THE "UMOT’ PROJECT YACOV ZAHAVI AUGUST 1979 PREPARED FOR US. DEPARTMENT OF TRANSPORTATION — MINISTRY OF TRANSPORT, FED. REP. OF GERMANY Washington, D.C. BonnTechnical Report Docomentotion P [Toes o] B Bavernett Recor Nay T Retna er } DOT-RSPA-DPR-20-79-5 Was ood STE 3 Regan Bait August 1979 The IMT Project © Parlorming rgpnitanan Cade wee [a Petoing Sraon cation Re Yacov Zahavi |B Pevoraing Oguniiaian Wane wd Allene TE Wan Do No RATT Yacov Zahavi = a Bethesda, Maryland DOTS 80079 Tar Type ot Repart end Paved Covered U.S. Department of Transportation Final Report Research and Special Programs Administration | Washington, D.C. 20590 TE. Spensarag Aaeney Cote RSPA| TE Soprlowanry Notes This report presents the results of two complementary research studies, conducted by the author for the U.S. Department of Transportation and for the Federal Republic of Gernany Ministry of Transport. Te Avavas? ‘A nex urban travel nodel te being developed to aid transportation policy aimlysts, beset on a theory of consuner behevior, The nedel 1s based on the theory that travelers atteapt, to saxinize their utility of spattal and econoale opportunities, represented by the total dally travel distance, subject to constraints of tine and money budgets. ‘The application of explictt constrints is a powerful tool, since the constratnte do avay with the need for much of the coefficient caliteation of conventional motels. Thus, ence the constraints ani unit coste of all alternative nodes ara known, the nodel profiices estinates of ouch travel characteristics as dally travel distance, modal shares, and car ownership levels. These outputs are then consared with the observed values of ‘the variatles, and not calltrated to then, for the model's validation. This approach enables the aodel to be transferable both between cities within the sane country, and over tine. The first part of the report develops the analytical framework of the rotel, which te found to be consistent with established econozic principles, The second part exanines ‘the evidence of predictable remularities 1n tine and money expenditures on travel by households and travelers, with favorable results. The thiml part develops the opera~ tonal franework of the todel, incluling travel generation, mode cholce, and car ower ship levels, 212 of shich interact by a fondtack process, to converge rapidly to the observe! travel characteristics. The report conciuies with ceccameniations for further Geveloprents of the model, especially with respect to tho integration of urban etmucture within the feodtack process. Th Kay Warde TE" Draniborien Stetamoot Travel Time and Money Budgets ‘This document is available to the public Travel Denand Models through the National Technical Inforna- Urban Structure Service, Springfield, Virginia 22161 W9. Seauaty Cloant Calm BW Beaviiy Cia Toe We Wea Pages PE Pree Unclassified Unclassified Form DOT F 1700.7 3-72) Reproduction of comsSpecial appreciation is extended to the two directing officers of this research study, Mr. Robert W. Crosby of the U.S. Department of Transportation, Research and Special Frograms Administration, and Dr. Hans -P. Weber of the Federal Republic of Germany Ministry of Transport, who with foresight and faith have encouraged and guided the two research studies to their successful completion. Special appreciation is also extended to the following officials of the World Bank: Mr. B.V.K. Jaycox, Director of the Urban Pro- jects Department, Mr. H.B. Dunkerley and Mr. G.J. Roth, for their personal interest in this project and for providing the travel data in cities of developing countries, Thanks are also extended to Mr. J.M. Ryan, of the Urban Planning Division, FHWA, U.S. DOT, for his helpful comments and for pro~- viding data on population distributions in Washington, D.C. The author also wishes to thank Dipl. Ings, Dirk Zumkeller and Manfred Poeck, representatives of Kocks Consult GMBH, of Koblenz, for their assistance in carrying out parts of the research in Germany. last, tut not least, the consultants to this study should be praised for their devoted efforts to assess and further develop the UMCT approach: Prof. Martin Beckmann, Dr. Thomas Golob and Prof, Antti Talvitie, who have accompanied the research to its conclusion; and Prof. Chong Liew and Prof. Richard Weissbrod, who were involved in the initial stages of the research.FOREWORD ‘The rosults presented in this report are part of long range research progran at DOT which has as its objective the inprovenent of transportation planning nethodologies, The eaphasis 1s on providing analytical means to aid decision makers In exploring long term policy alternatives for major system change ‘The initial foous of the progean ts on urban passenger transportation, with particular attention being given to estimating transportation energy consuap- tion. Great steps have been taken since the pre-1950 era of intuitive methods, and there are now available sophisticated, computerized planning tools that are widely used throughout the U.S. While these tools have been employed quite successfully for short-tera, increnental changes in transportation systens, thoy have beon found to be inadequete for assessing long-Lerm socio-economic effects, In addition, they have been hanpered by difficulties in transfer between cities, The underlying reason for this, I believe, ie in the percep~ tion of the urten systen as a cause and effect process, whore the causes are calibrated to the observed effects under no explicit constraints, The UNOT process, the subject of this report, is tased upon a fundanentally afferent perception of the urtan systen, It 1s seen as a dynamic feedtack Process, Linking people, transportavion and urtan form under explicit const ralnts. Given an urban form and transportation supply, the UMOI process generates all observed travel characteristics as direct outputs within the time and money constraints. This makes the UMOT a powerful tool for the evaluation of both short and long-term policy options under a wide range of scenarios. Of course, much more needs to be dono than is presented in thts report to achieve these goals. The DO? progran envisions further research in several important toptes, including: 1. Further stratification of travel groups, geographically at least. 