Waardering van kwaliteit en betrouwbaarheid maart 2004
04003
De waardering van kwaliteit en betrouwbaarheid in personenen goederenvervoer
een rapport voor AVV door RAND Europe
Waardering van kwaliteit en betrouwbaarheid
Voorwoord
In opdracht van de Adviesdienst Verkeer en Vervoer (AVV) en gefinancierd door de Directie Besturing en Strategie van het Ministerie van Verkeer en Waterstaat heeft RAND Europe een literatuurstudie uitgevoerd naar de waardering in geldeenheden van betrouwbaarheid van de reistijd en andere kwaliteitsaspecten in personen- en goederenvervoer. Deze aspecten worden momenteel niet meegenomen in kostenbatenanalyses van transportprojecten volgens de systematiek van het Overzicht Effecten Infrastructuur (OEI). De probleemstelling voor het hier gerapporteerde onderzoek is als volgt: Wat is in de Nederlandse en internationale literatuur beschreven over de waardering van kwaliteit en betrouwbaarheid van infrastructuur en hoe kan dat worden toegepast binnen de bestaande OEI-leidraad? Hierbij gaat het om personen- en goederenvervoer en om de vervoerwijzen weg, spoor en water. Dit rapport bevat een synthese van de bevindingen in de afzonderlijke rapporten en artikelen uit de literatuur. Beschrijvingen per rapport/artikel zijn te vinden in Bijlage 1 van dit rapport. Het rapport is bedoeld voor economen, verkeerskundigen en andere transportonderzoekers die zich bezighouden met kosten-batenalyses in het kader van OEI of invoer (effecten, waarderingen) leveren voor dergelijke analyses. Het projectteam van de zijde van RAND Europe bestond uit Gerard de Jong (projectleider), Eric Kroes (programmadirecteur), Ronald Plasmeijer en Peter Sanders. Dit rapport heeft de kwaliteitscontrole doorlopen volgens de richtlijnen van RAND Corporation (zie: http://www.rand.org/about/standards) en kan daarom beschouwd worden als een RAND Europe produkt. Voor nadere informatie over dit onderzoek kunt u contact opnemen met: RAND Europe: Gerard de Jong Newtonweg 1 2333 CP Leiden Tel.: 071-5245151 Fax: 071-5245191 e-mail:
[email protected]
Adviesdienst Verkeer en Vervoer: Pim Warffemius Postbus 1031 3000 BA Rotterdam Tel.: 010-2825764 Fax: 010-2825643 e-mail:
[email protected]
Pagina iii
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina iv
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Samenvatting
In deze studie is de nationale en internationale literatuur over onbetrouwbaarheid van reistijden en andere aspecten van kwaliteit in personen- en goederenvervoer bestudeerd om de volgende vragen te beantwoorden: • • • • • •
Welke definities hanteert men? Welke aspecten zijn bestudeerd? Welke kengetallen bestaan er voor de monetaire waardering? Welke methoden daarbij zijn gehanteerd? Zijn deze kengetallen toe te passen in kosten-batenanalyses volgens het OEIkader in Nederland? Welke lessen kan men trekken voor het opnemen van betrouwbaarheid en andere kwaliteitsaspecten in transportmodellen?
In totaal zijn in dit onderzoek 40 rapporten en artikelen bestudeerd en samengevat. Vervolgens zijn de uitkomsten geïntegreerd om de bovenstaande vragen te beantwoorden. Er zijn geen monetaire waarden voor betrouwbaarheid en andere kwaliteitsaspecten gevonden die in officiële nationale kosten-batenanalyses worden gebruikt. Wel wordt daar in sommige landen (behalve in Nederland ook in het Verenigd Koninkrijk, Zweden) over nagedacht en zijn er onderzoeken uitgevoerd in diverse landen die waarderingen in geldeenheden of reistijd leveren voor betrouwbaarheid van reistijden1. Zowel voor de definities als de kengetallen voor betrouwbaarheid zijn drie operationele varianten gevonden: •
Onbetrouwbaarheid gemeten als de standaardafwijking (of variantie) van de reistijdverdeling. Gegevens voor de waardering kunnen verkregen worden door in een stated preference (SP) onderzoek zowel een weergave van de variantie als de gemiddelde reistijd als attributen op te nemen.
•
Onbetrouwbaarheid gemeten via het verschil tussen het 80ste of 90ste percentiel van de reistijdverdeling en de mediaan. Ook hier kan de waardering volgen uit SP experimenten onder reizigers.
•
Onbetrouwbaarheid gemeten als het aantal minuten dat men vroeger of later vertrekt of aankomt dan meest gewenst (schedule delay). Ook dit kan als
1
Met waardering voor betrouwbaarheid van reistijden in geldeenheden of reistijd worden uitkomsten bedoeld zoals: 10 minuten langere reistijd dan verwacht is equivalent aan X euro of Y minuten verwachte reistijd. Pagina v
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
attribuut in een SP experiment worden aangeboden, naast andere kenmerken als reisduur en reiskosten. De kengetallen volgens de derde definitie zijn zeer lastig toe te passen binnen het huidige OEI-kader, omdat in de KBA de vertaling van reistijd (=reisduur) naar tijdstip van de dag (en afwijking van gewenste aankomsttijdstip) niet gemaakt wordt. Dit is pas mogelijk als in de KBA ook naar de tijdstipkeuze zou worden gekeken (in plaats van alleen naar bestaande reizigers met een bepaalde vervoerwijze en verandering van vervoerwijze) en als daarbij ook de gewenste aankomsttijdstippen bekend zouden zijn. Binnen de vigerende methodiek zijn er meer mogelijkheden voor de eerste twee methoden, maar kengetallen waarvan met een redelijke mate van vertrouwen kan worden gezegd dat ze representatief zijn voor met name het Nederlandse autoverkeer ontbreken. De bestaande kengetallen komen uit zeer specifieke onderzoeken en worden in het land van herkomst ook niet voor landelijke of regionale kosten-baten analyses gebruikt (met name omdat het kengetallen voor specifieke groepen en/of gebieden zijn). Wel geven alle gevonden studies naar de waarde van betrouwbaarheid van de reistijd aan dat dit een factor van substantieel belang is: geen enkele studie kwam tot de conclusie dat deze factor te verwaarlozen is. Aanbevolen wordt om een nieuw landelijk representatief SP onderzoek op te zetten onder Nederlandse automobilisten, waarin de attributen van de keuze-alternatieven zijn: •
• •
een reeks van mogelijke reisduren (evt. ook aankomsttijden), mogelijk met een grafische weergave; dit is een weergave van de mate van onzekerheid van de reistijd, die de respondenten beter begrijpen dan de variantie of standaardafwijking (die de onderzoeker bij ieder alternatief kan berekenen); een gemiddelde reisduur; reiskosten.
De uitkomsten van deze interviews kunnen dan gebruikt worden voor het bepalen van de waarde van onbetrouwbaarheid in het autoverkeer, zowel in de vorm van de standaardafwijking als percentielen van reistijd. Deze aanbevolen SP studie wijkt af van de SP studie uit De Jong et al. (2003), omdat daar ieder keuze-alternatief beschreven werd in termen van o.a. reistijd, reiskosten en een vertrektijd, terwijl het hier gaat om een set van mogelijke reisduren (bijvoorbeeld 10 verschillende reisduren), naast gemiddelde reistijd en –kosten. Voor het openbaar vervoer zijn er voor Nederland kengetallen afgeleid in Rietveld et al. (2001) en zou kunnen worden bekeken of de Klantenbarometer Openbaar Vervoer (AVV, 2003) uitgebreid kan worden met monetaire waarderingen in vervoer per bus, tram en metro. Het reistijdwaarderingsonderzoek dat RAND Europe momenteel uitvoert voor AVV kan monetaire waarderingen bieden voor betrouwbaarheid en enkele andere kwaliteitsaspecten (frequentie, schadekans) in het goederenvervoer over de weg, en met het spoor, binnenvaart, zeevaart en luchtvaart.
Pagina vi
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Kengetallen voor andere kwaliteitsaspecten dan betrouwbaarheid blijven in het personenverkeer en –vervoer beperkt tot waarderingen voor diverse componenten van reistijd (ook in samenhang met frequentie) en zitplaatskans in het openbaar vervoer. Het beeld hier is diffuser dan van betrouwbaarheid: niet alle studies vinden dat frequentie en zitplaatskans significante factoren zijn in het keuzeproces van de reiziger. Diverse bronnen geven aan dat tijd in voor- en natransport, looptijd wachttijd en overstaptijd minstens zo zwaar meewegen als tijd in het voertuig. Met name voor looptijd en wachttijd zijn herhaaldelijk factoren in de orde van grootte van 1,5 gevonden (1,5 maal zo belangrijk als tijd in het voertuig, per minuut). Bestaande transportmodellen in Nederland bevatten geen expliciete variabelen voor betrouwbaarheid en andere aspecten van kwaliteit. Wel kunnen deze variabelen de keuzen van de besluitvormers hebben beïnvloed, die gebruikt zijn als te verklaren variabelen. De invloed van deze kenmerken komt dan waarschijnlijk met name tot uiting in de constanten voor de diverse vervoerwijze-alternatieven in de vervoerwijzekeuzemodellen (maar mogelijk ook in de parameters voor reistijd en reiskosten, voorzover hier correlatie mee optreedt). Simulaties voor het effect van veranderingen in kwaliteit kunnen uitgevoerd worden door te bepalen hoeveel bijvoorbeeld de equivalente verandering in reistijd of reiskosten zou zijn (op basis van SP uitkomsten) en die door te rekenen, of door de alternatief-specifieke constanten aan te passen.
Pagina vii
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina viii
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Summary
In this study the national and international literature on reliability of travel times and on other aspects of quality in passenger and freight transport have been investigated to answer the following questions: • • • • • •
Which definitions have been used? Which aspects have been studied? Which monetary values have been obtained? Which methods were used for this? Can these values be applied in cost-benefit analysis in The Netherlands according to the OEI-guidelines? Which lessons can be learnt for including reliability and other quality aspects in transport models?
In total, 40 reports and articles were studied and summarised in this review. After that, the results were integrated to answer the above questions. No monetary values for reliability and other aspects of quality were found that are used in official national cost-benefit analyses. But the possibilities of doing this are being investigated in some countries (besides in the Netherlands also in the United Kingdom and Sweden), and studies have been carried out in several countries that yield values in money units or time units for the reliability of travel times. For the definitions as well as for the money values for reliability, three operational definitions were found: •
Unreliability measured as the standard deviation (or variance) of the travel time distribution. Data for the valuation of the standard deviation can be obtained by including in a stated preference (SP) survey both a representation of the variance and the mean travel time as attributes.
•
Unreliability measured as the difference between the 80th or 90th percentile of the travel time distribution and the mean. Again the valuation can be derived from SP experiments among travellers.
•
Unreliability measured as the number of minutes that one will depart or arrive earlier or later than preferred (schedule delay). This can also be offered as an attribute in an SP experiment, together with other attributes such as journey duration and travel cost.
The monetary values following the third definition are very difficult to implement in the present OEI-framework, because the link to travel time period choice is not made in Pagina ix
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
the cost-benefit analysis. In the current methodology, there is more scope for applying the first two definitions, but money values that can be regarded as representative for The Netherlands, especially for car traffic, are lacking. The money values in the present literature come from very specific investigations and are not even used in cost-benefit analyses in the respective countries of origin (the main reason for this is that these money values are group and/or region-specific). All the studies on the value of reliability that were reviewed yield that this is a factor of substantial importance: there were no studies that concluded that this factor can be neglected. We recommend to set up a new nationally representative SP survey among Dutch car drivers with choice alternatives described in terms of the following attributes: •
• •
a series of possible journey durations (could also include arrival times), possibly represented graphically; this is a measure of the degree of uncertainty in travel times, which respondents find easier to understand that the variance or standard deviation (the researcher can calculate these for each choice alternative); a mean journey duration; travel costs.
The outcomes of these interviews then can be used to calculate the value of unreliability in car traffic, both measured as the standard deviation and as percentiles of travel time. This recommended SP study differs from the SP study in De Jong et al. (2003), in which every choice alternative was described in terms of travel time, costs and departure time. In the recommended survey, besides the mean travel time and travel costs, a set of possible journey durations (e.g. ten different values) would be used. For public transport, money values have been derived for The Netherlands by Rietveld et al. (2001) and it could be investigated whether the Klantenbarometer Openbaar Vervoer (AVV, 2003) could be extended to include monetary valuation in bus, metro and tram trips. The freight transport value of time study that RAND Europe is carrying out at the moment for AVV can offer monetary values for unreliability and some other aspects of quality (frequency, probability of damage) in goods transport by road, rail, inland waterways, sea and air. Published recent money values for other quality aspects (than reliability) in passenger transport are restricted to values for the different components of travel time (also in connection to frequency) and probability of a seat in public transport. The results are more diffuse than for reliability: not all studies find that frequency and the probability of a seat are significant factors in the choice process of the travellers. Several sources show that access and egress time, walking time, waiting time and transfer time recieve a weight at least as high as in-vehicle-time. Especially for walking and waiting time factors of about 1.5 have been obtained in a number of studies (1.5 times as important as in-vehicle-time, in minutes).
Pagina x
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Existing transport models in The Netherlands do not include explicit variables for reliability and other aspects of quality. But these variables could have influenced the choices of the decision-makers that the models seek to explain. The influence of these attributes will probably be expressed in the mode-specific constants in mode choice models (but possibly also in coefficients for travel time and costs, if there is correlation with these variables). Simulations of the effects of changes in quality can be carried out by determining the effect of an equivalent change in travel time or costs (on the basis of SP outcomes). Alternatively, the mode-specific constants can be adjusted.
Pagina xi
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina xii
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Inhoudsopgave
Voorwoord .....................................................................................................................iii Samenvatting .................................................................................................................. v Summary........................................................................................................................ ix Inhoudsopgave.............................................................................................................xiii 1. Inleiding ...................................................................................................................... 1 1.1 Achtergrond .................................................................................................... 1 1.2 Doel en afbakening van het onderzoek ......................................................... 2 1.3 Inhoud van dit rapport .................................................................................. 5 2. De gevolgde methodiek en fasering .......................................................................... 7 3. Betrouwbaarheid: definities en bestudeerde aspecten ......................................... 11 4. Kwaliteit: definities en bestudeerde aspecten........................................................ 17 5. Betrouwbaarheid: kengetallen................................................................................ 21 6. Kwaliteit: kengetallen .............................................................................................. 31 7. Witte vlekken............................................................................................................ 35 8. Conclusies ................................................................................................................. 37 Lijst van bestudeerde rapporten en artikelen ........................................................... 41 Referenties .................................................................................................................... 45 Bijlage 1. Beschrijving van de bestudeerde artikelen en rapporten........................ 53
Pagina xiii
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina xiv
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
1. Inleiding
1.1 Achtergrond De leidraad voor de kosten-baten analyse (Eijgenraam et al., 2000) uit het Onderzoeksprogramma Economische Effecten Infrastructuur (OEEI) is de afgelopen twee jaar toegepast op diverse, met name grote, infrastructuurprojecten. De ervaringen met deze OEEI-leidraad zijn geëvalueerd in opdracht van de ministeries van Economische Zaken en Verkeer en Waterstaat (Buck Consultants International, 2002). Naast een beschrijving van ervaringen met en meningen over het toepassen van de leidraad, heeft deze evaluatie ook een lijst met procesmatige en inhoudelijke verbeterpunten opgeleverd. Vanwege het kader dat de leidraad biedt voor integrale afweging van effecten, is besloten voortaan te spreken over het Overzicht Effecten Infrastructuur (OEI). In de Actieagenda OEI geven de betrokken partijen aan welke verbeterpunten de komende tijd prioriteit krijgen. Binnen de directe effecten worden in de OEI-systematiek kengetallen aangereikt voor het meenemen van de reistijdwinsten voor het verkeer. Deze vormen vaak een belangrijke batenpost. Door middel van reistijdwaarderingen worden de reistijdwinsten omgezet in geldeenheden, zodat deze kunnen worden opgenomen in de kosten-baten analyse (KBA). Voor de reistijdwaarderingen in het personenvervoer zijn kengetallen beschikbaar gebaseerd op een uitgebreid onderzoek dat Hague Consulting Group (HCG, intussen opgegaan in RAND Europe) in 1997/1998 heeft uitgevoerd voor de Adviesdienst Verkeer en Vervoer (AVV). Voor de reistijdwaardering in het goederenvervoer is er het onderzoek naar de reistijdwaardering in het goederenvervoer van Hague Consulting Group voor de Dienst Verkeerskunde (nu AVV) uit 1992. Dit onderzoek wordt momenteel door RAND Europe geactualiseerd, in opdracht van AVV. Diverse infrastructuurprojecten verkorten niet alleen de reistijd, maar verhogen ook de kwaliteit (anders dan de reistijd zelf) en de betrouwbaarheid (in de zin van spreiding van de reistijd rondom de verwachte reistijd) van het vervoer. Deze aspecten behoren ook tot de directe effecten, maar worden doorgaans niet in de KBA opgenomen, omdat de OEI-leidraad geen kengetallen biedt voor het kwantificeren en monetariseren van deze baten. AVV heeft RAND Europe gevraagd een literatuuronderzoek uit te voeren naar de waardering van kwaliteit en betrouwbaarheid in personen- en goederenvervoer. Het onderhavige rapport bevat de uitkomsten van deze studie.
Pagina 1
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
1.2 Doel en afbakening van het onderzoek Het uiteindelijke doel van het project ‘Waardering van kwaliteit en betrouwbaarheid’, is (conform het Projectplan van AVV): Het ontwikkelen van goede kengetallen voor de waardering van kwaliteit en betrouwbaarheid bij KBA’s. Het betreft zowel personen- als goederenvervoer en verder vervoer over de weg, het spoor en het water. De kengetallen moeten nauwkeurig aansluiten bij de bestaande OEI methodiek. De probleemstelling voor het hier gerapporteerde onderzoek is als volgt: Wat is in de Nederlandse en internationale literatuur beschreven over de waardering van kwaliteit en betrouwbaarheid van infrastructuur en hoe kan dat worden toegepast binnen de bestaande OEI leidraad? Hierbij gaat het om personen- en goederenvervoer en om de vervoerwijzen weg, spoor en water. Deze probleemstelling is tweeledig: enerzijds gaat het om het bestuderen van de literatuur over kwaliteit en betrouwbaarheid van vervoerssystemen; anderzijds gaat het om het confronteren van de uitkomsten uit de literatuur met de OEI systematiek: past wat men bijvoorbeeld in het buitenland doet binnen het kader dat OEI biedt? Het projectplan bevat een uitsplitsing van de probleemstelling in zes deelvragen. 1. Hoe worden kwaliteit en betrouwbaarheid gedefinieerd (in de Nederlandse en internationale literatuur)? 2. Welke aspecten van kwaliteit en betrouwbaarheid worden beschreven (in de literatuur)? 3. Welke methoden voor het waarderen van kwaliteit en betrouwbaarheid worden beschreven (in de literatuur)? 4. Wat zijn goede kengetallen voor het waarderen van kwaliteits- en betrouwbaarheidsaspecten? 5. Hoe kunnen deze kengetallen worden toegepast voor KBA’s die worden opgezet volgens de bestaande OEI leidraad? 6. Welke richtlijnen bestaan er (in de literatuur) voor het opnemen van kwaliteitsen betrouwbaarheidsaspecten in verkeers- en vervoersmodellen? De deelvragen 1, 2, 3 en 6 betreffen het onderzoeken en beschrijven van de literatuur (de toevoeging ‘in de literatuur’ is steeds van onze hand). Verkeers- en vervoersmodellen (deelvraag 6) zijn van belang omdat die de meeste invoer, zoals de reistijdwinst per vervoerwijze en verplaatsingsmotief in minuten, leveren voor de directe effecten in een KBA. Net als bij de factor reistijd, gaat het in dit onderzoek dus niet alleen om de monetaire waardering van kwaliteit en betrouwbaarheid, maar ook om het ramen van de mate van verbetering in aspecten hiervan die op zal treden als gevolg van een infrastructuurproject. Alleen als de effecten op kwaliteit en reistijd Pagina 2
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
kunnen worden gekwantificeerd en gemonetariseerd kunnen deze worden opgenomen in een KBA. De kengetallen voor monetarisering moeten tevens aansluiten op wat de verkeers- en vervoersmodellen (nu en in de toekomst) kunnen leveren. De vragen 4 en 5 betreffen een reflectie op de uitkomsten van het literatuuronderzoek: voorzover er kengetallen gebruikt worden voor de waardering van kwaliteit en betrouwbaarheid, kunnen er uitspraken gedaan worden over de kwaliteit van deze waarderingen en over hoe dit al dan niet past binnen het OEI-kader? Van belang hierbij is zowel het streven naar integrale evaluatie (alles meenemen) aan de ene kant als het voorkómen van dubbeltellingen aan de andere kant. In de OEI-systematiek worden reistijdwinsten al meegenomen en daarbij komen sommige kwaliteitsaspecten reeds aan de orde. Al naar gelang de bevindingen in dit literatuuronderzoek kan het onderzoek eindigen met aanbevelingen in de vorm van kengetallen voor toepassing in het OEIkader in Nederland of aanbevelingen voor gericht empirisch onderzoek in Nederland ter bepaling van monetaire waarderingen van ontbrekende aspecten van kwaliteit en betrouwbaarheid. Het projectplan noemt als voorbeelden van kwaliteitsaspecten in het personenvervoer: • • • • • • • • • •
Overstaptijd Tijd nodig voor zoeken (b.v. van een parkeerplaats of van het juiste perron) Looptijd Frequenties Aantal (bereikbare) bestemmingen Beschikbaarheid van een zitplaats De mogelijkheid om tijdens een reis te kunnen werken (b.v. op laptop) Reiscomfort Informatievoorziening tijdens het wachten (en tijdens de reis in het voertuig) Betrouwbaarheid (maar deze wordt apart bestudeerd in dit project).
De eerste drie aspecten zijn in wezen componenten van de totale reistijd van deur tot deur. Frequenties hebben ook weer effect op de wachttijd, maar frequentie bevat ook andere kwaliteitsaspecten (b.v. een hoge frequentie kan gewaardeerd worden vanwege de verkleining van de consequenties van vertraging). De reistijdwaardering in de OEIleidraad is gebaseerd op het stated preference (SP) onderzoek van HCG (1998). Hierbij zijn variaties in de reistijd aangeboden, gebaseerd op de totale verwachte reistijd van deur tot deur. Dus de gevonden reistijdwaarderingen gelden ook hier. Aan de andere kant is bekend dat wachttijd en overstaptijd een hoger disnut2 opleveren dan tijd in het voertuig (b.v. Van der Waard, 1989). In dit literatuuronderzoek zal ook nader op deze componenten van de totale reistijd worden ingegaan, waarbij het doel is om alle onderdelen van de reistijd met hun meest juiste waardering mee te nemen zonder dubbel te tellen. De overige aspecten van kwaliteit gaan niet over de reistijd, maar over de omstandigheden tijdens deze reistijd (zitplaats, werken, comfort, informatie) en het aantal bestemmingen dat men bereiken kan. 2
Het woord ‘disnut’ gaat uit van een nutsfunctie die aangeeft hoe het nut dat een reiziger ontleent aan een bepaalde reis met een bepaalde vervoerwijze afhangt van de reiskosten, de reistijd en andere kenmerken van de reis met die vervoerwijze. Langere reistijden leiden hierbij tot een lager nut, met andere woorden ze brengen ‘disnut’ met zich mee. Pagina 3
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Overigens beoogt de lijst slechts een aantal voorbeelden te geven, en is niet uitputtend. Bij de beantwoording van de deelvragen 1 en 2 zullen meer aspecten aan de orde komen. Bij betrouwbaarheid in het personenvervoer gaat het om de mate van spreiding rondom de verwachte reistijd en de effecten daarvan voor de reiziger. De nadruk ligt op vertraging (te laat aankomen op de bestemming), maar ook aan te vroeg komen kunnen kosten verbonden zijn (RAND Europe, 2001). De reistijd die in het OEI-kader wordt gebruikt en de bijbehorende reistijdwaardering, betreffen de verwachte reistijd (inclusief eventueel verwacht oponthoud), net als in het onderliggende reistijdwaarderingsonderzoek (HCG, 1998). Betrouwbaarheid gaat dus om het zoveel mogelijk vermijden van onverwachte afwijkingen in termen van reistijd. Onverwachte vertragingen leiden tot extra wachttijd (met een hoger disnut), tot stress bij de reizigers, tot het hanteren van veiligheidsmarges en tot afwijkingen van de geprefereerde aankomsttijden (‘scheduling cost’, gebaseerd op de vertrektijdkeuzetheorie ontwikkeld door met name Vickrey, 1969 en Small, 1982, zie HCG, 2000 voor een overzicht van vertrektijdkeuzemodellen). In het goederenvervoer worden kwaliteitsaspecten genoemd: • • • • • •
in
het
projectplan
als
voorbeelden
van
Overslag- en wachttijden in de transportketen Transportfrequentie Aantal (bereikbare) bestemmingen Informatie over de zending tijdens het transport: tracking and tracing Kans op beschadiging (inclusief bederf) of diefstal tijdens het transport Betrouwbaarheid (maar deze wordt apart bestudeerd in dit project).
Het eerstgenoemde aspect bevat componenten van de totale reistijd van deur tot deur. Frequentie heeft hier weer invloed op, maar ook op de partijgrootte. Net als bij het personenvervoer, geldt hier dat het onderzoek waar de doorgaans toegepaste reistijdwaarderingen uit komen (HCG, 1992a) de verwachte reistijd van de deur tot deur betreft. Deze waardering is in principe dus reeds geldig voor vermindering van overslag- en wachttijden. Ook hier zal in de voorgestelde studie gekeken worden naar empirisch materiaal voor een verschillende behandeling van reistijd in het voertuig en overslag- en wachttijden. De overige kwaliteitsaspecten in het goederenvervoer betreffen de omstandigheden tijdens het transport (beschadiging, informatie) of het aantal bestemmingen. Onbetrouwbaarheid in het goederenvervoer kan leiden tot gemiste aansluitingen en wachttijden, maar tevens tot een suboptimaal gebruik van transportpersoneel -en materieel en tot gemiste kansen op het gebied van voorraadbeheer en distributiesystemen (HCG, 1997). Gezien de bredere toepassing van tijdkritische logistieke systemen (b.v. Just-in-Time), is het belang van aflevering op tijd naar verwachting steeds belangrijker geworden.
Pagina 4
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
1.3 Inhoud van dit rapport In hoofdstuk 2 van dit rapport wordt de gevolgde onderzoeksaanpak beschreven. Hoofdstuk 3 betreft uitkomsten voor wat betreft de definities en bestudeerde aspecten uit de literatuur over betrouwbaarheid in personen- en goederenvervoer. De definities inzake de overige kwaliteitsaspecten die in de diverse artikelen en rapporten zijn bestudeerd, komen aan de orde in hoofdstuk 4. Het vijfde hoofdstuk behandelt de gevonden kengetallen en waarderingsmethoden en hun toepasbaarheid in het kader van OEI voor betrouwbaarheid (personen- en goederenvervoer), en het zesde hoofdstuk doet het hetzelfde voor de andere kwaliteitsaspecten. In deze twee hoofdstukken komen ook de lessen voor het opnemen van betrouwbaarheid en kwaliteit in verkeers- en vervoersmodellen aan de orde. Hoofdstuk 7 beschrijft de witte vlekken waarover geen (recente) literatuur is gevonden. Tenslotte volgen de conclusies in hoofdstuk 8. Beschrijvingen (in het Engels) van de bestudeerde artikelen en rapporten zijn opgenomen in Bijlage 1.
Pagina 5
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 6
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
2. De gevolgde methodiek en fasering
RAND Europe heeft deze studie middels de volgende stappen uitgevoerd. I. Opstellen van een lijst van literatuur die in het onderzoek zal worden besproken Het gaat in dit onderzoek zowel om de internationale als de Nederlandse literatuur op het gebied van kwaliteit en betrouwbaarheid van het vervoer. Deze literatuurstudie richt zich niet specifiek op reistijdwaardering. Hierover bestaat een schat aan literatuur. In deze studie beperken we ons tot kernpublicaties cq. overzichtsstudies, waarbij met name als er iets wordt gezegd over de onderlinge verhoudingen van de componenten van reistijd dit wordt meegenomen. Er zijn veel studies over betrouwbaarheid van infrastructuur met een technischverkeerskundige invalshoek. Alhoewel er raakvlakken zijn met deze studie, is in overleg met AVV besloten dat deze niet een centraal onderwerp zijn van dit onderzoek. Dit onderzoek richt zich meer op de waardering van de gebruikers. RAND Europe heeft op de volgende wijze naar relevante literatuur gezocht: •
Bestuderen van nationale en internationale tijdschriften op het gebied van transport: o Transportation Research A (Policy and Practice), B (Methodological), C (Emerging Technologies), D (Transport and Environment), E (Transportation and Logistics Review) en F (Traffic Psychology and Behaviour) o Journal of Transport Economics and Policy o Transportation o Transportation Policy o Transport Reviews o Traffic Engineering and Control o Journal of Transport Geography o Journal of Transportation and Statistics o Tijdschrift Vervoerswetenschap o Verkeerskunde o Economisch Statistische Berichten.
•
Bestuderen van bundels en CD-ROMS van conferenties op het gebied van verkeer en vervoer: o European Transport Conference (jaarlijks) o Annual Meeting of the Transportation Research Board (jaarlijks) o World Conference on Transport Research (om de drie jaar) Pagina 7
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
o International Association of Travel Behaviour Research (om de drie jaar). o Colloquium Vervoersplanologisch Speurwerk, OV-colloquium, Transportlogistieke werkdagen (jaarlijks). •
Bestuderen van rapporten (voorzover publiek beschikbaar) van/voor AVV, het Projectbureau Integrale Verkeers en Vervoersstudies, Connekt, De Nederlandse Spoorwegen, het stads- en streekvervoer, ministeries in het Verenigd Koninkrijk, de Verenigde Staten, Frankrijk, Duitsland en Scandinavië.
•
Raadplegen van RAND Europe’s internationale netwerk.
•
Zoekacties op het worldwide web.
In de selectie van de literatuur is de nadruk gelegd op recente artikelen en rapporten (de laatste vijf jaar, indien nodig de laatste tien jaar). Een concept literatuurlijst is aan AVV aangeboden en door AVV aangevuld. II. Beschrijving van de relevante literatuur en beoordeling op toepasbaarheid binnen het OEI-kader in Nederland De uiteindelijke literatuurlijst bestond uit 72 artikelen en rapporten. Hierbinnen zitten enkele doublures, omdat soms over dezelfde studie meerdere artikelen (bijvoorbeeld in het Nederlands en Engels, of een korte en een lange versie) zijn geschreven. Na verwijdering van de doublures is een prioriteitenvolgorde aangebracht, omdat het tijdsbudget van het project het niet toestond om alle artikelen en rapporten te bestuderen en samen te vatten. Uiteindelijk zijn in de tweede fase 39 artikelen en rapporten bestudeerd en afzonderlijk samengevat en geduid. Hierbij zijn voor ieder artikel/rapport de volgende aspecten aan de orde komen: • • • •
•
•
Welke definities van kwaliteit en betrouwbaarheid worden gehanteerd? Welke aspecten van kwaliteit en betrouwbaarheid worden bestudeerd? Welke methoden voor het waarderen van aspecten van kwaliteit en betrouwbaarheid worden gebruikt of voorgesteld? Worden er kengetallen aanbevolen voor de waardering van aspecten van kwaliteit en betrouwbaarheid? Zo ja, van welke aard (absoluut, procentueel, enz.), welke concrete waarden en onder welke veronderstellingen en voorwaarden en voor welke situaties? Indien er kengetallen worden aanbevolen: zijn deze ook toe te passen in Nederland in een KBA volgens de OEI systematiek (waarin reistijdwinsten al worden gemonetariseerd: kengetallen voor kwaliteit en betrouwbaarheid moeten geen aspecten weergeven die al meegenomen worden in de waardering van reistijdwinst). Zijn er richtlijnen af te leiden voor het opnemen van aspecten van kwaliteit en betrouwbaarheid in verkeers- en vervoermodellen?
