l l three choice p a i r s , whi le boat ownership i s associated with the choice of a i r i n a l l but choice s i tua t ions invo lv ing a i r and r a i l . The ins ign i f i cance of these resu l t s are fur ther emphasized by very poor r e l i a b i l i t y . Dwell ing Ownership Whether an adult owns h i s home without a mortgage, with a mortgage, or rents appears to have a negl igeable re la t ionsh ip with h i s choice of mode. I f the resu l t s for t h i s var iab le may be taken to ind ica te anything at a l l , i t would be suggested that dwel l ing ownership, e i ther with or without mort-gage, i s associated wi th a greater tendency to choose a i r i n both A i r vs Auto and A i r vs R a i l than dwell ing r e n t a l . However, the s i tua t ion i s r e -versed i n A i r vs Bus, wi th dwel l ing ren ta l being associated with choosing a i r . The re la t ionsh ips between dwel l ing ownership and mode choice are 144 generally characterized by poor r e l i a b i l i t y . However, as indicated by examination of Appendix D, the results for Air vs Auto are an exception with f a i r l y moderate r e l i a b i l i t y . Dwelling Type Dwelling type exhibits the same negligeable relationship with the choice between a i r and auto as that discussed above for the dwelling ownership variable. However, as illu s t r a t e d i n Graph XXII, i t s associa-tions are somewhat greater i n the cases of Air vs Rail and Air vs Bus. In the case of Air vs Rail, adults l i v i n g i n single-attached dwellings have a greater tendency to choose air than other adults, while adults l i v i n g i n single-detached dwellings are those most l i k e l y to choose r a i l . In the case of Air vs Bus, adults l i v i n g i n an apartment or f l a t have a greater tendency to choose air than other adults, whereas those most l i k e l y to choose bus l i v e i n single-attached dwellings. The s t a t i s t i c a l r e l i a b i l i t y of the relationship between dwelling type and mode choice i s generally poor. Lif e Cycle Stage L i f e cycle stage shows only a slight relationship with mode choice. This i s ill u s t r a t e d i n Graph XXIII. In situations of choice between a i r and auto or between air and bus, adults most l i k e l y to choose air are under 45 and single or married without children. Adults having the greatest tendency to choose auto are married with children, while those most l i k e l y to choose bus are over 45 and single or married without children. When the situation involves a 145 GRAPH XXII DWELLING TYPE AND MODE CHOICE Adjustment To Mean Choice Auto vs Air Rail vs Air Bus vs Air Single-Detached i • Single-Attached Apt./Flat l/5" = .10 Deviation From Mean Choice choice between air and r a i l , air has the greatest tendency to be chosen by married adults with children, while r a i l i s most likely to be chosen by adults who are under 45 and single or married without children. The; relationships between family l i f e cycle and Air vs Auto show a high degree of reliability. This i s somewhat better than the reliability shown by the results for Air vs Bus and considerably better than the results for Air vs Rail. Duration of the Trip The duration of the trip (the time away, between leaving Vancouver and returning to Vancouver) appears to have very l i t t l e significance in choice situations involving air and auto or air and bus. This i s illustrated in Graph XXIV. 146 GRAPH XXIII LIFE CYCLE STAGE AND MODE CHOICE Adjustment To Mean Choice Auto vs Air Rail vs Air Bus vs Air Under 4 5 , Single or Married Without Children Married With Children Over 4 5 , Single or Married Without Children 1 / 5 " = . 1 0 Deviation From Mean Choice If these results may be attributed any significance, i t would be suggested that the choice of air i s associated with t r i p s of short duration (6 days or less), while bus i s associated with medium duration t r i p s (7 to 14 days), and auto with t r i p s of long duration (15 days or more). In the case of Air vs Rail, however, the duration of the t r i p appears to be somewhat more significant. Adults going away for a very short period have a significantly greater tendency to choose air than other adults, while those going away for a long period are more l i k e l y to choose r a i l than other adults. The relationships between t r i p duration and the Air vs Auto and Air vs Bus mode choice pairs are characterized by poor r e l i a b i l i t y . However, as indicated by perusal of Appendix D, the Air vs Rail results 147 GRAPH XXIV TRIP DURATION AND MODE CHOICE Auto vs Air Adjustment To Mean Choice • Rail vs Air Bus vs Air 6 Days or Less 7-14 Days 15 Days or Over 1/5" = .10 Deviation From Mean Choice show a f a i r l y high degree of r e l i a b i l i t y . Inverse of the Number Who Went A variable was included i n the Air vs Auto regression which measured the inverse of the number of people who went together on the t r i p . As the t o t a l cost of travel by a i r increases i n proportion to the number who travel i n the party, whereas the marginal cost for additional passengers i n an auto i s negligeable, i t was expected that adults travel-ling i n groups of 5 or 6 would be less l i k e l y to choose a i r than adults travelling alone or i n smaller groups. The results support this expec-tation. Graph XXV shows that groups of five or six persons have a greater tendency to travel by auto than groups of three, four, seven or more. A 148 GRAPH XXV i INVERSE OF NUMBER WHO WENT ON TRIP & CHOICE BETWEEN AIR & AUTO Adjustment To Mean Choice Auto vs A i r Three, Four, Seven or More F i v e , S i x One 1/5" = .10 D e v i a t i o n From Mean Choice p o s s i b l e r a t i o n a l i z a t i o n f o r t h i s i s t h a t an automobile (w i t h room f o r f i v e o r s i x persons) w i l l g i v e a lower cost o f t r a v e l per person f o r groups of f i v e or s i x than f o r groups of t h r e e or f o u r . Groups of seven or more people would i n v o l v e the a d d i t i o n a l c o s t s o f running a second automobile and, thus, would mean a higher cost per person than a group of f i v e o r s i x ( u n l e s s , o f course, the group c o n s i s t s of 10-12 persons, i n which case the c a p a c i t y o f both automobiles would be f u l l y u t i l i z e d ) . In keeping w i t h t h i s r a t i o n a l i z a t i o n , t h e Graph a l s o i n d i c a t e s t h a t a d u l t s t r a v e l l i n g i n . p a r t i e s of two are more l i k e l y t o choose a i r than a d u l t s t r a v e l l i n g i n l a r g e r groups, and a d u l t s t r a v e l l i n g alone have even a g r e a t e r tendency t o choose a i r than p a r t i e s of two. The r e l i a b i l i t y of the r e l a t i o n s h i p between the number who went on 149 the t r i p and the choice between a i r and auto i s extremely good. Distance of The Trip Trip distance ( i . e . the one-way distance to the furthest point from Vancouver) i s very s i g n i f i c a n t i n a l l the mode pair situat i o n s . Graph XXVI i l l u s t r a t e s . The r e s u l t s f o r A i r vs Auto and A i r vs Bus suggest a direct relationship between t r i p distance and the propensity to choose a i r ; the greater the distance of the t r i p , the greater i s the tendency for the adult to choose a i r rather than auto or bus. However, t h i s same direct r elationship i s not apparent i n A i r ys R a i l . A l l else considered, adults are more l i k e l y to choose a i r over r a i l on short distance t r i p s (less than 1,000 miles) than they are on medium distance t r i p s (1,000-2,000 miles). Of course, t r i p s to destinations over 4,000 miles from Vancouver ( i . e . Overseas) must involve the choice of a i r , purely because of the natural l i m i t a t i o n s of auto, r a i l and bus. The r e l i a b i l i t y of the relationships between distance and mode choice i s extremely good i n a l l three mode choice p a i r s . Frequency of Travel As indicated by Graph XXVII, the frequency with which an adult undertakes personal t r a v e l has very l i t t l e significance i n his choice of mode. This i n s i g n i f i c a n c e i s p a r t i c u l a r l y apparent i n situations of choice between a i r and auto. In the cases of A i r vs R a i l and A i r vs Bus, the r e s u l t s give a s l i g h t suggestion that a i r t r a v e l i s most l i k e l y to be chosen by adults taking three or more personal t r i p s a year, while those 150 GRAPH XXVI DISTANCE AND MODE CHOICE Adjustment To Mean Choice Auto vs A i r R a i l vs A i r Bus vs A i r Less Than 700 Mi les 700-1,000 M i les K 1,000-2,000 M i les p 2,000-4,000 M i les 4,000 M i les and Over W/A 1/5" = .10 Deviat ion From Mean Choice most l i k e l y to go by r a i l tend to take two personal t r i p s a year and those most l i k e l y to t r a v e l by bus take only one such t r i p a year . The r e l i a b i l i t y of the resu l t s fo r Tr ip Frequency seems to match the magnitude of i t s re la t i onsh ips ; i t s r e l i a b i l i t y i s very poor i n the case of A i r vs Auto, somewhat bet ter i n A i r vs Bus, and qui te good i n the case of A i r vs R a i l . Purpose of The Tr ip The purpose of the t r i p appears to have a f a i r l y s ign i f i can t re la t ionsh ip wi th the mode of t r a v e l chosen. This i s shown i n Graph XXVII I . 