2, The determination of travel fields. 3, Inprovenents In the xeprescntation of inportant elenents of urban form, such as housing, job locations, and facilities. 4, The introduction nf dynamics to the cquilitriun process. Im carrying ont this progran, steps have been taken to coordinate the effort in urban transportation with similar work done in the Federal Republic of Germany on intercity travel, and at the World Bank tn transportation aystens for developing countries. i 0 kerdut W Oe Robert W. Grosty Office of Systens Engineering Research and Special Frograns Administration‘The Unified Mechanism of Travel (UMO?) Project 2 EXECUTIVE SUMMARY 9 ‘This report presents the results of two complenentary research studies, conducted for the U.S. Dopactnent of Tronsportation and for the Federal Republic of Gemmany Ministry of Transport. The forner study investigated and developed the theoretical framework of a proposed new transportation planning methodology, first conceptualized for the World Bank, and called the Unified Mechanism of Travel, or UMOT for short. The latter study invol- Ved testing several of the UMOF principles using observel data in Germany end supplenented by data fron other countries. Purpose The purpose of the research reported herein is twofold. First, to analyze the consistency of previously observed phenonena that trave= ders within urban areas appear to have avorage daily tine and money expenditures on travel which display predictable regularities both betwoon cities and over tine in the sane country. Second, if such expenditures are indeed found to be predictable, the study intended fo develop the conceptual framework of a new transportation planning tool which takes advantage of such regularities in relating travel denand and transportation systen supply to urtan form. The UMOT Concent The UOT concept is based on the assunption that when the datly mean expenditures on travel, per traveler and per househola(*), in tine and noney terms, display predictable regularities that can te attributed te such factors as the socioecononic characteristics cf the household, (zansport systen supply and urtan structure, and when these regulari- tes are found to be transferable both between cities and over tine in the ‘sane country, then these expenditures can be regarded as "travel budgets". Furthermore, under certain conditions these travel. budgets may be applied as constraints on travel behavior, ‘The application of explicit constraints in modeling travel behavior is a Powerful tool, since the constraints do avay with the need for much of the coeffictent calibration of conventional nolels, within this context, “traveler” and “household” refer to a zepresen= tative traveler or household within a socioeconomic group, and not to a specific, identifiatle traveler or household uho say, over tine, move fron one socioecononic group to another.(34) Current. travel denand nodeis(*) are calitrated to the travelers observed choices. Since such choices are nany and varied, separate models have to be calibrated to their components, such as trip rate ty purpose, trip Atstance ard tine, node choice, and trip distribution. ‘The UNC? process, cn the other hand, is tased on the money and tine ‘budgets under which travel choices are made, Thus, all travel con- ponents are unified ty trade-ofi's within one system. The implications of this concept can be stated in the following wy: (1) The Travel Money Budget © The household's expenditure on travel is strongly related to its socioeconomic characteristics, and is stable over time, © Hence, the total generated travel distance per household will depend on the allocated travel money budget and the unit costs, by node. (2) The Travel Tine Budget © The daily travel time per traveler is a function of such fac~ tors as travel spood and urban structure, and the function 4s stable over tino. © Hence, the total daily travel distance per traveler will depend on his allocated travel tine budget and the travel speed, ty node, (3) The Interaction Process o Assuming that a household allocates to travel a money budget of M dollars por day ant a time budget of T minutes per day, and it has m available modes with different travel speeds and costs, it then must choose a certain combination of modes to maximize its total spatial and economic opportunities, as rep- resented by the total travel distance, Solving this problem results in the following outputs: « Daily travel distance, by mode (i.e., from which nodal shares are direct derivation). The car ownership levels required to satisfy the denand for travel Aistance by car. (o) Through conmon usage, nodels which develop estinates of travel are called “travel demand medels", although this is technically imprecise. Nonetheless, the terns travel model and travel donand nodel are used interchangeably in this report. See Section 1.2, on p.6 of the report, for a further exposition of this point,(assy © Expenditures of money and tine allocated to each node and, hence, the expected revenues for public transit operators. «Many other travel conponents, such as trip rates vs, trip distan- ces, within the total travel distances trip zates vs. trip tine within the total travel tines and trip distribution vs, urban structure, vithin the totel travel money and tine budgets. These outputs are then validated by comparing then with the observed choices, not calibrated to then, ‘The following