Pagina 8
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
•
Volgen er aanbevelingen uit het artikel/rapport voor concreet empirisch onderzoek naar de waardering van aspecten van kwaliteit en betrouwbaarheid in de Nederlandse situatie voor gebruik in het OEI kader?
Fase 2 is afgerond met het aanbieden van de afzonderlijke beschrijvingen van relevante artikelen en rapporten (in het Engels; zie Bijlage 1) aan het projectteam van AVV, waarin wordt ingegaan op de hierboven genoemde punten. Deze beschrijvingen zijn met het projectteam van de opdrachtgever besproken. De integratie en synthetisering van de afzonderlijke beschrijvingen van de artikelen en rapporten volgde in fase 3. III. Opstellen concept eindrapportage: integratie en synthese Het materiaal uit fase II vormt de basis voor het eindrapport. In fase III is een overkoepelend gedeelte toegevoegd. Hierin is het geheel van de bestudeerde literatuur in de beschouwing betrokken. In deze integratie en synthese is ingegaan op de volgende vragen: • • • •
•
•
•
Welke definities van kwaliteit en betrouwbaarheid komen in de literatuur voor? Is er een categorisatie mogelijk? Kan aangegeven worden of sommige definities beter zijn dan andere, c.q. beter toepasbaar in het Nederlandse OEI raamwerk? Welke aspecten van betrouwbaarheid worden in de literatuur behandeld? Zijn er ook witte of nagenoeg witte vlekken? Welke methoden voor het waarderen van kwaliteit en betrouwbaarheid worden in de literatuur gehanteerd? Zijn deze methoden toegepast of toepasbaar en Nederland en passen de uitkomsten dan in het OEI kader? Welke kengetallen voor de waardering van kwaliteit en betrouwbaarheid biedt de literatuur? Is hier een categorisering mogelijk? Welke concrete uitkomsten zijn mogelijk relevant voor toepassing in het Nederlandse OEI kader (overdraagbaarheid)? In hoeverre worden de aspecten van kwaliteit en betrouwbaarheid reeds verdisconteerd in andere aspecten die meegenomen worden in de KBA’s volgens de OEI systematiek, zoals met name de reistijdwinsten? Weten we zeker of er geen dubbelteling optreedt. Wat zijn de consequenties van het bovenstaande voor de verkeers– en vervoersmodellen die invoer leveren voor de directe effecten in de KBA? Kan volstaan worden met wat momenteel als invoer in de KBA’s gebruikt wordt? Dienen additionele uitkomsten te worden gebruikt die de huidige modellen wel kunnen leveren? Of zijn er uitbreidingen nodig van de verkeers- en vervoermodellen? Voor de witte vlekken voor wat betreft de waardering van aspecten van kwaliteit en van betrouwbaarheid in het OEI-kader: zijn er in de literatuur aanknopingspunten te vinden voor onderzoeksmethoden (zowel wijze van dataverzameling als analyse) die in vervolgonderzoek in Nederland kunnen worden toegepast om tot waarderingen te komen? Het betreft hier aspecten waarvoor in de literatuur geen concrete kengetallen zijn te vinden en aspecten waarvoor de overdraagbaarheid van de kengetallen naar de Nederlandse situatie uitgesloten moet worden of ter discussie staat.
Pagina 9
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 10
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
3. Betrouwbaarheid: definities en bestudeerde aspecten
In Tabel 1 staat een overzicht van de definities van (on)betrouwbaarheid in de literatuur, voorzover het personenverkeer- en vervoer betreft. Tabel 2 bevat de definities die voor (on)betrouwbaarheid in het goederenvervoer worden gehanteerd. Beide tabellen bieden tevens een overzicht van de aspecten van (on)betrouwbaarheid die in de artikelen en rapporten behandeld worden. De definities van betrouwbaarheid of onbetrouwbaarheid in de literatuur kunnen in drie categorieën worden ingedeeld: • • •
Brede definities van (on)betrouwbaarheid; Toegespitste definities, met name van (on)betrouwbaarheid van de reistijd; Operationele invullingen voor het meten en/of waarderen van (on)betrouwbaarheid.
Bij de brede definities van betrouwbaarheid (b.v. Berdica, 2002; Hilbers en Blijie, 2003, Peeters et al., 1998, projectorganisatie Benutten en Bouwen, 2002; Rietveld, 2003) gaat het erom dat voldaan wordt aan de van te voren gemaakte afspraken of de verwachtingen. Volgens een dergelijke ruime definitie, betreft betrouwbaarheid ook de prijs, het comfort en de veiligheid. Voor gebruik binnen het kader van OEI is zo’n ruime definitie niet aan te bevelen, omdat deze elementen reeds afzonderlijk in de evaluatie van een project worden meegenomen (of zouden kunnen worden meegenomen in het geval van comfort). Een brede definitie van betrouwbaarheid zou hier tot dubbeltellingen leiden. In dit onderzoek gaat het bij betrouwbaarheid om betrouwbaarheid van de reistijden. Definities die hierop zijn toegespitst zijn daarom van groter belang dan meer algemene definities van betrouwbaarheid in het transport. De toegespitste definities van (on)betrouwbaarheid in de literatuur hebben de volgende elementen: • • •
Het gaat om de afwijking van de werkelijke reistijd van de verwachte reistijd (dit is niet noodzakelijkerwijs de ongehinderde of ‘free flow’ reistijd; reizigers kunnen uitgaan van regelmatig terugkerende congestie); Zowel de frequentie van deze afwijking als de omvang ervan zijn van belang; De consequenties van deze afwijkingen van de verwachting bestaan uit vroeger of later dan verwacht aankomen (soms ook vroeger dan gemiddeld nodig vertrekken om maar op tijd te komen) en mogelijk ook de stress van de onzekerheid op zich;
Pagina 11
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Tabel 1. Betrouwbaarheid in personenverkeer: definities en behandelde aspecten Studie Accent en HCG, 1995 AVV, 2003
Bates et al, 2001
Bell, 2002
Berdica, 2002
Bonsall, 2003 Brownstone en Small, 2002
Definitie Behandelde aspecten Onbetrouwbaarheid is de kans op een Waardering door onverwachte vertraging. automobilisten van onbetrouwbaarheid van de reistijd. In de Klantenbarometer Openbaar Vervoer is betrouwbaarheid ingevuld als het op tijd rijden van de bus (tram, metro). Onbetrouwbaarheid is de Bespreking van mogelijkheid van te laat op de onderzoeksmethoden van bestemming aankomen, langer dan onbetrouwbaarheid van de gewenst of verwacht bepaalde reistijd. activiteiten uitvoeren, en de stress van de onzekerheid op zich. Een netwerk is betrouwbaar als de Methoden voor het meten en verwachte reiskosten acceptabel zijn, analyseren van zelf als de gebruikers uiterst betrouwbaarheid van een pessimistisch zijn over de staat netwerk. waarin het netwerk verkeert. Betrouwbaarheid is de kans dat een Betrouwbaarheid in transport: instrument functioneert zoals bedoeld 1) van de verbinding: kans om voor de geplande tijdsduur en onder de bestemming te bereiken; 2) de heersende werkomstandigheden. van de reistijd: kans om binnen een bepaalde tijd de bestemming te bereiken; 3) van de capaciteit: kans dat het netwerk de vraag aankan. Indicatoren voor het meten van onbetrouwbaarheid en aanverwante begrippen. Hoe gaan mensen om met onzekerheid? consequenties voor modellen. Onbetrouwbaarheid is de variatie in Waardering door de reistijd van dag tot dag en wordt automobilisten van ingevuld als het 90ste min 50ste onbetrouwbaarheid van de percentiel van de reistijdverdeling, of reistijd. als het 80ste min 50ste percentiel van de reistijdverdeling. Dit is ongeveer de kans om substantieel later te zijn dan verwacht.
Pagina 12
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie Bruinsma et al., 1999
Definitie Betrouwbaarheid is brede zin is de kans om veilig aan te komen. Meer specifiek is de betrouwbaarheid van reis- en aankomsttijden. Onbetrouwbaarheid hangt samen met de mate van variatie in de waargenomen of gepercipieerde reistijdverdeling. Copley et al, Onbetrouwbaarheid van de reistijd 2002 heeft twee dimensies: frequentie van vertragingen en de omvang van de vertragingen. Hilbers en Blijie, Betrouwbaarheid van een 2003 verplaatsing is de kans dat een verplaatsing verloopt volgens de van te voren verwachte karakteristieken voor wat betreft reistijd, reiskosten, reiscomfort en veiligheid. De Jong et al., Schedule delay: aantal minuten dat 2003 men later aankomt (of vertrekt) dan gewenst; aantal minuten dat men eerder aankomt dan gewenst. Lam en Small, Onbetrouwbaarheid is de variatie in 2001 de reistijd van dag tot dag en wordt ingevuld als het 90ste min 50ste percentiel van de reistijdverdeling. MVA, 2000 Onbetrouwbaarheid wordt in verband gebracht met onregelmatigheid van de reistijd en gemeten als de standaardafwijking van reistijd. Noland en Polak, Onbetrouwbaarheid komt overeen 2002 met de variatie van de reistijden van dag tot dag. Praktische invullingen: standaardfout van de reistijd en schedule delay. Peeters et al., Betrouwbaarheid in een 1998 transportsysteem is de kans om de bestemming te bereiken, gegeven bepaalde kenmerken.
Behandelde aspecten Methoden om de verdeling van de vertragingen in het openbaar vervoer te schatten.
Waardering door automobilisten van onbetrouwbaarheid van de reistijd. Objectieve versus subjectieve betrouwbaarheid. Betrouwbaarheid hangt af van variatie en voorspelbaarheid. Diverse maatstaven voor betrouwbaarheid Waardering van schedule delay versus reistijd voor autobestuurders en treinreizigers Waardering door automobilisten van onbetrouwbaarheid van de reistijd. Waardering door busreizigers van onbetrouwbaarheid van de reistijd. Bespreking van onderzoeksmethoden van onbetrouwbaarheid van de reistijd Betrouwbaarheid van reistijd, reiskosten en comfort. Betrouwbaarheid van de reistijd is de variatie in de reistijd zelf en in vertrek- en aankomsttijd. Objectieve en subjectieve betrouwbaarheid. Meten van betrouwbaarheid in het openbaar vervoer.
Pagina 13
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie Projectorganisatie Benutten en Bouwen, 2002
Rietveld, 2003
Rietveld et al., 2001 SACTRA, 1999
Senna, 1991
•
Definitie Betrouwbaarheid van spoorvervoer is de kans dat de produkten en diensten geleverd worden volgens de overeengekomen kwaliteit. Praktische invulling voor personenvervoer: niet meer dan 3% van het totale aantal treinen komt te laat aan. Onbetrouwbaarheid betekent dat de reiziger de exacte uitkomsten voor reistijd, reiskosten en comfort niet van te voren kent. Onbetrouwbaarheid wordt ingevuld als de kans op een vertraging van een bepaalde omvang (bijvoorbeeld: 15 min, 2 min). Onbetrouwbaarheid wordt geïnterpreteerd als de variabiliteit van reistijd, met consequenties voor de aankomsttijden. Onbetrouwbaarheid van reistijd is de onzekerheid of de werkelijke reistijd overeen zal komen met de verwachting. Ingevuld als de standaardafwijking van reistijd.
Behandelde aspecten Methode voor het bepalen van betrouwbaarheid in het railvervoer.
Methoden voor het bestuderen van de effecten van onzekerheid op reisgedrag. Waardering door openbaar vervoerreizigers van onbetrouwbaarheid van de reistijd. Onbetrouwbaarheid door ongelukken versus andere oorzaken. Methoden voor de waardering van de onbetrouwbaarheid van de reistijd
Soms worden alleen de langer dan verwachte reistijden bestudeerd: de vertragingen. Korter dan verwachte reistijden leveren een reistijdwinst op, maar mogelijk ook ‘scheduling cost’. Te laat komen wordt doorgaans wel als hinderlijker beschouwd dan te vroeg.
In de operationele definities van onbetrouwbaarheid komen we drie varianten tegen: • De variantie of standaardafwijking van de reistijdverdeling; • De afstand tussen het 80ste of 90ste percentiel en de mediaan van de reistijdverdeling; • Het verschil tussen het feitelijke en de gewenste aankomst- of vertrektijdstip (scheduling delay). De derde variant wijkt af van de beide eerste doordat er niet zozeer naar de duur van de reis wordt gekeken, maar naar de gevolgen van langere of kortere reistijden voor de kloktijd, de dagindeling. Dit is momenteel moeilijk in te passen in de OEI-praktijk (zie ook hoofdstuk 5). Er wordt momenteel in evaluaties geen informatie gebruikt over het reistijdstip en veranderingen daarin. Als er op een gegeven moment evaluaties plaats zouden vinden met netto baten voor de reizigers op basis van de verandering in de logsom uit het Landelijk Model Systeem (LMS), waarin ook de verandering van vertrektijdstip van de reiziger tot uitdrukking komt, dan zou de derde variant gelden. Een voordeel van het meenemen van betrouwbaarheid op deze wijze is de aansluiting bij de scheduling theorie (b.v. Small, 1982). Het effect van onbetrouwbaarheid wordt Pagina 14
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
op deze wijze echter niet geheel meegenomen, omdat het huidige en het nieuwe vertrektijdstip-keuzemodel (De Jong et al, 2003) uitgaan van de verwachte congestie.3 Ook het effect van stress door onbetrouwbaarheid zit niet in deze modellen. Ook de standaardafwijking (c.q. variantie) of de afstanden tussen percentielen van de reistijdverdeling zijn niet eenvoudig in te bouwen in de huidige OEI-systematiek. Waarderingen voor onbetrouwbaarheid in deze dimensies bestaan voor sommige specifieke situaties (zie hieronder) en zouden ook voor Nederland in gericht onderzoek kunnen worden bepaald. Nodig is tevens dat er bepaald wordt wat de reistijdverdeling is in de toekomst zonder en met het te evalueren project. Beide maatstaven kunnen vervolgens berekend worden uit die reistijdverdeling. Tabel 2. Betrouwbaarheid in goederenvervoer: definities en behandelde aspecten Studie Accent en HCG, 1995 Bruzelius, 2001
Fowkes et al, 2001
HCG, 1992a,b; De Jong et al., 2001; RAND Europe, 2003 HCG, 1997
Definitie Onbetrouwbaarheid ingevuld als de kans op een vertraging van 30 minuten of meer. Onbetrouwbaarheid ingevuld als de frequentie van vertragingen of als schedule delay. Onbetrouwbaarheid ingevuld als de tijd waarop 98% van de leveringen aankomt min de vroegste aankomsttijd (‘spreiding’); Tevens als afwijkingen van de vertrektijd (schedule delay). Onbetrouwbaarheid ingevuld als het percentage niet op tijd geleverd (afgesproken tijdstip of tijdvenster).
Behandelde aspecten Waardering van onbetrouwbaarheid in het wegvervoer. Waardering van onbetrouwbaarheid in het wegvervoer, het spoorvervoer en de luchtvaart. Waardering van onbetrouwbaarheid in het wegvervoer
Waardering van onbetrouwbaarheid in wegvervoer, spoorvervoer, binnenvaart en zeevaart. Belemmeringen in goederen wegvervoer: vertragingen door congestie en grensformaliteiten. Projectorganisatie Betrouwbaarheid van spoorvervoer is Methode voor het bepalen van Benutten en de kans dat de produkten en diensten betrouwbaarheid in het bouwen, 2002 geleverd worden volgens de railvervoer. overeengekomen kwaliteit. Voor goederenvervoer hangt dit af van de afspraken met de klant.
3
Het is wel mogelijk om onverwachte congestie op te nemen in het theoretische scheduling model (Bates et al., 2001). Pagina 15
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Voor de definities in het goederenvervoer (Tabel 2) geldt ook dat onbetrouwbaarheid breed en toegespitst op reistijd gedefinieerd kan worden, alsmede operationeel kan worden ingevuld. De studies in de literatuur richten zich vooral op het laatste. De invullingen die in de praktijk voorkomen voor onbetrouwbaarheid in het goederenvervoer zijn: • •
de frequentie van of kans op vertraging (eventueel van een bepaalde minimale omvang); de schedule delay: afwijking in termen van kloktijd van de afgesproken/gewenste tijd.
In principe zou ook hier met de standaardafwijking of percentielen van de reistijdverdeling gewerkt kunnen worden, maar deze maatstaven zijn niet aangetroffen. Voor de kans op vertraging bestaan waarderingen uit diverse landen (zie hoofdstuk 5), maar toepassing in een KBA volgens OEI vereist bovendien dat de verandering in de kans op vertraging als gevolg van het te bestuderen project wordt berekend. Dit is geen eenvoudig te leveren invoer, maar de verandering in de schedule delay in het goederenvervoer door het project is waarschijnlijk nog lastiger te leveren (zie ook hoofdstuk 5).
Pagina 16
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
4. Kwaliteit: definities en bestudeerde aspecten
De bestaande definities van kwaliteit in het algemeen, zoals in Tabel 3 (Friman en Gärling, 2000; Johannson, 1998; Probert 2001) zijn zeer algemeen. Ook de prijs van de reis, de reistijd en de betrouwbaarheid van de reistijd kunnen er dan onder vallen. Voor dit onderzoek willen we een beperktere definitie hanteren. Reistijd en reiskosten zijn al separate posten in de KBA, en dienen om dubbeltellingen te vermijden niet onder kwaliteit te worden gerangschikt. De betrouwbaarheid van de reistijd valt conceptueel wel onder de definitie van kwaliteit (ook bij een beperktere definitie), maar als het gaat over kengetallen, dan dienen die voor betrouwbaarheid van de reistijd afzonderlijk te worden bezien. Daarom spreken we in dit onderzoek over ‘betrouwbaarheid’ en ‘overige aspecten van kwaliteit’. Tot het laatste behoort de vraag of verschillende componenten van de reistijd (looptijd, wachttijd, overstaptijd) anders gewaardeerd moeten worden dan tijd in het voertuig. Hier dient gewaakt te worden voor dubbeltellingen, aangezien de verkeersmodellen die invoer leveren voor een KBA volgens de OEI-systematiek doorgaans de totale reistijd (van deur tot deur) per hoofdvervoerwijze betreffen. Tabel 3 geeft ook aan welke aspecten van kwaliteit in de diverse artikelen en rapporten over kwaliteit in het personenverkeer- en vervoer bestudeerd worden. Tabel 4 doet hetzelfde voor goederenvervoer. Tabel 3. Overige aspecten van kwaliteit in personenverkeer: definitie en behandelde aspecten Studie AVV, 2003
Definitie
Friman en Kwaliteit van een aan een klant Gärling, 2000 geleverde dienst is de satisfactie of dissatisfactie met de geleverde dienst.
Gunn, 2001
Behandelde aspecten Faciliteiten op de halte, informatie en veiligheid, prijs, rijcomfort, tijd en doorstroming. Niet alleen zaken als comfort en informatievoorziening worden meegenomen, maar ook prijs, reistijd en betrouwbaarheid. Van groot belang voor de waardering van kwaliteit (in het openbaar vervoer) zijn de negatieve kritieke incidenten. Belang van looptijd en dienstregelingsinterval ten opzichte van tijd in het voertuig.
Pagina 17
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie Johannson, 1998
De Jong et al, 2003 Morface Int en Cambridge Systematics, 1999
MVA, 2000 Probert, 2001
Rietveld et al, 2001 Transek, 2002 Wardman, 2001a
Definitie De kwaliteit van het openbaar vervoer is goed als de perceptie van de klanten van de geleverde produkten en diensten minimaal gelijk is aan hun verwachtingen.
Behandelde aspecten Comfort in het voertuig (zitplaats, drukte, acceleratie, schokken en trillingen, bagageruimte, geluid, temperatuur, verlichting, omgeving, informatie, personeel, veiligheid, dienstverlening), bij in- en uitstappen, op station/halte (wachttijd, omgeving, informatie). Frequentie en zitplaatskans in de trein Kwaliteit van de dienstverlening in het openbaar vervoer omvat drukte, betrouwbaarheid, prijs, informatie, frequentie, schoonmaken, temperatuur, geur, privacy, weinig trillingen, zitplaats, vriendelijkheid, betaalgemak, tariefstructuur, haltecomfort. Waardering voor staan in de bus, bij volle en niet-volle bus. Kwaliteit van transportdiensten is de Vervoerwijze, comfort, tijd, mate van overeenstemming met de eisen veiligheid, beheersing, genot, van de klant of uiteindelijke gebruiker. status, kosten, gezondheid en gemak. Waardering voor zitplaatskans.
Wardman, 2001b
Waardering voor staan in bus, metro en forenzentrein. Belang van looptijd, voortransporttijd, wachttijd, dienstregelingsinterval en overstap ten opzichte van tijd in het voertuig. Belang van looptijd, wachttijd en overstaptijd ten opzichte van tijd in het voertuig.
Samenvattend kan geconcludeerd worden dat de volgende aspecten van kwaliteit (anders dan betrouwbaarheid) in de literatuur worden behandeld: • •
Faciliteiten op de halte Informatie Pagina 18
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
• • • • • • •
Veiligheid Comfort (voertuig, in- en uitstappen, station/halte) Looptijd Dienstregelingsinterval, frequentie, wachttijd, overstaptijd Drukte Zitplaats Tariefstructuur en betaalgemak.
Reistijd en prijs worden in dit rapport niet gerangschikt onder de kwaliteitsaspecten. Tabel 4. Overige aspecten van kwaliteit in goederenvervoer: definitie en behandelde aspecten. Studie HCG, 1992a HCG, 1992b De Jong et al, 2001 RAND Europe, 2003
Definitie
Behandelde aspecten Schadekans, frequentie. Schadekans, flexibiliteit Beschikbaarheid van logistieke services, flexibiliteit. Schadekans, frequentie.
Pagina 19
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 20
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
5. Betrouwbaarheid: kengetallen
In Tabel 5 staat een overzicht van de kengetallen (steeds met definitie) die in de literatuur gevonden zijn voor betrouwbaarheid. Ook de hierbij gehanteerde methode wordt benoemd en andere lessen, met name voor het opnemen van betrouwbaarheid in transportmodellen, worden beschreven. Tabel 5. Betrouwbaarheid in personenverkeer: kengetallen, gebruikte methode en andere lessen. Studie Accent en HCG, 1995
AVV, 2003
Kengetallen (+definitie) Verdubbeling van de kans op vertraging is equivalent aan 13 min. reistijd (woonwerkverkeer) of 20 min. (zakelijk, overig); Halvering van de kans op vertraging: 3 minuten (woon-werk) of 5 min. (zakelijk, overig). Betrouwbaarheid (op tijd rijden) is het belangrijkste aspect (belang 3,58 op schaal 1-5) voor bus,tram en metro en de geboden betrouwbaarheid krijgt een laag rapportcijfer (5,94 op schaal 1-10).
Methode Stated preference (SP) in het wegverkeer in het VK, met als attributen: reistijd, informatievoorziening en kans op vertraging.
SP onder 3387 gebruikers van bus, tram en metro in Nederland.
Bates et al, 2001
Andere lessen Voor sommige segmenten (b.v. zakelijk, reistijdwinsten) is de reistijdwaardering per minuut hoger in situaties met congestie dan in situaties zonder congestie. .
Achterliggende theorie en aanbevelingen voor onderzoeksmethoden in de praktijk: scheduling model, SP data. Achterliggende theorie: speltheorie.
Bell, 2002
Pagina 21
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie Bonsall, 2003
Brownstone en Small, 2002 Brownstone en Small, 2002
Bruinsma et al., 1999 Copley et al, 2002
Copley et al, 2002
HCG, 1998
Kengetallen (+definitie)
Methode
Andere lessen Achterliggende theorie: prospect theorie; Methoden: gemiddelde versus variantie aanpak, gerestricteerde keuzeverzameling. ste Waarde voor 90 min RP: reistijdmetingen Methode: kan ook 50ste percentiel van de in onderzoek naar met RP data (in reistijdverdeling: 12-15 State Route 91 in speciale gevallen), $/uur (mannen) en 30-32 Californië, met gebruik van $/uur (vrouwen). variabele tolheffing. percentielen. Waarde voor 80ste min RP (zie boven) en SP. Reistijd neemt 2/3 ste 50 percentiel van de van het reistijdverdeling: 28 kwaliteitsverschil $/uur. tussen tolroute en alternatieve route voor zijn rekening; betrouwbaarheid 1/3. Methode om de vertragingen in het openbaar vervoer te schatten. De waarde van de SP onder 167 standaardafwijking van autobestuurders in de reistijd is 1,3 maal die woon-werkverkeer in van de reistijd (beide per Manchester; minuut). gemiddelde versus variantie methode. 1 minuut te laat of te SP (zie boven); Methode voor vroeg is minder erg dan 1 scheduling model. waardering: minuut reistijd. gemiddelde versus variantie methode of scheduling model. Methode: SP; De reistijdwaardering kan variëren per wegtype en gemiddelde snelheid (afhankelijk van de mate van congestie op de weg).
Pagina 22
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie De Jong et al., 2003
MVA, 2000
Kengetallen (+definitie) Woon-werk, zakelijk, overig: 1 minuut te laat of te vroeg is 1-1,5 keer zo erg als 1 minuut reistijd Onderwijs: 1 minuut te vroeg of te laat in minder erg dan 1 minuut reistijd Alle motieven: 1 minuut langere of kortere activiteit op de bestemming is minder erg dan 1 minuut reistijd. De waarde van de standaardafwijking van tijd in de bus is 24% van de waarde van de reistijd in de bus (met zitplaats; minder als reistijd staand gebruikt; beide in minuten). De waarde van de standaardafwijking van wachttijd is 48% van die van wachttijd.
Methode Andere lessen SP onder ongeveer 1000 autobestuurders en treinreizigers in de spitsperioden in Nederland; Scheduling model.
SP onder 309 busreizigers in Frankrijk; Gemiddelde versus variantie methode.
Noland en Polak, 2002
Rietveld et al, 2001
Methoden: scheduling model of gemiddelde versus variantie aanpak (beide in combinatie met SP) Een vermindering van de kans op 15 min. Vertraging van 50% naar 0% is waard f 4,81 (30% van de waarde van een uur reistijd). Een vermindering van de kans op een vertraging van 2 minuten van 50% naar 2% is f 0,64 waard (dus 1 minuut vertraging is 2,4 keer zo erg als 1 minuut reistijd: risico-aversiteit).
SP onder 781 openbaar vervoer reizigers in Nederland, met als attributen: reistijd, kans op vertraging en kans op een zitplaats.
Rietveld, 2003
Achterliggende theorie over onzekerheid en risicohouding
Pagina 23
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie SACTRA, 1999
Kengetallen (+definitie) Het ontbreken van de variabiliteit van reistijd leidt tot een onderschatting van de economische baten van 550%(VK). Senna, 1991 Het disnut van de standaardafwijking van reistijd is ongeveer 2,5 keer zo hoog als voor reistijd.
Methode
Andere lessen
SP onderzoek onder 301 respondenten in Porto Alegre (Brazilië), met als attributen verschillende reistijden en kosten.
Achterliggende theorie: nutstheorie en risicohouding; Methode: SP en gemiddelde versus variantie aanpak.
Wardman, 2001a
Tijd in omstandigheden met congestie is 1,5 keer zo erg als tijd in de auto.
Uit het overzicht van kengetallen en methoden voor de waardering van betrouwbaarheid in het personenverkeer in Tabel 5 blijkt dat er drie verschillende methoden gebruikt worden: 1. De gemiddelde versus variantie methode. Hierbij worden zowel de gemiddelde reistijd als de variantie in de reistijd (of de standaardafwijking) in de nutsfunctie opgenomen en de parameters voor beide variabelen geschat. Meestal worden hierbij SP data gebruikt, uit interviews waarbij de respondenten zowel de gemiddelde reistijd als een bepaalde weergave van de variantie kregen te zien, en waarbij deze kenmerken onafhankelijk (of bijna onafhankelijk: geen of zeer geringe correlatie) varieerden. Dit betekent dat er een de waarderingskant geen dubbeltelling op zal treden, de waarde voor reistijd en die voor betrouwbaarheid kunnen worden opgeteld. Voor toepassing in een KBA is het wel nodig dat de omvang van de verandering in reistijd en die in betrouwbaarheid beide afzonderlijk worden vastgesteld. Hierbij dient betrouwbaarheid gemeten te worden als de standaardafwijking van de reistijd rond de verwachte reistijd. Senna (1992) heeft een model met gemiddelde en variantie van reistijd expliciet uit de nutstheorie afgeleid, en kwam uit op een niet-lineaire nutsfunctie (dit in tegenstelling tot het gebruik van lineaire nutsfuncties in de praktijk). 2. Een verwante methode is die waarbij met percentielen van de reistijdverdeling wordt gewerkt. Het 80ste of 90ste percentiel min de mediaan (het 50ste) is gebruikt voor de meting en waardering van onbetrouwbaarheid. Hiertoe dienen modellen geschat te worden op RP, SP of gecombineerde RP/SP data, waarbij de genoemde onbetrouwbaarheidsmaatstaf en reistijd afzonderlijke variabelen zijn. Ook deze waarderingen kunnen worden opgeteld.