151 GRAPH XXVII TRAVEL FREQUENCY AND MODE CHOICE Adjustment To Mean Choice Auto vs Air Rail vs Air Bus vs Air 1 Trip/Year 2 Trips/Year 3 Trips or More/Year 1/5" = .10 Deviation From Mean Choice As set out i n Graph XXVIII, the results for the Trip Purpose variable show a high degree of consistency between the three mode choice pairs. Adults most l i k e l y to choose a i r travel are those on trips involving personal emergency. Those having the greatest tendency to choose an alternative to air, whether auto, r a i l , or bus, are adults on vacation t r i p s . Personal t r i p s which are not concerned with an emergency and t r i p s which are-partly for business reasons are less l i k e l y to lead to a i r travel than personal emergency tr i p s , but more l i k e l y to involve this mode than vacation t r i p s . It i s also apparent that persons on non-emergency personal t r i p s have a slig h t l y greater tendency to choose a i r than those on partly business t r i p s , with the exception of tr i p s for which there i s only a choice between a i r and r a i l , i n which case there i s 1 5 2 GRAPH XXVIII TRIP PURPOSE AND MODE CHOICE Auto vs Air Adjustment To Mean Choice • Rail vs Air  Bus vs Air Vacation Personal: Emergency W777A Personal: Non-Emergency Partly Business 1 / 5 " = . 1 0 Deviation From Mean Choice only a negligeable difference. The r e l i a b i l i t y of the relationships between Trip Purpose and mode choice varies between the different Trip Purpose classifications. The results for the Vacation classification show a f a i r l y high degree of r e l i a b i l i t y . The Personal: Emergency classification i s characterized by a moderate degree of r e l i a b i l i t y . The r e l i a b i l i t y of the relationships for the Partly Business classification i s poor. Whether an adult i s male or female appears to have a negligeable association with his choice of mode. If the results for this variable are to be given any consideration, i t would be suggested that adults Sex 153 having the greatest tendency to choose a i r i n A i r vs Auto are female, whereas those most l i k e l y to choose a i r i n A i r vs R a i l and A i r vs Bus are male. However, i n contrast to t he i r smal l magnitude, these r e l a t i o n -ships are character ized by a f a i r l y high degree of r e l i a b i l i t y . C e l l The resu l t s for the re la t ionsh ips between c i t y c e l l and mode choice are i l l u s t r a t e d i n Table IX. The deviat ion values fo r the re la t ionsh ips between each c e l l and each mode choice pa i r are described by codes. The l e t t e r s i g n i f i e s the d i rec t ion of the re l a t i onsh ip : P fo r associat ions with choosing a i r and N f o r assoc iat ions wi th choosing the a l te rna t ive mode, auto, r a i l or bus, depending upon the pa r t i cu la r mode choice p a i r . The subscr ipt numbers Indicate the strength of the re l a t i onsh ip : 1 fo r negl igeable, 2 fo r weak, 3 fo r moderate, and 4 fo r strong. Thus a code of P3 fo r A i r vs R a i l i n C e l l 2 s i g n i f i e s that adul ts l i v i n g i n C e l l 2 have a moderately greater tendency to choose a i r over r a i l than other adu l ts . Table IX ind ica tes that the c e l l i n which an adult res ides has very l i t t l e importance i n h i s choice between a i r and auto. However, i n choice s i tua t ions invo lv ing a i r and r a i l or a i r and bus, the C e l l var iab le shows severa l s ign i f i can t re la t i onsh ips . In the case of A i r vs R a i l , adul ts l i v i n g i n C e l l 9 have a greater tendency to choose a i r . than other adu l ts , whereas those i n C e l l 4 are more l i k e l y than other adul ts to choose r a i l . Several other c e l l s show moderate importance i n t h i s mode choice p a i r : C e l l 2 i s re la ted to the choice of a i r and C e l l s 11, 14, 16 and 17 are associated wi th choosing r a i l . 1 5 4 TABLE IX RESIDENCE LOCATION AND MODE CHOICE C e l l A i r vs Auto A i r vs R a i l A i r vs Bus 1 2 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 P l N 2 Nl P2 0 P l P2 P l N l N2 P l P l N l N l 0 N l P2 P 3 N 4 N 2 P l P l P2 P4 N l N3 0 N l N3 N 2 N3 N3 P2 N 2 N i 0 P l P2 0 P4 N2 N3 N 2 P2 N l N 2 N 2 P l Codes Deviat ion Values 0 to + . 0 4 +.05 to +.09 +.10 to +.19 +.20 and Over Code Code Strength of Le t ter No. Relat ionship P 1 Negligeable P 2 Weak P 3 Moderate P 4 Strong Code Code Deviat ion Le t te r No. Values N 1 0 to + . 0 4 N 2 - . 0 5 to - . 0 9 N 3 - . 1 0 to - . 1 9 N 4 L - . 2 0 and Over 155 Two of the cells found to be significant in Air vs Rail also show considerable importance in Air vs Bus. In situations involving choice between air and bus, adults living in Cell 9 are more likely to choose air than other adults, while those living in Cell 11 have the greatest tendency to choose bus. The reliability of the relationships between Cell and mode choice varies widely, both between mode choice pairs and between the different cells, but must generally be labelled as poor. Examination of Appendix D indicates that the results for Air vs Bus are particularly unreliable.• Cells 9 and 11, which were shown above as having the greatest magnitude of relationship with mode choice, are not exceptions to this generally poor reliability. General Effectiveness Of The Regression Analyses This chapter, in addition to discussing the magnitude and direction of the relationships between each variable and mode choice, has also i considered the statistical reliability of the regression results. However, each variable has been considered separately from a l l the other variables. Consideration will now be given to the contribution of each complete regression equation to the explanation of the total variance in the data for the mode choice pair that i t represents. Table X sets out the; proportion of the variance for each mode choice pair which is explained by the regression equation for that pair. It will be noted that the explained variance ranges from 33.1$ in the case of Air vs Rail, to 41.2$ for Air vs Bus and 59.0$ for Air vs Auto. ITABLE X VARIANCE ANALYSIS OF REGRESSIONS FOR MODE CHOICE A i r vs Auto A i r vs R a i l A i r vs Bus Sum of Squares Explained By Regression 164.38 • 59.0 30.34 t 33.1 35.47 <k 41.2 Sum of Squares Unexplained By Regression 114.19 4L.0 61.21 66.9 50.59 58.8 Tota l Sum of Squares 278.57 100.0 91.55 100.0 . 86.06 100.0 O N 157 These resu l t s are considerably higher than those exhibi ted by e i ther the Tota l Travel Frequency regressions, or the regression fo r A i r Travel Frequency. The former ranged between 12.7$ and 28.6$ whi le the l a t t e r was an even poorer 8.8$. The s i g n i f i c a n t l y bet ter showing by the mode choice regressions would seem to be a t t r ibu tab le to the fac t that they included several very powerful var iab les measuring the charac te r i s t i cs of the t r i p s : Distance, Inverse of the Number Who Went, and Purpose. These va r i ab les , and i n pa r t i cu la r Distance, showed very large regression co-e f f i c i e n t s and very small standard e r ro rs . Conclusions The foregoing ana lys is has revealed several va r iab les which, by v i r t ue of the magnitude and r e l i a b i l i t y of t h e i r re la t ionsh ips with mode choice, might be termed c r i t i c a l i n the determination of the markets fo r t r a v e l by the var ious modes. Tr ip Var iables Distance The distance of the t r i p , from Vancouver to the point of des t ina-t i o n , has been shown t d b e the s ing le most important var iab le associated wi th mode choice, both i n terms of the magnitude and the r e l i a b i l i t y of i t 's r e l a t i onsh ips . In both the case of A i r vs Auto and the case of A i r vs Bus, there i s a d i rec t re la t ionsh ip between t r i p distance and the propensity to choose a i r ; the greater the distance involved i n the t r i p , the greater the p robab i l i t y that the adult w i l l choose a i r rather than auto. However, t h i s d i rec t re la t ionsh ip does not hold fo r s i tua t ions 158 invo lv ing a choice between A i r and R a i l . In t h i s case, t r i p s invo lv ing short distances ( less than 1,000 mi les) are more l i k e l y to lead to the choice of a i r than t r i p s invo lv ing distances i n the 1,000-2,000 mi le range. Tr ips to dest inat ions more than 4,000 mi les from Vancouver w i l l na tu ra l l y invo lve the choice of a i r rather than r a i l , auto or bus, by v i r t ue of the natura l l im i t a t i ons of the l a t t e r . Inverse .of the Number Who Went The inverse of the number who went on the t