sections present the theoretical foundation of the UMGT and the enpirical evidence for its application. Theoretical Foundation of the UNOT ‘The travel behavior concepts of the UMOT process were developed with Tespect to a particular theory of huran decision naking called utility theory. While this is not the only theory capable of providing a basis for the analytical structure of the proposed process, it is adequate to, ani nay in fact be, the best alternative currently available, particu- larly in light of its extensive end successful use in deserSbing many types of consumer choices in microecononics. The particular forns of utility functions applicable to travel situa~ tions were formulated, and it was found that these forms predict the enpirical regularities in travel decisions which forn the bases of the UMOT process. Forecasting individual travel demand can be accomplished by two alter- native approaches. The current approach is to calibrate a "demand model to individual travel behavior with an extensive number of inde- pendent variables in order to describe the variability of individual choices, This approach has the difficulty that each of the independent variables must be predicted separately for the future date. Therefore, operational planning models are typically restricted to a subset of the independent variables, at a sacrifice of explanatory power. The UMO? approach, on the other hané, ts based on the observation that the variabilities in the travel tudgcts anong individuals within groups are similar for all population segments. Therefore, assuming that such varlabilitios renain stable over time as weil, it is necessary to fore~ cast only the moon values of the budgets for éach group. These mean values, together with the variabilities eround then, provide the pro- babilities of individual travelers behaving in proileted mys. One advantage of thts process 1s that it needs only a few independent variables in omier to predict individval travel behavior under new conditions.a (ay) The Findings The Tindinga of this study were very encouraging and agreed with previous findings, Although they need further verification in more cases in orter to becone conclusive, their consistency in the avai- lable cases in Cermany, the U.S., the U.X., and even in developing countries, appears te be atove coincidence, as sunnarized below. The Travel Tine Budget (7T-budget) (1) The nean daily TT-budget per sotorized traveler (car, tus, train, taxi) of each of 55 different household types in Munich 1s stable over weekdays, and is similar to the TT-budget observed in other cities of developed countries. (2) The variations around the mean value of dally TT-tudgets by individual travelers within each population segnent in Munich and Nurenberg are very sinilar for all population segnents, and are consistent with those found in other cities, including cities of developing countries (Coefficient of variation about 0.6), (3) The mean daily TT-budget per traveler is an inverse function of speed, decreasing as speed increases, to an asymptote of about 1.1 hours per day. Thus, when low speeds increase, such as by transfor fron tus to car travel, part of the “saved" tine is actually saved, while the other part is traded-off for nore travel distance, However, at initially high speeds, practi- cally ei] speed increases are traded-off for more travel dis~ tance. (4) Te minimun mean daily TT-tudget per notorized traveler in an urtan area, at high travel speed, is just over 1 hour per day. The maximun mean daily T?-budget per motorized traveler, at low travel speed, is about 1.5 hours per day in cities of developed countries. However, in large and/or fast expanding cities in developing countries, the mean daily 77-tudget per motorized traveler at low travel speeds can reach the two-hour level per day. In conclusion, the mean TT-tudget per traveler wes found to be closely related to the speeds of the available transportation system supply, and to urban structure. High incone travelers, baving better oppor- tunities for traveling at higher speeds than low incone travelers, can both travel more daily distance (1.e., and have better spatial and econonic opportunities) than low incone travelers, and expend less travel tine. Low incone travelers, on the other hand, have to spend more tine in onder to travel less daily distance. Tais difference in mainly due to the travel meney budget, as detalled below.() The Travel Noney Bulget (TH-budget) (4) The nean TH-budget per household, for all households over tine, 4s found to be a relatively stable proportion of disposable income in Gernany, the U.S., the Usk. and Canaca, Yhile the proportion may vary ty definition and conditions between countries, the sta~ bility within one country resained unchanged even during the energy crisis and cost increases in 1973-75. (2) The daily M-budgot in an urban area is strongly related to the household's disposable income level and car availability. It is about 11-12 percent of disposable income for car oxning house- holds, and about 3-5 percent for carless households, for all in- gome levels. Such results in the Nurenberg region are similar to those in Washington, D.C. and Twin Cities in the U.. (3) The significant gap, or “quantum-junp", in the TH-budget of car and carless households suggests that purchasing a car is a najor household decision, closely related to other major decisions, Such as Tesidence location. However, the probability of @ house- hold to make such a decision inereases rapidly with incone. (4) The daily variations around the mean Ta-tudgets per household Within each segnont of the 55 household types in the Nurenbers region (as reflected ty the daily travel distance) are higher by about 50 percent than the varlations around the mean ‘?