Pagina 24
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
3. Tenslotte zijn er de scheduling modellen waarbij gegevens (meestal SP) over tijdstipkeuze gebruikt worden om te bepalen hoe erg in geld of reistijd een minuut te vroeg of te laat komen (of vertrekken) is. Betrouwbaarheid van reistijd wordt hier dus bestudeerd middels de consequenties op de dagindeling (scheduling). Combinatie van de waarderingen voor te vroeg of te laat komen met de OEI-systematiek is niet goed mogelijk, omdat in de KBA de vertaling van reistijd naar tijdstip van de dag (en afwijking van gewenste aankomsttijdstip) niet gemaakt wordt. Dit is pas mogelijk als in de KBA ook naar de tijdstipkeuze zou worden gekeken (in plaats van alleen naar bestaande reizigers met een bepaalde vervoerwijze en verandering van vervoerwijze) en als daarbij ook de gewenste aankomsttijdstippen bekend zouden zijn. Voor alle drie de methoden is het gebruik van SP data een logische keuze (dezelfde aanbeveling komt ook voor in Bates et al, 2001; Noland en Polak, 2002). Schatting van een model inclusief een variabele voor betrouwbaarheid op RP data is zelden mogelijk. In RP data (bijvoorbeeld voor verschillende dagdelen en dagen) zullen namelijk betrouwbaarheid, reiskosten en reistijd vaak sterk gecorreleerd zijn, wat schatting van afzonderlijke parameters zeer bemoeilijkt. Bovendien moet men voor deze variabelen waarden hebben voor de niet gekozen alternatieven (b.v. uit toedelingen). In SP onderzoeken heeft de onderzoeker controle over deze attributen en kan deze als nietgecorreleerd of licht gecorreleerd specificeren. De respondenten krijgen de attribuutniveau’s voor alle keuze-alternatieven gepresenteerd. Een uitzondering, waar RP gegevens wel een goede basis voor modellen vormen, is de keuze tussen een betrouwbare tolweg met variabele tol en een gratis route met kans op congestie, zoals bij State Route 91 in Californië. Wel zou het in principe mogelijk zijn om de kosten van transport, distributie en productie te berekenen, en op basis van ad hoc veronderstellingen over welke kostenposten samenhangen met de mate van onbetrouwbaarheid, de kosten van onbetrouwbaarheid te berekenen. Deze methode (de ‘factorkostenmethode’), waarbij SP noch RP modellen worden geschat, kan men zich het beste voorstellen voor goederenvervoer, maar ook voor woon-werk en zakelijk verkeer zou men kunnen proberen om de productiekosten van vertragingen op deze manier te berekenen. Het belangrijkste probleem is dat de empirische basis ontbreekt voor de benodigde veronderstellingen over welke kostenposten (en in welke mate) veroorzaakt worden door onbetrouwbaarheid. In het evaluatie-instrument voor openbaar vervoerprojecten in Nederland THOM/PIOV (NEI en HCG, 1998b), dat een combinatie is van KBA en multicriteria-analyse, wordt betrouwbaarheid (kans op afwijking van de verwachte reistijd) meegenomen als een ordinaal criterium. Er wordt dus een rangorde op dit aspect bepaald. Dit aspect is in de standaard gewichtenset bijna half zo belangrijk als reistijd. De standaardmethode voor projectevaluatie in verkeer en vervoer in Nederland is echter nu OEI, op basis van KBA. Voorzover ons bekend worden nergens in binnen- of buitenland monetaire waarderingen voor betrouwbaarheid gebruikt in officiële KBA’s. Wel wordt daar in sommige landen (behalve in Nederland ook in het Verenigd Koninkrijk, Zweden) over nagedacht en zijn er onderzoeken uitgevoerd in diverse landen die waarderingen in geldeenheden of reistijd leveren voor betrouwbaarheid van reistijden in het personenverkeer over de weg. In Zweden werd recentelijk geconstateerd dat de empirische basis voor het opnemen van een hogere waarde voor vertragingen (dan de Pagina 25
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
reistijdwaardering) onvoldoende was en werd nieuw onderzoek op dit gebied aanbevolen (SIKA, 2002). Op basis van een uitgebreid SP onderzoek in het VK zijn modellen geschat voor het wegverkeer (Accent en HCG, 1995). De volgende uitkomsten voor onbetrouwbaarheid werden verkregen Verdubbeling van de kans op vertraging is even erg als 13 minuten extra reistijd (woon-werk) of 20 minuten (zakelijk, overig). Halvering van deze kans is even goed als 3 minuten minder reistijd (woon-werk) of 5 minuten (zakelijk, overig). Deze waarden worden niet in de praktijk van KBA’s gebruikt. Brownstone en Small (2002) is een overzicht van diverse studies in Californië. Hiertoe behoort Lam en Small (2001). Zij gebruikten RP gegevens (reistijdmetingen) in studies naar de routekeuze rond de State Route 91 (SR 91) waar tol wordt geheven. Een andere studie uit het overzicht is Small, Winston en Yan (2002). Zij gebruikten SP en RP data voor de SR 91. Gevonden werden waarden voor onbetrouwbaarheid van 12-15 $/uur (mannen) en 30-32 (vrouwen) op basis van de RP. Onbetrouwbaarheid is hierbij gemeten als het 90ste percentiel van de reistijd verdeling min het 50ste percentiel. Deze operationele definitie maakt geen onderscheid tussen verwachte en onverwachte congestie; beide vormen van variatie kunnen de mate van onbetrouwbaarheid beïnvloeden, maar de gemiddelde congestie zal tot uitdrukking komen in de mediaan. Onze interpretatie is daarom dat de gevonden reistijdwaardering toegepast moet worden op de reistijd inclusief de verwachte congestie, en de waardering voor betrouwbaarheid op de variatie in de reistijd door onverwachte congestie. De SP/RP leverde 28 $/uur (mannen en vrouwen) voor onbetrouwbaarheid gemeten als het verschil tussen het 80ste en het 50ste percentiel. Het is zeer de vraag of de gevonden waarden zelfs maar representatief zijn voor Californië. Het gaat over een corridor met vooral reizigers met hogere inkomens. Voor de VS als geheel is dit zeker niet representatief. Copley, Murphy en Pearce (2002) hebben een Stated Preference (SP) onderzoek uitgevoerd in Manchester (VK), onder forenzen die als autobestuurder zonder passagiers naar het werk gaan (167 geslaagde interviews). Ze hebben op deze gegevens discrete keuzemodellen geschat. Daarbij werd gevonden dat de standaardafwijking van de reistijd 1,3 keer zo zwaar werd gewaardeerd als de reistijd (betrouwbaarheidratio van 1,3). Hierbij werd geen expliciet onderscheid gemaakt tussen verwachte en onverwachte congestie. Voorzover er regelmatig congestie optreedt voor de betreffende verplaatsingen, zal congestie ervoor zorgen dat de gemiddelde reistijd langer is dan de free flow reistijd. Volgens onze interpretatie zit de gemiddelde regelmatige congestie dan reeds in de gemiddelde reistijd, en de variatie daar omheen is het gevolg van onverwachte congestie en van de dagelijkse variatie in de verwachte congestie. Voor de gemiddelde reistijd inclusief de verwachte regelmatige congestie daarin geldt de factor 1, voor de standaardfout door onverwachte congestie en andere variatie geldt de factor 1,3. Modellen volgens de scheduling theorie op dezelfde data gaven echter aan dat een minuut te laat of te vroeg minder erg was als een minuut reistijd. De waarde van 1,3 wordt in het VK niet in evaluaties gebruikt; eerst wil men meer onderzoek doen. Senna (1991) vond in Brazilië een nog groter belang van de standaardafwijking ten opzichte van de gemiddelde reistijd, maar het is niet duidelijk welke vervoerwijze(n) dit betreft. De Jong et al. (2003) is een SP onderzoek naar de keuze van vertrektijdstip en vervoerwijze in het personenverkeer in Nederland. De steekproef van bijna 1000 autobestuurders en treinreizigers is niet representatief voor de gehele dag, maar richt Pagina 26
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
zich op spitsreizigers. Een model volgens de scheduling theorie is geschat. Gevonden is dat voor alle motieven behalve onderwijs de waardering voor een minuut te laat of te vroeg komen (of vertrekken) hoger is dan voor een minuut reistijd (doorgaans tussen de 1 en 1,5 keer zo belangrijk). Het SP experiment gaat uitsluitend over reacties op de bekende (verwachte) mate van congestie; onverwachte congestie wordt niet meegenomen. Gesteld werd door een commissie voor het Engelse ministerie (SACTRA ,1999) dat het weglaten van variabiliteit van de reistijd in evaluaties leidt tot een onderschatting van de economische baten met 5-50%. Dit betreft dus de verandering door een project in het produkt van waardering en omvang van de onbetrouwbaarheid. MVA (2001) heeft gevonden dat in het busvervoer het belang van een minuut standaardafwijking van de reistijd maar een kwart zo groot is als dat van de reistijd (in het voertuig) zelf. Rietveld et al. (2001) vonden dat een minuut vertraging 2,4 zo erg was als een minuut reistijd in het openbaar vervoer. Deze waarden zijn niet noodzakelijkerwijs met elkaar in conflict, omdat MVA met de standaardafwijking van reistijd werkt en Rietveld et al. (2001) niet. Tabel 6. Betrouwbaarheid in goederenvervoer: kengetallen, gebruikte methode en andere lessen Studie Accent en HCG, 1995
Kengetallen (+definitie) Een 1% toename van de kans op een vertraging van 30 of meer minuten is equivalent aan £ 0,25 – 1,00 (ongeveer 0,45 – 1,8 Euro van 2003).
Bruzelius, 2001
Zweden: voor het spoor is een 1% toename van de frequentie van vertragingen equivalent met 40-60 SEK (4,7-7,0 Euro van 2003) per wagon. Voor het wegvervoer: 30-280 SEK per transport (3,5-32,6 Euro van 2003). Zweden: waarde van risico op vertraging 63 SEK per promille per transport voor weg, 1142 voor spoor en 264 voor vliegtuig (in Euro’s van 2003: 6,1 resp. 111,3 resp. 25,7).
Bruzelius, 2001
Methode Andere lessen Stated preference (SP) wegverkeer in VK, met als attributen: reistijd, reiskosten, informatievoorziening en kans op vertraging. SP onderzoek onder verladers in Zweden in 1989/1990, met onder de attributen: kosten, transporttijd en kans op vertraging.
SP onderzoek onder verladers in Zweden in 1999, met onder de attributen: kosten, transporttijd en kans op vertraging.
Pagina 27
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie Bruzelius, 2002
Fowkes et al, 2001
HCG, 1992a
HCG, 1992b
De Jong et al, 2001
Kengetallen (+definitie) VS: Vermindering van de afwijking van de afgesproken levertijd (schedule delay) met 1 uur is $ 373 waard per transport (ongeveer 393 Euro’s van 2003). VK, wegvoervoer: de waarde van het verschil tussen de vroegst mogelijke aankomsttijd min de vertrektijd is gemiddeld £ 1,07 per minuut per transport (ongeveer de ongehinderde reistijd); voor de tijd waarop 98% van de leveringen aankomt min de vroegste aankomsttijd is de waarde £ 0,85 (‘spreiding’); voor afwijkingen van de vertrektijd (schedule delay) is de waarde £ 0,66 (1,81 resp. 1,44 resp. 1,12 Euro van 2003). Nederland: een verhoging van het percentage niet op tijd met 10% (b.v. van 10% naar 11%) is even erg als 58% hogere transportkosten.
Methode SP onderzoek onder vervoerders in de VS; Scheduling model.
Andere lessen
SP onderzoek onder 40 verladers en vervoerder in het Verenigd Koninkrijk in 1999, met attributen tijd, kosten. Laatst mogelijke vertrektijd, vroegst mogelijke aankomsttijd, aankomsttijd voor 90, 95 en 98%.
SP onderzoek in Methode: SP. 1991/1992 onder 119 verladers en vervoerders in wegvervoer, spoorvervoer en binnenvaart, met als attributen: tijd, kosten, percentage niet op tijd, schadekans en frequentie. Vermindering van de kans op SP onderzoek in 1992 Methode: SP. vertraging met 10 in Nederland, procentpunten (b.v. van 15% Duitsland en Frankrijk naar 5%) is waard 0,5 – 2 met ongeveer 50 Eurocent per ton-km. interviews per land, met attributen kosten, tijd, kans op vertraging, schadekans, frequentie en flexibiliteit. Methode: SP of gezamenlijk SP/RP model.
Pagina 28
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie RAND Europe, 2003
Kengetallen (+definitie) Nederland: een verandering van 10% in het percentage niet op tijd (b.v. van 10% naar 11%) is equivalent aan 1,15 tot 1,20 Euro per transport per gereden uur voor goederenwegvervoer.
Methode SP/RP onderzoek onder 194 vervoerders en verladers in wegvervoer, met als attributen: tijd, kosten, percentage niet op tijd, schadekans en frequentie.
Andere lessen Methode: SP of gezamenlijk SP/RP model.
Voor het goederenvervoer wordt voor waarderingen van betrouwbaarheid (Tabel 6) meestal een variant van de gemiddelde versus variantie methode gebruikt. Hierbij wordt wel de gemiddelde reistijd gebruikt, maar in plaats van de variantie neemt men de kans op vertraging. De reden hiervoor is dat in de SP interviews variantie wordt beschouwd als een te moeilijk begrip om aan de respondenten te presenteren. De kans op vertraging blijken respondenten in de praktijk wel te kunnen hanteren. Ook in SP onderzoek in het personenvervoer wordt variantie niet als zodanig aan de respondenten voorgelegd, maar bevat bijvoorbeeld ieder keuze-alternatief een reeks van mogelijke vertrek- of aankomsttijden (waarbij de onderzoekers later de variantie of standaardafwijking bepalen). Alle ons bekende studies die een waardering bieden voor betrouwbaarheid in het goederenvervoer zijn gebaseerd op SP gegevens.
Pagina 29
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 30
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
6. Kwaliteit: kengetallen
Ook alle kengetallen voor de waarderingen van andere kwaliteitsaspecten dan betrouwbaarheid in personenverkeer- en vervoer (Tabel 7) en goederenvervoer (Tabel 8) zijn afkomstig uit SP onderzoek. Er is aanzienlijk meer van dergelijk SP onderzoek uitgevoerd dan hier is gerapporteerd, met name voor spoorwegbedrijven en openbaar vervoerbedrijven van grote steden, maar de uitkomsten van dergelijke onderzoeken zijn vaak vertrouwelijk. Er zijn geen recente voorbeelden gevonden van nationale evaluatiemethoden waarin een monetaire waardering voor kwaliteitskenmerken anders dan betrouwbaarheid wordt gebruikt. Wel worden het aandeel zittend en de frequentie als kwantitatieve (maar niet-monetaire) indicatoren gebruikt in de Duitse Standardisierte Bewertung (NEI en HCG, 1998a) voor evaluatie van openbaar vervoer projecten. Het Nederlandse programma THOM/PIOV (NEI en HCG, 1998b), waarmee eveneens openbaar vervoerprojecten worden geëvalueerd kent drie ordinale indicatoren (dus rangorde van projecten of projectalternatieven) voor comfort: comfort in het voertuig, op de halte en bij in- en uitstappen. Tabel 7. Overige aspecten van kwaliteit in personenverkeer: kengetallen, gebruikte methode en andere lessen Studie AVV, 2003
Kengetallen (+definitie) Relatieve belang (schaal 1-5, met 1=zeer onbelangrijk en 5=zeer belangrijk): Faciliteiten: 2,24 Informatie en veiligheid: 3,06 Prijs: 3,02 Rijcomfort: 2,70 Tijd en doorstroming: 3,20.
Methode Interviews onder 3387 gebruikers van bus, tram en metro in Nederland.
Gunn, 2001
Looptijd weegt even zwaar als 0,8 – 1,7 maal de tijd in het voertuig; Dienstregelingsinterval 0,6 tot 1 keer zo zwaar.
ISIS, 1998
Voor- en natransporttijd ongeveer even belangrijk als tijd in het voertuig; Wachttijd 30-60% erger dan tijd in het voertuig.
Modellen voor vervoerwijze en bestemmingskeuze voor Nederland (twee versies van het LMS) en Parijs, op basis van RP data (rittenboekjes). Literatuuronderzoek
Pagina 31
Andere lessen De mate van stedelijkheid is niet van belang voor het relatieve belang van kwaliteitsaspecten in het openbaar vervoer.
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie Johannson, 1998
Johannson, 1998
De Jong et al, 2003
MVA, 2000
Rietveld et al, 2001
Transek, 2002
Wardman, 2001a
Wardman, 2001b
Kengetallen (+definitie) Acceleratie, temperatuur in het voertuig en geluid krijgen resp. 5, 10 en 5 van de 100 punten voor kwaliteit in het OV, tegen reistijd 10 (VS) Zweden: de kosten van staan in de bus bedragen 0,5 – 1,4 SEK/verplaatsing (0,06 – 0,17 Euro van 2003) plus 4-9 SEK per uur (0,48 – 1,07 euro) Metro : 0,9 - 2,1 SEK/verplaatsing (0,11 – 0,25 euro) plus 3,5 – 6 SEK/uur (0,42 – 0,71 euro). Frequentie en zitplaatskans niet significant in tijdstipkeuze en vervoerwijzekeuzemodel, na correctie voor effect van herhaalde metingen. De reistijdwaardering tijdens staan in de bus is 25% hoger dan zittend; voor staande in een volle bus zelfs 85% hoger. Een vermindering van de kans om geen zitplaats te hebben van 25% naar 0% is f 3,09 waard (1,52 euro van 2003).
De reistijdwaardering van tijd staand in de bus is 40% hoger dan zittend; voor metro 50% hoger en voor forenzentrein 60% hoger. Looptijd weegt 1,66 maal zo zwaar als tijd in voertuig; Voortransporttijd 1,81; Wachttijd 1,47 Dienstregelingsinterval 0,80; Overstap equivalent aan 17 – 35 minuten reistijd. Looptijd weegt ongeveer 1,5 keer zo zwaar als tijd in het voertuig; Wachttijd 2,5; Overstaptijd 1,75.
Methode Expert judgement
Andere lessen
SP onderzoek onder openbaar vervoer reizigers in Zweden in 1989.
SP onder ongeveer 630 autobestuurders en 360 treinreizigers in de spitsperioden in Nederland. SP onder 309 busreizigers in Frankrijk. SP onder 781 openbaar vervoer reizigers in Nederland, met als attributen: reistijd, kans op vertraging en kans op een zitplaats. SP onderzoek onder openbaar vervoer reizigers in Zweden. Gemiddelden uit groot aantal Britse waarderingsstudies (SP en RP).
Overzicht van diverse openbaar vervoerstudies (wereldwijd).
Pagina 32
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Diverse bronnen in Tabel 7 (Gunn, 2001 ISIS, 1998; Wardman, 2001a,b) geven aan dat tijd in voor- en natransport, looptijd wachttijd en overstaptijd minstens zo zwaar meewegen als tijd in het voertuig. Met name voor looptijd en wachttijd zijn herhaaldelijk factoren in de orde van grootte van 1,5 gevonden (1,5 maal zo belangrijk als tijd in het voertuig, per minuut). Verder zijn er in de literatuur over overige kwaliteitsaspecten in het personenvervoer (Tabel 7) diverse monetaire waarden of reistijdequivalenten te vinden voor wat betreft het zitten of staan (al dan niet in drukte) tijdens een openbaar vervoer reis. In het goederenvervoer zijn waarderingen afgeleid voor schadekans, frequentie en flexibiliteit (Tabel 8). Tabel 8. Overige aspecten van kwaliteit in goederenvervoer: kengetallen, gebruikte methode en andere lessen Studie HCG, 1992a
HCG, 1992b
De Jong et al, 2001
Kengetallen (+definitie) Nederland: een verhoging van de schadekans met 10% is even erg als 3-7% hogere transportkosten; Een verlaging van de frequentie is even erg als 02% hogere transportkosten.
Methode Andere lessen SP onderzoek in Methode: SP. 1991/1992 onder 119 verladers en vervoerders in wegvervoer, spoorvervoer en binnenvaart, met als attributen: tijd, kosten, percentage niet op tijd, schadekans en frequentie. Vermindering van de SP onderzoek in Methode: SP. schadekans met 10 pro mille 1992 in Nederland, (b.v. van 2% naar 1%) is Duitsland en waard 1 – 3 Eurocent per Frankrijk met ton-km; ongeveer 50 Flexibiliteit (binnen 24 uur interviews per land, aan een onverwachte order met attributen voldoen) is waard 5 Eurocent kosten, tijd, kans op per ton-km. vertraging, schadekans, frequentie en flexibiliteit. Methode: SP en gecombineerde SP/RP.
Pagina 33
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Studie RAND Europe, 2003
Kengetallen (+definitie) Nederland: een verhoging van de schadekans met 10% is even erg als 2-5% hogere transportkosten; Een verlaging van de frequentie is even erg als 2% hogere transportkosten.
Methode SP/RP onderzoek onder 194 vervoerders en verladers in wegvervoer, met als attributen: tijd, kosten, percentage niet op tijd, schadekans en frequentie.
Pagina 34
Andere lessen Methode: SP en gecombineerde SP/RP
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
7. Witte vlekken
Er is zowel literatuur voorhanden over de waardering van betrouwbaarheid als over andere aspecten van kwaliteit in het personen- en goederenvervoer. Wel blijkt er een aantal witte vlekken te bestaan van aspecten die nog niet of nauwelijks zijn bestudeerd. Deze worden hieronder besproken. Er bestaan diverse studies voor het wegverkeer (auto, bus, vrachtvervoer over de weg) en het railgebonden openbaar vervoer. Voor het goederenvervoer met de binnenvaart bestaan slechts enkele studies (en een enkele voor het luchtvervoer). Zeevervoer wordt nu meegenomen in het nieuwe reistijdwaarderingsonderzoek van RAND Europe voor AVV, maar in de bestaande literatuur is hier geen enkele verwijzing naar gevonden. Meer specifiek geldt in het personenverkeer en –vervoer dat betrouwbaarheid van reistijden zowel bestudeerd is voor autoverkeer als openbaar vervoer, maar dat de gepubliceerde literatuur over de andere kwaliteitsaspecten nagenoeg uitsluitend het openbaar vervoer betreft. Het belang van zoektijd naar een parkeerplaats en looptijd bij autoverplaatsingen ten opzichte van de tijd in de auto wordt zelden gerapporteerd. Hiervoor zouden aparte rittenboekjes en afzonderlijk SP onderzoek kunnen worden geformuleerd (ook in combinatie als een RP/SP onderzoek naar de omvang en de waardering van parkeerzoektijd en looptijd naar/van de auto). Binnen de kwaliteitsaspecten die in het openbaar vervoer zijn bestudeerd is met name aandacht besteed aan componenten van reistijd (ook wachttijd in samenhang met frequentie) en beschikbaarheid van een zitplaats. Over de volgende kwaliteitsaspecten is in de gepubliceerde literatuur weinig of geen informatie te vinden: • • • •
Aantal (bereikbare) bestemmingen; De mogelijkheid om tijdens een reis te kunnen werken (b.v. op laptop); Andere aspecten van reiscomfort dan zitplaatskans; Informatievoorziening tijdens het wachten (en tijdens de reis in het voertuig).
Wel geldt dat er naar de waardering van sommige van deze aspecten (mogelijk) wel vertrouwelijk onderzoek voor openbaar vervoerbedrijven is uitgevoerd. In een nieuw gericht SP onderzoek onder openbaar vervoer reizigers zouden deze aspecten als attributen van de keuze-alternatieven kunnen worden opgenomen. Wel zijn er aanwijzingen dat een minuut in de file als onaangenamer wordt beschouwd als een minuut rijtijd. Dit impliceert dat gewerkt zou kunnen worden met drie waarderingen, met oplopende kosten per minuut: • •
De waardering voor reistijd zonder congestie; De waardering voor reistijd door verwachte congestie;
Pagina 35
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
•
De waardering voor tijdverlies door onverwachte congestie (bijvoorbeeld gemeten als de standaardafwijking van de verwachte reistijd).
Ook wordt in de literatuur over onbetrouwbaarheid van reistijden doorgaans geen onderscheid gemaakt tussen verwachte en onverwachte congestie. Het is van groot belang dat in nieuw waarderingsonderzoek dit onderscheid wel wordt gemaakt. Daarbij kan verwachte congestie expliciet opgenomen worden in de verwachte reistijd, en onverwachte congestie behandeld worden als bron van variatie in de reistijden In de literatuur over het goederenvervoer zijn waarderingen gevonden voor betrouwbaarheid, frequentie, schadekans en flexibiliteit, maar niet voor overslag- en wachttijden in de transportketen, aantal (bereikbare) bestemmingen en voor informatie over de zending tijdens het transport: tracking and tracing. Ook deze witte vlekken zouden met gericht onderzoek onder verladers en vervoerders kunnen worden opgevuld. De belangrijkste witte vlek is echter dat er geen representatieve monetaire waarden bestaan voor betrouwbaarheid en andere kwaliteitsaspecten, met name voor het personenverkeer over de weg, die direct in Nederland toegepast kunnen worden. Hiervoor bestaan er slechts kengetallen uit buitenlands onderzoek (met uitzondering van de schedule delay waarden uit De Jong et al., 2003, die moeilijk in OEI zijn in te passen), waarvan de representativiteit voor de Nederlandse situatie alles behalve vaststaat. De opzet van het onderzoek van Copley et al. (2002) zou de basis kunnen vormen voor een nieuw op Nederland gericht SP onderzoek over de waardering voor betrouwbaarheid onder automobilisten. Voor het goederenvervoer loopt een onderzoek van RAND Europe voor AVV, waarin monetaire waarden voor betrouwbaarheid en enkele andere kwaliteitsaspecten, zoals schadekans en frequentie, worden vastgesteld. In RAND Europe (2003) staan reeds uitkomsten voor de betrouwbaarheid in het goederenwegvervoer. Voor het openbaar vervoer is er voor Nederland enig houvast te vinden in Rietveld et al. (2001) en zou kunnen worden bekeken of de Klantenbarometer Openbaar Vervoer (AVV, 2003) uitgebreid kan worden met monetaire waarderingen. Ook valt te denken aan een SP onderzoek zoals dat van MVA (2000) of het onderzoek dat RAND Europe momenteel uitvoert voor STIF (openbaar vervoersorganisatie in Parijs en omgeving). Hierbij is één van de attributen van ieder keuze-alternatief een reeks van meerdere reistijden.
Pagina 36
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
8. Conclusies
De nationale en internationale literatuur over onbetrouwbaarheid van reistijden en andere aspecten van kwaliteit in personen- en goederenvervoer is bestudeerd om na te gaan: • • • • • •
welke definities men hanteert; welke aspecten zijn bestudeerd; welke kengetallen voor de monetaire waardering zijn gevonden; welke methoden daarbij zijn gehanteerd; of deze kengetallen toe te passen zijn in kosten-batenanalyses volgens het OEIkader in Nederland; welke lessen voor het opnemen van betrouwbaarheid en andere kwaliteitsaspecten er volgen.
Er zijn geen monetaire waarden voor betrouwbaarheid en andere kwaliteitsaspecten gevonden die in officiële nationale kosten-batenanalyses worden gebruikt. Wel wordt daar in sommige landen (behalve in Nederland ook in het Verenigd Koninkrijk, Zweden) over nagedacht en zijn er onderzoeken uitgevoerd in diverse landen die waarderingen in geldeenheden of reistijd leveren voor betrouwbaarheid van reistijden. Zowel voor de definities als de kengetallen voor betrouwbaarheid zijn drie operationele varianten gevonden: •
Onbetrouwbaarheid gemeten als de standaardafwijking (of variantie) van de reistijdverdeling. Gegevens voor de waardering kunnen verkregen worden door in een stated preference (SP) onderzoek zowel een weergave van de variantie als de gemiddelde reistijd als attributen op te nemen.
•
Onbetrouwbaarheid gemeten via het verschil tussen het 80ste of 90ste percentiel van de reistijdverdeling en de mediaan. Ook hier kan de waardering volgen uit SP experimenten onder reizigers.
•
Onbetrouwbaarheid gemeten als het aantal minuten dat men vroeger of later vertrekt of aankomt dan meest gewenst (schedule delay). Ook dit kan als attribuut in een SP experiment worden aangeboden, naast andere kenmerken als reisduur en reiskosten.
De kengetallen volgens de derde definitie zijn zeer lastig toe te passen binnen het huidige OEI-kader, omdat de koppeling met tijdstipkeuze hier niet wordt gemaakt. Binnen de vigerende methodiek zijn er meer mogelijkheden voor de eerste twee methoden, maar kengetallen waarvan met een redelijke mate van vertrouwen kan worden gezegd dat ze representatief zijn voor met name het Nederlandse autoverkeer Pagina 37
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
ontbreken. De bestaande kengetallen komen uit zeer specifieke onderzoeken en worden in het land van herkomst ook niet voor landelijke of regionale kosten-baten analyses gebruikt (met name omdat het kengetallen betreft voor specifieke groepen en/of gebieden). Aanbevolen wordt om een nieuw landelijk representatief SP onderzoek op te zetten onder Nederlandse automobilisten, waarin de attributen van de keuze-alternatieven zijn: •
• •
een reeks van mogelijke reisduren (evt. ook aankomsttijden), mogelijk met een grafische weergave; dit is een weergave van de mate van onzekerheid van de reistijd, die de respondenten beter begrijpen dan de variantie of standaardafwijking (die de onderzoeker bij ieder alternatief kan berekenen); een gemiddelde reisduur; reiskosten.
De uitkomsten van deze interviews kunnen dan gebruikt worden voor het bepalen van de waarde van onbetrouwbaarheid in het autoverkeer, zowel in de vorm van de standaardafwijking als percentielen van reistijd. Deze aanbevolen SP studie wijkt af van de SP studie uit De Jong et al. (2003), omdat daar ieder keuze-alternatief beschreven werd in termen van o.a. reistijd, reiskosten en een vertrektijd, terwijl het hier gaat om een set van mogelijke reisduren (bijvoorbeeld 10 verschillende reisduren), naast gemiddelde reistijd en –kosten. Voor het openbaar vervoer zijn er voor Nederland kengetallen afgeleid in Rietveld et al. (2001) en zou kunnen worden bekeken of de Klantenbarometer Openbaar Vervoer (AVV, 2003) uitgebreid kan worden met monetaire waarderingen in vervoer per bus, tram en metro. Het reistijdwaarderingsonderzoek dat RAND Europe momenteel uitvoert voor AVV kan monetaire waarderingen bieden voor betrouwbaarheid en enkele andere kwaliteitsaspecten (frequentie, schadekans) in het goederenvervoer over de weg, en met het spoor, binnenvaart, zeevaart en luchtvaart. Uitkomsten voor het wegvervoer zijn gerapporteerd in RAND Europe (2003). Kengetallen voor andere kwaliteitsaspecten dan betrouwbaarheid blijven in het personenverkeer en –vervoer beperkt tot waarderingen voor diverse componenten van reistijd (ook in samenhang met frequentie) en zitplaatskans in het openbaar vervoer. Het beeld hier is diffuser dan van betrouwbaarheid: niet alle studies vinden dat frequentie en zitplaatskans significante factoren zijn in het keuzeproces van de reiziger. Diverse bronnen geven aan dat tijd in voor- en natransport, looptijd wachttijd en overstaptijd minstens zo zwaar meewegen als tijd in het voertuig. Met name voor looptijd en wachttijd zijn herhaaldelijk factoren in de orde van grootte van 1,5 gevonden (1,5 maal zo belangrijk als tijd in het voertuig, per minuut). Bestaande transportmodellen in Nederland bevatten geen expliciete variabelen voor betrouwbaarheid en andere aspecten van kwaliteit. Wel kunnen deze variabelen de keuzen van de besluitvormers hebben beïnvloed, die gebruikt zijn als te verklaren variabelen. De invloed van deze kenmerken komt dan waarschijnlijk met name tot Pagina 38
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
uiting in de constanten voor de diverse vervoerwijze-alternatieven in de vervoerwijzekeuzemodellen (maar mogelijk ook in de parameters voor reistijd en reiskosten, voorzover hier correlatie mee optreedt.). Simulaties voor het effect van veranderingen in kwaliteit kunnen uitgevoerd worden door te bepalen hoeveel bijvoorbeeld de equivalente verandering in reistijd of reiskosten zou zijn (op basis van SP uitkomsten) en die door te rekenen, of door de alternatief-specifieke constanten aan te passen.
Pagina 39
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 40
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Lijst van bestudeerde rapporten en artikelen
1. Accent en Hague Consulting Group (1995); The value of travel time on UK roads – 1994; Final report; Rapport voor het Department of Transport; Accent en HCG. 2. Adviesdienst Verkeer en Vervoer (2003); Klantenbarometer Openbaar Vervoer: Het meten van belang; AVV, Rotterdam. 3. Arcadis Infra (2003); Vertragingen in het binnenlands distributievervoer 2 (VERDI 2); Rapport in opdracht van het Ministerie van Verkeer en Waterstaat, DGG, afdeling wegvervoer, Amersfoort. 4. Basbas, S. en C. Taxilaris (2001); The quality of an urban public transport system as perceived by the users; European Transport Conference – 2001, Cambridge. 5. Bates, J., J. Polak, P. Jones en A. Cook (2001); The valuation of reliability for personal travel; Transportation Research E (Logistics and Transportation Review), 37-2/3, 191-229. 6. Bell, M.G.H. (2000); A game theory approach to measuring the performance reliability of transport networks; Transportation Research part BMethodology, 34-6, 533-545. 7. Berdica, K. (2002); An introduction to road vulnerability: what has been done and should be done; Transport policy, 9, 117-127. 8. Bonsall, P. (2003); Traveller behaviour: decision-making in an unpredictable world; Institute for Transport Studies, University of Leeds. 9. Brownstone, D. en K.A. Small (2002); Valuing time and reliability: assessing evidence from road pricing demonstrations; University of California, Irvine. 10. Bruinsma, F.R., P. Rietveld en D.J. van Vuuren (1999); Unreliability in public transport chains; Vrije Universiteit, Amsterdam. 11. Bruzelius, N. (2001); The valuation of logistics improvements in CBA of transport investments – a survey; SAMPLAN, SIKA, Zweden.