r i p was shown to have a very s i gn i f i can t and h igh ly r e l i a b l e assoc iat ion with the choice between a i r and auto. I t was found that the propensity to choose auto over a i r increases with the s i ze of the party t r a v e l l i n g together, up to the point where those t r a v e l l i n g together f i l l the capacity of an automobile. Thus, assuming an automobile capaci ty of f i ve or s i x passengers, adul ts t r a v e l -l i n g i n a group of f i v e or s i x are more l i k e l y to choose auto than adults t r a v e l l i n g i n a group of l ess than f i v e , or greater than s i x . Purpose Although l e s s important than e i ther of the var iab les discussed above, both i n terms of the magnitude and the r e l i a b i l i t y of i t s r e l a t i o n -sh ips, T r ip Purpose was found to have a s i gn i f i can t assoc iat ion wi th mode choice. In a l l three mode choice pa i r s , adul ts having the greatest tendency to.choose a i r are those on t r i p s invo lv ing personal emergency whi le those having the greatest tendency to choose the a l te rna t ive mode (auto,, r a i l or bus) are those on vacat ion t r i p s . Personal t r i p s which are not for an emergency and t r i p s which are pa r t l y for business are l ess 159 l i k e l y to lead to a i r travel than personal emergency trips, but are more l i k e l y to involve this mode than vacation t r i p s . It was also found that persons on non-emergency personal t r i p s have a somewhat greater tendency to choose a i r than those on partly business trips, with the exception of trip s for which there i s only a choice between a i r and r a i l , i n which case there i s only a negligeable difference. Market Variables The analysis also investigated the relationships of certain market variables with mode choice. These were: income, auto ownership, boat ownership, cottage ownership,- dwelling ownership, dwelling type, travel frequency, sex, and c e l l . However, none of these variables were found to have associations with mode choice which could be considered significant, either i n terms of magnitude or i n terms of r e l i a b i l i t y . CHAPTER IX CONCLUSIONS The purpose of t h i s study i s to determine the c r i t i c a l var iab les i n the Vancouver personal t r a v e l market and to suggest add i t iona l surveys that could be undertaken. The foregoing chapters have ind icated the importance of market informat ion, reviewed past t r ave l market s tud ies , described the research methods used i n the analys is of the Vancouver Survey, and presented the resu l t s of the ana lys i s . To summarize and conclude, t h i s f i n a l chapter w i l l b r i e f l y ind ica te the advantages of the methodology used i n t h i s study, h igh l igh t the resu l t s of the ana lys is , point out b r i e f l y the methodological shortcomings of the study, and make suggestions as to fur ther surveys which might be undertaken. Advantages of the Methodology As a point of departure fo r the ana lys is of the Vancouver Survey, a review of past t r ave l market studies was undertaken. While t h i s review was probably not representat ive of the studies made by a i r l i n e s , due to t he i r con f iden t ia l nature, some general conclusions were suggested con-cerning the studies ava i l ab le . Aside from the l im i t a t i ons i n terms of scope or the types of data co l l ec ted , a ser ious shortcoming i n most of the studies was found to be i n the nature and extent of t he i r analyses. Even some of the most exhaustive studies have not gone beyond simple b i va r ia te l i s t i n g s of each separate t r i p and market va r i ab le , with the number or percentage of respondents as the quant i fy ing va r i ab le . This type of ana lys is provides only a loose general character sketch of the 161 ent i re survey sample. The market i s described i n terms of the proportion of the t o t a l respondents exh ib i t ing each cha rac te r i s t i c , with no i n d i -cat ion of the di f ferences i n the cha rac te r i s t i cs of the t r i p s taken by d i f fe rent segments of the market. Only a few studies have ventured to consider the in te r re la t ionsh ips between t r i p var iab les and market va r iab les . This i s the primary ing re -dient i n t rue market research. By matching t r i p cha rac te r i s t i cs to market cha rac te r i s t i cs the f i rm can get some idea of the type of serv ice , adver-t i s i n g appeal, p r i c ing p o l i c y , et cetera, that are most e f fec t i ve for pa r t i cu l a r segments of the market as defined by such charac te r i s t i cs as l i f e cyc le stage, education, income, e tc . Another important shortcoming to much of the current t r ave l r e -search i s i t s confinement to the considerat ion of only those persons who are already t r a v e l l e r s . Firms undertaking studies such as these would appear to be making the assumption that t h e i r best l i n e of expansion i s i n get t ing present customers to t r a v e l more. In actual f ac t , i t may be that those who do not t r a v e l at present provide the greatest po ten t ia l fo r expansion. At any ra te , i t would be desi rable fo r the f i rm to use some sort of empir ica l research to determine the degree to which the non-t r a v e l l e r provides a po ten t ia l market and the type of marketing strategy to which he might be sens i t i ve . The most advanced studies of the t r a v e l market have been under-taken by the Survey Research Center at the Univers i ty of Michigan. These studies have used mul t ivar ia te s t a t i s t i c a l methods to go even beyond a simple cross- tabula t ion of market cha rac te r i s t i cs and t r i p cha rac te r i s t i c s . 162 The use of multiple regression has enabled these studies to separate the direct relationship between two variables from their indirect relation-ships with other variables. Thus, i t was possible to considerably reduce the chances of attributing significance to one variable which may actually be the result of the interacting influences of other variables. The analysis of the Vancouver Survey undertaken as part of this study has closely paralleled the type of analysis used i n the Michigan studies. Besides providing the advantages just described, this approach has also enabled comparability i n the results. C r i t i c a l Variables The analyses of tot a l travel frequency, a i r travel frequency and mode choice have revealed that certain variables have a particularly significant relationship with personal travel behaviour. While i t may not be inferred that changes i n these variables actually cause differences i n travel behaviour, the results of the analyses have definitely shown that' there are significant associations between an adult's travel behaviour and certain other variables. The variables investigated are of two types: market variables, which have been here defined as those which describe the person taking t r i p s , and t r i p variables, which are the characteristics of the t r i p s themselves. The f i r s t two areas of the analysis, t o t a l travel frequency and ai r travel frequency, were concerned exclusively with market variables. The third area of analysis, mode choice, considered several t r i p variables while foregoing investigation of some of the market variables considered i n the f i r s t two phases of the analysis. The analysis of t o t a l travel frequency by a l l modes combined was 163 s t r a t i f i e d by fami ly l i f e cycle, stage. The f i v e stages considered were: S ing le , Under 45; Marr ied, Under 45 Without Chi ld ren; Married With Ch i ld ren ; Marr ied, Over 45 Without Chi ld ren; and S ing le , Over 45. Even wi th the mul t ip le regression mechanism holding a l l other var iab les i n the analyses constant, the d i f ferences i n t r a v e l behaviour between adults i n these d i f fe ren t l i f e cyc le stages were found to be qui te s i g n i f i c a n t . Those showing the greatest frequencies of t r a v e l are married adu l ts , under 45 without ch i ld ren , whi le those leas t l i k e l y to t r ave l are s ing le and over 45. Although the a i r t r a v e l frequency and mode choice regressions were not s t r a t i f i e d , fami ly l i f e cyc le stage was included as an independent va r i ab le . In the case of a i r t r a v e l frequency, i t s re la t ionsh ip was found to be s i gn i f i can t but very d i f fe rent from i t s associat ions with t o t a l t r a v e l frequency. Adults showing the greatest frequencies of a i r t r ave l are those over 45 and s ing le or married without ch i ld ren , whi le those leas t l i k e l y to t r a v e l by a i r are under 45 and s ing le or married without ch i l d ren . In the case of the Mode Choice regressions, the fami ly l i f e cyc le var iab le was found to have l i t t l e s ign i f i cance . Among a l l the market var iab les studied, income showed the most s i gn i f i can