-tudgets, One oxplenation for this is that if Is easier to transfer money between dave than to transfer travel tines daily travel tines are constreined by 2h-howrs per day, while the travel money is const- Fained by the nonthly or the yearly incone. The point to note, however, is that the coefficient of variation of the TH-buiget, sintlar to the case of the T?-budget, is.very siniler for all population segnents. In conclusion, the IN-budgets of car and carless households appear to be strongly relatet to incone and car omership, and stable over tine. Hence, the Tw-hudget, together wlth the T?-budget, can he applied as two simultaneous bidgets unter which all travel choices take place, as sunnariged in the following section, It may thus be inferrel that travelers have preferred levels of expen- ditures for travel in both tine and money, In other words, travelers prefer to expend certain amounts of time and meney for travel in onder to derive the tenefits that St brings, Hovever, actual expenditures may or may not accord with preferred expenditures, depending on incone, urban forn and trensportation supply. When such a difference occurs, @ social discquilibriun may be sald to exiet, despite the fact that a. teal equilibrium in travel denand ani supply has teen achieved, The Social discquilibriun 2an te reardel as a force for change in urban structure or transportation supply.5. (va) Inplications of the TT and TM tuigets Given the houssholds' socioscononic characteristics and the transpor- tation systen characteristics, the UMOT process proceeds along the following, simplified, steps: (4) Household incone affects the upper Limit of the TM-budget, alio- cated to travel. Houeehold size and incone level affect the nunber of travelers per household and, hence, the initial TT- budget per household, (2) Applying the travel-distance maximization process under the TM ani TT budgets, and the unit costs of available or planned nodes, in money and tine toms, results in the denand for travel distance by each mode. (3) Car omership is generated by the denand for car travel distance. (4) The interaction between the estimated number of cara and a given road network results in new unit costs of travel. These new unit costs are fed tack into the travel denand phase, affecting both the TT-tuigets (which are sensitive to speed) and the denand for travel distance by node, and the process is repeated by iterations until convergence of the demand for travel and systen supply te reached. Tests have show that convergence of the process 1s rapid, and results in the observed travel characteristics, Tt should be noted at this stage thats (a) The process converges to the observed travel characteristics even when starting the process with extrene assumptions, such as every household ows, say, five cars; the constraining budgets and unit costs ensure that the process will converge rapidly to the observed characteristics, thus suggesting that the process is robust. (2) The variability between individual households ie expressed ty the variabilities in the TT and TM budgets, thus resulting in the pro- tability of a household with certain socioeconomic characteristics to ow 0-1-2-3+ cars and make certain travel choices. (c) Since the process 1s not dependent on calibration to observed travel characteristics, it is responsive to policy options beyond observed conditions, such az to new modes or to new travel costs. Furthermore, since the constraints are more stable than the choices (1se., choices change when conditions change because of the rela. tively stable constraints), there is more assurance in the trans— ferability of the process between citios and over tine than with nodels which are calibrated to the observed choices.(vis) (a) The process gives equal attention to all modes, generating esti- nates of travel ty all defined nodes, such as walking, car, tus, urban and interurban train, and even air modes, all of which are expressed by their operational characteristics, Further developments, as suggested below in Section 7, tncluie the addition of other characteristics, such aa confort and safety, in order to make the process nore sensitive to small differences between similar nodes. Urban Structure Operational urtan form models and travel denand nodels currently are developed ani operated separately, where each one requires the assis~ tance of the other in order to deal with the complexities of the uxtan syaten, The UNOT process has the potentiality of conbining toth aspects in one process, Prelininary tests suggest that given systen supply, the process can generate the protatilities of different household types to reside in certain areas of a city. In this case, however, the process will have to include the housing supply as an’ explieit component, and the travel utility functions will have to include Roney expenditures on housing. One important advantage of the UNOT process in dealing with urten structure is the question of equilibriun, Practically all urban form models that deal with dynamic changes in urban structure are Tased on the principle of equilibrium between denand and supply, 80 that all testei alternative urban structures always reach, oF at least approach, equilitriun conditions, such as between popu~ lation and job spatial distributions, In'the UMOT process, on the other hand, urtan structure may impose on the population long trip distances or trip tines/costs which may becone the binding gonstraints, overriding the preferrel TT and TM budgets, Put ano- thor way, travelers in such cities ray have to spend 4 hours per day and 25 percent of their income in order to travel to and fron Work, as is the case for sone population segments in several cities of developing countries. While conventional models regard such a case as being in equilttriun, tased on the definition that observed Genand is always in equilibriun with supply, the UMOT process mes- sures the anount of diseguilibeiun between the preferred TT and TM tudgets under stable urban conditions and the Tf and TM budgets that travelers will be forced to spend in order to satisfy their minimun denand for travel, such as nondiscretionary travel. Mhis subject is one of the areas sugested for further developnent, as proposed below.