Pagina 41
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
12. Copley, G. P. Murphy en D. Pearce (2002); Understanding and valuing journey time variability; European Transport Conference – 2002, Cambridge. 13. Fowkes, A.S., P.E. Firmin, A.E. Whiteing en G. Tweddle (2001); Freight road user valuations of three different aspects of delay; European Transport Conference – 2001, Cambridge. 14. Friman, M. en T. Gärling (2000); Satisfaction with public transport related to service performance attributes; Karlstad University/Göteborg University. 15. Gunn, H. (2001); Spatial and temporal transferability of relationships between travel demand, trip cost and travel time; Transportation research part ELogistics and transportation review, 37-2/3, 163-189. 16. Hague Consulting Group (1992a); De reistijdwaardering in het goederenvervoer, Rapport Hoofdonderzoek; Rapport 142-1 voor Rijkswaterstaat, Dienst Verkeerskunde, HCG, Den Haag. 17. Hague Consulting Group (1992b); Study into the social benefits of goods transport by road; the conjoint analysis; rapport HCG 170-2 voor de International Road Transport Union (IRU); HCG, Den Haag. 18. Hague Consulting Group (1997); Economic costs of barriers to road transport, Report for the IRU, HCG-report 7040, The Hague. 19. Hague Consulting Group (1998); Value of Dutch travel time savings in 1997 – volume 1; Rapport voor de Adviesdienst Verkeer en Vervoer, HCG, Den Haag. 20. Hilbers, H. en B. Blijie (2003); Accessibility in the Netherlands – the growing importance of reliability; Framing land use dynamics; 17 april 2003, Utrecht. 21. ISIS et al. (1998); Final report for publication, QUITS Quality Indicators for Transport Systems; Submitted to the European Commission DGVII in April 1998, Rome.
22. Johansson, B. (1998); The influence of comfort on public transport ridership; Chalmers University of Technology. 23. Jong, G.C. de, C. Vellay en M. Houee (2001); A joint SP/RP model of freight shipments from the region Nord-Pas-de Calais; European Transport Conference - 2001, Cambridge. 24. Jong, G.C. de, A.J. Daly, M. Pieters, C. Vellay, M.A. Bradley en F. Hofman (2003); A model for time of day and mode choice using error components logit; Transportation Research E (Logistics and Transportation Review), 39, 245-268. 25. Lam, T.C. en K.A. Small (2001); The value of time and reliability: measurement from a value pricing experiment; Transportation research part E-Logistics and transportation review, 37-2/3, 231-251. Pagina 42
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
26. Morface Int. en Cambridge Systematics (1999); A handbook for measuring customer satisfaction and service quality; Transportation Research Board, Washington D.C. 27. MVA Report (2000) Etude de l’impact des phénomènes d’irregularité des autobus – Analyse des resultants. 28. Noland, R.B. en J.W. Polak (2002); Travel time variability: a review of theoretical and empirical issues; Transport reviews, 22-1, 39-54. 29. Peeters, P., P. Rietveld, F. Bruinsma, D. van Vuuren en A.J. Rooijers (1998); Hoe laat denk je thuis te zijn? Een onderzoek naar de betrouwbaarheid van vervoersystemen en de invloed daarvan op het verplaatsingsgedrag; PbIVVS, Den Haag. 30. Probert, A. (2001); Quality – It is not what you think; European Transport Conference – 2001, Cambridge. 31. Projectorganisatie Benutten en Bouwen (2002); Project Benutten en Bouwen 2003-2015; Projectorganisatie Benutten en Bouwen, versie 1.0, 24 januari 2002. 32. RAND Europe (2003); Hoofdonderzoek naar de reistijdwaardering in het vervoer van goederen over de weg; Rapport voor de Adviesdienst Verkeer en Vervoer, RAND Europe, Leiden. 33. Rietveld, P., F.R. Bruinsma en D.J. van Vuuren (2001); Coping with unreliability in public transport chains: A case study for the Netherlands; Transportation research part A, 35, 539-559. 34. Rietveld, P. (2003); Valuation of travel time and traveller information in multimodal personal travel under certainty; Vrije Universiteit, Amsterdam. 35. SACTRA (1994); Trunk roads and the generation of traffic; The Standing Advisory Committee on Trunk Road Assessment Department for Transport, London. 36. SACTRA (1999); Transport and the economy: full report; The Standing Advisory Committee on Trunk Road Assessment Department for Transport, London. 37. Senna, L.A.D.S. (1991); Risk of delays, uncertainty, and travellers’ valuation of travel time variability; 19th PTRC Summer Annual Meeting, Brighton. 38. Transek (2002); Evidence on subjective factor values; EXPEDITE project, Transek, Solna. 39. Wardman, M. (2001a); A review of British evidence on time and service quality valuations; Transportation research part E-Logistics and transportation review, 37-2/3, 107-128. Pagina 43
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
40. Wardman, M. (2001b); Public transport values of time; ITS Working paper, University of Leeds, Leeds.
Pagina 44
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Referenties
Deze literatuurlijst bevat de bestudeerde artikelen en rapporten, de in de hoofdtekst aangehaalde literatuur en de in de beschrijvingen in Bijlage 1 geciteerde bronnen. Accent en Hague Consulting Group (1995); The value of travel time on UK roads – 1994; Final report; Rapport voor het Department of Transport; Accent en HCG. Adviesdienst Verkeer en Vervoer (1998); Advies inzake reistijdwaardering van personen; AVV, Rotterdam. Adviesdienst Verkeer en Vervoer (2003); Klantenbarometer Openbaar Vervoer: Het meten van belang; AVV, Rotterdam. Allen, W.B., M. Mahmoud en D. McNeil (1985); The importance of time in transit and reliability of transit time for shippers, receivers and carriers; Transportation Research B 19, 447-456. Arcadis Infra (2003); Vertragingen in het binnenlands distributievervoer 2 (VERDI 2); Rapport in opdracht van het Ministerie van Verkeer en Waterstaat, DGG, afdeling wegvervoer, Amersfoort. Asakura, Y. en M. Kashiwadani (1991); Road network reliabilitycaused by daily fluctuation of traffic flow; Proceedings of the 19th PTRC Summer Annual meeting, Brighton, 73-84. B&A groep (2000); Synthese bereikbaarheid – eindrapport; B&A groep, Den Haag, i.o.v. Connekt. Basbas, S. en C. Taxilaris (2001); The quality of an urban public transport system as perceived by the users; European Transport Conference – 2001, Cambridge. Bates, J., J. Polak, P. Jones en A. Cook (2001); The valuation of reliability for personal travel; Transportation Research E (Logistics and Transportation Review), 37-2/3, 191-229. Bell, M.G.H. (2000); A game theory approach to measuring the performance reliability of transport networks; Transportation Research part B-Methodology, 34-6, 533-545. Bell, M.G.H. en Y Iida (1997); Transportation network analysis; Wiley, Chichester, West Sussex.
Pagina 45
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Berdica, K. (2002); An introduction to road vulnerability: what has been done and should be done; Transport policy, 9, 117-127. Bonsall, P. (2000); Travellers’ response to uncertainty. In: Bell, M.G.H. and C. Cassir (eds.) Reliability of Transport Networks; Research Press Ltd., Baldock, U.K. Bonsall, P. (2003); Traveller behaviour: Decision-making in an unpredictable world; Institute for Transport Studies, University of Leeds. Booz-Allen & Hamilton Inc. (1998); 1998 California Transportation Plan: Transportation System Performance Measures: Final Report; California Department of Transportation, Transportation System Information Program, Sacramento, California, USA. Brattgård, S.-O. en Petzäll, J. (1982); Anpassning av landsvägsbussar till rörelsehindrade och andra handikappade; Gothenburg University, Department of Hanicap Research Brownstone, D. en K.A. Small (2002); Valuing time and reliability: assessing evidence from road pricing demonstrations; University of California, Irvine. Bruinsma, F.R., P. Peeters, P. Rietveld en D.J. van Vuuren (1999); Betrouwbaarheid van de openbaarvervoerketens; Tijdschrift Vervoerswetenschap, 2/99. Bruinsma, F.R., P. Rietveld en D.J. van Vuuren (1999); Unreliability in public transport chains; Vrije Universiteit, Amsterdam. Bruzelius, N. (2001); The valuation of logistics improvements in CBA of transport investments – a survey; SAMPLAN, SIKA, Zweden. Bryan, M. E. (1976); A tentative criterion for acceptable noise levels in passenger vehicles; Journal of Sound and Vibration. Vol 48. No 1976:4. pp 525-535. Buck Consultants International (2002); Evaluatie OEI-leidraad, Rapport in opdracht van Ministerie van Economische Zaken en Ministerie van Verkeer en Waterstaat, Den Haag mei 2002. Bullard, D. en D.L. Christiansen (1981); Level-of-Service Concept Applied to Public Transportation; Texas Transportation Institute. Technical report 1067-1F. Bureau of Transport Economics (1982) The Value of Travel Time Savings in Public Sector Evaluation. Occasional Paper 51, Bureau of Transport Economics, Canberra, Australia. Bus Partnership Forum (2003); Understanding customer needs – Final report; Bus Partnership Forum (UK). Chen, A., Z.W. Ji, en W. Recker (2002); Travel time reliability with risk-sensitive travelers; Transportation network modeling 2002 Transportation research record, 1783, 27-33. Pagina 46
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Chen, C., A. Skabardonis en P. Varaiya (2002); Travel time reliability as a measure of service; Transportation Research Board, Washington, USA. Chen, A., H. Yang, H.K. Lo en W.H. Tang (2002); Capacity reliability of a road network: an assessment methodology and numerical results; Transportation Research part B-Methodology, 36-3, 225-252. Cohen, H. en F. Southworth (1999); On the measurement and valuation of travel time variability due to incidents on freeways; Journal of transportation and statistics, 2-2, 123-131. Cook, A.J., P. Jones, J.J. Bates, J. Polak en M. Haigh (1999); Improved methods of representing travel time reliability in SP experiments; European Transport Conference – 1999, Cambridge. Copley, G. P. Murphy en D. Pearce (2002); Understanding and valuing journey time variability; European Transport Conference – 2002, Cambridge. Dale, H.M., S. Porter en I. Wright (1996); Are there quantifiable benefits from reducing the variability of travel times?; European Transport Conference – 1996, London. Eijgenraam, C.J.J., C.C. Koopmans, P.J.G. Tang en A.C.P. Verster (2000); Evaluatie van infrastructuurprojecten: leidraad voor de kosten-batenanalyse; Onderzoeksprogramma Economische Effecten Infrastructuur; CPB en NEI, Den Haag en Rotterdam. Elmberg, C. M. en D.A. Quarmby (1981); Interchanges in public transport. 44th UITP International Congress, International Commission on Traffic and Urban planning. Dublin Fowkes, A.S., P.E. Firmin, A.E. Whiteing en G. Tweddle (2001); Freight road user valuations of three different aspects of delay; European Transport Conference – 2001, Cambridge. Fowkes, A.S. en Johnson, I. (1985) Value of Time – Phase II: Small Study on InVehicle, Walk, Wait and Interchange Time. Institute for Transport Studies, University of Leeds. Friman, M. en T. Gärling (2000); Satisfaction with public transport related to service performance attributes; Karlstad University/Göteborg University. Gille, J. en S. Rienstra (2002a); Reistijdwinst: wat is snel OV je waard? – Het gebruik van reistijdwaarderingen in Kosten Batenanalyses; ECORYS Transport, Rotterdam. Gille, J. en S. Rienstra (2002b); Winst in cijfers – Kortere reistijd voor investeringen; Verkeerskunde, nummer 4-2002.
Pagina 47
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Gunn, H. (2001); Spatial and temporal transferability of relationships between travel demand, trip cost and travel time; Transportation research part E-Logistics and transportation review, 37-2/3, 163-189. Gunn, H.F. en Rohr, C. (1996) The 1985-1996 Dutch Value of Time Studies. Paper presented at PTRC International Conference on the Value of Time. Hague Consulting Group (1992a); De reistijdwaardering in het goederenvervoer, Rapport Hoofdonderzoek; Rapport 142-1 voor Rijkswaterstaat, Dienst Verkeerskunde, HCG, Den Haag. Hague Consulting Group (1992b); Study into the social benefits of goods transport by road; the conjoint analysis; rapport HCG 170-2 voor de International Road Transport Union (IRU); HCG, Den Haag. Hague Consulting Group (1997); Economic costs of barriers to road transport, Report for the IRU, HCG-report 7040, The Hague. Hague Consulting Group (1998); Value of Dutch travel time savings in 1997 – volume 1; Rapport voor de Adviesdienst Verkeer en Vervoer, HCG, Den Haag. Hague Consulting Group (2000) Re-estimation of the LMS time-of-day module: model structure and data; Rapport voor de Adviesdienst Verkeer en Vervoer, HCG, Den Haag. Helberg, W. and E. Sperling (1941); Verfahren zur Beurteilung der Laufeigenschafen von Fahrzeugen; Organ für die Fortschritte der Eisenbahnwehsens. Vol 78. No 1941:12. pp 176-187. Hilbers, H. en B. Blijie (2003); Accessibility in the Netherlands – the growing importance of reliability; Framing land use dynamics; 17 april 2003, Utrecht. Holmberg, B. (1977); Standard i lokal kollektivtrafik – Metoder för mätning och beskrivning; Nordiska institutet för samhällsplanering. R 1977:1. Ikharta, H. en P. Mitchell (1997); Technical Report of Southern California Association of Governments’ Transportation Performance Indicators; Transportation Reasearch Record, 1998, (1606), pp. 103-114. ISIS et al. (1998); Final report for publication, QUITS Quality Indicators for Transport Systems; Submitted to the European Commission DGVII in April 1998, Rome. Johansson, B. (1990); Comfort in Mass Transit; XII Meeting of Bus and Coach Experts, Budapest, 6-9, September 1990. Johansson, B. (1992a); Social Policy Toward Public Transport Vehicles – Developing Attractive Public Transport Vehicles; 6th World Conference on Transport Research, Lyon, 29 June-3 July 1992. Johansson, B. (1992b); Comfort requirements of rail vehicle passengers; Second International Conference on Vehicle Comfort, Bologna, 14-16, October 1992. Pagina 48
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Johansson, B. (1998); The influence of comfort on public transport ridership; Chalmers University of Technology. Jong, G.C. de (2000); Value of freight travel time savings; In: D.A. Hensher en K.J. Button (red.): Handbook of Transport Modelling; Elsevier Science, Amsterdam. Jong, G.C. de, C. Vellay en M. Houee (2001); A joint SP/RP model of freight shipments from the region Nord-Pas-de Calais; European Transport Conference - 2001, Cambridge. Jong, G.C. de, A.J. Daly, M. Pieters, C. Vellay, M.A. Bradley en F. Hofman (2003); A model for time of day and mode choice using error components logit; Transportation Research E (Logistics and Transportation Review), 39, 245-268. Kazimi, C,. D. Brownstone, A. Ghosh, T.F. Golob en D. van Amelsfort (2000); Willingness-to-pay to reduce commute time and its variance: evidence from the San Diego I-15 congestion pricing project; 79th Annual Meeting of the Transportation Research Board, Washington, D.C. Kazuya, K. (1999); Commercial vehicle value of time and perceived benefit of congestion pricing; University of California, Berkeley. Koopmans, C.C. en E.P. Kroes (2003); The real cost of congestion in The Netherlands; European Transport Conference – 2003, Straatsburg. Lam, T.C. en K.A. Small (2001); The value of time and reliability: measurement from a value pricing experiment; Transportation research part E-Logistics and transportation review, 37-2/3, 231-251. Lomax, T. en D. Schrank (2002); 2002 Urban Mobility Report; Texas Transportation Institute, TAMU, College Station, Texas, USA. Lomax, T., S. Turner en R. Margiotta (2001); Monitoring Urban Roadways in 2000: Using achieved operations data for reliability and mobility measurement; Federal Highway Administration, Operations Core Business Unit, Washington D.C., USA. May, A.D., D. Coombe, en T. Travers (1996); The London Congestion Charging Research Programme 5: Assessment of the Impacts;, Traffic Engineering and Control, 37. Mayr, R. (1959); Comfort in railway travel; The Railway Gazette. No 1959 9 Oct. pp 266-269. McFarland, R. A. (1969); Human Factors in High Speed Ground Transportation with Special Reference to Passenger Comfort and Safety; In: High Speed Ground Transportation. Transportation Research Institute. Carnegie Mellon University, TRI Research Report 3.
Pagina 49
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Mede, H.J. van der, H. Palm en H. Flikkema (1996); Travel time variability as a new quality indicator; European Transport Conference – 1996, London. Morface Int. en Cambridge Systematics (1999); A handbook for measuring customer satisfaction and service quality; Transportation Research Board, Washington D.C. NCHRP (2001); Economic implications of congestion; NCHRP report 643, TRB/NRC, Washington D.C., USA. NEI en HCG (1998a); Werknotitie 2 inzake MIT instrumentarium OV projecten; Rapport voor AVV; NEI, Rotterdam. NEI en HCG (1998b); THOM/PIOV – gebruikershandleiding; Rapport voor AVV; HCG, Den Haaag Nicholson, A.J. en Z.P. Du (1997); Degradable transportation systems: an integrated equilibrium model; Transportation Research B 31 (3) 209-223. Noland, R.B., K.A. Small, P.M. Koskenoja en X.H. Chu (1998); Simulating travel reliability; Regional science and urban economics, 28-5, 535-564. Noland, R.B. en J.W. Polak (2002); Travel time variability: a review of theoretical and empirical issues; Transport reviews, 22-1, 39-54. Peeters, P., Y. van Asseldonk, A.J. van Binsbergen, T.J.H. Schoemaker, C.D. van Goeverden, R.G.M.M. Vermijs, P. Rietveld en S.A. Rienstra (1996); Mag het ietsje minder snel? Een onderzoek naar de maatschappelijke economische kosten en baten van verlaging van snelheden van personenauto’s; PbIVVS, Den Haag. Peeters, P., F. Bruinsma en P. Rietveld (1998); Betrouwbaarheid van openbaar vervoer ketens: het effect van maatregelen; Verkeerskunde, november 1998. Peeters, P., P. Rietveld, F. Bruinsma, D. van Vuuren en A.J. Rooijers (1998); Hoe laat denk je thuis te zijn? Een onderzoek naar de betrouwbaarheid van vervoersystemen en de invloed daarvan op het verplaatsingsgedrag; PbIVVS, Den Haag. Pettersson, H-Å. (1985); Trafikledning och trafikantinformation; Speach given at SLTF:s congress in Luleå, Sweden, 1985-06-27. Pommer, J. en J. Schunselaar (1998); De paradox van de dienstregeling; PbIVVS, Den Haag. Probert, A. (2001); Quality – It is not what you think; European Transport Conference – 2001, Cambridge. Projectorganisatie Benutten en Bouwen (2002); Project Benutten en Bouwen 20032015; Projectorganisatie Benutten en Bouwen, versie 1.0, 24 januari 2002. RAND Europe (2001); Re-estimation of the LMS time-of-day module: estimation results; Rapport voor de Adviesdienst Verkeer en Vervoer, RAND Europe, Leiden. Pagina 50
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
RAND Europe (2003); Hoofdonderzoek naar de reistijdwaardering in het vervoer van goederen over de weg; Rapport voor de Adviesdienst Verkeer en Vervoer, RAND Europe, Leiden. Rietveld, P., F.R. Bruinsma en D.J. van Vuuren (1999); Coping with unreliability in public transport chains; Vrije Universiteit, Amsterdam. Rietveld, P., F.R. Bruinsma en D.J. van Vuuren (2001); Coping with unreliability in public transport chains: A case study for the Netherlands; Transportation research part A, 35, 539-559. Rietveld, P. (2003); Valuation of travel time and traveller information in multimodal personal travel under certainty; Vrije Universiteit, Amsterdam. Rooijers, T., P. Peeters en W. Oosterwijk (1997); Betrouwbaarheid verplaatsingsgedrag; een theoretisch kader; PbIVVS, Den Haag. Rooijers, T. en W. Oosterwijk (1998); vervoersystemen; PbIVVS, Den Haag.
Subjectieve
betrouwbaarheid
en van
SACTRA (1994); Trunk roads and the generation of traffic; The Standing Advisory Committee on Trunk Road Assessment Department for Transport, London. SACTRA (1999); Transport and the economy: full report; The Standing Advisory Committee on Trunk Road Assessment Department for Transport, London. Scannella, G. en M. Beuthe (2003); Valuation of road projects with uncertain outcomes; Transport Reviews, Vol. 23, No. 1, pp. 35-50, 2003. Senna, L.A.D.S. (1991); Risk of delays, uncertainty, and travellers’ valuation of travel time variability; 19th PTRC Summer Annual Meeting, Brighton. Shaw, T., D. Jackson, G. Morgan, A. Vandervalk en D. McLeod (2000); Estimation of reliability for Florida’s Mobility Performance Measures Program. Florida Department of Transportation, Transportation statistics Office, Tallahassee, Florida, USA. SIKA (2002); Review of cost benefit calculation, Methods and valuations in the transport sector (Summary in English); SIKA report 2002:4, Stockholm. Small, K.A. (1982); The Scheduling of Consumer Activities: Work Trips, American Economic Review, 72, June 1982, 467-479. Small, K.A., C. Winston en J. Yan (2003); Uncovering the distribution of motorists’ preferences for travel time and reliability: implications for road pricing; Department of Economics, University of California, Irvine. Swedish Board of Transport. (1989); Regulations for adapting public transport vehicles for use by disabled persons. Stockholm. Pagina 51
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Temming, J. (1981); Comfort requirements for Heating, Ventilation and Air Conditioning in Motor Vehicles; International Conference on Ergonomics and Transport, Swansea, 1980. Oborne D J, Levis J A (Eds.). Human Factors in Transport Research. Vol 2: User Factors. Academic Press. Whitstable. pp 67-75. TFD (Transportforskningsdelegationen) (1982); Kunskaper om kollektivtrafik – resultat från intervjuundersökningar. Infok r5:1982. Torell, G. (1986); Lokalisering och utformning av bussterminaler – Delrapport: Resenärers och förares syn på bussterminaler. Chalmers University of Technology, Div. of Urban Transport Planning. Rapport 1986:2. Transek (2002); Evidence on subjective factor values; EXPEDITE project, Transek, Solna. Vickrey, W.S. (1969); Congestion theory and transport investment; American Economic Review (Papers and Proceedings) 59, 251-261. Waard, J. van der (1989) Onderzoek weging van tijdelementen, deel 3; TU Delft. Adviesdienst Verkeer en Vervoer (2003); Probleemstellende notitie hoofdonderzoek reistijdwaardering goederen wegvervoer; AVV, Rotterdam. Wakabayshi, H. en Y. Iida (1992); Upper and lower bounds of terminal reliability in road networks: an efficient method with Boolean algebra. Journal of Natural Disaster Science 14 (1), 29-44. Wardman, M. (2001a); A review of British evidence on time and service quality valuations; Transportation research part E-Logistics and transportation review, 37-2/3, 107-128. Wardman, M. (2001b); Public transport values of time; ITS Working paper, University of Leeds, Leeds. Widlert, S. et al. (1989); Värdering av kollektivtrafikens standard. Swedish Transport Research Board. TFB-rapport 1989:2. Wilmink, I.R., P.L.C. Eijkelenbergh, W. Ploos van Amstel en L.A. Tavaszy (2002); Baathebbers van een verbetering van de bereikbaarheid; TNO Inro, Delft. Wunderlich, K., M. Hardy, J. Larkin en V. Shah (2001); On-time reliability impacts of advanced traveler information services (ATIS): Washington D.C., Case study Federal Highway Administration, Washing D.C., USA. Zeithaml, V. et al. (1990); Delivering quality service; balancing customer perceptions and expectations. Free Press. New York. Zuylen, H.J. van, P.J. van Rheenen en M.M. Minderhoud (2003); Het belang van een betrouwbare, stabiele weginfrastructuur; TUD, Delft / EVO, Zoetermeer.
Pagina 52
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Bijlage 1. Beschrijving van de bestudeerde artikelen en rapporten
Pagina 53
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 54
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The value of travel time on UK roads – 1994, Final report. Accent and Hague Consulting Group, 1995. The focus This report for the UK Department for Transport does not focus on reliability or quality, but on the value of time. Values of time (VOT) for the use of UK roads are presented for both passenger and freight transport, based on stated preference (SP) surveys carried out in 1994. Nevertheless, in some of the SP experiments, a measure of reliability was included. Furthermore, a distinction could be made between time spent in congested and in uncongested conditions. Definition The measure of reliability included in one of the passenger SP experiments was: the chance of unexpected delay. Three levels for this attribute were presented: 1. Chance of delay the same as now 2. Chance of delay half as much as now 3. Chance of delay twice as much as now. In the SP survey among freight operators the chance of an unexpected delay of 30 minutes of more was included as an attribute. Method For car drivers, the chance of delay was included in an SP experiment among car drivers together with travel time and provision of delay information on variable message signs. For freight operators, the reliability attribute had to be traded off against transport cost, transport time and provision of information. The data were used to estimate discrete choice models. Findings Accent and HCG (1995) found that in the United Kingdom a doubling of the chance of delay when translated into VOT terms is equivalent to an increase in travel time of approximately 20 minutes for both business and other purposes, but only 12.7 minutes for commuting. Halving the chance of delay translated into VOT terms is equivalent to a decrease in travel time of approximately 5 minutes for both business and other purposes, and approximately 3 minutes for commuting. For freight transport by road, the cost of the journey increases by 0.25 – 1 pound (depending on the vehicle type and the type of operator) for a 1% increase in the chance of a 30-minute or greater delay. Car drivers and passengers were asked to report both time spent in congested conditions and total travel time. The ratio of congested time to total travel time was inserted into the model as one of the adjustments to the time parameter. This variable was highly significant, but stronger for time savings than for time losses. The value of time for business travel time savings is almost 8 pence/minute higher (125% higher than the base value for business) for a trip spent entirely in congested conditions than in a trip with no congestion. For other purposes, the effect is only significant and positive for time savings.
Pagina 55
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Conclusion This study for the UK has produced monetary values for some measure of reliability and values of time that distinguish between congested and uncongested conditions.
Pagina 56
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Klantenbarometer Openbaar Vervoer: Het meten van belang, Adviesdienst Verkeer en Vervoer, 2003. The focus The focus in this report is on determining the valuation (on a scale from 1 to 10, not monetary) of different quality aspects of public transport for the user of public transport. Second, the importance of the quality aspects is addressed. Definition The authors describe 32 quality aspects divided over 5 quality indicators: 1. Facilities: facilities at stops, possibilities to buy food/drinks 2. Information and security: route information at stop and in the vehicle, information provided by signs at the stop/station and in the vehicle, safety at stop and in the vehicle, lights at the stop. 3. Price: clarity of ticket structure, price of public transport, clarity of where to purchase a ticket. 4. Comfort: level of noise inside, level of noise outside, noise disturbance, height of the vehicle, seat availability, cleanliness of the vehicle, driving style, lights in the vehicle and friendliness. 5. Time: number of interchanges, time spent before/after the actual trip, frequency, information concerning delay, walking distance at interchanges, reliability of travel time, travel time public transport and wait time at interchange. Method The method used in this report is based on data from interviews with 3387 users of bus, tram and metro in The Netherlands. The survey consists of the different quality aspects, which the respondents must value between 1 and 10 (high score) and rate their importance between 1 and 5 (very important). The hypothesis formulated in the research is that the level of urbanisation forms a factor of importance in explaining the average score of the importance of the different quality aspects. Three levels of urbanisation are defined; no/little urbanisation, average urbanisation, high level of urbanisation. For each level, 5 areas are selected to conduct the survey. Findings The most important conclusion in the report is that the level of urbanisation does not have a big influence on the average score for the importance of the different quality aspects. In other words, the initial hypothesis must be rejected. The four other important findings: 1. Strong influences of sex and age are found in the survey. Women value almost all the aspects more important than men. Second, if people are getting older they value the importance of the different quality aspects higher. 2. The scores for the importance of the different quality aspects cannot be explained based on the valuation of these quality aspects. 3. The most important quality aspect to improve is reliability of the travel time and the information provision in general. 4. The average valuation for all quality aspects of the public transport is 6,60 (on a scale from 1 to 10).
Pagina 57
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Conclusion This report illustrates the relative importance of different quality aspects of public transport. The report provides evidence for the fact that reliability of the travel times is perceived as a very important quality aspect. However, the valuations of the quality aspects are not translated to a monetary value, which makes the results less useful.
Pagina 58
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Vertragingen in het binnenlands distributievervoer 2 (VERDI 2), Arcadis Infra, 2003. Focus The focus is this report is on the delays in goods transport. Three different delays are identified; delays caused by congestion, delays caused by the business processes and delays caused by the restrictions in the regulations. Definition For our study only the delays related to traffic are important. These delays are defined as the delays during the driving time. The driving time is the time spent by the vehicle on the road. Method The method used to determine the delays (in roundtrips) related to traffic is field research. The field research consisted of accompanying the trucks of 26 companies. During 5 weeks, 168 roundtrips consisting of 1323 trips are followed. A roundtrip is defined as the sum of the trips between the place of departure and the endpoint (usually equal to the place of departure). A trip is defined as a movement between two places that is not interrupted by the business related activities. Findings The important finding in light of our study is summarized in table 1. Table 1: Distribution of the roundtrips based upon the traffic related delays Delay No delay 0 to 5 percent 5 to 10 percent 10 to 15 percent 15 to 20 percent 20 to 25 percent More than 25 percent Total
Amount 103 36 18 5 3 2 1 168
Percentage 61% 21% 11% 3% 2% 1% 1% 100%
The table indicates the difference between the normal and the actual travel time as a percentage of the normal travel time. Only 7% of the trips is delayed 10% or more. Conclusion This report is not very useful for our study because it only provides evidence concerning the distribution of the travel time around the normal travel time. No evidence is presented indicating the value of the different quality and reliability aspects.
Pagina 59
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The quality of an urban public transport system as perceived by the users. Basbas, Taxiltaris, 2001. The focus The main issue is to define what the public perceives as important elements in an urban public transport system. The definitions of quality and reliability There paper does not give explicit definitions for quality or reliability in urban public transport system. Instead the paper summarizes the characteristics of the public transport system in four Greek cities. These characteristics are number of buses, number of bus lines, total number of tickets, % of trips made by bus, average speed and average journey time. Methods to measure quality and reliability aspects and the assumptions The method to examine the development of the characteristics of the public transport system is questionnaire surveys. The aim of these surveys is to identify people’s opinions, preferences and proposals concerning the public transport system in their areas. Conclusions This article is not very interesting in terms of the purpose of this study.