t and s t a t i s t i c a l l y r e l i a b l e associat ions with t r a v e l behaviour. This i s p a r t i c u l a r l y t rue i n the case of t o t a l t r a v e l frequency and a i r t r ave l frequency. In both of these cases income was found to be i n d i rec t re la t ionsh ip to t r a v e l ; the greater an adu l t ' s annual income, the more frequent ly he tends to t r a v e l , both by a i r and by a l l modes combined. Domicile of Chi ldren or Parents was found to be the second most 164 s ign i f i can t va r i ab le . This va r iab le described the geographical area i n which the adu l t ' s parents or ch i ld ren l i v e . The va r iab le was given four c l a s s i f i c a t i o n s , corresponding roughly to d i f fe rent ranges of distance from Vancouver; ch i ldren and parents wi th in 250 mi les of Vancouver, ch i ld ren or parents i n Western Canada or the Western United States, ch i ld ren or parents i n Eastern Canada or the res t of the United States, and ch i ld ren or parents outside Canada and the United Sta tes. Consid-erat ion was given to t h i s va r iab le only i n the t o t a l t r a v e l frequency and a i r t r a v e l frequency regress ions. Although there was no s t r i c t cons is -tency between the resu l t s for the. d i f fe rent l i f e cyc le s t ra ta considered i n the t o t a l t r a v e l frequency regressions or between these regressions and the a i r t r a v e l frequency regress ion, several genera l izat ions could be made. Adults wi th no ch i ld ren or parents l i v i n g more than 250 mi les from Vancouver tend to t r a v e l l e s s frequent ly by a i r and by a l l modes combined than other adu l ts . Adults showing the greatest frequencies of t r a v e l are those having ch i ld ren or parents l i v i n g outside Canada and the United States. The s ign i f i cance of the re la t ionsh ips shown by the Domicile of Chi ldren or Parents var iab le was not found to be the case fo r Place of Schooling or Place of B i r t h . The only instance i n which these l a t t e r var iab les were revealed to have any s ign i f i cance was i n the case of the t o t a l t r a v e l frequency of s ing le adu l ts , under 45 and married adu l ts , under 45 without ch i l d ren . Adults i n these l i f e cyc le stages showing the greatest t r a v e l frequencies are those born outside Canada and the United States and those having gone to school wi th in Canada and the United 165 States, but more than 250 miles from Vancouver. Those least likely to travel are adults born inside British Columbia and adults having gone to school outside Canada and the United States. The study also considered a variable which indicated the geographic area of Vancouver in which the adult lives. Seventeen cells were drawn up on the basis of their rough homogeneity in terms of average income, occupation, education, immigration, age and dwelling type. The purpose of the Cell variable was to reflect differences in travel behaviour which are associated with location of residence; or possibly associated in an in-t direct way with those factors, other than the variables specifically con-sidered elsewhere in the regression analysis, which led an adult to choose a particular location. However, the results for this variable, although significant in terms of magnitude, are characterized by a relatively poor degree of statistical reliability. This suggests that there i s really not much homogeneity in the travel behaviour of persons living in particular geographic areas of Vancouver, other than that which can be attributed to homogeneity in terms of income, education, occupation and the other variables considered separately in the regression equations. However, several partial exceptions are worthy of note. Cell 2 was found to have reasonably reliable associations with total travel frequency. Adults living in this cell who are married with children or single and over 45 tend to travel more frequently than adults in the other cells. On the other hand, those in this cell who are married without children or single and under 45 are likely to travel less than adults in other cells. In the air travel frequency regression, Cell 4 showed the most significant 166 and s t a t i s t i c a l l y r e l i a b l e re la t ionsh ips -with t r a v e l behaviour. Adults l i v i n g i n t h i s part of Vancouver tend to t r a v e l much l ess frequent ly by a i r than other adu l ts . In the mode choice regressions, C e l l 9 and C e l l 11 were most s i g n i f i c a n t . Adults l i v i n g i n C e l l 9 have the greatest tendency to choose a i r , whi le those most l i k e l y to choose an a l te rna t ive mode of t ranspor tat ion l i v e i n C e l l 11. I t was expected that both education and occupation would show l i t t l e re la t ionsh ip to t r a v e l behaviour. I t i s commonly recognized that both education and occupation are h ighly in tereor re la ted with income. Thus, i t was f e l t that these two var iab les would show very s l i g h t , i f any, assoc ia t ion with t r ave l behaviour when income was simultaneously taken in to account, as i s cha rac te r i s t i c of the mul t ip le regression method. However, t h i s expectation was not en t i re l y upheld by the resu l t s of the ana l ys i s . Although not found to be var iab les of great importance, both var iab les showed some s ign i f i cance i n pa r t i cu l a r areas of the study. Education was s i gn i f i can t i n the t o t a l t r a v e l frequency ana l ys i s . Adults i n the process of un ivers i t y education have a considerably greater t en -dency to t r a v e l than other adu l ts . Occupation was found to be s i g n i f -icant i n i t s associat ions wi th a i r t r a v e l frequency. As would be expected, adul ts i n the managerial occupations tend to take personal a i r t r i p s more than other adu l ts . Adults l i k e l y to have the lowest frequencies of personal t r a v e l are those employed i n blue c o l l a r occupations. Several var iab les were included i n the analys is to measure the s ign i f i cance upon t r a v e l behaviour of f i n a n c i a l in te res t i n cer ta in major assets . These var iab les were auto ownership, boat ownership, cottage 167 ownership, dwel l ing ownership, and dwell ing type. This data was used to i tes t the hypothesis that these var iab les would represent a dra in upon the adu l t ' s d iscre t ionary income, so that l ess than an average number of t r i p s would be taken by adults who. had f i n a n c i a l i n te res t i n a s i n g l e -detached home, a car , a cottage or a boat. However, the resu l t s of the analyses d id not uphold t h i s hypothesis. In f ac t , i t was revealed that , i n general , adul ts who invest i n these major assets also undertake personal t r a v e l more frequent ly than other adu l ts . Whether an adult i s male or female was found to be i ns i gn i f i can t as fa r as t o t a l t r a v e l frequency and mode choice are concerned. However, a surpr is ing reve la t ion was that females tend to undertake personal a i r t r a v e l more frequent ly than males. Tr ip va r iab les , although only considered i n the analyses of mode choice, were found to have general ly more s i gn i f i can t and s t a t i s t i c a l l y r e l i a b l e associat ions with t r a v e l behaviour than market va r i ab les . Three t r i p var iab les are p a r t i c u l a r l y s i g n i f i c a n t : t r i p d is tance, inverse of the number who went, and t r i p purpose. Tr ip d is tance, i n s i tua t ions of choice between a i r and auto or a i r and bus, was shown to be i n d i rec t r e l a t i o n to the propensity to choose a i r . t r a v e l . However, t h i s d i rec t re la t ionsh ip does not hold fo r s i tua t ions invo lv ing a choice between a i r and r a i l . A l l e lse considered, adul ts are more l i k e l y to choose a i r over r a i l on short distance t r i p s ( less than 1,000 mi les) , than they are on medium t r i p s ( 1 , 0 0 0 - 2 , 0 0 0 m i l es ) . A var iab le was. included i n the A i r vs Auto mode choice regression which measured the inverse of the number of people, who went together on 168 the t r i p . This var iab le was expected to r e f l e c t the fact that the cost of t r a v e l l i n g by a i r increases i n d i rec t proport ion to the number of adul ts t r a v e l l i n g together, whi le the cost of running an auto i s l a rge ly f i xed and can be spread over the number of i nd i v idua ls i n the veh ic le . In support of t h i s expectat ion, the resu l t s of the analys is ind ica te that adul ts t r a v e l l i n g i n groups of f i v e or s i x have a s i g n i f i c a n t l y greater tendency to choose auto than groups of three or four or l e s s . This i s r a t i ona l i zed by the fact that an automobile (with room for f i v e or s i x persons) w i l l g ive a lower cost of t r a v e l per person fo r a group of f i v e or s i x than fo r a smaller