(vats) Conclusions and Recommendations ‘The conclusions nay be summarized as follows: (1) The analyses conducted in Germany corroborate previous observa- tions, tn that households and their travelers tend to exhibit consistent and predictable travel tine and noney expenditures, closely related to the socioeconoaic characteristics, levels of systen supply and urban structure. The data aleo display very state variations around mean TT and TH budgets for all population segnents, which reflect individual departures fron the groups" mean values, such as daily variations and personal Freferences in travel behavior. (2) The UOT process appears to be theoreticslly consistent and Fesses enpirical tests by being able to produce observed travel characteristics independently, even without being calibrated to then. Furthernore, all travel components are closely inter- Linked by trade-offs within the process. (3) The UMOT process appears to be able to address many important. policy issues, such as increasing fuel prices, land use control, Population migration, and changing transportation infrastructure, when based on the travel budgets under which decisions are made. The resultant changes in travel behavior can be predicted for each policy altemative without having te assure that each new choice 1s tased on the interpolation or extrapolation of previous relationships. Indeed, extrene policy changes can tring about extensive, but interrelated, changes In the various travel choices. ‘The Recommendations nay be summarized as follows: (1) Further developments of the UMOT process should include: (a) Further verification of the regularities of travel budgets over time. (b) Determination of whether additional attributes besides travel tine and cost are needed to derive node choice from the bud- gets, These attributes are allowed for in the utility fune- tions of UMOT, A further part of this investigation would determine what modal characteristics, such as confort and safety, would be required. (c) To further relate trip rate with trip distance for different purposes and nodes. (a) Further elaboration of the car ownership model, as a link to travel denand, systen supply and urtan structure.(2) @) @) G) (ix) The relationship of travel and urban structure should be nore fully developed. This would include the determination of aspects such as the probability distribution of residence and job locations. A further goal of this work is to include the dynanic processes of urtan development and travel as integral parts of the model. Bven at this stage of development of the UNO process, the mil- tiple interactions anong the travel conponents require ite deve- lopaent into a computerized model. Such development would enable At to both be exenined and checked under more complex situations and to be tested on nore data that vary over tine. A regional and interurban UMOT madel, not being dependent on urban structure, can be developed directly as e computerized nodel, inc- luding car travel and all najor public transport nodes, ground and air,CONTENTS Foreword GHAPTER 1. INTRODUCTION st The Approach 1.2 The Measurement of Travel Demand SHAPTER 2, UMOT MODELS OF TRAVEL BEHAVIOR ‘Introduction Sunmary ‘The Concept of Utality A General Utility Mode? ‘The Logarithmic Utility Model Decision Variables for Travel A Model of Travel with One Budget A Wodel of Travel with Two Budgets Sinplifications for Short-Run Behavior © Extensions for Nondiscretionary Travel and Price Indices 1 Gar Oimership Decisions 2 Estimation Considerations Sota on BURL Be HAPTER 2. [He TRAVEL SUNGETS 3.4 Introduction A, The Travel Time Budget ‘The Use of Time: An International Comparison Activity Tine Allocation in the Fed. Rep. of Germany Baily Travel Time per Person in the U.K. Daily Travel Tine per Person in French Cities Daily Trevel Times in the U.S, 8} Washington, D.C. and Twin Cities (2) St. Louis (3) The TT-budget per Person in ? Cities (4) The Bffect of an Expressway on the ‘TT-budget 7 Sone Connents on Disagerogate Tata 8 Daily Variations in the Use of Cars 9 taily Travel Time per Traveler in Bogota, Colombia 10 Daily Travel Tine per Traveler in Santiago, Chile it Daily Travel Tine per Traveler in Singapore 12 Mily Travel Tine per Traveler in Salvador, Brazil +13 Daily Variations in TT-tudgets in Munich, Germany: 3.i4 Daily Travel Tine per Traveler in the Nurenberg Region On a 73 76 2B 87 93 100 (cont. }(>) B, The Travel Money Budget 3.15 The Travel Money Budget 3,16 The TH-tudget Between Countries 3.47 The Money Budgots of Kouscholds 3.18 The Th-budgets in Two U.5. Cities 3,19 The TH-budget in the Nurenberg Region 3.20 General Comments on the TM-tudgots CHAPTER 4s, IMPLICATIONS OF THE TRAVEL BUDGETS 1 The Dally Travel Distance 2 Travel Distance per Traveler va, Speed in Two U.S. Cities 3 Travel Distance per Traveler vs. Speed in Singapore 4 Travel Distance per Traveler vs, Speed in the Nurenberg Region 5 Travel Distance per Traveler vs. Speed in Munich & Comparisons of the Distance vs. Speed Relationships 7 The T-budget vs. Speed Relationship 8 Variations in [nily Travel Distance Between Days 9 Dsily Travel Distance vs, Household Income 10 Modal Choice 11 Minimun Daily Travel Distance 12 The Measure of Mobility 1. Introduction 2, Nobility Neasure of Road Networks 3. Nobility Measure of Travelers 4, The Alpha Relationship in Germany 5. Mobtlity in the Nurenberg Region 4,13 Proportions of Households Generating Travel, Nurenberg 4.1 travelers per Households FERERESSEEES CHAPTER 5, THE UMOT PROCESS 5.1 Introduction 5.2 The UNOT Gar Owmership Model 4, Introduction 2. The Car Ounersh3p Process 3. An Example 3. Convergence of the UNOT Process 4 Denand vs. Supply 5 Urtan Structure & ‘Te Differential Accumulation Process 7 Travel Prowatility Fields Al 1 TER 6, CONCUISIONS AND RECOMMENDATIONS Conelustons Reconnenda tons RSPERENCES Page 104 4104 106 113 115 122 185 192 195 197 209 243 226 235 245,AFPENDIGES Appendix I. Appendix 11, Appondix TIT. Appendix IV. (e) The Munich 3-day Data Le By 3 im 5 6 2 a. ‘Travelers per Household Daily Travel Time per Traveler Inily Travel Distance per Traveler Tnily Travel Tine per Treveler Dnily Travel Distance por Traveler Deily Travel Tine per Traveler ist Day ist Day 2nd Day 2nd Day 3rd Day Daily Travel Distance per Traveler - 3rd Day Inily Average Speed The Nurenberg Data 1 2 3. 