Pagina 60
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The valuation of reliability for personal travel. John Bates, John Polak, Peter Jones, Andrew Cook, 2001. Focus The focus in this article is on personal travel including public transport and private transport. The key theoretical results relating to the highway mode are discussed, an expanded to take in the additional complexity of scheduled public transport services. Definitions In this article, reliability is referred to as the possibility of late arrival, spending longer in certain activities than desired or expected, as well as the stress associated with the uncertainty per se. By its nature, reliability implies a notion of repetition. Repeating a journey can illustrate the variation in travel time. The essence of any measure of variability relates the variations to the expected value. The article further states that the difference between car journeys and journeys by public transport is not very important in the light of reliability. The reliability is calculated around the expected time for both; this introduces a further measure in reliability: the average discrepancy between timetable (for public transport) and outturn. The importance of variability of travel times constitutes of two elements. First travellers are sensitive to the consequences associated with travel time variability (prolonged waiting time, missed connection, late or early arrival). The second element is that travellers place some sort of value on the variability (actually on the uncertainty induced by variability), independently of its consequences at the origin/destination in terms of the other components of generalized costs. Methods The aspects of reliability studied in this research are schedule delay, travel time variability and choice. The aspects are studied for both public transport and private transport. The method for assessing the schedule delay, travel time variability and choice are not fundamentally different besides some small complicating factors when addressing public transport. Therefore only the methods for individual transport are summarized. Schedule delay The major impact of travel time variability is on the choice of departure time. For most travel purposes, change in departure time is considered to be the most sensitive response to changes in generalized costs. The generalized costs determine the choice of mode as the outcome of the maximum utility over all possible departure times. The most advanced method by Noland and Small (1995) decomposes the costs associated with travel time variance into the following components at the optimal departure time, to derive the optimal expected cost: The authors take PAT as the preferred work arrival time (PAT). The issue is to choose the time of departure from home, th. It is assumed that for any possible th there is an associated utility U(th : PAT) which depends on the preferred arrival time. The rule for the departure time choice is then Pagina 61
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
MAX U(th : PAT) with respect to th Without being explicit about he form of the utility function, it is reasonable to expect certain quantities to play a role: 1. the total travel time T (this may well depend on when the journey begins, so that T=T(th)); 2. whether the arrival time is before or after (or equal to) the preferred arrival time: this depends on the sign of the quantity PAT- [th +T(th)]; 3. the amount by which the actual arrival time differs from the preferred arrival time: this is given by the absolute value of the quantity PAT- [th +T(th)]. Empirical research suggests a possible travel (dis-) utility for the trip (NB excluding utility associated with the destination) of
where SDE is early schedule delay defined as Max(0, PAT- [th +T(th)]), SDL the late schedule delay defined as Max (0, [th +T(th)] - PAT), DL a dummy variable which is equal to 1 if SDL > 0 and 0 otherwise and α, β, γ, θ are model parameters, assumed to be negative. Travel time variability The method discussed above does not include travel time variability. The specification of the utility function is independent of whether the travel time is known certainty. The travel time variability is introduced by defining an additional element of travel time Tr which is a random quantity greater than or equal to zero. The distribution of Tr is in general dependent on th. Thus, travel time T is composed of T = Tf + Tx(th)+Tr(th) Hence, for a given departure time th , the traveler will arrive late if th+Tf + Tx(th)+Tr(th)> PAT (preferred arrival time) This is no longer know with certainty because it becomes possible to speak of the probability of late arrival, which is written as PL(th) and implies that the traveler will estimate the probability distribution function of Tr(th). Choice The general body of theory on how travel choices are made relies on the concept of generalized cost, which itself is directly related to utility theory. Utility theory is reasonably well developed to deal with choices under uncertainty. The common criticism that people will develop simplifying rules and assumptions does not invalidate this approach according to the authors. The article explains how the utility theory can be applied to decision making under uncertainty. Pagina 62
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Data In the article, the methods for the measurement of valuations of reliability are compared. There are three methods namely revealed preference (RP), stated preference (SP) and ‘quasi-RP data’. The authors conclude that special cases aside SP is the only realistic possibility for data collection. The major problem with this method is the quantity of information, which needs to be conveyed, and how to extract this information from a particular respondent. Conclusion This article is useful in answering the research questions of the study to quality and reliability of transportation. It describes in practical ways to assess different aspects of reliability.
Pagina 63
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
A game theory approach to measuring the performance reliability of transport networks. Michael Bell, 2002. Definition of network reliability Network reliability has two dimensions. The first relates to the connectivity of a network. When links fail in unfavourable configurations it may no longer be possible to reach a given destination from a given origin, in which case the network becomes disconnected. However, even a connected network may fail to provide an adequate level of service. For example, events may cause unacceptable variation in origin-to-destination travel times, making it difficult for travellers to arrive at their destinations on schedule. The second dimension of reliability is referred to as performance reliability. The author defines network reliability as: a network is reliable if the expected trip costs are acceptable even when users are extremely pessimistic about the state of the network. This definition is a product of the method used in this study, which is explained in the next section. Method to measure network reliability The conventional approach to the study of network performance reliability (see Bell et al., 1999) is to obtain statistical distributions for link performance (typically link travel time, link delay or link capacity) and then to study the impact of link performance variation on network performance (typically the means, medians or percentiles of origin-to-destination travel times or costs). However, there are many circumstances where there is insufficient data on link performance to reliably derive statistical distributions. Moreover, even where sufficient data may exist, there are doubts about the stability of such distributions over time. The paper describes a game theoretic approach to the assessment of network reliability from a network user perspective. Hereby, the expected cost of a trip from its origin to its destination corresponds to the performance of the network. The procedure is nonparametric in the sense that frequency distributions for link performance are not required. Instead, a hypothetical two-player game is envisaged between a network user on the one hand and an evil entity on the other. The game is non-cooperative, in that the network user does not know with certainty which of a set of possible link costs they will encounter and the evil entity imposing link costs on the network user does not know what route the user will choose with certainty. Conclusion The results from this article are very interesting in the sense that it addresses the user response to uncertainty in the link costs. There no indicator values mentioned that are useful for the study into quality and reliability of transport.
Pagina 64
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
An introduction to road vulnerability: what has been done, is done and should be done. Katja Berdica, 2002. The definitions of quality: In this article, vulnerability is regarded as a problem of reduced accessibility and service level of a (road) transport system. The quality of transport is therefore defined by the definition of accessibility and serviceability of the road system. Accessibility is a function of mobility, meaning the ability of an individual to move about by means of private and/or public transport. Mobility consists of (1) the effectiveness of the transport system in connecting spatially separated locations, and (2) individual characteristics influencing the extent to which people are able to make use of the transport system. Serviceability of a link/route/road network describes the possibility to use that link/route/road network during a given time period. The events of interest are in the case of vulnerability in the road transportation system are the ones causing disturbances in traffic and thereby have a negative influence on the serviceability. These events are referred to as incidents: an event, which directly or indirectly can result in considerable reductions or interruptions in the serviceability of a link/route/road network. These incidents causing the road network to malfunction cover a wide range of combinations where probabilities and consequences are concerned. Risk is a composite of (1) the probability for an incident to occur and (2) the resulting consequences should the incident occur. A combination of the previous definitions explains the meaning of the vulnerability in a road transportation system. Vulnerability in a road transportation system is the susceptibility to incidents that can result in considerable reductions in the road network serviceability. The definition of reliability Reliability is the probability of a device performing its purpose adequately for the period of time intended under the operating conditions encountered. Vulnerability in the road transportation system is the opposite of reliability, meaning adequate serviceability under the operating conditions encountered during a given time period. In transportation systems, reliability describes the possibility of successfully traveling from one place to another. It considers three mains aspects: (1) reliability of connectivity meaning the probability of at all reaching a chosen destination, (2) reliability of travel time, meaning the probability of reaching a chosen destination within a given time (3) capacity reliability, meaning the probability of he network being able to ‘swallow’ a certain amount of traffic. In our study, the second aspect of reliability, reliability of travel times, is emphasized. Methods to measure quality and reliability aspects and the assumptions Network reliability in general (Wakabayashi and Iida 1992, Bell and Iida 1997). The focus is on reliability of connectivity. The stochastic variable xi represents the state of the link i with the value of 1 if the link functions and 0 otherwise. Pagina 65
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The system reliability is calculated from either: (1) the minimal routes, which is the number of successive links needed to connect a pair of nodes. As long as one of the routes functions, the system functions. (2) The opposite of the same coin. The minimal number of links needed to disconnect a pair of nodes. Reliability and fluctuation of traffic flow (Asakura and Kashiwadani 1991). The main issue is a modified traffic assignment model for simulating day-to-day fluctuations of traffic flow during ‘ordinary’ traffic conditions that is when capacity is not reduced. The basic assumption is that travel demand in origin-destination (OD) relations fluctuates stochastically from day-to-day around its (observed/estimated) mean value and that this random variation can be described by two mutually independent, normally distributed random variables. The results of the simulation model are used to define and estimate two reliability measures; (1) a connection measure, the probability of travel in an OD-relation without encountering congestion beyond a certain level and (2) a time reliability measure, the probability of a travel time in an OD-relation will not exceed the prescribed travel time. Degradable transportation systems (Nicholson and Du). Nicholson and Du use an integrated equilibrium model, which allows components to have degradation levels between full and zero capacity and demand. This concept is further developed and leading to the introduction of the capacity reliability, meaning the probability that a certain traffic demand can be accommodated at an acceptable level of service.
Pagina 66
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Performance indicators Measure
Calculation
Travel time, tt Travel rate, tr
Tt/ segment length = (speed-1)-1
Delay rate = dr
(atr-dtr = att-dtt)/length
Total delay Relative delay rate
Dr*people vol.*length = (att-dtt)* people vol. Dr/atr = 1 – dtr/atr
Delay ratio
Dr/atr = 1- dtr/atr
Speed of person movement, spm Corridor mobility index
Passenger vol. * average travel speed spm/standard value
Accessibility
Average tt to objectives or percentage of objectives reachable within a specified time ∑ (congested segment length * people vol.)
Congested travel
Comment att= actual tt, dtt = desired tt atr= actual tr, dtr = desired tr, easier to use than speed in statistical analysis Illustrates the intensity of the congestion problem to travellers Illustrates impact of improvements on transportation systems Dimensionless measure for comparing systems operation to a standard Dimensionless measure to illustrate the magnitude of the mobility problem in relation with the actual conditions Measure of travel efficiency Standard value = e.g. freeway lane operating at capacity with typical urban vehicle occupancy rate Specially for joint performance of transportation and land-use
Conclusion This article provides a good overview of the different definitions relating to the vulnerability of road transportation. The definitions are generic and can be applied in studies of vulnerability of different modes of transport (with minor adjustments).
Pagina 67
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Traveller behaviour: decision-making in an unpredictable world, Bonsall, 2003 The focus This presentation gives an overview of how to handle uncertainty (e.g. of travel time) in transport analysis. Findings Variability over time can be regular variation (either stepwise or continuous variation) or seemingly chaotic or random variation. The seriousness of the consequences of delay increase with the amount of delay, but in a discontinuous fashion. The classic modelling approach just multiplies the utility of outcome o of an action a with the probability of outcome o of action a. But in practical decision-making, people have shown to exaggerate low probabilities, to be inable to distinguish between low probabilities and sometimes to seek exposure to risk. There is risk averse and risk seeking behaviour, not just the risk neutral behaviour of classic decision theory. Prospect theory, developed especially by Kahneman and Tversky, has found: • Gains and losses are identified w.r.t. a reference point ; • When the outcomes involve gains, risk-averse behaviour dominates; • When the outcomes involve losses: risk-seeking behaviour dominates. To deal with the perceived uncertainty travellers can use several strategies: • Accept the uncertainty and try to make the best of it (e.g. choose on the basis of the most probable outcome, or the most pessimistic outcome); • Seek to reduce the uncertainty by searching for additional information; • Seek to reduce the uncertainty by changing travel behaviour (e.g. change speed, route, mode); • Seek to reduce the consequences of uncertainty (e.g. safety margins); • Seek to capitalise on uncertainty (e.g. see it as an enjoyable game). Two ways of including uncertainty in modelling are: • Include not only the mean value as explanatory factor, but also the variation around it (e.g. the standard deviation of the distribution); • Use a choice from a restricted choice set: only include outcomes above a certain threshold value. Bonsall also points towards models of learning and of information acquisition. Conclusion This presentation does not provide monetary values of reliability and quality, but does give a framework for thinking about uncertainty of travel time, as well as some recommendations for models including uncertainty.
Pagina 68
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Valuing Time and Reliability: Assessing the Evidence from Road Pricing Demonstrations. David Brownstone and Kenneth A. Small, 2002. Definition Since 1994, a series of demonstration projects in the United States has provided some real-life experience with congestion pricing. These projects provide an opportunity to study some behavioural parameters that are central to the evaluation of transportation projects. The most important is the value of time (VOT), i.e. the marginal rate of substitution of travel time for money in a travellers’ indirect utility function. Another is the value of reliability (VOR), which measures willingness to pay for reductions in the day-to-day variability of travel times facing a particular type of trip. These projects provide particularly useful data for a number of reasons. First, the choice to pay to use the toll facilities is relatively independent from other travel choices because very little transit service exists on these corridors. Second, one of the projects contains sources of independent variation of travel time, reliability, and cost that makes it possible to sort out their separate effects on travel choices. Third, local travellers’ familiarity with these projects, through direct experience or media coverage, makes it possible to collect “stated preference” (SP) data on hypothetical choices in a setting where travellers are likely to understand the context. Method The ability to collect meaningful stated preference (SP) data presents an opportunity to examine reasons for the rather large differences seen in recent literature between estimated values of time from revealed preference (RP) studies and those from SP studies. It also opens the possibility of combining RP and SP data so as to simultaneously take advantage of the realism of the former and the controllability of the latter. The research collects data on revealed preferences (RP) and stated preferences (SP) from the same or overlapping subsets of respondents. The authors have then attempted to compare values of time from these two sources in a controlled manner. In both cases, the median SP estimates are about half the median RP estimates. Findings indicator values (value of reliability) In this section, a number of studies of value of time (VOT) and value of reliability (VOR) are selected and the results are summarized in table 1.
Pagina 69
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table 1: Comparison of selected model results
Notes to Table 1: NA: not applicable (variable not included in model). NR: not reported (variable included but resulting distribution not calculated). * Calculated by drawing from asymptotic distribution of parameter estimates as well as from other sources. a Unreliability is measured as 90th minus 50th percentile of travel-time distribution except where otherwise noted.
b c
Lam and Small (2001), Table 6, Model 1f..
Lam and Small (2001), Table 11, Model 4d.
Calculations for this paper, from coefficients of Small, Winston, and Yan (2002), Table 2, “RP Only” model. d
e Model was estimated on combined RP and SP data, with separate coefficients for RP, SP observations. Observed heterogeneity is not reported in SWY, but was calculated for this paper. f
SWY, Table 3, “RP Estimates.”
g Error components include constants, repeated SP observations with correlation between SP and RP observations for the same individual, and random coefficients of cost, travel time, and unreliability. h i
Unreliability was measured as the 80th minus 50th percentile of travel-time distribution. SWY, Table 3, “SP Estimates.”
Additional models estimated for this paper: same specification and data as in lines 4-5 except with fewer error components. j
Pagina 70
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Includes an error component for repeated SP observations to allow them to be correlated, and allows for correlation between the SP and RP observations for the same individual. k
m
Brownstone et al. (forthcoming), VOT calculations described in text.
Steimetz and Brownstone (2002), Table 4. Uncertain travel times are multiply imputed based on predic-tion models using loop detectors and other variables. n
From same model as previous row, but VOT distribution recomputed for this paper after reweighting individuals in I-15 sample to have same joint distribution of income and distance as the SR91 sample. p
Ghosh (2001), Table 6, p. 78, weighted models, parts 1 and 3. Logit error structure with 5 alternatives: (1) Free lanes, solo driver, no transponder; (2) Free lanes, solo, with transponder; (3) Express lanes, solo, with transponder; (4) Express lanes, carpool, no transponder; (5) Express lanes, carpool, with trans-ponder. Estimated on data from wave 5 of the I-15 panel study. q
r Approximated as 1.35 times the reported standard deviation of drawn values of time. The factor 1.35 is the difference between the 75th and 25th percentiles of the standard normal distribution.
Ghosh (2001), Table 16, p. 113. Estimated on data from wave 5 of the I-15 panel study. s
t Calfee, Winston, and Stempski (2001), Table 5, Scenario 1. We have approximated the inter-quartile range IQR of the value of time from the reported estimates of the means µ and standard deviations σ of price p and time t according to the following formula, which would apply if µp and µt were each normally ad independently distributed:
In this formula, the factor 1.35 is the difference between the 75th and 25th percentiles of the standard normal distribution. Calfee et al. report the 90th and 10th percentiles, whose difference is $2.00. Results from the SP portion of the SR91 data are compared to a binary logit model with no random parameters to one with a random constant and another with a random constant and two additional random parameters. The results are shown in Table 2. (These include some variables solely for comparison with other models not described here.) Accounting for random parameters produces great improvements in goodness of fit, and the estimated standard deviations of the parameters are strongly statistically significant. Yet these results do not show substantial differences in either median VOT or median VOR.
Pagina 71
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table 2: Comparison of SP Models with and without Random Parameters (SR91)
Notes: Dependent variable: choice of toll road. Standard errors in parentheses. All travel variables describe a one-way trip. a Fraction of trips with unexpected delays of 10 minutes or more (SP question) b c Number of workers at work site, in 1000s d One-way commute distance, in units of 10 miles e Dummy variable equal 1 if work arrival time or work departure time are flexible Dummy variable equal 1 if have bachelor degree or higher Conclusion The focus in the research project described is on the value of time, which is outside the scope of the study of quality, and reliability of transport. The interesting element for us is the valuation of reliability. The study presents evidence that that the travel time accounts for about two-thirds, and reliability one-third, of the service quality differential between the free and express lanes. A second interesting observation is that the valuation of reliability is much higher for women than for men – roughly twice as high. One possible explanation for this result is that women have more child-care responsibilities, which reduce their scheduling flexibility. This illustrates the importance of the measurement of valuation of reliability. However the research describes big differences (100%) in the values obtained by the different methods, stated preference and revealed preference. The author concludes that since SP Pagina 72
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
methods are increasingly used for project evaluation; our results suggest that this use of SP value-of-time estimates will undervalue projects to reduce congested travel time. Explaining the large differences between SP and RP estimates is clearly an important topic for future research. For the study into quality and reliability of transport it is important to take into account the big differences between stated preference and revealed preference estimates for aspects of service quality. Furthermore the ratio between value-of-time and value of reliability (2:1) illustrates their relative importance.
Pagina 73
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Reliability of public transport chains. F.R. Bruinsma, P. Peeters, P. Rietveld, D.J. van Vuuren, 1999. The focus The focus is on public transport in the Netherlands. The aspect of quality of transport that is emphasized is reliability of chain travelling times. The article is based on the research: “What time do you think you will be home? A study to the reliability of transportation systems and its influence on the travel behaviour”. A more detailed description is provided in the summary of this research included in our study. Conclusion This article does not provide additional results. For the summary of the result we refer to the article mentioned earlier.
Pagina 74
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Unreliability in public transport chains. F.R. Bruinsma, P. Rietveld, D.J. van Vuuren, 1999. The focus Two components are analyzed in detail in the paper: deviations from the official timetable and the impacts of reliability on transfers. The focus is the reliability of public transport chains. Definition Reliability may relate to various quality aspects of transport such as the quality of vehicles and the quality of infrastructure as determinants of the likelihood that one will arrive safely. Other reliability factors concern the probability of obtaining a seat in public transport, that certain facilities are open in railway stations, etc. The paper focuses on a specific aspect of reliability, i.e. reliability of travel times as we do (or arrival times). Three notions are introduced: 1. official travel time 2. actual travel time distribution 3. perceived travel time distribution In public transport the official travel time is the travel time according to the published timetables. The actual travel time distribution represents the travel time outcomes of actual trips. Delays may occur due to a large number of factors. The perceived travel time distribution refers to travel times as perceived by the traveller. The paper focuses on the actual travel, departure and arrival times. This is not to say that perceptions are unimportant. Perceptions operate significantly in actual decisions in transport, such as route choice and transport mode. However, an analysis of the role of perceptions lies outside the scope of this paper. Method A substantial number of trips made by public transport are chains consist of more than one link. Therefore, whenever unreliability in public transport is studied from the unimodal perspective of individual suppliers the result is an incomplete picture, because the probability that travellers miss connections was not included. A sample of 300 journeys by public transport chains are used and made operational for three different periods of the week (peak hour, off-peak and during the weekend) according to the official timetables as published by the public transport companies. The distribution of arrival and departure times - specified for each public transport mode – are used to simulate disturbances of the travel times compared to official travel times as published by the public transport companies. The three distribution functions used in the paper are the gamma, the log-normal and the Weibull distribution. These distributions have a common feature in that they have a unique mode and finite expectation and variance. Whereas the standard gamma, lognormal and Weibull distributions are all defined on the positive half axis, a third parameter is introduced to all distributions; the location parameter m. This parameter is chosen such that it is identical to the mode of the particular distribution. If the shape parameters are rightly chosen (i.e. smaller than 1), the modes of the gamma and the Weibull distribution are not equal to the minimum value. For the log-normal Pagina 75
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
distribution this is always the case. In practical situations this is often a desirable property - both for departure and arrival distributions -, whereas it may allow early departures and arrivals. The formal expressions for the cumulative distribution functions are as follows:
Findings Irrespective of the period of the day, there is an expected delay in travel time of about 10 percent compared to the official scheduled travel time (see table 1). The total average travel time is shortest during the morning peak and longest on Sundays. According to the official time table and the simulations, this is mainly caused by the increase in waiting time, due to the frequencies in services that are high during the morning peak and low on Sundays. Regardless of the time of day in about 30 percent of the chains passengers arrive before the scheduled arrival time. However, the number of times one arrives over two minutes before the scheduled arrival time are rare. More important is the fact that about 30 % of the trips have a delay of over five minutes. Table 1: Average scheduled and simulated arrival times in the morning peak, offpeak hours, and on Sundays
Pagina 76
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table 2: Travel times and distribution of arrival times by different periods of the day
The table shows that irrespective of the period of the day on a one-hour trip one should expect a delay of over five minutes every fourth or fifth trip. There are only a few chains 14 where there are large differences in the percentage of trips causing more than five minutes delay depending on the period of the day the trip is taken. For example, the train/bus chain shows a relatively high percentage of trips in the morning peak with a delay of over five minutes compared to off-peak hours and in particular Sundays. Conclusion This article provides insight in how to estimate the distribution of delays around the official travel time in public transport chains and answers the research question which method can be used to measure aspects of transport quality and reliability.
Pagina 77
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The valuation of logistics improvements in CBA of transport investments – a survey. Bruzelius, 2001 Introduction Cost-benefit analysis (CBA) has been used to appraise infrastructure investments in the road sector in Sweden since the late 1960s and in the railway sector since the late 1980s. Important components of logistics cost have been taken into account, but never in a comprehensive way. Originally only the cost of transport was recognized, later on the cost of capital tied down in goods being transported have been taken into account, and more recently attempts have been made to also account for improved quality, including expected delays or risk of delays. Methods Empirical studies concerning the goods dimension of logistics costs focus on three aspects, viz. the time duration of transport, reliability, and damage and loss. Given that the costs related to the mode used to transport the goods are handled separately, as is the case in the CBA methodology used in Sweden, the remainder of the benefits associated with improvements in logistics should adequately be reflected through these three variables and the prices or (unit) values associated with them. Two approaches may be used to estimate the values associated with these three variables, viz. market prices and using values derived from models of choice between transport alternatives in which the three variables are explanatory factors. The CBA methodology used in Sweden today is, in effect, based on a mixture of these two methods. The two approaches give very different results. The approach based on market prices (also referred to as the capital value approach) results in low prices, indeed much lower values than the ones currently in use in Sweden. Even taking into account reliability, which is currently not recognized in the CBA methodology used in the road sector, is not likely to change that assessment. It appears that the value of reliability using the market price approach would yield values per time unit (minutes) saved, which are of the same order as the 'pure' value of freight time (in minutes) based on the market price approach. No estimates have been prepared for damage and loss using this approach on account of lack of data on the probability of damage and loss per km or time unit. This is a void that should not be all that difficult to fill. . There are two methods for obtaining data sets for deriving unit values by way of econometric models: data on observed choices (revealed preference (RP) data), and data on hypothetical choices obtained through interviews (stated preference (SP) data). Findings CBA is normally based on market prices. There must therefore be a reason - related to the issue of validity - for not using the capital value approach. According to Bruzelius the literature does not appear to offer any such reasons. SP- and RP-models make use of variables, which cannot be provided within performing cost-benefit analyses of transport schemes. For example, reliability is in these models often measured in terms of the portion of shipments that arrive late. When appraising Pagina 78
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
transport interventions by way of CBA it would normally not be feasible to measure their impact on the portion of shipments arriving late. The SP- and RP-approaches are based on methods which raise a number of issues as concerns what is being measured, and therefore whether estimated values are valid from an economic point of view. There is a need to address these issues through research before the estimates of these models are accepted for use in officially sanctioned CBA methodology. Conclusions This article is interesting in light of the purpose for the study into quality and reliability of transport because it reviews the literature on values of freight time and freight time reliability. Further it illustrates how the time duration of transport, reliability, and damage and loss can be included in the CBA methodology in Sweden. Analogies with cost-benefit analysis in the Netherlands can be observed. Findings indicator value (value of reliability) Table 1: Results of SP studies, values of reliability
Pagina 79
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Understanding and valuing journey time variability. Copley, Murphy, Pearce, 2002. The focus The main issue is to explore what travel time variability means to people. The definition The authors state that journey time variability is an important factor in influencing travel behaviour and that it has two dimensions – a frequency dimension – how often delays occur and a magnitude dimension – how big the delay is when it occurs. Method The research includes two methods, qualitative methods and quantitative methods. The qualitative part involves in-depth interviews and focus groups to explore the perception of journey time variability by travellers. The results of the qualitative studies are used as an input for the quantitative part consisting of a stated preference techniques. The two approaches are the traditional travel time variability modelling approach (= the mean variance approach), and activity approach. The former measures the relative valuation of travel time variability compared with mean journey time expressed as a ratio (β/α): U=αTT+βSDTT Where U is the utility, TT is the travel time and SDTT represents the standard deviation of travel time (e.g. both in minutes). The activity approach assumes there is an ideal arrival time for arriving for a particular activity. Any departure from this ideal time imposes scheduling costs to the traveller (Small 1995). The following utility function is estimated: E(U)=γcE(T)+δE(SDE)+εE(SDL)+ηPL Where expected utility E(U) is dependent on expected (or mean) travel time E(T), expected schedule delay-early, E(SDE), expected schedule delay-late E(SDL) and the probability of late arrival. Findings qualitative methods Respondents were asked to describe their experience with and response to journey time variability. This revealed little that was unexpected. The key findings were that journey planning does not allow for extreme incidents, for example accidents and vehicle breakdowns. Travellers, especially for journeys with fixed appointments, value predictability of journey time. Where journey time variability exists considerable buffers are built into schedules in order to avoid being late. In the focus groups it was established that most respondents can cope with the task of comparing data relating to journey choices and do understand the information presented. Furthermore respondents are able to understand journey time information presented as a distribution histogram based on large amount of data with no training.
Pagina 80
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Findings quantitative methods Modelling the data using the mean variance approach has shown that the standard deviation of journey time is valued 30% higher than journey time, which produces a reliability to time ratio of 1.3. However, the scheduling modelling approach, in which parameters for scheduled delay early (SDE), scheduled delay late (SDL) and probability of being late produced unexpected results. While SDL time is valued higher than SDE, both are valued less than journey time. The inconsistency between the mean variance and the activity approach requires further investigation. Conclusion First this article reveals the perception of people concerning travel time variability. Secondly it shows that different approaches for modelling the data can lead to different results, which are inconsistent. The second observation can be of use in the study into quality and reliability of transportation and deserves attention. The other evidence presented in this article is basic knowledge.
Pagina 81
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Freight road user valuations of three different aspects of delay. Fowkes, Firmin and Tweddle, 2001. The focus The paper looks at the valuation of different delays from the perspective of firms using the roads for freight transport. Definition The paper considers three types of delay. The first of these was due to increased freeflow journey times. The second was due to increased spread of actual arrival times and the third was schedule delay in undertaking the journey. Method The research has utilized the Leeds Adaptive Stated Preference survey methodology. This methodology consists of a computerized stated preference (SP) survey, which is adaptive to earlier SP choices made by the respondents. The four attributes included in the survey are the three different forms of delay and a cost component. The former is particularly interesting in light of the research questions formulated in the study into quality and reliability of transportation. Findings The study is based on 40 interviews of which 39 respondents are used effectively. Table 1 shows the results and the good t ratios indicate that all three types of have significant positive costs equivalents. Table 1: Valuations of delay time (VDT), arrival times spread (VSP) and schedule delay (VSH) expressed as pence per minute, end-2000 prices (table contains averages, t ratios in brackets) n
Cost (£)
Whole sample
40
285.8
Distance (km) 281.6
Own account
11
227.3
237.2
(haulier
19
298.2
286.8
(shipper
10
326.8
320.6
25
310.2
281.0
15
245.3
282.7
n
Cost (£)
J.I.T / QR
27
277.9
Distance (km) 297.1
Not J.I.T/QR
13
302.4
286.8
Articulated
33
306.8
291.6
Not articulated
7
186.8
234.7
Distance less than 250 km
14
179.4
132.9
Third party interviewed) Third party interviewed) Distribution Not distribution
VDT (p/min) 107.1 (15.7) 169.3 (15.4) 155.1 (9.2) 37.2 (3.7) 183.6 (14.4) 76.2 (9.5) VDT (p/min) 128.6 (15.6) 61.0 (5.1) 98.4 (12.0) 126.6 (10.2) 89.9 (9.5)
Pagina 82
VSP (p/min) 85.3 (13.7) 89.5 (10.0) 167.6 (8.3) 61.5 (6.5) 128.7 (13.0) 56.9 (7.1) VSP (p/min) 101.8 (13.7) 46.8 (4.1) 90.2 (11.2) 78.1 (8.1) 93.8 (11.3)
VSH (p/min) 65.8 (26.3) 126.0 (25.0) 86.8 (15.7) 31.3 (9.3) 104.2 (23.6) 47.7 (15.7) VSH (p/min) 75.9 (26.3) 35.6 (7.2) 63.4 (22.6) 74.7 (13.4) 59.0 (17.5)
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid Distance greater than 250 km
26
343.2
361.7
Chemicals, Chem. products
8
397.3
285.0
Food, drink, grocery
15
288.7
298.0
Other commodities
17
230.9
265.6
Rail possible
13
301.4
300.5
Rail not possible
27
278.3
272.6
Daytime movement only
32
283.7
268.2
Some night time movement
8
294.7
325.5
North east based interviewer
18
321.1
275.8
Huddersfield based interviewers
22
257.0
286.4
125.0 (12.9) 224.7 (6.10 90.9 (11.6) 145.7 (9.8) 77.9 (6.4) 120.5 (146) 97.3 (14.0) 431.5 (10.7) 50.5 (4.1) 131.4 (16.1)
74.5 (7.9) 126.6 (6.5) 77.5 (10.6) 93.3 (6.2) 60.4 (5.4) 96.2 (12.9) 72.0 (10.6) 159.0 (10.0) 104.7 (7.5) 80.3 (11.6)
74.1 (19.8) 94.3 (10.0) 48.4 (15.6) 97.0 (20.9) 56.3 (12.2) 69.6 (23.5) 61.4 (15.7) 173.9 (13.7) 49.2 (13.4) 80.0 (23.5)
The authors conclude that the results represent plausible responses (compared with the values cited in the literature) given the composition of the sample. This sample was not intended to be representative of any particular population, e.g. the commodity distribution in the sample will not reflect the population. However, an attempt was made to obtain sufficient responses in each dimension of interest. The reasons that freight transporters and logistics operators value a high level of journey predictability according to the evidence presented in the paper are; on the supply side the road freight transport industry face an upward pressure on costs. More efficient methods of operating are necessary to survive and these methods require sophisticated scheduling techniques, which depend upon a high level of journey time predictability. On the demand side, various sectors have adopted modern logistics and supply chain management techniques. Just-in-time management is the most significant technique adopted. This technique relies heavily on the predictability of journey times, especially when a number of sequential activities are involved. Conclusion The paper contains evidence for the valuations of different forms of delay. The SP experiment gives an indication of the valuations. One should be careful in adopting these results because of limitations in the sample of the SP survey. Nevertheless the paper illustrates quite well, the factors that play a role in freight decision-making and most importantly a translation to costs is included.