group. The purpose of the t r i p was also found to be a var iab le s i g n i f -i c a n t l y and r e l i a b l y associated wi th t r a v e l behaviour. Adults on t r i p s invo lv ing personal emergency are most l i k e l y to t r a v e l by a i r , whi le those on vacat ion t r i p s are more l i k e l y to choose a slower and l ess cos t l y mode of t ranspor ta t ion . Personal t r i p s which are not fo r an emergency and t r i p s which are par t l y fo r business are l ess l i k e l y to lead to a i r t r a v e l than personal emergency t r i p s but are more l i k e l y to involve t h i s mode than vacat ion t r i p s . Another t r i p var iab le considered i n the mode choice analys is was t r i p durat ion. However, t h i s var iab le was not found to be p a r t i c u l a r l y s i g n i f i c a n t . General Ef fect iveness of the Regression Analys is At f i r s t glance, the resu l t s of the variance analyses fo r the d i f fe rent regression equations are d isappoint ing. In the case of the t o t a l t r a v e l frequency regressions, the explained variance ranged between 169 only 28.6$ and 12.7$. The a i r t r a v e l frequency regression appears even worse, with only 8.8$ explained var iance. However, even t h i s 8.8$ explained variance does not seem surpr is ing when several factors are considered. F i r s t , i t must be remembered that these regression equations are being f i t to very inexact , s o c i a l and emotional phenomena rather than phys i ca l , chemical or inherent ly mathematical phenomena. Second, t h i s i s compounded by the c lose l y re la ted fac tor that the var iab les considered i n the equations are only those which were considered l i k e l y to be most important among an almost i n f i n i t e number of var iab les having some ef fect upon the i n d i v i d u a l ' s t r a v e l behaviour. Th i rd , both of these factors are compounded by the large number of observat ions, each add i t iona l observa-t i o n i n j ec t i ng in to the scat ter of data some add i t iona l va r i a t i on to be explained by the equation. The variance analys is resu l t s fo r the mode choice regressions are considerably"better than those exhib i ted by e i ther the t o t a l t r a v e l f r e -quency regressions or the a i r t r a v e l frequency regress ion. The A i r vs R a i l regression explained 33.3$ of i t s t o t a l var iance, whi le the A i r vs Bus and A i r vs Auto regressions explained 41.2$ and 59.0$ respect ive ly . The s i g n i f i c a n t l y bet ter showing by these mode choice regressions would seem to be a t t r ibu tab le to the fac t that they included several very powerful var iab les measuring the cha rac te r i s t i cs of the t r i p s : Distance, Inverse of the Number Who Went, and Purpose. These va r iab les , and i n pa r t i cu la r Distance, showed very large regression coe f f i c i en ts and very smal l standard e r ro rs . These resu l t s compare very c lose ly to those achieved i n the M i c h i -170 gan Studies. There i s l i t t l e doubt that the performance of both studies, i n terms of explaining the va r ia t ions i n t he i r data, would have been improved by inc lud ing more va r iab les . There were s t r i c t budget and computer l im i t a t i ons which confined t h i s study to the number of var iab les considered. I t i s also qui te poss ib le that the performance of the ana lys is might have been bet ter i f some of the var iab les considered had been replaced by other va r i ab les . However, the choice of var iab les was based upon the a v a i l a b i l i t y of data, the resu l t s of past s tud ies , and the spec i f i c aims of the study. I t i s hoped that the resu l t s from t h i s study w i l l a s s i s t future studies i n choosing var iab les .which w i l l enable better performance i n the explanation of t r a v e l behaviour. However, a very high explanation of variance should not be expected, fo r the human nature of t r a v e l i s undoubtedly in f luenced to no small extent by such nebulous fac tors as personal i ty t r a i t s , i n t u i t i o n , i d iosyncrac ies , e tc . There i s also l i t t l e doubt that the performance i n the ana lys is of the Vancouver Surrey could have been improved by the implementation of modi f icat ions to the basic mul t ivar ia te method. However, the extent of t h i s ana lys is was la rge ly constrained by budget and time l i m i t a t i o n s . The shortcomings i n the methods of analys is are pointed out below. Shortcomings i n the Ana lys is Despite the success of t h i s study i n deal ing with many of the shortcomings and problems found i n past t r a v e l market s tud ies , i t has a lso been the subject of several shortcomings which should be considered i n future research. With the exception of the l i f e cyc le stage s t r a t i f i c a t i o n i n the 171 ana lys is .of t o t a l t r a v e l frequency, the study produced only one parameter fo r the re la t ionsh ip between each c l a s s i f i c a t i o n of each var iab le and the t r a v e l behaviour of the ent i re sample. Thus, the study does not point out spec i f i c sect ions of the Vancouver population ( i . e . more than a random pattern of i nd iv idua ls ) who may show t r a v e l behaviour contrary to that shown by the ent i re sample when analysed as a whole. For example, whi le the parameter fo r the re la t ionsh ip between auto ownership and a i r t r a v e l i s shown to be pos i t i ve , there may we l l be a s i gn i f i can t port ion , of the sample who own automobiles and who ac tua l l y have a very low frequency of a i r t r a v e l . A cha rac te r i s t i c of the mul t ip le regression method used i n the ana lys is i s that i t produces a parameter for each var iab le which i s i n -dependent of the in te rac t ion ef fects of any: of the other var iab les i n the i ana l ys i s . While t h i s i s one of the chief advantages of the mul t ip le regression method, i t a lso gives r i s e to an important weakness as fa r as t h i s study i s concerned. The mul t ip le c l a s s i f i c a t i o n conversion of the mul t ip le regression resu l t s assumes that the ef fect of several var iab les together may be measured simply by adding the regression coe f f i c ien ts fo r t h e i r separate e f fec ts . This assumption of a d d i t i v i t y does not consider the i n te rac t i on ef fects which were removed when the i n i t i a l mul t ip le regression ana lys is produced parameters fo r the independent e f fects of each va r i ab le . Suggestions fo r Add i t iona l Surveys Having undertaken a general survey, fur ther research of the 172 Vancouver personal t r a v e l market should progress in to a deeper and more spec i f i c i nves t i ga t i on . This deeper inves t iga t ion should be ef fected both through the types o f data co l lec ted and the types of analys is appl ied to the data. The data co l lec ted i n the A i r Canada Vancouver Survey concerned the general aspects of the t r i p—dis tance , durat ion, purpose, e t c . ; the general cha rac te r i s t i cs of the t ravel ler—income, occupation, domici le of ch i ldren and parents, e ta ; and the reasons given by the respondent con-cerning h i s choice of mode, a i r l i n e , t r a v e l agent vs t i c ke t o f f i c e , e tc . Having invest igated these basic areas, the next step i n the research might consider more deta i led and searching questions on those var iab les which were found to be important i n general terms. For example, vacat ion t r ave l was found to have s i gn i f i can t re la t ionsh ips with mode choice. Further, research might invo lve questions such as : how and when the decis ion to i take a vacat ion i s made?; who i n the fami ly decides what mode i s used?; what a c t i v i t i e s are most important on the vacation? Questions might a lso be asked which would have a more d i rec t bearing on the t ransportat ion f i r m ' s adver t is ing p o l i c i e s than the general questions asked i n the Vancouver Survey. For example, a survey might simply enquire about the t r a v e l l e r ' s and non - t rave l l e r ' s a t t i tudes towards t r a v e l l i n g by each mode. Resul ts from questioning along t h i s l i n e might suggest smal l but important ways the t ranspor tat ion company could change i t s se rv i ce , or i t might ind ica te emotions to which the company might d i rec t i t s adver t is ing appeals. The ana lys is of the market could a lso be made deeper by mod i f i -173 cations to the ways in which the data are analysed. One modification would simply be to_ increase the number of sub-classifications in the variables under further investigation. For example, the managerial classification of the occupation variable might be divided into separate classifications for managerial, professional, and proprietary occupations. As an i n i t i a l survey of the Vancouver travel market, this study has concentrated on the