4 5 & ‘Travelers per Household Tnily Travel Time per Traveler Thily Travel Distance per Traveler Daily Average Speed Taily Travel Tine per Kousehold Daily Travel Distance per Household ‘The Munich Complete 1-day Sample te & Summary Table per Traveler Summary ‘lable per Householé ‘Travel Time and Money per Household 1 Nurenberg and Munich Page 250 252 253 255 256 257 258 259 260 264 262 263 264 265 266 267THE_UNOT PROTECT FOREWORD 4, This report presents the results of two research studies, conducted. during 1978-1979 in two different countries: (@) "Further Development of the UNOT Process"), carted out for the U.S. Department of Trensportation, Research and Special Progmins Administration, Office of Systens Engineering, Washington, D.C.; (b) "Verification of the VUSI Components" ™*), carried out in collatoration with Kocks Consult GHB, Koblenz, for the Ministry of Transport of the Federal Republic of Germany. The forner study investigated the conceptual franework of a pro- posed trensportation planning methodology called the UMOT process, concentrating on its theoretictl tases. The latter study involved testing several of the UMOT principal assumptions using observed travel data, While the tno studies were initiated as separate projects, their complenentarity was soon recognized and - to the full eredit and foresight of the two authorities - an interchange of progress reports ensued, culninating in this integrated final report. ‘The presentation in this report proceeds fron the geneal to the specific, and from the theoretical structure of the UMOT to its testing and application. The report is divided into six parts: Ghapter 1, Introduction, outlines the UMOT process and discusses the tasic principles of the approach, eee (*) UMOT - Unified Mechanism Of Travel (**) WUSI - Yerkehr Und Stadt ale Interakttonsnechanisnus (USOT)Chapter 2, The Theoretical Rusts, presents underlying principles and mthenstical formulations of models used within the UMOT process, Various model components are defined and analyzed for internal consistency and potentsal applice bility. Chapter 3, The Travel Budgets, sunnarizes available German travel data, supplexented by data from other parts of the world. ‘The data are analyzed and compared, toth cross-sectionally and over time, in light of compatability with UMOT assump- tons and model results, Ghapter 4, Inplications of the Travel Budgets, explores the effects the rolatively stable tine and money budgets have on velationships among travel distances and speeds by nodes Ghapter 5, The UMOT Nodel, details the structure of the UMOT process, with severs] example applications and sensitivity tests. Ghapter 6, Conclusions, summarizes the conclusions from the research study and recommends directions for further developments. Roaders not interested in detailed theoretical nodel developments night find it appropelate to skip sone of the materiel presented in Chapter 2. ‘The approximate shares of the two projects in the cix chapters, in percent, aros Gapter US. DoT FBG. MOT 1 50 50 2 30 io 3 10 90 4 20 80 5 80 20 6 50. 50.‘Tis report is the outcome of concerted efforts of a devoted tean ef consultants, who bave put mich thought and time into this re- search study, Walle 1t ts not possible to isolate the personal contributions of each consultant in this report, the general fields of expertise ares Prof. Wartin Beckramn - mathematical economics and microeconomic theory; De, Thomas Golob - consurer theory and transportation analyses; Prof, Antti Talvitie - systems performance and nunerical analyses, ‘Two additional consultants particlmted in the early stages of the study: Prof, Chong Liew - econonetricss Prof, Richard Weissbrod - urten geography, The author is deeply indebted to all consultants for their helpful assistance, However, the final responsibility for the opinions, analyses and conclusions presented in this report zest with the author only,GUAPTER 1: INTRODUCTION 1a ‘The Approach ‘Te UMOT was conceptualized for the World Bank, Urban Projects Department, as a tool for the rapid estination of urten travel and the evaluation of altemative transportation plans (Zahavi, 1977). Tt was further developed for the U.S. Department of ‘Transportation and tested with actual travel data for the Ministry of Transport of the Fed. Rep. of Germany, as presented in this report. ‘Tne UMOT process differs from the mjority of transportation planning procedures, whether aggregate or disiggreente, recursive or siml- taneous, In several aspects. Those differences are elaborated upon in the following chapters but can be summarized here as follows. (1) Galibration: Most conventional models are calibrated to the observed choices. Such models are then validated by their ability to reproduce the same observed choices, such as car ownership, trip rates and modal choice. The UMCT process, on the other hand, 1s based on paramoter ostimation of the constraints under which the choices are made, The process is then validated ty ite ability to produce the observed choices as independet