Pagina 83
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Satisfaction with Public Transport Related to Service Performance Attributes. Margareta Friman (Karlstad University) and Tommy Gärling (Göteborg University), 2000. Focus The primary objective of the current research is to investigate if there is any relationship between service performance in public transport (busses and streetcars), frequency of negative critical incidents, and user satisfaction. Definitions In this paper quality is defined as the satisfaction or dissatisfaction with the service. One approach in researching the customers satisfaction emphasizes transaction-specific or encounter satisfaction, that is, satisfaction with single transactions or encounters with a product or service. Another approach focuses on cumulative satisfaction, in which satisfaction is determined by satisfying or dissatisfying encounters with a product or service over time. In both approaches satisfaction is either defined as an overall judgment of satisfaction or decomposed into satisfaction with performance or performance attributes. A critical incident is an encounter that is particularly satisfying or dissatisfying. Its occurrence is assumed to have a significant impact on satisfaction. In public transport services, negative critical incidents is assumed to have more impact since, in contrast to some other services, avoiding dissatisfaction may be more important than achieving satisfaction. An important question to ask is how critical incidents are coded in memory. One explanation is that the frequencies of remembered negative critical incidents are the sources of attribute cumulative satisfactions, which in turn have direct effects on overall cumulative satisfaction. An example is that the remembered frequency of delays is coded in memory as a certain degree of unreliability of the service, which have a negative effect on traveller’s overall cumulative satisfaction with the service. The frequency of negative critical incidents may be affected by the service performance offered by the public transport company. Methods In order to investigate the relationships a survey was conducted of people in Göteborg, Sweden that use public transport to go to work (N=228). They were asked questions to measure attribute-specific satisfaction (ATTSAT) and overall satisfaction (SAT) with the service. For each question respondents checked a nine-point scale ranging from "very dissatisfied" to "very satisfied". The questions were intended to tap the following performance attributes: Users’ satisfaction: 1. Information 2. Treatment by other employee; 3. Punctuality at departure; 4. Comfort in the vehicle; 5. Employee knowledge; 6. Punctuality at arrival; 7. Clean front; 8. Treatment by driver; 9. Manage critical incidents; Pagina 84
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
10. Design of stops; 11. Price; 12. Engagement; 13. Co-ordination; 14. Manage lack of quality; 15. Travel time; 16. Comfort of stop; 17. Reliability; 18. Manage dissatisfaction; 19. Trust; 20. Departures; 21. Number of ticket retailers; 22. Safety in traffic; 23. Overall satisfaction. A set of descriptions of negative critical incidents (FNCI) was presented which are based on analysis of actual experiences in public transport obtained from customer complaints and interviews. Respondents were asked to read each description carefully and to judge whether they themselves had experienced, or heard that somebody else had experienced, a similar incident. Questions were asked about each incident including if respondents had never, a few times, occasionally, or regularly experienced a similar one; if they did not remember when it occurred or if it was last week, last month, some time last year, or earlier; how negatively they experienced the incident; the number of people they had told about it; and if they had heard about somebody else who had encountered a similar incident, a few times, occasionally, or frequently. They also rated how negatively they experienced it. Negative critical incidents: 1. Not permitted to embark the vehicle; 2. Careless driving; 3. Not permitted to get off the vehicle; 4. Driver did not stop at the stop; 5. Early departure; 6. Late departure; 7. Cancelled trip without notice; 8. Scheduling; 9. Incorrect display; 10. Crowding; 11. Comfort in the vehicle. Service performance (SERVPERF) was measured by asking respondents about place of residence and work so that information could be collected about the public transport service used by each respondent to work and to downtown. This information was retrieved from the public transport company, and included travel time, frequency of services, wait time, and number of changes. Service performance: 1. Wait time; 2. Travel time; 3. Frequency of services; Pagina 85
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
4. Number of changes. Results A model of the relationships between service performance, frequency of negative critical incidents, and user satisfaction was estimated on the survey data (see for standardized estimates and t-statistics figure 1 and 2). A direct relationship from satisfaction with specific attributes to overall satisfaction for both trips to work and to downtown was produced. Furthermore, an indirect relationship was observed from the frequency of negative critical incidents to overall satisfaction through attribute-specific satisfaction. Service performance had a direct effect on overall satisfaction through frequency of negative critical incidents.
FNCI Travel time Wait time
1.00
.06 (0.83) -.55 (-9.84)
.80 (10.64) .63 (8.89)
SERVPERF
Frequency of service
.06 (0.99)
ATTSAT .23 (3.73) .54 (10.03)
Changes
.50 (7.04)
SAT
Figure 1. Estimated structural model specifying causal relations between service performance, FNCI, ATTSAT, and SAT for trips to work.
FNCI Travel time
Wait time
1.00 .78 (11.97) .37 (5.52)
Frequency of service
Changes
.19 (2.85)
SERVPERF
-.60 (-10.96)
.08 (1.48)
ATTSAT
.11 (2.02)
.96 (13.72)
.56 (10.37)
SAT
Figure 2. Estimated structural model specifying causal relations between service performance, FNCI, ATTSAT, and SAT for trips to downtown. It is suggested that expectations about a given level of service performance determine whether travellers frequently perceive negative critical incidents. However, service performance may also have an independent direct effect on satisfaction. The present results suggest that in public transport the latter may be the strongest. On the other hand, there may be other measures of service performance that are correlated with the frequency of negative critical incidents such as of the actual reliability of the service or the number of passengers (crowdedness). It is still important to realize that measuring Pagina 86
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
users’ satisfaction by means of the critical incident technique will most likely fail to reveal all attributes of service performance that are important. A multi-method approach is called for. Conclusion This article is useful in answering the research questions on how to define and measure quality in (urban) public transport.
Pagina 87
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Spatial and temporal transferability of relationships between travel demand, trip cost and travel time. Hugh Gunn, 2001. The focus The focus of this article is on trends which appear in time and cost trading as between different time periods and different countries. The paper provides evidence for the stability of observed relations in revealed preference (RP) and stated preference (SP) over time and place. Methods The paper reviews six major studies assembled with specific of testing spatial and temporal transferability. In all studies SP and RP models are used. In the next section the interesting findings in light of the purpose of the study into quality and reliability of transport are summarized. Findings The findings are particular interesting to answer the research question: ‘can guidelines for incorporating quality and reliability aspects in transport models be derived?. The first finding is that many relationships seem to be stable and transferable over time and between regions. The RP studies together lend further support to the use of a logcost variable in mode-and-destination models, with the implication that value of travel time savings (VTTS) should increase with journey duration/cost/distance, as a result of absolute travel cost differences becoming less important. The SP studies confirm that experimental VTTSs increase with journey duration. Secondly, the SP studies confirm the stability of the major relationships between VTTS and explanatory variables of income, personal characteristics, household characteristics, free time and so on, between time periods and across regions. The danger of comparing average VTTS between experiments with different designs has been emphasized. For example, suppose “distance” is an explanatory factor in metaanalysis of many studies. Suppose some studies are of short-distance choices, some long-distance. If all the studies examined which had long-distance contexts used large time variations and all the studies examined which had short-distance contexts used small time variations, then a meta-analysis could associate an effect which was truly related to size-of-time-savings wrongly to distance. The third finding is that from SP work, factors are now available to adjust time coefficients in the utility functions, adjusting for effects such as population age, family structures, congestion levels extent of free time and so on. ‘Generalized cost’ models should then be adjusted for income differences in forecasts by decreasing the influence of cost, not increasing the value of time. This is a result of some importance, since the cross-sectional equi-finality of generalized time and generalized cost makes it impossible to predict whether travel would increase or decrease as incomes rise. Findings The important findings in this article are summarized in table 1. Table 1 sets out the resulting co-efficients for the cost and time variables, along with two of the most important of the other variables in this model specification, being the effect of ‘carPagina 88
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
competition’ within the household on the likelihood of being either a driver or a passenger. The importance of the results is that they indicate the relative value of different components and the transferability in time and place. Table 1
Conclusion There are findings in this article that are of some importance. However the findings do not deal with the core of the literature review of quality and reliability of transportation. The deal with aspects associated with quality and reliability of transportation. Therefore, this article may become relevant in future research. Especially, the evidence concerning the transferability of relationships is relevant.
Pagina 89
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
De reistijdwaardering in het goederenvervoer. Hague Consulting Group, 1992. The focus This study was carried out for DVK (now: AVV). The objective was to provide values of time for freight transport by road, rail and inland waterways in The Netherlands. In the Stated Preference (SP) experiments, reliability and probability of damage were included as attributes. Definition Reliability was measured as the percentage of deliveries that do not arrive at the specified time or do not arrive within the specified time interval. Other quality aspects included in the SP were probability of damage to the goods during transport (in pro mille) and frequency. Method Shippers and carriers were asked to trade between hypothetical alternatives for their shipments, defined in terms of the following attributes: transport cost, transport time, reliability of transport time, probability of damage during transport and frequency. Data from the SP experiments were used to estimate discrete choice models with all attributes measured in percentage changes (observed level=100). The ratios of coefficients (e.g. reliability to costs) were called ‘trade-off ratios’. Multiplication of these ratios with the transport cost gives the monetary value of a percentage change in some attribute. Findings It was found that a 10% increase in the percentage of deliveries that were not on time was valued at the same level as a 5-8% (depending on the mode and commodity type) increase in transport cost. A 10% increase in the probability of damage was equivalent to a 3-7% increase in transport cost. Frequency was not significant in some segments; for the segments were it was significant, a 10% increase was equivalent to a 2% decrease in transport costs. Conclusions The study has provided relative values for reliability, probability of damage and frequency in freight transport. At the moment RAND Europe is carrying out an update of this study for AVV.
Pagina 90
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Study into the social benefits of goods transport by road. Hague Consulting Group, 1992. The focus In a study for the International Road transport Union (IRU), Stated Preference (SP) experiments were carried out among shippers in The Netherlands, France and Germany. These experiments included reliability, probability of damage and flexibility as attributes. Definition Reliability was measured as the percentage of deliveries that do not arrive at the specified time or do not arrive within the specified time interval. Other quality aspects included in the SP were probability of damage to the goods during transport (in pro mille) and flexibility (an unexpected door-to-door transport can be organised within 24 hours: short lead time). Method Shippers and carriers were asked to trade between hypothetical alternatives for their shipments, defined in terms of the following attributes: transport cost, transport time, reliability of transport time, probability of damage during transport and flexibility. Data from the SP experiments were used to estimate discrete choice models with all attributes measured in percentage changes (observed level=100). The ratio of coefficients (e.g. reliability to costs) were called ‘trade-off ratios’. Findings A 10% point decrease (e.g. from 15 to 5%) in the percentage of deliveries not on time was found to be just as good as a 0.5-2 Eurocent increase in transport cost per tonnekilometre. Flexibility was worth about 0.5 Eurocent per tonne-kilometre. A decrease in the probability of damage by 10 pro mille points (e.g. from 20 per thousand to 10 per thousand) was worth 1-3 Eurocent per tonne-kilometre. Conclusions The study has provided relative values for reliability, probability of damage and flexibility in freight transport.
Pagina 91
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Economic costs of barriers to road transport. Hague Consulting Group, 1997. The focus In a study for the International Road transport Union (IRU), shippers and carriers in road transport in various countries were asked to identify the types of impediments they encountered during the last year of operation Findings Respondents in Italy and the UK said that traffic congestion was by far the most important impediment. For shippers and haulers in the Czech Republic and Poland, traffic congestion also ranked highest, but delays at borders were also identified as a major impediment. In France, however, strikes and blockades were identified as the impediment that caused the biggest time loss (note, however, that the interviews were collected shortly after a long French haulers’ strike in 1997). Follow-up in-depth interviews with some shippers identified that congestion leads not only to higher transport cost (because the trucks and drivers cannot be used for other transports), but also to disruptions in the production process (or emergency shipments) and higher inventory and distribution costs. A substantial share of the firms would have chosen different (more efficient) inventory sizes, depot structures and distribution systems if these impediments (especially congestion, but also border delays) had not existed. Some firms would even have changed their production location and production process, were it not for congestion. Conclusions The study provides insight in the choices of shippers and carriers that are influenced by transport time and transport quality.
Pagina 92
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Value of Dutch travel time savings in 1997 – volume 1. Hague Consulting Group, 1998. The focus This report for AVV focuses on the value of time for passenger travel (car and public transport, all purposes) in The Netherlands. Reliability is not included in the Stated Preference experiments or models, not is quality. Method SP surveys were carried out on 1997 and used to estimate discrete choice models. Findings The findings on the value of time are not only in this report. The main outcomes are also in a memo on recommended values (AVV, 1998) and in the manual for costbenefit analysis (Eijgenraam et al., 2000; see part II, Appendix F). Distinctions are made by travel purpose, transport mode ands income group. In the Dutch 1988/1990 VOT survey (HCG, 1990), a higher value of time was found for commuting and business travellers on motorways compared to urban traffic. This finding was confirmed in a 1997/1998 update of the survey (HCG, 1998), but not included in the recommended values (AVV, 1998). The 1988/1990 study also found that the VOT increases with decreasing speed. This result is in line with expectations -travel time becomes more unpleasant as traffic flows become congested (thus, if average speed decreases). The value of time for commuting on motorways with speeds below 90 km/h is 68% higher than the value of time in urban traffic. The corresponding percentage for business travel is 33%. For other travel purposes, the difference is not significant. Conclusion SP surveys in The Netherlands have produced different values of time for different road types and average speed classes.
Pagina 93
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Accessibility in the Netherlands. Hans Hilbers and Berry Blijie (The Netherlands Institute for Spatial Research/RPB), 2003. Focus The study presented in the paper focuses mainly on the reliability for automobile users on roadway networks. It is part of a research project with the concept of reliability of accessibility, reliability in the Netherlands and the relationship with spatial development as main topics. Definitions The reliability of a trip is defined by B&A groep (2000) as the chance that a trip is made with the pre-expected characteristics regarding travel time, travel costs, travel comfort and the safety of the trip. Two types of reliability can be defined: objective and subjective. Objective reliability is (an indicator of) the chance that a trip can be made with predefined time, costs and comfort. In this context not only the odds of a deviation is important, but the size of the deviation (variation) is as well. In their study, the objective reliability is regarded in terms of travel times by car. Subjective reliability is the perceived chance (experience) that a trip can be made within a certain time and with certain costs and comfort. Subjective reliability consists of two elements: 1. The reliability image: some travel modes have, justified or not, a reliable or unreliable image; 2. The experienced reliability: accounts for the match between de expectations of a trip and the realised course of it. The variation and the predictability of the trip are import factors of this element. The variation and its predictability are important factors of the reliability of trips. The following rules apply for the relation between variation and reliability: 1. When the variation is small, the reliability is large 2. When the variation and the predictability are large, the reliability is large 3. When the variation is large and the predictability is small, the reliability is small. This generates a need for incorporating uncertainty margins. Reliability and variation in travel time are important for two reasons, namely the consequences of unreliability (e.g. arriving late, missing a connection, longer travel time, waiting time) and the negative experience of unreliability itself, apart from the consequences. It is proven that a band of indifference exists around the average travel time. Only when the travel time exceeds this band, unreliability is being experienced. The literature speaks of a reliability ratio, which indicates how travellers respond to changes in reliability in comparison with changes in the average travel time. Noland and Polak indicated that on average this ration is about 1.3 (found in the U.S.), and somewhat higher for public transport. Travellers have several strategies to cope with unreliability. An overview is given by Bonsall (2000): Pagina 94
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
1. Minimising the variability; choose a different route or period of travel with more stable conditions, adapting speed. 2. Use the available knowledge optimally; choose known routes, leave at a familiar time. Increase knowledge; experiment, apply all available sources of information. 3. Minimise the consequences; build safety margins into the travel scheme, inform the destination by phone about the estimated time of arrival. 4. Make a game of it; challenge the unreliability, as it was a gambling game. The reliability of travel time has probably the biggest influence on freight transport, so even more than the average travel time (Allen et al., 1985). Allen et al. define reliability as the variance of the travel time. Based on a model, they conclude that unreliability is a more important part of the costs than travel time itself, due to keeping up a safety stock or stopping deliveries or production when the safety stock is depleted. The NCHRP (2001) states that (in the USA) non-recurring congestion is more important than recurring congestion. The business costs of congestion can be divided in four categories: 1. Higher transport costs. 2. Higher costs for logistics and Just-in-time processes caused by larger safety stocks and/or uncertainty in delivery. 3. Costs of scale, because of the smaller operation area. 4. Business costs of commuting, which is compensated through higher wages in the USA. The objective approach of reliability is more or less independent of the characteristics of a transport network user. It depends mostly on the time and place of the trip; time reflecting the period of day (e.g. peak or non peak hour) and place as the type of area and direction of the trip (e.g. urban or non urban and into or out of the city). In general this approach can be divided into two groups: theoretical and empirical. The theoretical studies of the reliability of accessibility focus mainly on the primary origin of travel time delay, namely road traffic congestion. The empirically orientated approach focuses more on the consequences of congestion: the travel time for links and corridors and their implications for the travellers. The empirical approach of determining the reliability of a transport network is mostly targeted on the operational aspects of a network. In this context, reliability can be defined as "the probability of a device performing its purpose adequately for the period of time intended under the stated operating systems". From the literature, the following definitions of reliability of transportation systems can be found (Shaw and Jackson, 2002): 1. The likelihood of a traveller's expectations being met. Reliability is measured as the variability between the expected travel time (based on scheduled or average travel time) and the actual travel time (due to the effects of non-recurrent congestion); 2. The range of travel times experienced during a large number of daily trips; and 3. The impact of non-recurrent congestion on the transport system, estimated as a function of the variation in the duration, the extent, and the intensity of traffic congestion on a system.
Pagina 95
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Aspects Regarding the definitions, the following major aspects of determining reliability can be distinguished: 1. Type of congestion: Since the traveller already reckons with recurrent congestion from previous trips, the congestion studied is usually non-recurrent. Of course, local or regional situations influence this expectation. For example, under bad weather conditions one can assume that travel times are longer than normal and during holidays, when commuting traffic flows will be smaller. 2. Travel times: A collection of travel times ought to provide enough information to perform statistic analysis and create insight into the reliability of the network. This means that, for a certain situation (e.g. commute in peak hour into the central business area), a sufficient amount of observations is needed. Ideally, the travel times used to calculate the reliability of transport systems are measured in "realtime". However, this is a very expensive and time intensive process. Generating travel time estimates based on the empirical (stochastic) chance on congestion could be an option as well. 3. The consequences for the traveller. If unreliability is noticed and measured, what does it imply for the traveller? This is an important issue, since the extra travel time, incorporated in a trip due to less reliable travel times, has a great influence on the activity pattern of users. It is also an important measure for the quality of a transport network Methods From the literature a wide range of methods for measuring reliability is available: Reliability measure Range of travel times (Shaw et al. 2000)
Clarification Range of travel times during a large number of daily trips, obtained by calculation of the mean and standard deviation. The range is based on a fixed benchmark, using the 85th percentile. This technique also involves a two-tail test, although a one-tail-test (the right tail) should be better: this is the area of unreliable travel times that are too long. Total Segment Difference in delay experienced on Delay From incident days versus non-incident days. Incidents The total delay is the amount of time (Lomax et al. 2001) lost due to congestion on a road segment in vehicle-minutes. Reliability defined as the probability Reliability that users will arrive at their destination Performance within the expected travel time. Indicator (R) (Ikharta & Mitchell 1997) California Reliability as variability between the Reliability Method expected travel time (based on (Booz-Allen & scheduled or average travel time) and Hamilton 1998) the actual travel time (due to the effects of non-recurrent congestion). A smaller standard deviation means a larger reliability. Percent variation Amount of variability in relation to the
Remarks Can be used for a variety of roadway systems, including roadways, corridors and area wide networks. Not very meaningful at itself though, since approximately 15 percent of the travel time observations would always be considered unreliable, regardless the number of observations. Thus, it can mainly be used to make comparisons of conditions along the same facility. Does not consider both recurrent and non-recurrent delay: the travel time on non-incident days may be far from acceptable. Theoretical in nature. Indicator ranges almost always between 0.9 - 1.1. R reflects the variation in travel times more than the acceptability of travel times to the user. Describes dispersion or variability of travel time. Does not really indicate how well conditions on corridors meet the travellers' expectations.
Percentage value is distance and time Pagina 96
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid (Lomax et al. 2001)
average travel rate: (standard deviation / average travel time). Florida Reliability Percent of travel on a corridor that takes no longer than the expected travel Model time plus a certain acceptable additio(Shaw et al. 2000) nal time. This acceptable travel time is the sum of the median travel time and a percentage of the median travel time. Reliability buffer index that estimates TTI Buffer Index (Lomax & Schrank the difference between average travel time and the 95th percentile travel time 2002) (equivalent for arriving one time per month too late). The extra time that has to be budgeted for a trip is compared to the average travel rate to define a reliability index. HOWLATE Model HOWLATE (Heuristic On-line Web(Wunderlich et al. Linked Arrival Time Estimator) that applies dynamic programming 2001) techniques to archived observed roadway travel times data to quantify the impact of regular ATIS (Advanced Traveller Information Systems) utilisation by urban travellers. Three reliability measures are distinguished, all derived from a proportion of simulated yoked trials: On-time, Justin-time and Rate of late shock. Misery index Measures the intensity of delay for only (Lomax et al. 2001) the worst trips. The average travel rate is subtracted from the upper 20 percent of travel rates to get the amount of time.
neutral. Combines variation and acceptability, but very difficult to define the acceptable percentage that represents "acceptable" delay. Close to the users' perception of reliability. Assumes only trips that are unreliable at the last five percentage trips and indirectly reports on the reliability. Operations research has shown that the percent of trips accomplished within the acceptable time is a more direct measure of experienced reliability. Shows that ATIS reduce in-vehicle time marginally, but realise more effective time management, improved on-time reliability and trip predictability. Nevertheless, the HOWLATE model is needed to calculate these measures.
Rare extreme travel times have less influence on the measure. An indication is given on how severe the late arrivals are. However, this measure gives no information about the reliability on time.
The Florida Reliability Model is the most suitable for facility based measures, and the HOWLATE Model for trip-based analysis. Conclusion This article gives a useful overview of the definitions and aspects of reliability and methods for measuring reliability at the transport network level. It does not provide values of reliability or quality.
Pagina 97
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Final report for publication, QUITS Quality Indicators for Transport Systems, ISIS et al, 1998. The focus This report covers all transport modes and deals with both passengers and goods transport and addresses the quality of transport from an internal and external dimension on inter urban European routes. Definition The internal quality dimension includes two macro-aspects; respectively travel time and direct costs. A comprehensive valuation framework of the global quality of the transport system is described; emphasizing the need of homogeneity of the units of measure, the result of such a comprehensive valuation is expressed in the form of a “total cost of the travel”, which is nothing but the straightforward sum of the costs of: 1. travel time (or value of the travel time) 2. direct costs, and 3. external costs. The external quality indicators that are included in this report are: • air pollution, • climate change, • traffic noise, • accidents (as far as they are not internalised by insurance premiums), • transport infrastructure costs (as far as they are not covered by charges) and • subsidies for infrastructure use. Findings The findings that are of interest in our study are given in table 1. The table illustrates the relative importance of the different components of travel time for passenger transport. Table 1: The valuation of the different components of travel time Disposition time Access-egress time Wait-search time Pure travel time Disposition time Access-egress time Wait-search time Pure travel time
Value of the travel time for business travel 8,5 ECU/PERSON-HOUR 17 ECU/PERSON-HOUR 21,8 ECU/PERSON-HOUR 17 ECU/PERSON-HOUR Value of the travel time for leisure travel 2,1 ECU/PERSON-HOUR 4,3 ECU/PERSON-HOUR 8,5 ECU/PERSON-HOUR 5,3 ECU/PERSON-HOUR
The findings for the external quality indicators are not particular interesting because of the purpose of our study. The perspective in our study is more on the user side, which makes the external quality indicators less important. Conclusion The report gives a clear description of internal and external quality indicators. However the chosen perspective of this report is not in line with the purpose of the study of quality and reliability. Pagina 98
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The Influence of Comfort on Public Transport Ridership. Börje Johansson (Chalmers University of Technology), 1998. Focus One of the objectives of the paper is to define the concepts of quality, standard, comfort and total comfort in public transport and how they relate to each other. Another objective is how comfort can be assessed. Definitions In this paper, public transport quality is defined as: "Public transport has (the right) quality when the public transport customer's perception of (products and) services equals or exceeds his/hers expectations" If the customer thinks the services are complying with his/hers expectations, then the customer perceives the service as having (the right) quality. If the services are below expectations, the quality is unacceptable and if services are over expectations, the quality is high. A customer comparing two services equal in all except quality, both having quality as or above expectation, but with one service having higher quality than the other, will choose the service with the higher quality. According to Zeithaml et al. (1990) there are four key-factors that influence customers' expectations: word-of-mouth communications, personal needs, past experience and external communications. The standard of a transportation system is defined by Holmberg (1977) as: the system's ability to satisfy a certain travel demand regarding the qualities that are demanded by the travellers. He divided standard into: accessibility, comfort and safety. Accessibility comprises travel time, frequency, number of transfers and reliability. Safety comprises traffic safety and safety against assaults. There are a number of definitions of comfort given in the paper. Common for all definitions is that they include that a traveller should reach some state that can be characterised as an "extreme value"; maximum of comfort, minimum of discomfort or maximum of positive and minimum of negative sense impressions. An operational definition could then be a definition of this "extreme value". Mayr (1959) defined comfort as follows: "...the sum total of all those measures which maintain and improve the well being of a person and reduce his fatigue" He divided a public transport journey into three stages; preliminary, main and concluding stage and assigned the concept travelling comfort to the comfort experienced during the whole journey. According to Mayr travelling comfort consists of: 1. Riding comfort: the comfort experienced in the vehicle itself; 2. Local comfort: the comfort experienced at stations, airports, interchange points, etc.; Pagina 99
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
3. Organisational comfort: a further group of factors which have a considerable effect on travelling comfort, such as good connections, frequency of service, reliability in terms of time and place, independence of weather and climate, legal obligation to provide transport facilities, customs clearance, etc. Total comfort is introduced in this paper and defined as: "The overall outcome of the passenger's comparison of offered and expected (products and) services during the total journey (door-to-door)." Aspects That comfort is an important quality factor was a starting-point for the paper. An overview of how several aspects of comfort can be assessed is given. Holmberg (1977) showed that comfort was the second most important quality factor for preferring car to public transport. Also Bullard and Christiansen (1981) tried to assess the importance of comfort compared with other quality factors. They defined a level-ofservice (LOS) concept consisting of eight indicators. The authors used a weighting system for the LOS indicators. Passenger comfort was given the weighting factor 20. This was higher than any of the other seven LOS indicators. LOS indicators 1 – Accessibility 2 - Travel time 3 - Directness of service (Transfers) 4 – Delay 5 - Frequency of service 6 – Reliability 7 - Passenger density 8 - Passenger comfort A) Acceleration B) Temperature C) Noise
Weight 10 10 10 5 15 15 15 5 10 5
The following aspects of comfort are described: Riding comfort: 1. Seating place: Johansson (1990) showed that having a seat is the most important comfort factor in all kinds of vehicles. The need to be seated increases in importance with the journey length, but it is great even in local traffic. 2. Room for standing passengers: Widlert et al. (1989) found that bus travellers in Stockholm valued the inconvenience of standing in buses at 0.50-1.40 SEK per trip plus 4-9 SEK per hour. Standing in the subway with a few other people standing were valued at 0.892.06 SEK per trip plus 2.50-5 SEK per hour. In a crowded subway, the valuation raised to 1.31-3.05 SEK per trip plus 3.50-6 SEK per hour. The figures were also influenced by the length of the trip; standing (not crowded) less than 10 minutes was valued at 1-2 SEK per hour, more than 10 minutes at 3-6 SEK per hour. Pagina 100
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
3. Acceleration/retardation: Limits due to comfort could be put on acceleration and retardation based on Johansson (1992a and 1992b). If e.g. at least half the travellers should be satisfied (=grade 50), this restricted the retardation when braking to 2.14 m/s² for seated passengers and to 1.25 for standing. For lateral acceleration when driving through switches the maximum was 0.84 m/s² when seated and 0.40 when standing. 4. Jerks: Limits due to comfort could also be put on longitudinal and lateral jerks (changes in acceleration level) based on Johansson (1992a and 1992b). For the grade 50, the highest jerk allowed when braking was then 2.14 m/s³ for seated and 0.96 for standing. The lateral jerk when driving through switches should not exceed 0.81 m/s³ for seated passengers and 0.22 for standing. 5. Vibrations: Vibrations with frequencies between 0.5 and 80 Hz are important for comfort, but 4-8 Hz is the area where man is most sensitive to vertical vibrations. Horizontal vibrations are most noticeable between 0.5 and 2 Hz (Peters et al. 1992). The standard of "Evaluation of human exposure to whole-body vibration" (ISO 2631) was set in 1974, but has been criticised mainly for being based on insufficient data. A proposal for revision of the ISO 2631 standard has the following classification of vibrations: Vibration levels < 0.315 m/s2 0.315 - 0.63 m/s2 0.5 - 1.0 m/s2 0.8 - 1.6 m/s2 1.25 - 2.5 m/s2 > 2.0 m/s2
Perception No discomfort Slightly uncomfortable Rather uncomfortable Uncomfortable Very uncomfortable Extremely uncomfortable
With regard to railways, a measuring number (Wz, Wertungsziffer) is used for the riding quality of railway cars. The Wz gives a measure of the vibrations a traveller is exposed to when riding in a certain car at a certain speed on a certain track. The basis is acceleration amplitudes measured in the car at the speed studied. The Wz is based on a study by Helberg and Sperling (1941). Wz ranges from one to five, where one means very good ride quality and five very bad. Modern passenger cars on good tracks rate below three. 6. Luggage: Johansson (1990) showed that the factor "luggage" increased in importance with the length of the journey, reflecting the same relation for the amount of luggage. There are indications that public transport is not considered as a good mode of transport for journeys where luggage is involved, e.g. shopping journeys. Williams (1977) found that railway travellers have a strong preference for close, personal supervision of their luggage at all times, and that they favour retention of present handling systems, especially if they were improved. 7. Noise: Bryan (1976) dealt with noise in passenger cars. He gave the following criteria for noise:
Pagina 101
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Subjective rating Quiet Noticeable Intrusive Annoying Very annoying
Noise level not exceeding (dB(A)) 67 73 79 85 91
The Swedish Road and Traffic Research Institute (1981) recommended that the noise level at the driver's seat should not exceed 75 dB(A) and the level of the infrasound should not exceed 110 dB(IL). The infrasound level should however not exceed the noise level by more than 35 dB. The project mentioned above has been continued in two stages. In the second (Peters et al 1992), the recommendations were tightened up to 72 dB(A). The noise should be measured at three places, two of them in the passengers’ compartment. 8. Thermal climate: Temming (1981) gave recommendations for the inner temperature at varying outdoor temperatures. Kjeldstad (1978) recommended for seated passengers with outdoor clothes 17 - 21 °C in winter and 25 - 29 °C in summer. McFarland (1969) stated that the air velocity in the vehicle should be between 0.1 and 0.3 m/s. 9. Lighting: There are conflicting demands on lighting in vehicles. Some travellers want good lighting at the seat, others, especially on longer journeys, want it dark to be able to rest. A security requirement is sufficient lighting in stairs, aisle etc. 10. Surrounding: In Johansson (1990), the factor "surrounding" had a medium rank. This can be interpreted as that the passengers considered the factor as fairly important but well cared for. 11. Information during the journey: TFD (1982) found that most of the interviewed in a survey in two Swedish "counties” were satisfied with the information given outside or inside the bus. There was however a widespread disapproval on one point; the announcement of the stops. 12. Personnel: Johansson (1990) showed that "personnel" were important only on long-distance journeys. 13. Traffic safety: In Johansson (1990), traffic safety was placed rather far down the ranking-list, probably because it was considered good. 14. Service: Johansson (1990) showed that "service" were important only on long-distance journeys. "Service" was more important for train passengers than coach, which can be explained by differences in the length of the journey but also by differences in values between train and coach passengers. Boarding/alighting comfort: Brattgård and Petzäll (1982) made experiments that showed that the steps should have a height of 150 - 200 mm. The regulations issued by The Swedish Board of Transport (1989) were mainly based on these results. Local Comfort Pagina 102
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
1. Waiting time: The travellers' assessment of the different components of the travel time clearly show that changing vehicles and waiting time for an interchange is perceived as very uncomfortable (Elmberg and Quarmby 1981). It should be noted that it is the subjective, not the real waiting time that determines the travellers' degree of discomfort. In Nantes, a system was installed that showed real time left to the arrival of the bus to the stop. This reduced the travellers’ overrating of their waiting time from 77 to 12 percent over real time (Pettersson 1985). 2. Environment: Torell (1986) studied the travellers' opinions of six bus terminals in the Gothenburg area. Slightly more that 60 percent of the travellers perceived the terminals as pleasant and comfortable. For a terminal to be considered comfortable it should be kept clean and there should be plantations, painted surfaces and enough seats. A full half of the travellers thought the individual stops were well designed, although it was stated that there were not enough seats at the stops. 3. Information: According to Torell (1986), 80 percent of the travellers stated that the information at the stops was good. Timetables were considered to be the clearly most important information source. Other necessary or valuable information at terminals that were mentioned were information of destinations of the buses, direction, marking of the stops, clocks and information about alterations and delays. Conclusion This article is useful in answering the research questions on how to define quality, and especially comfort, in (urban) public transport.