separate effects of each variable. Further research might involve consideration of the joint effects of several variables together, which, including their interactions, might be substantially greater or less than the simple addition of their separate effects. For example, i t would be interesting to determine the combined effect of having income over $10,000 and being a college graduate. It will be remembered that, while income over $10,000 was found to be associated with high frequencies of travel, college graduation was associated with surprisingly low frequencies. A potential value of travel market research is as an assist to forecasting future travel patterns. While further surveys of the Vancouver personal travel market should involve progressive changes toward greater depths of analysis, i t would also be of value to repeat questions asked in the earlier surveys. This would enable intertemporal comparisons, which, by reflecting past trends, might suggest future changes in the travel market. With results from similar questions in two surveys taken about ten years apart, i t would be possible to compare the travel behaviour of the same age group or l i f e cycle stage at the two points in time. Another possibility would be to look at the same generation or cohort age group at the two points in time. This would involve comparison of the 1 7 4 30-40 age group i n a 1 9 7 6 survey with the 20-30 age group i n the 1 9 6 6 survey. Thus, i t would be poss ib le to learn something of the e f fec ts a t t r ibu tab le to the unique circumstances of the period i n which a person a t ta ins a given age l e v e l . Lansing and Blood have obtained in te res t ing and use fu l r esu l t s from t h i s type of ana l ys i s . ^ The nature of fur ther t r a v e l market research w i l l depend to an overr id ing extent upon the pa r t i cu la r object ives involved i n continuing the research. However, t h i s present study has value as a bas is fo r fur ther study of the Vancouver personal t r a v e l market i n pa r t i cu l a r , and as a point of departure fo r fur ther inves t iga t ion in to the use of mu l t i var ia te s t a t i s t i c a l ana lys is i n t r a v e l market research i n general . John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1 9 6 4 ) . 175 APPENDIX A DEFINITION OF INDEPENDENT VARIABLES IN MULTIPLE REGRESSION ANALYSIS Symbol or Var iab le \ Abbreviat ion Range of Values Income Under $5,000 1 i f adult f a l l s i n t h i s group • 0 otherwise Income $5,000-9,999 Y2 1 i f adult f a l l s i n t h i s group 0 otherwise Income $10,000 & Over Y3 1 i f adult f a l l s i n t h i s group 0 otherwise Owns a car AUTO 1 i f fami ly owns one or more cars 0 otherwise Owns a boat BOAT 1 i f fami ly owns one or more boats 0 otherwise Owns a cottage . COT 1 i f fami ly owns cottage i 0 otherwise Owns home with no mortgage °1 1 i f i n t h i s group 0 otherwise Owns home with mortgage °2 1 i f i n t h i s group 0 otherwise Dwel l ing: single-detached D l 1 i f i n t h i s group 0 otherwise Dwel l ing: a p t . / f l a t D 3 1 i f i n t h i s group 0 otherwise Occupation: managerial, OCC 1 i f i n t h i s group profess iona l or propr ietary 0 otherwise Occupation: blue c o l l a r 0CC3 1 i f i n t h i s group 0 otherwise Occupation: r e t i r e d 0CC4 1 i f i n t h i s group 0 otherwise APPENDIX A—Continued Var iable Symbol or Abbreviat ion .Range of Values Education: grade school , some high school or l ess EDi Educat ion: high school grad- ED 2 uate or some un ive rs i t y Education: co l lege graduate ED^ Chi ldren or parents i n western Canada or western U.S.A. Chi ldren or parents i n eas-te rn Canada or rest of U.S. Chi ldren or parents outside Canada & U.S.A. Schooling i n western Canada or western U.S.A. Schooling i n eastern Canada or res t of U.S.A. Schooling outside Canada and U.S.A. Born i n western Canada or western U.S.A. Born i n eastern Canada or res t of U.S.A. Born outside Canada or U.S.A. Sex: male C i ty C e l l 1 Ci ty C e l l 13 A. C / P 2 C / P 3 > C / P 4 sc2 sc3 sc4 B 2 B 3 B 4 SEX '13 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f male 0 i f female 1 i f l i v e i n c e l l 1 0 otherwise 1 i f l i v e i n c e l l 13 0 otherwise 177 APPENDIX A—Continued Var iab le Symbol or Abbreviat ion Range of Values C i ty C e l l 15 C i ty C e l l 16 C i ty C e l l 17 L i f e cyc le stage: s ing le or married under 45 wi th no ch i ld ren L i f e cyc le stage: married with ch i ld ren Time away on t r i p : 6 days or l ess Time away on t r i p : 15 days or more c 15 Cl6 c 1 7 FLC-j_ FLC 2 DUR]_ DUR3 DIS-L DISo Distance: l e s s than 700 mi . Distance: 700-1,000 mi . Distance: 1,000-2,000 mi . DIS3 Distance: 2,000-4,000 mi . DIS4 Frequency of t r a v e l : 1 t r i p per year Frequency of t r a v e l : 2 t r i p s per year 1 i f l i v e i n c e l l 15 0 otherwise 1 i f l i v e i n c e l l 16 0 otherwise 1 i f l i v e i n c e l l 17 0 otherwise 1 i f i n t h i s stage Purpose: vacat ion 0 otherwise 1 i f i n t h i s stage 0 otherwise 1 i f t r i p t h i s long 0 otherwise 1 i f t r i p . t h i s long 0 otherwise 1 i f went t h i s distance 0 otherwise 1 i f went t h i s distance 0 otherwise " 1 i f went t h i s distance 0 otherwise 1 i f went t h i s distance 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 1 i f i n t h i s group 0 otherwise 178 APPENDIX A—Continued Variable Symbol or Abbreviation Range of Values Purpose: personal emergency- • P2 1 i f in this group 0 otherwise Purpose: partly business p 4 1 i f in this group 0 otherwise Inverse of number who went on trip NOx N02 NO3 N04 10. 1 person 05. 2 people 03. 3, 4, 7 or more, people 02. 5 or 6 people APPENDIX B ESTIMATES OF REGRESSION EQUATIONS FOR TRAVEL FREQUENCY BY ALL MODES, STRATIFIED BY FAMILY LIFE CYCLE Stage i n Family L i f e Cycle . Under 45 Over 45 Under 45 Married Married Married Over 45 Variable Single No Chi ldren Chi ldren No Chi ldren Single C .75082 i .60044 .80300 .8020 .5535 (.0775) .0829 (.0943) -.0910 (.1822) -.1210* (.0595) .0929 (.0665) -.1489 Y2 .1115 (.1039) .1230 (.1268) .0144 (.0422) .1416* (.0656) .0957 (.1227) Y 3 .3032* (.1509) .3952* (.1766) .3155** (.0584) .2758* (.0968) .1700 (.3210) AUTO .1210 (.0780) .2419 (.1676) .2032** (.0660) .2790** (.0593) .1418 (.0829) BOAT .0471 (.1746) .1161 (.1892) .0457 (.0670) -.0153 (.0981) -.1075 (.2774) COT .1219 (.1532) .4890 (.2909) .0854 (.0765) -.1854 (.1190) .2723 (.1943) Ol .1330 (.2188) -.1147 (.1905) -.1027 (.0566) -.0195 (.0848) .1791 (.0913) 02 -.1229 (.1635) -.2925 (.1742) -.0820 (.0473) -.1290 (.0918) .1719 (.1181) Dl -.0529 (.1523) .2536 (.2149) -.0571 (.0620) -.0557 (.1008) .0233 (.1375) D 3 .0059 (.1671) .1100 (.2055) -.0877 (.0984) -.0850 (.1248) .2991* (.1419) OCCl .2203* (.1002) .1394 (.1498) .1175* (.0465) .0361 (.0865) -.1181 (.1466) occ3 .0846 (.1061) -.0747 (.1289) .0198 (.0450) -.0428 (.0803) -.0451 (.1428) QCC4 -.1957 (.1635) -.4212 (.4145) -.0354 (.1432) .0146 (.0828) -.1233 (.1059) EDi -.3356* (.1233) -.1540 (.3327) -.2150 (.1598) -.4011 (.6244) .0028 (.1640) ED 2 -.1288 (.1060) -.0667 (.3246) -.1557 (.1591) -.2900 (.6235) .1218 (.1593) ED 3 (.3738) .8044 (.6345) m,k -.2207 (.1808) -.0749 -.2378 (.1667) -.3720 (.6301) C/?2 .3148* (.1346) -.0821 (.2054) .1897** (.0507) .2677** (.0606) .1822* (.0869) C / P 3 .2325 (.2188) .0340 (.2735) .0065 (.0923) .1785* (.0814) .1108. (.1192) C / P 4 .4041* (.1725) .2303 (.2207) .0003 (.0628) .0873 (.1146) .2201 (.1875) SC 2 .1452 (.1250) .2502 (.2070) .0667 (.0550) .0437 (.0808) .1417 (.1136) SC 3 .1337 (.2431) .3038 (.3357) -.0278 (.1097) .0044 (.1257) .0345 (.1445) SC4 -.2207 (.2119) -.3673 (.3455) -.1706* (.0841) .0823 (.0945) -.0178 (.1347) B2: .3257* (.1121) .1390 (.1819) -.0038 (.0529) -.0546 (.0961) -.1675 (.1441-) APPENDIX B—Continued Stage i n Family Life Cycle  Under 45 Cver 45 Under 45 Married Married Married Over 45 Variable Single No Children Children No Children Single B 3 .1850 (.1739) -.0429 (.2538) .0652 (.0954) -.1132 (.1224) -.1351 < M477) six .3076 (.1639) .3269 (.3263) .1685* (.0769) -.2034* (.1000) -.0455 ( M473) -.0713 (.0863) .0761 (.1020) .0407 (.0348) .0005 (.0497) -.0851 < [.0864) Cl -.2579 (.2484) -.1299 (.3121) -.2025* (.0892) .1714 (.1258) -.1997 ( M795) C2 -.3729 (.3232) -.5941 (.3413) .0245 (.1304) -.3056 (.1938) .3235 ( ;.2857) C4 -.0413 (.2613) .0068 (.2946) -.0391 (.0953) .2901 (.1622) -.5053 ( :.2839) C5 -.3147 (.2037) -.0639 (.3128) -.1300 (.2162) .0353 (.1728) -.2958 ( [.1703) C6 -.1220 (.2247) -.5137 (.3184) -.0656 (.1071) .2586* (.1256) -.0882 ( :.1776) c7 -.0140 (.2061) -.0304 (.3075) -.2760* (.0933) .1688 (.1260) .0354 ( [.1629) -.1592 (.1881) -.1718 (.2570) -.1922* (.0935) .1213 (.1202) -.0232 < ,.1518) c9 -.1925 (.3153) -.2470 (.3610). -.1390 (.0940) .0605 (.2159) -.0759 ( [.3037) ClO .0278 (.2205) .0503 (.2866) -.1521 (.0830) .1953 (.1073) -.0289 ( [.1672) C l l -.0044 (.2145) -.3396 (.4054) -.2069 (.1131) .0336 (.1298) -.3038 ( [.187D C12 -.3582 (.2039) -.4598 (.2777) -.2502* (.0902) -.0610 (.1148) -.1009 ( [.1547) C13 .0504 (.2550) .0008 (.3085) .