cutpute. (2) Dravel Derand: In most conventional nodels travel denand 1s expressed by trip rates, Such models start with trip rates and conclude with passenger and vehicle travel distance, The UMOT process starts with the “denand for travel" as expressed by desired passenger and vehicle travel distance that can be gene- rated within the travel constraints and, after a feedback process, concludes with trip rates, (3) Car Omership: In most conventional models car omership 1s treated as @ long-tern decision that affects short-term trip generation and mode choices, In the UMCT process long-term and short-term decisions interact, for instanoa, it is pro- posed that denand for high-speed personal travel generates demand for cars, The number of cars, estimted to satisfy ‘the demand then interacts «ith cystem cupply, resulting in revised unit costs of trivel. These costs then feed backinto the first phase, resulting in a revised denand for car travel and for each alternative mode. The process is iterated until convergence ts reached, Tests have shown that thie convergence is rapid. In sunnary, a major difference between conventional planning processes and the UMOT process 4s that conventional nodels are generally calibrated to observed choices, such as car ownership, trip rates and nodal split, choices which may change beyond the range of observations if conditions change. A tasic assunption, therefore, is that such models can predict future choices, even beyond the range of tase-year observations. The UNOP process, on the other hand, can be initiated even with extreme assunptiona, such as every household In an urben area owning 5 cars, and the nodel outputs will nonetheless converge to the observed choices. This is achieved by shifting the enphasis from the observed choices to the constraints under which the choices are nade. The UMOT process presupposes that most travel choices are made Decause travelers are under constraints, such as those of tine and money. Current transportation supply conditions dictate how travelers can allocate their constrained resources so as to best achieve travel benefits, Jf these conditions change, the choices are likely to change, while the constraints renain relatively stable. A purpose of this study te to identify such travel const- raints, test their stabilities over space and tine, and - if they are verified - explore their implications to travel modeling. ‘The Neasurenent of Travel Denand "Travel demand", in the conventional meaning, is the part of travel that is realized and observed after interaction of the "demand for travel" xith system supply, Travel denand is currently measured mostly by trip rates, based on the reasoning that the direct "uti- Rity", or benefit, of making a trip to a certain destination fora certain purpose by a certain mode at a certain tine, should surpass its "disutility", such as its cost in tine and money tems. ‘The performance and productivity of system supply, on the other hand, are meacured by the passenger-km., ton-kn., or vehicle-ka, "produced" at certain specds, Therefore, the only possible way to establish comnon denominator between denand and supply is to tase it, in both cases, on either trips or travel distance. ‘There is an Increasing number of independent indications to suggest that the daily travel distance per traveler/houschold is a better measure of travel demand than trips, especially when aggregating travel over tine and travelers, While more on this subject is detailed in Section 2.6, the following example itlust~ rates the problen and its possible solution. s Consider, for example, a traveler who is observed te make 6 trips per day in his small hone-tom, tut after moving toa large city As found to make only 3 trips per day: Did his travel demnd decrease 7; Is his utility of travel, and nobility, less in a Jarge city than in a small tom ?; Or aid the utility of each trip change ? Let us now edd another travel component to the above example, nazely trip distance, and it is found that the trip distance is 5 km, in the former case, and 10 km. in the latter case. Tt now becones evident that the traveler covers 30 km. per day in both cities, within which he trades off trigs vs. trip dis- tances, Put another way, city size, which affects trip dis- tances, can also affect trip rates, while the total travel demand, 1f measured by the daily travel distance, may remain unchanged by city size. The question, of course, 1s whether travelers do behave in this way ?Table 1.3-t presents one such example, from England, for the relatively smal town of Kingston-upon-Hull and the large metropolitan area of Tendon, Although the size of the study area of Kingston-upon-Hull is only l,l percent that of Tondon, the daily travel distance per average car is sinilar in both cities, while the trip rate and trip distance are inversely related within the total travel distance, The sane petterns were also observed ina wide selection of cities in the U.S., Europe and developing countries, (Zaha, 1979-2), Table 1.3-1: Travel Characteristics in Two UK. Gities Garacteriztie Keresten-wpan-fuld| tendon Year 1967 1962 Population 3,890 |8,857,000 Area, Squta, 107 2,450 ars 43,185 |L.289,450 cars/t00 Persons 325 wt car Trips 238,000 [4,119,000 car Trip Rate TT Tehp Distance, £2, 738 Trip Mee, May Be GE Daily Distance, Kay Ded Ger Matty Tio, Bre. °. 