Pagina 103
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
A model for time of day and mode choice using error components logit. De Jong, Daly, Pieters, Vellay, Bradley and Hofman, 2003. The focus In this project for AVV, RAND Europe has designed a Stated Preference (SP) survey for persons travelling by car or train in the (extended) peak periods in The Netherlands. The objective of this research project was to gain insight into the factors that influence departure time choice and to develop a new time of day module for the Dutch National Model System (LMS). Method The population from which respondents were recruited consists of persons travelling in the extended peak periods (6.00-11.00 and 15.00-19.00 hours during working days) as car drivers or train passengers within The Netherlands. Respondents were recruited for participation in the actual stated preference survey from an existing panel or from short recruitment interviews at train stations and at a petrol station beside a motorway. The estimation sample contains information on more than 1,000 travellers. The attributes presented include: 1. departure time from home; 2. arrival time at destination; 3. departure time from destination; 4. arrival time at home; 5. tour travel time; 6. duration of stay at destination; 7. travel cost not including (extra) peak charge; 8. peak charge (second experiment only); 9. probability of a seat (train only); and 10. frequency (train only). Simultaneous error components logit models for time of day choice for the outward trip and activity duration were estimated, with penalties for departing earlier or later than preferred and for shorter or longer than preferred activity duration (following the scheduling theory as developed by Vickrey, 1969 and Small, 1982). Findings For commuting, most of the ratios of the schedule delay penalty coefficients, both for too early and too late, to travel time are between 1 and 1.5; half an hour earlier or later at work gives the same disutility as 30-45 minutes travel time. In the previous 1989 time of day stated preference survey in The Netherlands, these ratios were generally between 0.5 and 1 for commuting. Time of day shifting appears to be less sensitive now, perhaps because many travellers have already shifted to less preferred time of day periods in response to increasing congestion. The disutility from arriving early is now very similar to that of being late. The above discussion referred to the outward leg. For the participation time decision, working too long or too short is generally preferred to an equivalent amount of travel time. In the estimated models for business travel and other purposes, arriving 30 minutes too late or too early at the destination is also valued to be worse than 30 minutes of travel time. For education tours, the opposite is found. Longer than preferred activity Pagina 104
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
participation time is generally valued to be less important than an equivalent amount of travel time. The variable for seat availability in public transport was significant and has the right sign only for commuting. After applying the Jackknife method to get proper t-ratios, this variable was no longer significant. The frequency variable for public transport was not even significant before applying the Jackknife method (RAND Europe, 2001). Conclusion The paper includes models that permit to calculate the travel time equivalent (in minutes) of departing/arriving early or late (outward leg of the tour) and of shorter or longer activity duration (return leg of the tour). The ratios for the outward leg are above one (one minute schedule delay is more important than one minute travel time) for most purposes.
Pagina 105
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
A joint SP/RP model of freight shipments from the region Nord-Pas-de-Calais. De Jong, Vellay and Houée, 2001. The focus In this project for the French Ministry of Transport (MELTT), RAND Europe has designed a Stated Preference (SP) survey for shippers in the region Nord-Pas de Calais and estimated discrete choice models on the resulting SP data and on Revealed Preference (RP) data from a shippers’ survey. The objective of this research project was to gain insight into the factors that influence the mode choice of shippers. Definition Reliability is measured in the SP as the percentage of shipments not delivered on time. Aspects Aspects of quality that were included are availability of logistic services, flexibility and frequency. Method The RP survey was carried out in the region Nord-Pas de Calais by INRETS in 1998 and contains interviews with 215 shippers focussing on 3 recent shipments. A subsample of these were re-contacted for the SP survey. The SP interviews were carried out by ISL in 2000 among 98 shippers. These SP interviews included experiments on one or two shipments. The SP contains information on 149 shipments in total, by road own account transport, road hire and reward, train, combined transport and maritime transport. Both within-mode and between-mode SP experiments were included. Disaggregate logit models have been estimated, for the within-mode and the between mode SP experiments, the RP data as well as on the joint SP/RP data. Findings According to the estimation results on the ‘between-mode’ SP data, the following shipment variables (besides attributes of the shipments and the shipper) are important for the choice of mode in freight transport: 1. Transport cost 2. Transport time 3. Probability of delay 4. Availability of logistic services suitable to the commodity 5. Capability of quick reaction to unforeseen requests 6. Frequency. Conclusion Reliability (probability of delay) and several quality aspects were found to be important for mode choice in freight transport in North-West France.
Pagina 106
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The value of time and reliability: measurement from a value pricing experiment. Terence C. Lam, Kenneth A. Small, 2001. The focus The article provides measurements on the values of time and reliability from 1998 data on actual behaviour of commuters on State Route 91 in Orange County, California, where there is a choice between a free and variably tolled route. The value of reliability The value of reliability (VOR) has received little attention in the literature. Virtually all the work on it has used data related to hypothetical scenarios, for two reasons: measuring the variability of travel times facing actual travellers is difficult, and traveltime variability is highly correlated with mean travel time. Definition In the article the reliability of travel time is defined as the 90th percentile of travel time minus the median travel time in minutes. Data In the paper, the value of time (VOT) and value of reliability (VOR) are measured using data on actual travel behaviour in a real pricing context. People face a choice between two parallel routes, one free but congested and the other with time-varying tolls (on the State Route 91 in California). There are two main difficulties with estimating VOT and VOR from such data. First, the main variable of interest – differences across the two routes in time, reliability and cost – vary across time of day and days of week, but in a highly correlated manner by design. Second, the survey responses must be supplemented by other data in order to accurately measure travel time and its uncertainty. The latter difficulty is solved by using data laboriously extracted from loop detectors placed by the California Department of Transportation in both the free and the express lanes. The first difficulty is covered in several ways: 1. by measuring travel time for relatively narrow (15 min) time-of-day intervals 2. by including work trips that occur during off-peak periods, so that additional independent variation is obtained. 3. the measurements reveal that mean travel time and variability in travel time are only imperfectly correlated across time-of-day intervals. 4. carpooling introduces additional variation into the cost per person of the toll, because of a discount that is given (on the SR-91) to a car with more then three persons. 5. VOT and VOR depend on certain measurable socio-economic and travel characteristics, by specifying interactions between travel variables and these characteristics additional independent variation in the variables entering the model can be obtained. The results obtained are that many model specifications fit well and result in statistically significant coefficients on all three of the key travel variables. Thereby credible estimates of VOT and VOR, which vary in plausible ways with traveler characteristics, are found. Pagina 107
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Method The method consists of a combination of stated preference and revealed preference research. The sample contains 162 observations from the 1997 survey plus 371 from the 1998 survey. The loop detector data record traffic volumes and vehicle density on each lane at every 30 seconds. From this information a standard engineering algorithm is used to estimate the average travel times on both the free lanes and the tolled lanes, for either 5 or 15 min time-of-day intervals. There are four models specifications described in the article: 1. Route choice only 2. Route and time-of day choice 3. Route and mode choice 4. Transponder choice (4a + 4b) The basic model rests on the assumption that traveller n chooses route i (i=1,2) by maximizing the following conditional indirect utility function: where t, v, and c are the measures of travel time, variability in travel time and cost, respectively (just a single measure of each in a given specification); x is a vector of observable socio-economic or other characteristics (including time of day and car occupancy), assumed exogenous; and ε is a random utility component whose distribution is extreme value (i.e. double-exponential). These assumptions lead to a binomial logit model of choice of route. The values of time and reliability are defined as:
These quantities could in principle depend on route i, but the specifications in the article they do not, so they are omitted the i subscript. Findings The model described above is the basic model specification. In the four other models the assumptions are relaxed. In table 2 the results of interest are summarized. Table 1: Implied values of travel time and reliability.
With route choice alone, the value of the median travel-time is about $19/h, or 61% of the sample average wage rate. This applies to congested travel, for which the value probably has a higher value than for uncongested time. The VOR, defined as the 90th percentile travel-time minus the median, is 38% of this average wage for men and 91% for women. Including time of day as one of the endogenous decisions, as in Model (2d), greatly reduces the estimates of VOT and VOR. Unfortunately, the accuracy of these estimates Pagina 108
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
is doubtful because of the heroic assumptions that were made to compute how the travel times vary across time-of-day alternatives. The other models show that most of the results are reasonably robust to how the simultaneous decisions about mode and transponder choice are handled. Accounting for mode choices raises VOT by about 28%, with little effect on VOR. Accounting explicitly for transponder choice reveals that the transponder installation decision has its own determinants, distinct from those of the daily decision of whether or not to use the transponder; but accounting for this does not affect VOT and VOR very much. The authors regard model (4d), which accounts explicitly for both transponder and mode choice, as the most trustworthy of those presented. This model produces a VOT of $22.87per hour and VOR of $15.12 per hour for men and $31.91 per hour for women, all from a sample with weighted average wage rate equal to $31.69 per hour. All models show interesting and mostly plausible variations in the propensities for various choices with respect to personal characteristics. In particular, several factors are brought to light by the unusual opportunity to observe route choice when one route is subject to time-of-day pricing. Income, gender and language seem especially to affect the willingness to undertake the fixed cost of installing a transponder, whereas workhour flexibility and total trip distance seem to influence the daily decision of which route to take. Conclusion This article is very interesting in light of the study of quality and reliability in transportation. It provides an indicator for reliability in terms of costs and shows the effect of several factors on the value of this indicator.
Pagina 109
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
A handbook for measuring customer satisfaction and service quality. Morface Int. and Cambridge Systematics, 1999. The focus The primary focus of this handbook is how to measure customer satisfaction and how to develop transit agency performance measures in response to research findings. These are key elements of an ongoing customer satisfaction monitoring process. The handbook proposes a new "impact score" or problems encountered approach. This approach determines the relative impact of service attributes on overall satisfaction, when a recent problem with the attribute is reported. Since the primary way transit agencies can improve customers' overall satisfaction with service is to reduce customers' problematic experiences, the goal is to identify those attributes, which have the greatest negative impact on overall satisfaction and the greatest number of customers encountering a problem. These "driver attributes" can be identified and prioritized in a three step process. Large sample and sub sample sizes, and multivariate analysis techniques, are not required. The determinants of quality of determinants The authors suggest that, within most service industries, consumers use basically similar criteria in evaluating service quality. These criteria seem to fall into 10 key categories labelled "service quality determinants". These determinants are listed below. Overlap among the 10 determinants may exist.
Pagina 110
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table1: Determinants of service quality 1 2 3 4 5 6
7 8 9 10
RELIABILITY involves consistency of performance and dependability. RESPONSIVENESS concerns the willingness or readiness of employees to provide service. It also involves timeliness of service. COMPETENCE means possession of the required skills and knowledge to perform the service. ACCESS involves approachability and ease of contact. COURTESY involves politeness, respect, consideration, and friendliness of contact personnel. COMMUNICATION means keeping customers informed in language they can understand and listening to them. It may mean that the company has to adjust its language for different consumers — increasing the level of sophistication with a welleducated customer and speaking simply and plainly with a novice. CREDIBILITY involves trustworthiness, believability, and honesty. It involves having the customer's best interests at heart. SECURITY is the freedom from danger, risk, or doubt. UNDERSTANDING/KNOWING THE CUSTOMER involves making the effort to understand the customer's needs. TANGIBLES include the physical environment and representations of the service.
Research in other service industries indicates consumers "group" a wide array of attributes of service under one of the 10 dimensions noted when judging service quality. However, this research is preliminary and also suggests that it is advisable to determine, within the industry of study, whether identifiable service quality segments exist — and whether, and in what ways, consumer expectations differ across industry segments. Investigating how transit customers aggregate attributes of service into collapsed quality dimensions is important to understanding how customer satisfaction should be measured within an industry. Method The Impact Score approach determines the relative impact of attributes on overall satisfaction, by measuring customers' relative decreases in overall satisfaction, when a recent problem with an attribute is reported. This makes sense because, within the delivery of quality service framework, the primary way transit agencies can improve customers' overall satisfaction with service is to reduce customers' problematic experience with those attributes, which have the greatest negative impact on overall satisfaction. These driver attributes can be identified and prioritized in a three-step process. Step One is to determine which attributes have the most impact on overall customer satisfaction. For each attribute, the sample is divided into those respondents who have had a recent problem with the attribute and those respondents who have not recently experienced a problem with the attribute. (Those who have not experienced the attribute within the past 30 days are grouped with those who have, but have not had a problem.) The mean overall satisfaction ratings of the two groups are compared. The difference between the two mean overall satisfaction ratings is called the "gap score". Gap scores are computed and the attributes are then ordered by the size of their gap scores. A t-test can be used to determine where statistical significance lies among gap scores. The magnitude of an attribute's gap score should not change significantly over Pagina 111
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
time. The relationship between a service quality attribute and overall satisfaction with transit service can be assumed to be structural. That is, once it is determined that an attribute is a driver of customer satisfaction it will probably remain so, unless significant societal changes occur, i.e., graffiti comes to be viewed as an art form. Step Two lists the attribute problem incidence rate for each attribute in a column next to its gap score. (The percent of customers who experienced a problem with the service attribute within the past 30 days). It will be important to take into account the rate at which a problem with an attribute occurs within the customer base. It may be that a particular attribute has a large gap score (and thereby a significant impact on overall satisfaction), but the percent of customers reporting a problem with the attribute is relatively small. In this case, it probably is not worth a transit agency's time and expense to attempt to further lower the problem occurrence rate for the attribute. On the other hand, if an attribute's gap score (impact on overall satisfaction) is moderately low, while the rate at which customers experience a problem with the attribute is high, the effect of the attribute on overall satisfaction is magnified and will require attention. Whether future increases or decreases in problem incidence rates are statistically significant can be validated by statistical tests (e.g., chi-square test, z-test of proportions, etc.). Step Three creates a composite index by multiplying the attribute's overall satisfaction gap score by the attribute's problem incidence rate. The result is an attribute "impact score". The attributes are then placed in descending order of their impact scores. The top attributes are the drivers of customer satisfaction. To summarize, impact scores are computed as shown in the following example: Table 2: Impact Score Approach
The impact score data analysis can be implemented using just a spreadsheet program. The spreadsheet can be structured so that the relevant inputs reside in one worksheet, the data analysis is conducted in a second worksheet, and the results summarized in a third worksheet. Inputs from the survey can be fed into simple formulas to determine mean ratings by group, gap values, percentages of respondents who had a problem with transit service, impact scores and t-tests to determine the statistical significance of identified differences. If this data analysis system is constructed in the benchmark year, transit agencies can input their own tracking data (from on-board surveys) during subsequent years.
Pagina 112
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Findings In the handbook, the results for the customer satisfaction of four rail lines are described. In the following tables (3,4,5,6) the ranking in order of importance of the different attributes are summarized. Table 3 and 4 show the computed impact score for two railway lines and table 5 and 6 provide insight in the impact scores for two bus services. Table 3: CTA Red line service (N=300)
Table 4: CTA Blue line service (N=302)
Table 5: Sun tran (N=303)
Pagina 113
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table 6: GLTC (N=63)
Conclusion The report explains how customer satisfaction indices can be measured and computed in an easy way. Further, the report ranks different attributes of service quality based on their relative importance to the overall customer satisfaction. In relation to our study, especially the different attributes of service quality are interesting. This report gives the most detailed division of service quality of transport in the literature known at this point. The relevance of this level of detail should be judged, because the study does not provide indicators that are useful in a cost-benefit analysis.
Pagina 114
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Etude de l’impact des phénomènes d’irregularité des autobus – Analyse des resultants. MVA Report, mars 2000. Focus The main focus is on developing an operational methodology to represent the impact of irregularity inherent in travel by bus (variation in arrival time, in-vehicle travel time). The study derives empirical valuations for these factors using Stated Preference (SP) experiments, and develops a methodology to apply the results in a demand forecasting model system. Definition Irregularity has been represented, based upon a literature study and qualitative research, by the following variables: 1. 2. 3. 4. 5.
Amount of waiting time at bus stop; Irregularity of waiting time at bus stop; Amount of in-vehicle travel time in bus; Irregularity of in-vehicle travel time in bus; Comfort while travelling in the bus.
The representation of the phenomenon of irregularity has been based upon qualitative research. Irregularity was presented by showing 5 different times, representing different levels of variation around the given mean amount of time. For instance: Mean waiting time (minutes) 7 Possible waiting times (minutes) 2 5 7 9 12 For the estimation of the valuation of irregularity two coefficients were retained (after testing many different specifications): a coefficient for mean time and a coefficient for standard-deviation of waiting time. Aspects As indicated above the study addressed two aspects of irregularity: waiting time and invehicle travel time. The study did not look at any consequences of irregularity for follow-up interchanges to other public transport, for those passengers that had to transfer when they had stepped out of the bus. Method The study used literature search and qualitative research to guide the representation of the phenomenon of irregularity in the SP experiments. In order to obtain valuations for all five variables mentioned above three-stated preference “games” were carried out including the following variables: 1. Game 1: mean waiting time and standard deviation of waiting time; 2. Game 2: mean waiting time, mean in-vehicle travel time, comfort in bus; 3. Game 3: mean in-vehicle travel time and standard variation of in-vehicle travel time. Pagina 115
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The SP experiments were presented as simple pair-wise comparisons between two situations, where 10 comparisons were offered to the respondent for each game. In total 309 passengers were interviewed. The analysis of the completed questionnaires was carried out using ALOGIT to obtain utility weights for the entire sample and for different segments of passengers. All coefficients could be estimated with acceptable accuracy for the entire sample and for almost all segments. The research also investigated how the results could be applied. By using results from previous RATP research and theoretical waiting time modelling (based upon the assumption of random arrival of passengers at the bus stop) the report describes how the average passenger waiting time and the standard deviation of the waiting time can be computed when the mean bus interval and its standard deviation are known for a given bus service, together with the mean total travel time and its standard deviation. Finally the report describes three types of application of the results: 1. Modelling to evaluate the impact of transport policy measures on demand; 2. Modelling for detailed local impact analyses; 3. Project evaluation: estimation of the consumer benefits obtained from specific measurements to improve bus regularity. Findings The main findings in terms of valuations of travel time (total population) are as follows: Variable (minutes) Average waiting time In-vehicle travel time seated In-vehicle travel time standing In-vehicle travel time standing in crowded bus Standard deviation waiting time Standard deviation in-vehicle travel time
Relative valuation 1.00 0.50 0.63 0.92 0.48 0.12
Results for segments of passengers are also available. Using the above table effects of measures to improve the regularity of buses can be quantified by expressing them in generalised time. The generalised time effects, in turn, can be input to the demand-forecasting model to estimate the impact on patronage. Conclusion This is a very comprehensive and extremely useful report, providing many of the elements (both methodology and results) required for the STIF Punctuality Research. Key difference relative to the suburban trains is the fact that the study investigated buses operating at frequencies up to 10-12 minutes, where the majority of the passengers showed (according to the authors) behaviour of random arrival, as opposed to the more time-table oriented arrival pattern that can be expected for the suburban trains. Pagina 116
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Travel time variability: a review of theoretical and empirical issues Robert B. Noland and John W. Polak, 2002. The focus The research reviews both the theory and empirical results of several projects that estimated coefficients on various measures of variability of travel time using stated preference techniques. The main issue is the behavioural reactions to travel time reliability. Definition Travel time variability in the research review is defined as day-to-day variability in travel times. It is important to recognize that this definition is independent of congestion effects. That is, a congested system may exhibit very stable day-to-day travel times that travellers anticipate in advance. Variability introduces uncertainty for travellers such that they do not know exactly when they will arrive at a destination. . Methods to measure travel time variability and the assumptions Scheduling models (Small 1982) Scheduling decisions are a key component of the behavioural reaction associated with the choice of departure time. Small (1982) originally specified a model of scheduling choice as follows: U=αT+β SDE+γ(SDL)+θDL where utility, U, is a function of travel time, T, schedule delay-early, SDE, schedule delay-late, SDL, and a fixed penalty for any late arrival, DL. SDE is defined as the amount of time one arrives at a destination earlier than desired, while SDL is the amount one arrives later. Noland and Small (1995) decompose the costs associated with travel time variance into the following components at the optimal departure time, to derive the optimal expected cost:
where Tf is the free flow travel time, Tx is extra travel delay due to recurrent congestion, b is the mean travel time due to non-recurrent congestion (also equal to the variance in an exponential distribution), and ∆ is the change in the profile of recurrent congestion. PL* is the optimal probability of arriving late, which is equal to:
In most cases, the scheduling cost formulation captures the behavioural reactions of travellers. This hinges to a large extent on the value of the coefficient associated with the probability of late arrival and the time varying nature of recurrent congestion. Pagina 117
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Travel time variability with fixed service intervals Public transport services are provided at a fixed interval. This results in the selection of departure times in discrete intervals and implies that a traveller is very unlikely to be lucky enough to match a given service with their preferred arrival time (though, in theory, preferred arrival times may change in response to schedule constraints). This means that most travellers utilizing public transport will arrive either earlier or later than desired, irrespective of variability. Bates et al. (2000) add additional terms to the model of Noland and Small (1995) to include the quantities ASE and ASL, defined as adherence to schedule, early and late respectively. AS is defined as the absolute difference between actual and scheduled arrival time:
where Ts is the scheduled travel time, dependent upon the departure time th. Then the expected utility of travel will be:
where the terms subscripted with `s’ represent costs associated with adherence to schedule and PLs is the probability of not adhering to schedule. The article concludes, that behavioural changes due to variability in public transport are highly dependent on the structure of the network and the alternative choices available to travellers. Limitations and recommendations Empirical estimations of how travellers react to changes in travel time variability have mirrored the two theoretical approaches discussed previously. Early approaches focused on trading off the value of travel time with a measure of travel time variance, such as the standard deviation of travel time. More recent studies have attempted to measure the scheduling costs associated with variability. The evidence is clear that there is a fundamental link between the response to travel time variability and scheduling behaviour. However, very little is known about how travellers perceive reliability. In particular, to what extent do subjective probability distributions exist, and what relationship might these have to objective distributions. The authors of the article could not identify any research that analysed this question. A clearer understanding of the behavioural response to reliability changes is essential for improving cost-benefit assessment of transport projects. Inclusion of reliability effects in a generalized cost function, whether through a reliability ratio or scheduling effects is likely to increase the benefits of many transport projects, since such projects are likely to reduce not only the mean, but also the variability in travel time.
Pagina 118
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Conclusion This article provides answers to the research questions of the study to quality and reliability of transportation. The research provides insight in the difference between public transport and individual transportation in terms of methods to measure travel time variability. Further the research presents evidence that clearer understanding of the behavioural response to reliability changes is essential for improving cost-benefit assessment of transport projects.
Pagina 119
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Hoe laat denk je thuis te zijn? Een onderzoek naar de betrouwbaarheid van vervoersystemen en de invloed daarvan op het verplaatsingsgedrag (= What time do you think you will be home? A study to the reliability of transportation systems and its influence on the travel behaviour). P. Peeters, P. Rietveld, F. Bruinsma, D. van Vuuren and A.J. Rooijers 1998. The focus The focus is on personal travel, mainly on public transport (chains) in the Netherlands. The aspect of quality of transport that is emphasized is reliability of travel time. Definition In the report, a distinction between a journey and a trip is made. A trip is defined as a movement, where only one mode of transportation is used. A journey, or a transport chain, consists of one or more trips and forms the movement between the origin and the destination, where location-related activities are based. The distinction is important when one is considering the perspective. The provider of the transport service is concerned with the reliability of the trip and the traveller is concerned with the reliability of the total travel movement. The Reliability of a transportation system is defined as the probability of reaching a chosen destination given certain characteristics. The characteristics are reliability of travel time, reliability of comfort and reliability of costs. Reliability of travel time is the variation in travel time and the variation in departure time and arrival time. The reliability of comfort includes information on trains and busses, quality of the seats, the seat availability, the quality of the road and information for road-users. The reliability of costs contains the unexpected costs associated with parking, toll, traffic tickets, theft, damage and so on. Another distinction that is made in the report is between in subjective and objective reliability. The objective reliability is the objective probability of reaching a destination according to the objective norms of travel time, costs and comfort. The subjective reliability of a transportation system is perceived probability (by a traveller or a service provider) of reaching a destination according to the anticipated (by a traveller or a service provider) travel time, costs and comfort levels. Method The report describes the assessment of objective reliability and of the subjective reliability for public transport. First, the objective reliability assessment is summarized. The models for the objective reliability assessment study are based on the data provided by the National Travel Survey (OVG, 1994) of Statistics Netherlands. The sample contains 300 public transport origin-destination relations. 5 categories for level of urbanisation and 4 categories for distance bands are specified, for each combination 15 trips are drawn leading to a total sample size of 300 relations. Then, the observed distribution of the data is estimated by the distribution functions; the gamma, the log-normal and the Weibull distribution. This is done for the following Pagina 120
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
variables deviation from the departure time, deviation from the travel time and deviation from the arrival time for different times-of-day and modes of transportation. The findings for the distribution functions are summarised in the next three tables. Table 1: Deviation from the departure time according to the timetable (in minutes) Mode of transport City bus - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour Tram - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour Metro Regional transport Slow train - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour Intercity train - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour
Distribution
Average
Variation
Normal (0.2337, 2.2142) Normal (0.2666, 2.6768) Normal (0.8572, 2.7463)
0.23 0.27 0.86
4.9 7.2 7.5
Normal (0.8572, 2.6547) Normal (1.1166, 2.6489) Normal (0.7672, 2.0667) Exponential (2) Lognormal (2.4813, 0.2047, 0.1697)
0.86 1.12 077 0.50 0.91
7.0 7.0 4.3 0.3 0.5
Lognormal (0.9665, 0.7315, 0.3938) Lognormal (0.8390, 0.6749, 0.2144) Lognormal (0.6936, 0.5365, 0.0627)
2.29 1.65 0.87
2.4 1.7 0.8
Lognormal (0.9388, 0.7213, 0.2616) Lognormal (0.9185, 0.7187, 0.1980) Lognormal (0.7594, 0.5182, 0.1113)
2.06 1.95 0.92
2.3 2.2 0.8
Average -0.20 0.42
Variation 11.5 2.9
Distribution
Average
Variation
Normal (-0.8349, 2.4158) Normal (-07715, 2.9106) Normal (0.1236, 2.8175)
-0.83 -0.77 0.12
5.8 8.5 7.9
Normal (0.2521, 2.5145) Normal (0.5202, 2.5309) Normal (0.1688, 2.21228)
0.25 0.52 0.17
6.3 6.4 4.5
Lognormal (2.7347, 0.2557, -0.7440)
0.74
1.1
Lognormal (1.1518, 0.4934, 0.7047) Lognormal (1.5292, 0.4242, 0.4169) Lognormal (1.2975, 0.3688, 0.1087)
2.28 1.16 0.83
1.4 1.0 0.6
Lognormal (1.5373, 0.5014, 0.1895) Lognormal (1.6936, 0.4448, 0.1572) Lognormal (1.7951, 0.2629, 0.0567)
1.85 1.70 0.67
1.5 1.3 0.4
Table 2: Deviation from the travel time (in minutes) Mode of transport Regional transport Metro
Distribution Gamma (0.4403, 2.2381, -2.4665) Lognormal (0.2021, 0.9593, -1.0364)
Table 3: Deviation from arrival time Mode of transport City bus - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour Tram - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour Metro Regional transport Slow train - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour Intercity train - Peak 7.30 hour - Off-peak 13.00 hour - Sunday 11.00 hour
A big limitation of the model is that it assumes that the measured delays are randomly distributed over all trips, however in reality delays are not independent. This especially important in rail transport, one delayed train will cause delays for other trains. Pagina 121
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The findings of the simulations show two interesting observations. First, the travel time does not vary a lot between the different time-of-day (table 4). Second, the travel time variation depends to a large extent on the number of changes between modes of transportation (table 5). The numbers for the variation do not sum up till 100%, because the figures for public transport chains arriving early are not included. Table 4: Arrival times in peak, off-peak periods and Sunday. Travel time data Travel time simulation Additional travel time simulation Arrival on time or to early 1 til 5 minutes too late 6 til 30 minutes too late > 30 minutes too late
Min Min % % % % %
Peak 59.9 65.9 10.0 43 30 22 5
Off-peak 61.8 67.9 9.9 42 30 22 6
Sunday 65.1 72.5 11.4 45 31 16 8
Table 5: Travel time and probability of delay for different public transport chains Chain
Peak on-time
+ 5 min
B 64.5 3.0 B/b 48.1 17.9 T/b 46.9 28.7 B/t 28.0 24.9 B/t/b 33.0 36.6 B/t/t/b 38.9 31.8 Ta/t/b 17.0 59.8 B/t/ta 33.3 33.9 Total 42.6 27.1 b = Bus, t = train, ta= tram
+30 min 0.0 2.4 5.1 3.7 8.3 6.1 3.1 0.2 5.1
Off-peak on-time + 5 min 56.8 46.5 47.5 39.1 28.8 34.0 27.6 30.5 41.8
6.7 20.3 19.3 27.7 41.4 39.4 52.2 37.5 28.5
+30 min 0.1 4.0 4.0 6.1 12.1 12.7 4.1 6.1
Sunday on-time
+ 5 min
62.0 46.0 47.2 49.1 37.9 35.6 25.2 34.1 45.2
5.6 20.1 14.6 18.0 31.1 31.2 48.2 30.4 23.6
+30 min 0.0 6.4 3.9 2.1 9.9 14.2 25.6 12.1 8.0
The assessment of the subjective reliability consists of an in-depth interview study and a survey of different types of travellers. The research underlines the importance of the reliability of travel time and comfort of road transport, cycling and public transport with as purpose commuting between home and workplace. The results of both studies show that the perceived reliability of public transport is rated the lowest and the perceived reliability of cycling is the highest. Road transport is rated between these two. The valuations vary greatly between the different users. The reliability is perceived the highest by regular users, than by incidental users and is perceived the lowest by the people who never use a certain mode of transport. The research presents evidence that there is little influence of changing reliability on the mode choice. The majority of the respondents will not change their travel behaviour and those who do, change their time of departure or route. The difference between the types of users is again significant. If there is influence of change in reliability on the travel behaviour, this is the largest for incidental users of a mode of transport. However it was not possible to quantify the behavioural effects based on the data that is available.