-.2297 (.1027) -.1370 (.1516) -.0332 ( [.3064) C14 -.0980 (.2935) -.5048 (.3337) -.0880 (.1058) -.2176 (.1778) .0052 ( [.3267) C15 -.0790 (.2121) -.0662 (.2548) -.0080 (.0838) .1557 (.1333) -.0956 ( .1977) Cl6 .0707 (.2832) -.0431 (.4579) -.2093 (.1347) -.0256 (.2212) .0009 ( [.3065) R2 .4804 .5348 .3563 .4206 .4065 N 361 198 1366 676 368 I .8443 .9040 .7142 .7192 .5625 * Coefficient i s 2 times i t s standard error. ** Coefficient i s 3 times i t s standard error. 181 APPENDIX C ESTIMATES OF REGRESSION EQUATIONS FOR AIR TRAVEL FREQUENCY " Variable Regression Coefficient Standard Error c .0913 Y l -.0085 (.0181) Y 2 .0146 (.0162) Y 3 .1390** (.0236) AUTO .0239 (.0169) BOAT .0538* (.0260) COT .0607* (.0291) 01 -.0001 (.0188) o2 -.0117 (.0181) ^.0294 (.0238) D 3 .0093 (.0290) OCCl .0548* (.0185) occ3 - .0 2 90. (.0176) OCC4 -.0142 (.0228) ED! .0005 (.0383) ED 2 .0037 (.0377) ED4 .0271 (.0428) C / P 2 .0520** (.0176) C/P 3 .0847** (.0259) C/P4 .0794** (.0237) SC 2 .0112 (.0202) sc3 .0375 (.0339) SC4 .0262 (.0280) B 2 .0233 (.0208) B 3 .0137 (.0303) B4 .0294 (.0266) SEX -.0271* (.0128) Cl .0004 (.0349) c 2 -.0248 (.0499) C4 -.0778 (.0390) C5 .0125 (.0431) C6 .0149 (.0375) c7 .0000 (.0349) C8- .0152 (.0332) c9 -.0148 (.0400) ClO -.0162 (.0319) C l l -.0592 (.0392) Cl2 -.0102 (.0331) C13 -.0619 (.0405) c 14 -.0530 (.0433) 182 'APPENDIX C—Continued Variable Regression Coefficient Standard Error C15 -.0056 ' (.0333) Cl6 -.0460 (.0521) FLCl -.0707** (.0163) FLC 2 -.0084 (.0257) R2 .2974 N 2981 X .12529 * Coefficient i s 2 times i t s standard error. Coefficient i s 3 times i t s standard error. APPENDIX D ESTIMATES OF REGRESSION EQUATIONS FOR MODE CHOICE Paired Choice Variable Air vs Auto Air vs Rail Air vs Bus C .69073 1.20684 .8990 Y l - . 0 2 5 7 (.0198) - . 0 7 9 7 ( .0573) -.0235 < :.0517) Y2 .0018 ( .0156) .0186 (.0518) .1238* [.0492) Y 3 .0422* (.0200) .0264 (.0617) .0845 [.0602) AUTO - . 0 9 0 4 * * (.0198) .0443 (.0490) .1366** [.0445) BOAT .0414 (.0229) - . 0 2 4 5 (.0736) .0059 < [.0716) COT .0465 (.0251) .0222 (.0686) .0852 ;.0713) Ol .0345 (.0183) .0431 (.0544) -.0140 :.0523) 02 .0337 (.0174) .0400 ( .0559) -.0161 [.0533) DI -.0152 (.0232) -.1380 (.0751) .0277 [.0648) D 3 -.0059 (.0289) -.0558 (.0850) .1059 ( [.0736) FLCl .0557** (.0156) -.0473 (.0501) .0905* ( [.0493) FLC 2 -.0536* ( .0217) .0209 (.0731) .0538 [.0692) DURi .0059 (.0150) .1431* ( .0601) .0715 [.0517) DUR3 - .0107 (.0159) -.1395** (.0469) .0341 [.0427) NO .0578 (.0029) — . . — • ' — — DISi -.8136** (.0266) -.3636** (.0804) - . 5 9 2 6 * * [.0621) DIS 2 -.7391** (.0261) -.4863** (.0552) - . 3 6 9 4 * * [.0506) DIS 3 -.7323** (.0347) -.7393** (.0673) -.2653** [.0837) DIS4 -.5666** (.0282) -.4203** (.0509) -.2560** [.0494) F l .0103 (.0161 -.0992. ( .0574) -.0853 [.0509) *2 .0173 (.0178) -.1413* (.0614) -.0129 ( [.0562) Pl -.0971** (.0134) -.1148* (.0409) -.1921** ( [.0368) P2 .2545** (.0460) .2036* (.0830) .0288 < [.0780) P4 -.0709 (.0376) .0177 (.1090) - .1555 ( [.0911) SEX -.0244* (.0122) .0120 (.0399) .0503 ( .0370) Cl .0811* (.0305) .2119 (.1149) .0704 ( .1216) c 2 -.0059 (.0470) .2694 (.1891) -.0789 ( .1744) APPENDIX D—Continued Paired Choice Variable Air vs Auto 'Air vs Rail Air vs Bus C4 .0027 (.0336) -.0743 (.1503 -.0441 (.1617) C5 ..0978 (.0421) .0815 (.1261) -.0203 (.1290) C6 .0389 (.0315) .1419 (.1162) .0081 (.1227) c 7 .0558 (.0313) .1501 (.1144) .0355 (.1220) cs .0987** (.0320) .1888 (.1112) -.0221 (.1173) c 9 .0526 (.0358) .3405* (.1466) .1855 (.1507) ClO -.0019 (.0287) .0955 (.1192) -.0938 (.1252) C l l -.0071 (.0372) -.0591 (.1353) -.2062 (.1421) C12 .0595 (.0340) . .1284 (.1190) -.0779 (.1233) Cl3 .0504 (.0370) .1097 (.1556) .0367 (.1592) C14 .0289 (.0430) .0181 (.1611) -.0560 (.1670) C15 .0230 (.0288) .0657 . (.1187) -.0882 (.1252) C16 .0385 (.0518) .0310 (.1693) -.0664 (.1847) R2 .7682 .5757 .6420 N .1846 468 459 X .1854 .7323 .7483 * Coefficient i s 2 times i t s standard error. Coefficient i s 3 times i t s standard error. 185 APPENDIX E MULTIPLE DISCRIMINANT ANALYSIS IN SCALING THE DEPENDENT VARIABLE Basically, multiple discriminant analysis involves the c l a s s i -fication of observations, for which a common set of measurements i s available, into one or more groups with a minimum of error due to mis-classification.^" The optimality of the classification criterion provided by the linear discriminant function comes as a result of finding that linear combination of variables which maximize the ratio of the variance among group means to the tot a l variance. In writing the scale for the to t a l travel frequency dependent variable, the procedure involves specifying a predetermined number of travel frequency classes and then f i t t i n g a single linear multiple d i s c r i -minant function to the data. The general form of the discriminant function i s : 2 = a l X l + a 2 X 2 + . . . . + a ^ Fitting of this function to the data involves estimating the co-efficients aj_, &2, . . . a n i n such a manner as to f u l f i l the optimality criterion that the among-group means are maximized i n relation to the to t a l variance. A value of Z i s calculated for each observation i n each frequency class. The scale value for each frequency class i s then obtained by finding the mean value of 2 of a l l the observations i n that class. During the calculation of the multiple regression equations each obser-•H/illiam W. Cooley and Paul R. Lohnes, Multivariate Procedures For  The Behavoiral Sciences (New York: John Wiley & Sons Inc., 1962), pp. 162-163. 186 vation i s assigned the scale value or mean value of £ for the frequency-class of which i t i s a member. The mathematical basis for using multiple discriminant analysis to scale dependent variables i s set out at the end of this Appendix. The number of,trip frequency groups specified prior to the f i t t i n g of the linear discriminant function i s constrained by the scatter found i n the data. One of the recent Michigan Studies found that the data could support five groups for to t a l travel frequency and three groups for a i r 2 travel frequency. Two frequency groups were automatically defined; a class for those who took no t r i p s and a class for those who took only one t r i p . For t o t a l t r i p frequency the additional three groups were defined as: 2-4 t r i p s , 5-15 trips, and 16 trips and over. The third group for a i r travel frequency was defined as two t r i p s and over. In order to preserve the logic of the relationships, the zero and one t r i p groups i n each case were preassigned the scale values of 0 and 1. The scale values determined by the multiple discriminant analysis, for the other frequency groups were then subjected to a linear transformation so that their numerical value-appeared i n relationship to the designated value for these two groups. It could be shown that the regression coefficients obtained from equations estimated by regressing the newly defined scale upon the same set of independent variables that were used i n estimating the scales are identical to the original "a" coefficients i n the discriminant function equation except for the scale factor. Thus, the multiple discriminant ^John B. Lansing and Dwight M. Blood, The Changing Travel Market (Ann Arbor, Michigan: Survey Research Center, March 1964), pp. 162-163. 187 procedure provides a basis for using the original data to calculate scale values which, when used as the values for the dependent variable i n a regression equation, w i l l maximize the amount of the explained variance i n comparison with alternative scales produced by other methods of linear estimation.^ It would have been possible to base the multivariate analysis of the Vancouver Survey purely upon a presentation of the coefficients obtained from a multiple discriminant analysis. As indicated above, the coefficients produced by the multiple discrimant analysis are mathe-matically proportional to those obtained from a multiple regression analysis. However, i t was desired to s t r a t i f y the analysis of total travel frequency so as to determine variations i n regression coefficients for the independent variables resulting from interactions of these variables with the family l i f e cycle stage variable. Although this could have been done by running a separate multiple discriminant analysis for each l i f e cycle stage, the results for each stage would have been based upon a different scale which would deter from the s t r i c t comparability of the coefficients between the l i f e cycle stages. Furthermore, i t , i s unlikely that the limited dispersion i n the data and the limited number of observations i n each l i f e cycle stage could support the use of multiple discriminant analysis on such a s t r a t i f i e d basis.^ Dwight M. Blood, "A Cross Section Analysis of The Domestic Inter-cit y Travel Market" (unpublished doctoral dissertation i n economics, University of Michigan, Ann Arbor, 1963), p. 39. %1.J. Williams, Regression Analysis (London: John Wiley & Sons Inc., 1959), pp. 176-177. 