2. An additional interesting result in the above exanple is that the sane inverse relationship also applies to the trip mate vs. trip ‘ime within the total daily travel tine per average car which, once again, is very sinilar in both cities, nancy 0.72 vs. 0.75 hours, One possible interpretation of the above example is that the daily travel per average car driver in the tvo cities, if measured by the total datly travel distance, and for which the traveler has to pay in both time and money terms, is pmctically the same, It nay, therefore, be inferred that if we wish to have # moasure of travel denand, that can ‘ve generated within certain amounts of allocated travel tine and money, and which is transferable botween cities of different sizes, it appears that the total daily travel distance per representative traveler/house- hold 1s a better measure than the daily trip mte, It is shown later in this report that this measure of travel allows ue to unify travel demand, systen supply, ear ownership, nodal choice and urtan structure within one operational system, with full feedtack between all components.‘That travel distance, at least for car travel, is transferable between chties, as well as even between countries, is shown also in Table 1.3-2, which summarizes the daily travel per average car (internal-interal, tzips of cars registered in the study area) in the metropolitan area of Washington, D.C., U.S. 1968, the Nurenberg Region in the Fed. Rep. of Gernany 1975, and the city of Kuala Lanpur, Malaysia 1973. Table 1,3-2: ‘Travel haracteristics of Washington, D,C., Nurenberg and Kusla Tampur Cmaracteristie Wachingten [ Nurenterg | Yala Lanpar Year 1968 1995 1919 Popslatton 2,558,190 | 1,160,000 912,690 Aron, Saka, 340 3,000 37 care 3,018,900 328,000 65,0 ars/100 Persons 39.8 28.3 ne ar Trips 3,342,000 _| 1,007,000 49,950 car Trip Rete 3 77 | S78) Gar Trip Distances, ka, 10.59 11.20 5.36 Oar Duly Distance, ke, a Bat 3603 It my be inferred from the above table that travel distance is indeed a relatively stable measure of trevel, transferable between cities, While the trip rate and the trip distance interact within it by trade- otts.(*) Much attention is given to such relationships already at this introductory stage of the report in order to emphasize two points; first, that most travel components interact with each other and, therefore, they should be treated simultaneously within one analytical system; and second, that Af a nodel 1s to be transferable over time, for prediction purposes in one city, 1t should first demonstrate transfembility over spice, between @ifferent cities, with no need for recalibration of the model to the isolated travel components, such as trip rates, This, then, is also & prerequisite for the UMOT process, (*) Variations atout mean values, ona disaggremte levels, are discussed in Section 4,8,10 An additional inportant conclusion from the above examples is that when travel conponents are sunmed up over a day (and preferably over ® more prolonged period, such as a week), several travel factora and pattems energe ag stable, While the observed stability of the daily travel distance per average car Tegirdless of city size has many implications for both planners and policy nakers, as discussed in Section 4.11 , it 4s quite obvious that 1t 4s otill only one yart of @ nore complex system, since different households my have three cars, or no car at all, and still generate travel by other nodes. Hence, when extending the search for travel behavioral factors which are stable, a2] results leod to the travel mudgets of tine and aoney that travelers/nouscholds are willing to allocate to travel, and within which 221 observed travel is generated, At this stage, however, several questions need to be addressed before proceeding with developnent of the UNO? process. Forenost are the questions: * Do such travel "budgets" exist in the veal world ? © If such travel tudgets exist, can they be derived from avatlable tamvel observations, and oan they be related to the socioeconomic characteristics of population segments ? * How stable are such budgets spatially and temporaly, and what are the distritutions around the neans of the budgets for diffe- rent population segnents ? ¢ How sensitive are the travel budgets te exogencus changes, such As changes in the cost of travel ? ¢ How can the “willingness” of population segnents to allocate travel budgets be inferred and estimated if they are not actually observed ? ¢ Finally, how can they te integrated within @ unified theory of travel ? A flret step in answering the above questions 4s to determine whether or not the requirements of the UMOT process are condusive to the Gevelopnent of travel demnd models which are consistent with estab- lished theories of human behavior, This 1s the subject of the next chapter.un CHAPTER 2: UNOT MODELS OF TRAVEL BEHAVIOR 2A Introduction The UMOT process is tased upon the strong evidence that travelers have two explicit and separate travel tudgets, of money and time, as detailed in Chapter 3, ‘he effects of such money and time budgets on travel behavior can be sunnarized as follows1 (1) The Travel Money Budget. © The household's expenditure on travel is strongly related to its income, and is relatively stable over time. @ Hence, the total generated travel distance will depend on the unit costs, ty mode, (2) The Travel Tine Budget © The daily travel tine per traveler is relatively stable over tine, © Hence, the total travel distance will depend on the travel speed, by mode. (3) The Interaction Process © Assuning that @ household allocates to travel 2 noney budget of M dollars per day and a tine tudget of T minutes per day, and it has m available nodes with different travel speeds and costs, the question is by what combinations of modes can the household maxinize its spatial economic opportunities, as represented by the total travel distance 7 The first purpose of the research reported in this chapter is to test whether principles underlying the UMOT process are consistent with established thoorles of behavior in economics, ‘he second purpose 1s to explore inpliettions of such economic theories to further formiation and verification of the UOT process, The yartioular eeonente theozy chosen to accomplish these purposes is consuner utility theory, as elaborated in the following sections,