Pagina 122
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Conclusion The research is an exploratory study and therefore provides insight in methods that could be used to assess the quality and reliability aspects of transport. The findings are preliminary and have not been validated in a way that firm conclusions could be drawn from the findings according to the writers.
Pagina 123
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Quality – It is not what you think. Andrew Probert, University of Cambridge, 2001. Focus The focus of this paper is on the quality of passenger transport. Definitions In this article it is stated that the conventional engineering meaning of quality, the ability of a product to perform as specified, is not sufficient. There are two reasons for this: transport is a service and is different every time it is performed, and passengers are applying values in their judgements of products and services. Thus, quality of transport service is its closeness to the requirements specified by the customer, or the end user. Finding transport quality requires a switch in emphasis. What matters are the factors that are important to each customer, not the factors that are important to the provider or a third party. Aspects To determine the aspects of transport quality brainstorm sessions and a Q-sort technique were used, which resulted in 10 domains: 1. mode functional attributes; 2. comfort; 3. time; 4. safety; 5. control; 6. fun and enjoyment; 7. status; 8. cost and value; 9. health and fitness; 10. convenience. Methods In order to measure these domains, SERVQUAL, a questionnaire-based technique usually used in marketing to test the quality of an organisation’s service quality, was used. In this research the task of the respondents (230 surveys in the UK covering 8 modes and 6 journey types) was to evaluate a specified past journey which has recently been undertaken using a specific mode. The 10 domains have been broken down into 50 attributes, and the respondents were asked to scale what level of these attributes they wanted for the journey, what level they would have tolerated, and what level of service they actually got. The difference between the desired and tolerated performance in a SERVQUAL survey is called the Zone of Tolerance. Complex mixtures of circumstances, social, physical and psychological factors, influence these levels, but the levels can be measured (indirectly) without resort to the reasons. Any product’s level of service must fall within the Zone of Tolerance for it to be purchased voluntarily. How far up the zone a mode is scored is a reflection of its performance against expectation and should give an indication of quality. Results Factor analyses of the exploratory data resulted in a significant regrouping and rejection of attributes from the original grouping (as described under “Aspects”). The results formed 8 significant factors, grouped into 3 groups: Pagina 124
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
1. Benefits: • Personal well-being (includes comfort and security); • Social well-being (includes status of the mode and fellow passengers); • Fun and enjoyment (the positive journey experience); • Concurrent occupation (include working, eating, sleeping, talking etc.). 2. Logistics: • Logistical assistance (e.g. availability of modes, information); • Logistical resistance (e.g. queues, problems, stops, long waits). 3. Costs: • Exchange costs (input by the traveller such as effort, time, money); • Consequential costs (negative consequences such as unpleased experiences, damage to the environment, danger tot others). The empirical research showed that the Zone of Tolerance in transport varied not only according to the type of journey wanted but also in anticipation of the mode of transport to be used (i.e. people raised and lowered their expectations according to the mode). Assessment of modes as well allowed a direct comparison of mode supply with Zones of Tolerance, showing the journey quality for specific journeys. Aeroplanes, for example, had a high level of requirement, and a high level of performance, whereas metros had low expectation yet barely adequate performance. Taxis had a very narrow gap between desire and acceptability levels, suggesting they are seen as very functional modes, with very little variation in tolerance. Incidentally, it has been observed that modes with physical similarities have similar service provisions and quality characteristics. To improve operating quality the operator must provide missing elements. The need is to find out what customers want, compare this with the current operation and then establish a new set of performance indicators. Conclusion This article is useful in answering the research questions on how to define and measure quality in passenger transport.
Pagina 125
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Project Benutten en Bouwen 2003-2015. Projectorganisatie Benutten en Bouwen, 2002. Focus The report focuses solely on the reliability of rail transport. The aspect of reliability within the scope is the reliability of a trip from station to station. This means that the reliability of the total travel time lies outside the scope of the study. The interchange and the trip to and from the railway station are not included in the reliability assessment. Definition of reliability The reliability of rail transport is the probability that products and services are delivered by the railway carrier according to the agreed quality: For personal transport this means no more then 3 % of the total amount of trains should be late. For freight transport this depends on the agreements made with the customer about the reliability. Data The data in this report are the irregularities documented in the system called ‘Geeltje’. For every irregularity the following elements are reported: 1. The place of the irregularity 2. The number of primary delay minutes 3. The number of secondary delay minutes Method The authors suggest queuing theory to illustrate the improvements in the performance of the rail network. This model was finished in the time available for the project. Conclusion The report is not very useful for our study of quality and reliability. It only looks at a very small aspect of reliability. Further, queuing theory described to assess the reliability of rail transport contains little insights of relevance.
Pagina 126
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Hoofdonderzoek reistijdwaardering in het vervoer van goederen over de weg. RAND Europe, 2003. The focus This study is an update of an investigation from 1992 (Hague Consulting Group, 1992), in which the segmentation has been adapted to reflect the increased importance of container transport It was carried out for AVV Transport Research Centre to establish monetary values for transport time and reliability in goods transport by road in the Netherlands. The new values are derived for use in cost-benefit analyses of infrastructure projects, that often have transport time savings as one of the major benefits. A similar study for freight transport by rail, inland waterways, sea and air will be completed in the spring of 2004. Definitions Reliability was measured as the percentage of deliveries that do not arrive at the specified time or do not arrive within the specified time interval. The observed amount of delay was registered as well. Other quality aspects included in the SP were probability of damage to the goods during transport (in pro mille) and frequency. Method The study consists of revealed preference (RP) and stated preference (SP) interviews among shippers and carriers in road freight transport. On the basis of these interviews, discrete choice models were estimated. These models provide trade-off ratios between transport time and transport costs and between reliability and transport costs of time, which in combination with the factor costs give the monetary values of transport time and reliability. Trade-off ratios for probability of damage and frequency are provided as well. Findings The analysis yields that a 10%4 change in reliability (measured as the percentage not delivered on time), e.g. an increase from 10% not on time to 11%, is equivalent to the following costs: • 0.67 Euro per transport per hour for low value raw materials and semi-finished goods; • 0.87 Euro per transport per hour for high value raw materials and semi-finished goods; • 1.78 Euro per transport per hour for final products with loss of value; • 1.67 Euro per transport per hour for final products without loss of value; • 1.90 Euro per transport per hour for containers; • 1.15 to 1.20 Euro per transport per hour for total freight transport by road. In the User Manual on the Value of Time in Freight Transport, produced as part of the same study, recommendations can be found for applying the values of time and reliability obtained in this study in the evaluation of transport projects.
4
The percentage used here (10% change) is an arbitrary example. Pagina 127
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
A 10% increase in the probability of damage was equivalent to a 2-5% increase in transport cost. For frequency, a 10% increase was equivalent to a 2% decrease in transport costs. Conclusion The study has provided monetary values not only for travel time in freight transport by road, but also for reliability. Probability of damage and frequency were also included in the models. Values for the other freight transport modes will follow in 2004.
Pagina 128
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Valuation of travel time and traveller information in multimodal personal travel under uncertainty. P. Rietveld, 2003. Focus This paper focuses on public transport and especially on multimodal travel. The relation between the supply of traveller information and the valuation of travel time is explained. Definitions The author claims that there are three ways in which policies can address reliability problems. First, at the supply side, infrastructure (road, rail) and public transport services may be extended and improved. Second, demand management may take place via pricing and other measures in order to reduce demand at times and places in the network. A third way is the provision of traveller information with the aim to reduce the negative effects of reliability problems. The paper emphasizes the role of uncertainty in travel behaviour. Consider a traveller who can choose among n alternatives, for example modes or routes. The utility (or generalised costs) of the alternatives depend on criteria such as the monetary costs, travel time and other comfort aspects. These features are denoted as X1,….., Xj. Uncertainty means that the traveller does not know the exact values of the choice characteristics. They are subject to variation. Method The consequences of introducing uncertainty in decision making of travellers who consider a public transport trip are described for various settings. The author assumes that a traveller considers a public transport trip consisting of 7 elements, but the approach would also apply to simpler or more complex trips: 1. walking to bus stop 2. waiting at bus stop 3. bus ride 4. walking from bus stop to railway station 5. waiting at railway station 6. train ride 7. walking to final destination. Six alternative specifications of uncertainty in transport and the implications for the marginal value of travel timesavings are considered. The various dimensions are indicated in table 1. Table 1: Alternative specifications of uncertainty Case
Certainty / Constant / variable mvots* Uncertainty 1 Certainty Constant 2 Uncertainty Constant 3 Certainty Variable 4 Uncertainty Variable 5 Uncertainty Constant 6 Uncertainty Constant * mvots: marginal value of travel time savings.
Risk attitude: expected utility vs risk aversion Not applied Expected utility Not applied Expected utility Risk aversion Expected utility
Pagina 129
Scheduling costs No No No No No Yes
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
The exact specification of the six alternatives is given in the report, in this summary the most interesting findings are provided. Findings In the paper it is demonstrated that uncertainty aspects play a role in actual behaviour but when these uncertainties are ignored in value of time estimations, a tendency may be expected that values of travel time are overestimated. The consequences of the provision of information on the travel behaviour are threefold. First, better information enables the traveller to adjust their choices so that generalised costs are lower compared to the situation without this information. Second, even when information provision does not lead to the choice of another alternative, it may be useful to know an adjusted travel time forecast so that people one wants to meet can be warned. This is especially relevant in the case of scheduling costs (case 6 mentioned in the previous section). Third, even when the above two factors do not play a role, travellers may appreciate information on the projected time. The author concludes that for these three reasons there is a willingness to pay for this information. The author further argues that the most important step forward in coping with uncertainty in public transport is to provide precise information. A traveller needs information about the probability that the interchange at B for his trip from A to C will run smoothly, or that he will miss the connecting bus or train. It is not very useful to know that 85 % of the trains arrives with a delay of less than 5 minutes. Conclusion The paper shows the influence of uncertainty on the travel behaviour. This is interesting because it illustrates the importance of this aspect of quality of transport. Unfortunately, no values for this indicator are reported. The results in the paper prove that there is an influence of uncertainty on the travel behaviour but do no provide quantitative insights.
Pagina 130
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Coping with unreliability in public transport chains: A case study for the Netherlands. P. Rietveld, F.R. Bruinsma and D.J. van Vuuren, 2001. Focus The article addresses the reliability of public transport from a multimodal perspective, implying a shift of attention towards the customer. The aim of the article is to analyse the reliability of arrival times in public transport with particular emphasis on delays due to missing connections. The article is the English version of the article: “Reliability of public transport chains” (Bruinsma, et al 1999, which was translated out of Dutch for the study into quality and reliability of transport). The article mentioned above and this article, are both based on the study: “What time do you think you will be home? A study to the reliability of transportation systems and its influence on the travel behaviour”. The summary of this study is included in our study. For this reason, only the findings that are not documented in the other two summaries are described. The data, the method and the definitions are exactly the same as documented in the summaries of the article mentioned earlier. Findings The findings of interest in this article are indicator values based on a questionnaire distributed among 781 respondents. Four attributes of quality are valued: scheduled travel time, probability of delay and probability of getting a seat. The valuations are given in table 1 Table 1: Estimation of valuation of travel time, reliability and probability of getting a seat in Dutch guilders
The estimate of reliability means that public transport travellers are prepared to pay an additional amount of Dfl 4.81 in order to reduce a 50 % probability of a 15 minutes delay to zero. This is a rather high extra cost given the reference value (price of a ticket) for the trip is Dfl 12.50. In order to make the in-vehicle time and reliability estimates comparable, the authors recomputed them for a standard time unit of 1 minute at the bottom of table 1, assuming that valuations are proportional to the length of the time units concerned. The valuation of a certain in-vehicle time loss of 1 min is 27 cents, whereas the valuation of 50% probability of a 2 min delay is 64 cents. Thus, in transport an “uncertainty minute” is weighted as a factor 64/27 = 2.4 higher than a certain minute. In the case of a riskneutral traveller, these values would be equal. The conclusion is that the respondents show a substantial dislike for unreliable transport services: are risk-averse.
Pagina 131
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Conclusion The results described in the section ‘findings’ are of interest to the study into quality and reliability of transport because it illustrates the importance of the topic very well.
Pagina 132
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
‘Trunk roads and the generation of traffic’, Standing Advisory Committee on Trunk Road Assessment (SACTRA), 1994. The focus The SACTRA report of 1994, prepared for the UK Department for Transport, focussed on the question whether new roads can generate additional traffic. The report includes empirical evidence on this issue from some 150 highway studies. Quality of transport and reliability are not studied in any detail. Findings: recommendations for cost-benefit analysis The SACTRA report concluded that for highway networks that are close to saturation or are expected to be at saturation in future years the likely amount of induced/suppressed traffic needs to be assessed. SACTRA distinguished eight elements of induced/suppressed traffic: 1. Route change 2. Journey re-timing 3. Travel to new destinations (trip redistribution) 4. Switch to another transport mode 5. Change vehicle occupancy 6. Travel more frequently (generated traffic) 7. Make entirely new journeys (also generated traffic) 8. Change in the pattern of land use and vehicle ownership. In the appraisal of new highway schemes, consideration should be given to some or all of the above elements. SACTRA recommended using transport models with a more detailed demand representation and/or elastic assignments instead of the often used fixed matrix approach (which only includes changes of route). Conclusion The SACTRA report of 1994 has had important implications on project evaluation in the UK, and the main lessons for cost-benefit analysis (see above) are also valid for The Netherlands. However, this report does not provide relevant and recent material on the importance of reliability and quality of transport.
Pagina 133
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Transport and the economy: full report. The Standing Advisory Committee on Trunk Road Assessment Department for Transport (SACTRA), London, 1999. Focus One of the questions addressed by the report is: “Are economic impacts fully captured in the procedures for estimating benefits and costs currently used by the Department of the Environment, Transport and the Regions?” In the report the (in-)completeness of transport appraisal in the UK is described in a broad sense. Quality of transport and reliability are not studied in any detail. Findings Travel time Generalised cost varies by mode and is usually a linear combination of the various components of a journey. For cars, generalised cost is a combination of: 1. in-vehicle travel time; 2. operating costs (related to distance travelled); 3. parking 'costs' (which notionally include time spent searching and queuing for a space and walking to the final destination); and 4. tolls or congestion charges. For goods vehicles, the components are similar, except that different vehicle operating costs and values of time are used. For public transport users, generalised cost is a combination of: 1. walking time from the origin to a stop or station (usually weighted relative to invehicle time by a factor of about two); 2. waiting time for the service (again, usually weighted relative to in-vehicle time by a factor of about two); 3. fare; 4. in-vehicle time; 5. penalty representing the inconvenience of changing between services; and 6. walking time to the destination (again, usually weighted relative to in-vehicle time by a factor of about two). Money costs are usually converted to time units using a value of time. Valuation of changes in travel time reliability It is clear that a range of policy interventions in the road network - for example, variable speed limits and other traffic calming measures, provision of traveller information, reallocation of road space to give priority to buses - are designed to reduce travel time variability and thereby achieve greater dependability and certainty of arrival time. Whereas it has been reasonable in the past to assume that the travel time savings calculated for road schemes were correlated with improved reliability, a range of policies may yield reliability benefits with little or no average travel time savings. Currently, the reliability benefits are incorporated into the appraisal by means of crude proxy indicators such as changes in highway stress levels and qualitative assessment of their likely significance. Improvement of the valuation of reliability benefits implies deriving the values of the variance of travel times as well as the mean. Pagina 134
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
It is likely that reliability benefits are of particular significance for commerce and industry. Journeys which are undertaken with scheduled arrival times mean that the schedule must be drawn up, not in relation to the average journey time, but in relation to what can be achieved on a high proportion - say 90, 95 or 99 per cent - of occasions. The Department thinks that, on the basis of research previously undertaken for it, "by ignoring travel time variability, the economic benefits of (trunk road) schemes are likely to be underestimated by 5% to 50%". There is also the question of the impact of unreliability on the magnitude of stock inventories held by firms. This raises even more difficult questions for appraisal, such as the value of the resources consumed by the extra inventory held to cover unreliability in the transport system. Modelling reliability Current work is focused on the increases in unreliability which are caused by incidents. SACTRA is concerned that unreliability stems from other causes as well as from incidents. For example, as traffic flows increase, so smoothly flowing traffic conditions break down and queues propagate quite quickly upstream from the initial point of overload. Also, certain kinds of traffic reduction mechanism will have quite different effects in terms of reliability. For example, measures to reduce or reallocate road capacity seem likely to increase unreliability for certain classes of vehicle (those not accorded priority) while improving reliability for other classes (those receiving priority). Measures which price demand off the network would have the opposite effect as they would generally reduce congestion levels. We note that traffic reduction measures are generally considered where conditions are congested and where unreliability is likely to be a key issue. See for example modelling outcomes for congestion charging in London. Table 1. Economic benefits of 1991 Inbound Cordon Charging for Central London (£m pa) Source: May, Coombe and Gilliam (1996) £4 Charge
£8 Charge
Car Time Reliability Costs Total
65.4 19.5 -114.9 -30
89.2 26.1 -242.7 -127.4
Public transport Time Reliability Costs Total
13.8 0.2 0 14
20.9 0.4 0 21.3
Freight & Taxis Time Reliability
74 23
99.1 30.5 Pagina 135
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid Costs Total Travellers total
-109.9 -12.9 -29
-217.9 -88.3 -194.5
Non-travellers Congestion charge Other Non-travellers total
246.6 -79.2 167.4
437.1 -109.5 327.6
Accidents Net economic Benefits
20.3 158.7
26.1 159.2
May, Coombe and Gilliam (1996) examined the potential incidence of monetised impacts. The general conclusion was that while both commercial vehicles and continuing car-users would obtain benefits through reduced travel times and improved reliability (how this is measured is not explained in this report), travellers would suffer net disbenefits as the payment of congestion charges would offset the benefits of reduced congestion. Nevertheless, there would be overall net economic benefits due to the generation of significant amounts of revenue. Recommendations The SACTRA report concluded that the calculation of the transport effects needs to take account of all sources of transport costs, and all the important direct and indirect behavioural responses of individuals and firms, in the short and long run, to changes in those costs. This is necessary in order to make an accurate assessment of the resulting pattern and conditions of travel and environmental effects. These conditions are not usually fulfilled: in practice, simplified assumptions are used which leave out some important responses. SACTRA therefore makes recommendations on the importance of improving conventional appraisal methods, with a special focus on the modelling and forecasting of freight movement and travel for business purposes whose connection with economic activity is more direct. They also comment on improvements in the treatment of travel time and reliability. Calculation of the transport impacts will need improved treatment of the patterns of behavioural responses of individual travellers and of companies (such as structural changes in land-use, production and employment) since these may themselves have further impacts on the volume and pattern of traffic, and the resulting costs. Conclusion This report contains general recommendations and lessons for cost-benefit analysis / transport appraisal that are also valid for The Netherlands. However, it does provide only little relevant and recent material on the importance of reliability and quality of transport.
Pagina 136
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Risk of delays, uncertainty and travellers’ valuation of travel time variability. Senna, 1991. Focus The focus in this article is on travel time variability in general, not on one specific mode of transportation. Definition The author assumes that uncertainty in travel times can be compared to a lottery. In this case individuals win when their actual travel time is less than the expected travel time. The expected value is the sum of the outcomes, each multiplied by its probability of occurrence. E(t) = pt1 + (1-p)t2 Where t1 and t2 are different travel times. The utility of the expected value is given by table 1, depending on the character of the individual (risk neutral, risk averse and risk prone). A traveller is said to be risk neutral if s/he equates the expected value and the utility of the expected value if U [pt1 + (1-p)t2] = pU(t1)+ (1-p)U(t2) The general utility function is specified according to following form: U =α tβ The value of β would be one for risk-neutral individuals. This is the most commonly used function in studies of choice behaviour. The individuals surveyed by Senna (from Porto Alegre in Brazil) appear to be risk prone. Table 1 Risk neutral Risk aversion Risk proneness
U [pt1 + (1-p)t2] = pU(t1)+ (1-p)U(t2) U [pt1 + (1-p)t2] < pU(t1)+ (1-p)U(t2) U [pt1 + (1-p)t2] > pU(t1)+ (1-p)U(t2)
β =1 >1 <1
Method The mean-standard deviation approach is the usual empirical approach for valuing variability. The objective of each traveller is to: Minimize E(tp) + τV(t) p ℮ Pab where τ is a parameter that measures the influence of variance in travel time; tp is the expected travel time for each origin-destination pair; V(t) is the variance of travel time; and Pab is the set of all paths from origin a to destination b.
Pagina 137
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
It is interesting to note that there exists a relationship between the expected utility approach and the mean-standard deviation approach. Applying the expectation operator to equation (U =α tβ ) leads to: E(U) =αE( tβ) If then is considered that tβ = tβ/2 * tβ/2, it is tempting to take the expectation and get: E(tβ)= E(tβ/2 * tβ/2) However, one property of the expectation operator when X and Y are not independent is that E(X Y) = E(X)*E(Y) + Cov(X,Y) Where Cov (X,Y) = rx,y*σx*σ y In the present case X=Y= tβ/2, which implies the following: rx,y = 1, σx*σ y = [σ(X)] 2, E(X)*E(Y) = [E(X)] 2 or [E(tβ/2)] 2 This leads to the following formulation: E(U) =αE( tβ) = α[E(tβ/2)] 2 + [σ (tβ/2)] 2 Note that this function is not separately in two terms involving mean and variance as seems to be typically estimated in practice. Data The data used in the paper is gathered through a stated preference experiment carried out in Porto Alegre, Brazil. The survey contains 301 individuals. Findings The model used to estimate parameters is the latter function including the cost variable. E(U) =αE( tβ) = α[E(tβ/2)] 2 + [σ (tβ/2)] 2 + τ Cost The values of time V(E(t)) are defined by the marginal cost of time, which means they can be calculated as the derivative of the utility function with respect to time divided by the derivative of the utility function with respect to cost. The definition of the value of variability is V(σ(t)) similar to the definition of the value of time. The value of time function is given by V[E(t)] = 2α*E(tβ/2)/ τ And the value of variability function is given by V[σ(t)] = 2α*σ (tβ/2)/ τ Pagina 138
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
This gives the following outcomes from the survey. Table 2: The valuation for travel time and travel time variability Travel time (min) 15 20 30 45 60
Value of (US$ / hr) 2.25 2.06 1.82 1.61 1.48
time
Value of variability (US$ / hr) Low Medium 3.62 5.50 4.34 6.69 4.92 7.53 5.36 8.11 5.79 8.84
High 7.39 8.84 10.00 10.87 11.74
The author concludes from the results in table 2 that the benefits of reducing one unit of variability (standard deviation) appears to be much higher than the benefits of reducing on unit of mean travel time. Conclusion The paper shows how an equation for estimating the value of travel time variability can be derived, which is explicitly based on a (general) utility function. A general utility function has been estimated, based on data obtained using SP techniques. The function allows for risk-aversion, risk-proneness or risk-neutrality. The author concludes that the results provide strong arguments to consider the inclusion of the benefits of reducing variability in travel time in the generalised cost function. This paper is of great relevance to the study into quality and reliability of transport, because it provides a generic method to valuate variability of travel times.
Pagina 139
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Evidence on subjective factor values, Transek, 2002 The focus This presentation prepared in the EXPEDITE project, carried out for the European Commission DGTREN, provides values (in money or travel time) for some quality aspects in transport, based mainly on Swedish studies. Findings Values obtained in the UK or The Netherlands already described in other parts of this study (the reliability factor of 1.3 found in the UK, Rietveld et al (2001) on public transport in The Netherlands) are quoted. In a Swedish Stated Preference (SP) study by Widell and others, the value of time when standing in the bus was found to be 40% higher than when seated. For subway and commuter train the value of time standing was 50% (subway) or 60% (commuter train) higher then when seated. Conclusion This presentation gives values of time for standing in public transport relative to the value of time seated, based on Swedish SP data.
Pagina 140
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
A review of British evidence on time and service quality valuations. Mark Wardman, 2001a. Definition of quality The research reports on an extensive review of the valuations of a wide range of service quality attributes based on a very large amount of empirical evidence. These service quality attributes include in-vehicle time, walk time, wait time, service headway, interchange departure time adjustments, search time, late time and time spent in congested traffic conditions. Methods to measure quality and reliability aspects and the assumptions Valuations of walking and waiting time and of other service quality attributes are often expressed in units of in-vehicle-time. In this research descriptive statistics for overall values of time and service quality expressed in units of in-vehicle time is presented (table 1). Furthermore, the valuations in the money values according to socio-economic characteristics are considered in detail. Table 1
Obs: Observation
A regression model explains the variation in the service quality indicators. The factors that might explain the variations that are included are GDP, retail prices, sample size, journey distance, the type of data upon which the model was estimated, journey purpose, choice context, mode, numeraire, location, the omission of non-traders and use of logic checks and the purpose of the studies. In short, the review is based on evidence drawn from British studies conducted between 1980 and 1996 where behavioural choice models were estimated of either revealed or stated preference data. Limitations and recommendations Research of this type is essentially an aggregate approach, providing insights into general relationships between average valuations and relevant socio-economic variables. However the model is able comprehend that the valuations of walk time, wait time and the service quality attributes are not constant and will vary across different situations. The maximum level of disaggregation of the values collected in this review was according to journey purpose and mode. Further analysis of the effect of income on monetary values is mentioned as very important. Segmenting by income group would
Pagina 141
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
allow analysis of both cross-sectional and inter-temporal variations in monetary valuations. The review found evidence that the values of out-of-vehicle time, headway and departure time shifts depend on the levels of these variables over time. Given that there has been relatively little research on how the value of in-vehicle time varies with the level of in-vehicle time, and even less on how the values of the other attributes depend on their levels. Further research in this area is strongly recommended. Conclusion This research offers insight in the relative importance of the different service quality attributes. These insights are based on a vast body of evidence presented in previous research projects in the UK.
Pagina 142
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Public Transport Values of Time. Mark Wardman 2001b The focus The paper aims to provide recommended valuations of public transport in-vehicle time (IVT), walk time and wait time/headway with appropriate modifiers according to key factors such as: mode user type, the mode to which the value relates, journey distance and inter urban or urban context and journey purpose. Definition This paper deals with values of walk time, access time, wait time and headway, in addition to the value of IVT. The meaning and interpretation of the values of walk time are straightforward. Walk time covers time spent walking to and from the main mode of the journey which is primarily a public transport mode but can be car. Time spent walking as a mode in its own right is not covered. The interpretation of the values is clear since in the data assembled for the purposes of this study the authors have to the greatest extent possible separated walk time from other aspects of out-of-vehicle time. The data that are analysed also contain values of access to and egress from public transport modes by other vehicular modes, and values which represent varying mixtures of the latter and walking. These are defined as access time. Headway represents the interval between public transport services and is a measure of how frequent the services are. Waiting time occurs either prior to the arrival of the first vehicle or during the course of a journey when interchange is required. The majority of wait time values are of the former type in the meta-analysis data set. Literature The paper reviews the literature on valuation of time components. In this section, the literature that gives insight in the relative value of the different component is summarised. Table 1. Walk and wait time values for North Kent commuters
Pagina 143
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table 2. Relationship between values of in-vehicle, walking, waiting time and transfer times.
Table 3. Public transport components of journey time
These three tables illustrate the relative importance of the different components of travel time, which is important for the study of quality and reliability of transport. Method The variables, about which information is collected, are either continuous or categorical. The form of model used to explain variations in the monetary values (V) takes the form:
where there are n continuous variables (Xi) and p categorical variables having q categories (Zjk). q-1 is specified as dummy variables for a categorical variable of q Pagina 144
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
levels and their coefficient estimates are interpreted relative to the arbitrarily omitted level. A logarithmic transformation of equation 1 allows the estimation of the parameters by ordinary least squares. The αi coefficients are interpreted as elasticity’s, denoting the proportionate effect on the valuation after a proportionate change in Xi. The exponential of βjk denotes the proportionate effect on the valuation of a level of a categorical variable relative to its omitted level. The τ term is a scale factor which applies to all of the values. Its absolute value will depend upon the scale used for the continuous variables such as GDP per capita and distance, although of course the scale used will not affect the elasticity’s estimated to the continuous variables or the output of ‘forecast’ values of time. Findings Table 4 provides the implied money values of in-vehicle-time (IVT) for a range of user types, modes and distances. Absolute values in pence per minute and 2000 quarter 3 prices are given as well as ratios of these values to car users’ values of car IVT. Table 4. Implied money values of IVT
Car users’ values of car are higher than for train and generally lower than for bus. Although car time does become more highly valued than bus time, this only occurs at long distances where in fact we have very few observations for bus travel. The author states that he was unable to test whether there is any positive incremental effect on the distance elasticity for bus journeys over long distances. The distance and journey purpose effects are readily apparent as are the low values of bus users and the high values of rail users. The figures are in stark contrast to currently recommended values in that they exhibit a considerable amount of variation. Pagina 145
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Table 5 presents the implied IVT values of walk and wait time. The values do not differ by journey purpose, but they will differ by distance, since the distance elasticity is lower for walk and wait than for IVT, and they will also vary by user type. The numeraire is the value of IVT for the same mode as user type. Hence the rail values reported (headed RAIL RAIL in the table) are rail users’ money values of walk and wait time divided by rail users’ values of rail IVT. Table 5. Implied IVT values of walk and wait time
UG: underground
The results for walk and wait time are strongly dependent upon the weight we attach to the RP evidence. The values in Table 2 are based on the RP evidence. The author states that if one relied solely on the SP evidence, the walk time values would all be 32% lower and the wait time values would all be 59% lower. The most noticeable feature of the IVT values of walk and wait is that they vary considerably. In part this is because of differences in the money value of IVT by user type and mode, but there are other strong influences at work. The increase in the IVT values of walk and wait time as the levels of walk and wait time (denoted in the first column) increase is quite clear, as is the fall in the values as distance increases. For corresponding levels of walk and wait time and the same journey distance, the values of wait time tend to be greater than the value of walk time. This is consistent with the Pagina 146
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
review of past evidence in the literature section. The figures do, however, suggest that the value of walk is more centered around the convention of twice the value of IVT than is the value of wait. Table 6 provides the implied IVT values of headway across distance and purpose which are the factors which influence it. The strong distance effect is very apparent, with the headway valuation being much higher for shorter distance trips. Table 6. Implied IVT values of headway
EB business travel
The results illustrate the value of the different travel time component relative to each other. Conclusion The article is valuable to get insight in valuation of different components of travel time. These components are aspects quality of transport and therefore important in the study of quality and reliability of transport.
Pagina 147
RAPPORT 04003 maart 2004
Waardering van kwaliteit en betrouwbaarheid
Pagina 148
RAPPORT 04003 maart 2004