188 THE MATHEMATICAL BASIS FOR SCALING DEPENDENT VARIABLES BY MULTIPLE DISCRIMINATION^ Problem Given a ca tegor ica l dependent var iab le which has no numerical sca le , and given a set of explanatory var iab les X which may be e i ther scaled or dummy, f i nd a l i n e a r combination of X ' s , Xb, such that the var iance among group means Xb i s maximized r e l a t i v e to the t o t a l variance Xb. Solut ion Measure a l l X»s as deviat ions from t h e i r mean. Constrain the t o t a l var iance of Xb: (1) b» X ' Xb = k Write the expression fo r the variance among group means. (2) A = b* X» Xb Maximize A subject to ( l ) , by const ruct ing: (3) A* = b'X'Xb - \ (b 'X 'Xb-k) D i f f e ren t ia t i ng wi th respect to b and se t t ing der iva t ives equal to zero g i ves : ' -(4) X ' X b = \ X » X b , or (5) Ab = X b > w h e r e A = ( x ' x ) - 1 X»X. ^Taken from W.H. Locke Anderson, "Mu l t ip le Discr iminat ion and Sca l i ng " (unpublished paper to the Research Seminar i n Quant i tat ive Economics, Un ivers i ty of Michigan). 189 premul t ip ly ing (4) by b f g ives (6) = b»X'Xb, so that i s the unadjusted for an analys is b 'X'Xb of var iance of Xb. We then f i nd the la rges t eigenvalue of A. Since the constraint k i s wholly a rb i t ra ry , the eigenvecter, which cons is ts of the values of b, may be scaled i n any convenient way, fo r example (7) . b'b = 1 This w i l l determine the value of k, the t o t a l sum of squares of Xb, according to ( l ) , and A, the among-group sum of squares, according to (2) . The res idua l sum of squares i s g iven by K-A. The problem may be recast as a regression problem by sca l ing the dependent var iab le i n each of the c lasses as : (8) Y. = X,- b fo r a l l c lasses . 1 X This may then be regressed on the observed values of X Y = Xa + u Least squares estimates of a are i d e n t i c a l l y equal to b Proof (9) a = ( X ' X ) - 1 X 'Y (10) = ( X ' X ) " 1 X»X b X (11) = (X»X) - 1 X»X b A By (4) above (12) (X»X) _ 1 X'Xb = b, therefore a=b A 190 The unadjusted R^ for the regression may be computed i n the standard way, and i s equal to Proof (13) R 2 = b'X'Xb Y'l (14) = X 2 b-X'Xb b»X*X b which by (6) above i s equal to \ , QED. The residual sum of squares, u*u, i s given by: (15) u»u = b'X'Xb (1- X ) \ The degrees of freedom are the number of points, minus the number of classes, minus the number of independent variables. BIBLIOGRAPHY BOOKS AND MONOGRAPHS Better Homes and Gardens. Travel Questionnaire Study ... A Report on ' Family Travel. New,York: Better Homes and Gardens, 1967. Cooley, W.W., and P.R. Lohnes. Multivariate Procedures for the  Behavioral Sciences. New York:' John Wiley and Sons, 1962. Crisp, R.D. Marketing Research. New York: McGraw-Hill Company, 1957. Day, Ralph L. (ed.). Marketing Models: Quantitative and Behavioral. Scranton, Penn: International Textbook Company, 1964. Ezekiel, Mordecai, and Karl A. Fox. Methods of Correlation and Regres- sion Analysis: Linear and Curvilinear. New York: John Wiley and Sons, 1959. Ferber, Robert. Statistical Techniques in Marketing Research. New York: McGraw-Hill Company, 1949. Ferber, R.,-.andH.C. Wales. Motivation and Market Behavior. Homewood, 111;: R.D. Irwin Inc., 1958. Freund, J.E. Mathematical Statistics. Englewood Cliffs, New Jersey: Prentice-Hall Inc., 1965. Goulden, Cyril H. Methods of Statistical Analysis. New York: John Wiley and Sons Inc., 1961. Green, P.E., and D.S. Tall. Research for Marketing Decisions. Englewood Cliffs, New Jersey: Pentrice-Hall Inc., 1966. Johnson, J. Econometric Methods. New York: McGraw-Hill Company, 1963. Johnson, P.O., and R.W.B. Jackson. Modern Statistical Methods:; Descriptive and Deductive. Chicago, 111: Rand McNally and Company, 1959. Kates, Peat, Marwick and Company. Survey and Analysis of Non-Resident  Travel. Report for the Department of Industry and Commerce of the Government of The Province of Alberta, April 1967. Klein, Lawrence R. An Introduction to Econometrics. Englewood Cliffs, New Jersey: Prentice-Hall Inc., 1962. Lansing, John B. The Travel Market: 1964-1965. Ann Arbor, Michigan: Survey Research Center, October 1965. 192 Lansing, J.B., and D.M. Blood. A Cross-Section Analysis of Non-Business  Air Travel. Ann Arbor, Michigan: Department of Economics, University of Michigan, 1958. Lansing, J.B., and D.M. Blood. Mode Choice in Intercity Travel: A Multi- variate Statistical Analysis. Report to the Under Secretary for Transportation, U.S. Department of Commerce. Ann Arbor, Michigan: Survey Research Center, 1964. Lansing, J.B.., and D.M. Blood. The Changing Travel Market. Ann Arbor, Michigan:- Survey Research Center, March 1964. Lansing, J.B., et al. The Travel Market: 1958. 1959-1960. 1961-1962. Ann Arbor, Michigan: Survey Research Center, 1963. Lucas, D.B., and S.H. Britt. Measuring Advertising Effectiveness. New-York: McGraw-Hill Company, 1963. Luck, D.J., H.C. Wales, and D.A. Taylor. Marketing Research. Englewood Cliffs, New Jersey: Prentice-Hall Inc., 1961. Martineau, Pierre. Motivation in Advertising. New York: McGraw-Hill Company, 1967. Morgan, James N., et al. Income and Welfare in the United States. New York: McGraw-Hill Company, 1962. Morrison, Donald F. Multivariate Statistical Methods. New York: McGraw-H i l l Company, 1967-Orcutt, Guy H., et al. Micro-Analysis of Socio-Economic Systems: A  Simulation Study. New York: Harper and Row, 1961. The Boeing Company. Characteristics of the U.S. Air Traveller: 1955-1965- 1980. Renton, Washington: The Boeing, Company, 1965. The Douglas Aircraft Corporation. Effect of Selected Demographic Charac- teristics on U.S. Citizen Travel Abroad. Los Angeles, California: The Douglas Aircraft Corporation, March 1967• The Douglas Aircraft Corporation. Measuring the 70's - An Air Travel  Market Analysis. Los Angeles, California: The Douglas Aircraft Corporation, November 1966. The New Yorker. A Five Year Study of Traffic to the Pacific (Through  June-30. 1966T. New York: The New Yorker, 1967. The New Yorker. A Five Year Study of Travel Between the U.S.A. and  Mexico, Central America, Panama and South America 1961-1965. New York: The New Yorker, 1966. 193 The Port of New York Authority. Forecast of The Overseas Air Passenger  Market Through New York 1965-1975. New York: The Port of New York Authority, May 1958. The Port of New York Authority. New York's Domestic Air Passenger Market  April 1963 Through March 1964. New York: The Port of New York Authority, May 1965. The Port of New York Authority. New York's Overseas Air Passenger Market  April 1963 Through March 1964. New York: The Port of New York Authority, June 1965. Williams, E.J. Regression Analysis. London: John Wiley and Sons, 1959. ARTICLES Clawson, C.J. "Quantifying Motivation Research" in L.H. Stockman (ed.) Advancing Marketing Efficiency. American Marketing Association (1959), pp. 54-70. H i l l , T.D. "An Analysis of the Distribution of Wages and Salaries in Great Britain," Econometrica (July 1959), pp. 335-381. Johnson, P.O. "The Quantification of Qualitative Data in Discriminant Analysis," Journal of the American Statistical Association (1950), pp. 65-66. Lansing, J.B., and Leslie Kish. "Family Life Cycle as an Independent Variable," American Sociological Review (October 1957). Lavidge, R.J., and G.A. Steiner. "A Model for Predictive Measurement of Advertising Effectiveness," Journal of Marketing (October 1961), pp. 60-62. Lee, G.H.C. "The New Mass Market in Air Travel," Institute of Transport  Journal (March 1965), pp. 101-103. Paranka, J. "Marketing Predictions from Consumer Attitudinal Data," Journal of Marketing (July I960), pp.. 46-51. Suits, Daniel B. "Use of Dummy Variables in Regression Equations," Journal of the American Statistical Association (December 1958). Yoder, W.0. "Behavioral Research and Marketing" in L.H. Stockman (ed.) Advancing Marketing Efficiency. American Marketing Association (1959), pp. 380-384. 194 UNPUBLISHED MATERIAL Air Canada." "The Atlantic Traveller - A Survey Among Canadian Trans-atlantic Travellers." Montreal: Air Canada, November 1966. Air Canada. "The California Study." Montreal: Air Canada, 1966. Air Canada. "Report on Operation Evergreen Research Surveys." Montreal: Air Canada, July 1965. Air Canada. "The Southern Traveller." Montreal: Air Canada, May 1966. Blood, Dwight M." "A Cross Section Analysis of the Domestic Intercity Travel Market." Unpublished Doctoral Dissertation in Economics, University of Michigan, Ann Arbor, 1963. O.R.C. International. "Air Canada Passengers Departing from Eight Canadian Cities: August-September 1966." Toronto: O.R.C. Inter-national, December 1966. Suits, Daniel B. "Interpreting Regressions Containing Dummy Variables." Technical paper for Research Seminar in Quantitative Economics, University of Michigan, Ann Arbor, May 1962.. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0102414/manifest

Comment

Related Items