Time valuation in traffic

Transcription

1 Time valuation in traffic Congestion costs, value of time & lost vehicle hours. RA-MOW K. Van Raemdonck, C. Macharis Onderzoekslijn Evaluatietechnieken DIEPENBEEK, STEUNPUNT MOBILITEIT & OPENBARE WERKEN SPOOR VERKEERSVEILIGHEID

2 Documentbeschrijving Rapportnummer: Titel: RA-MOW Time valuation in traffic Ondertitel: Congestion costs, value of time & lost vehicle hours. Auteur(s): K. Van Raemdonck, C. Macharis Promotor: Prof. dr. Cathy Macharis Onderzoekslijn: Evaluatietechnieken Partner: Vrije Universiteit Brussel MOSI T Aantal pagina s: 35 Projectnummer Steunpunt: Projectinhoud: Evaluatietechnieken Uitgave: Steunpunt Mobiliteit & Openbare Werken, december Steunpunt Mobiliteit & Openbare Werken Wetenschapspark 5 B 3590 Diepenbeek T F E [email protected] I

3 Samenvatting In dit rapport wordt een methode voorgesteld om de congestiekosten ten gevolge van een ongeval te berekenen. Alle kosten die verbonden zijn met een ongeval worden overlopen. Het gaat enerzijds over slachtoffergebonden kosten zoals medische kosten, kosten door productieverlies en immateriële kosten. Anderzijds zijn er ook de ongevalgebonden kosten, zijnde de materiële kosten, afhandelingskosten en de congestiekosten. Bij deze laatste blijkt dat enerzijds de voertuigverliesuren een belangrijke component zijn, langs de andere kant is ook de waarde van reistijd van fundamenteel belang voor de berekening ervan. Er wordt dan ook dieper ingegaan op de waarde van tijd. Het probleem van de variatie in deze reistijdwaardering wordt aangekaart en de verschillen tussen de waarderingen voor goederenvervoer, woon-werkverkeer, zakelijk verkeer en overig verkeer komen aan bod. Deze verschillen uiten zich in het feit dat goederenvervoer en zakelijk verkeer (verkeer tijdens betaalde uren) een hogere waarde van reistijd hebben dan woon-werkverkeer of overig verkeer zoals verkeer voor boodschappen of recreatie. Vervolgens wordt er nog dieper ingegaan op de voertuigverliesuren. Het bleek dat er voor België of Vlaanderen nog geen volledig redundant systeem met gesloten wegsegmenten en dubbele inductielussen om snelheden en intensiteiten te meten bestaat (de uitbouw hiervan is echter wel gepland tegen 2012). Hierdoor is ook de betrouwbaarheid van de gegevens niet altijd even groot, waardoor er door middel van een literatuurstudie wat dieper ingegaan is op de verschillende methoden voor het berekenen van voertuigverliesuren. Hieruit werd dan, op basis van de voordelen van de verschillende methoden, een nieuw framework opgesteld. Op basis van de leerpunten uit dit rapport kan dan in verder onderzoek, na verzameling van de nodige gegevens, een meer betrouwbare berekening van de voertuigverliesuren uitgevoerd kan worden. Tenslotte worden de congestiekosten besproken. Deze worden opgesplitst in kosten door uitwijkgedrag, kosten door de onbetrouwbaarheid van reistijden en kosten door gemiddelde extra reistijd. Onder uitwijkgedrag verstaan we onder andere het vertrekken op een ander tijdstip, met een andere transportmodus reizen, een andere route kiezen, of zelfs helemaal van de verplaatsing afzien. Van onbetrouwbaarheid van de reistijd is er sprake als de voorspelbaarheid van de reistijd klein is én wanneer de variatie in de reistijd groot is. De kosten door gemiddelde extra reistijd zijn alle kosten door tijdsverliezen opgelopen door weggebruikers op het wegennet gesommeerd over een bepaalde tijdshorizon, uitgedrukt in geld. Ze vertegenwoordigen slechts het tijdsverlies onderweg, en dekken dus niet de totale economische impact van congestie en vertragingen, noch de maximale voordelen die worden verkregen door het oplossen van een file. Deze kosten voor gemiddelde extra reistijd worden vervolgens berekend, gevolgd door een grondige bespreking van de beperkingen van deze berekening en de gebruikte data. Steunpunt Mobiliteit & Openbare Werken 3 RA-MOW

4 English summary Title: Time valution in traffic Subtitle: Congestion costs, value of travel time & lost vehicle hours. Abstract In this report a method to calculate the congestion costs as a result of a road accident is presented. First, all the costs associated with an accident are discussed. On the one hand there are the injury-related costs like the medical costs, the costs because of production losses and the intangible costs. On the other hand there are the accident-related costs, being the tangible or material costs, the handling costs and the congestion costs. With these last costs the lost vehicle hours seem to play an important role. On the other hand, the value of travel time is of great importance in the calculation of the congestion costs. This is why the value of travel time is studied in more detail. The problem of the variation in travel time valuation is raised, as well as the differences in the valuation of time of freight transport, commuting, business-related traffic and other traffic. These differences manifest themselves in the fact that freight transport and business-related traffic (traffic during paid working hours) have a much higher value of travel time compared to commuting or other traffic such as traffic for shopping or recreation. Thereafter, the lost vehicle hours are studied in more detail. It appeared that for Belgium or Flanders a fully redundant system with closed road segments and double induction loops to measure vehicle speeds and intensities does not yet exists (the extension of the current network to such a network however, is planned for 2012). Because of this, the data is not always as reliable as needed, which is why the different methods for calculating the lost vehicle hours are studied in a literature review. On the basis of this literature study, a new framework was derived, so that in further research, after collecting the necessary data, a more reliable calculation of the lost vehicle hours can be performed. Finally, the costs of congestion are being discussed. Congestion costs are divided into costs by alternative behavior, costs incurred by the unreliability of travel times and costs for the average extra travel time. Among alternative behavior we mean leaving at a different time, travelling with a different transport mode, using a different route, or even completely renouncing from the journey. Unreliability of travel time is present when the predictability of travel time is small and the variation in travel time is high. The costs for extra travel time include all costs by time losses incurred by road users summed over a certain time horizon, expressed in money. They only represent the waste of time while underway and do not cover the total economic impact of congestion and delays, nor the maximal benefits that could be obtained by solving a traffic jam. These costs for the average extra travel time are calculated, followed by a thorough discussion on the restrictions of this calculation and the used data. Conclusion finalizes the report. Steunpunt Mobiliteit & Openbare Werken 4 RA-MOW

5 Table of content 1. INTRODUCTION ECONOMIC COST OF ROAD ACCIDENTS VALUE OF TRAVEL TIME LOST VEHICLE HOURS st method: The Netherlands (AVV, 2004) Before After nd method: Method of Tampère et al (2007) th method: Method of Morales (1986) New Framework RELATION BETWEEN ACCIDENT COSTS AND THE VALUE OF TIME: CONGESTION COSTS Incidental traffic jams Congestion due to road accidents Costs of congestion Alternative behavior Costs by the unreliability of travel times The cost for average extra travel time Discussion CONCLUSION Policy Recommendations Future Research REFERENCES Steunpunt Mobiliteit & Openbare Werken 5 RA-MOW

6 1. I N T R O DUCTION Being stuck in traffic has a cost to the society that might not be underestimated. This cost depends, among others, on the travel motif of those who are stuck in traffic. For example, for a truck driver who has to deliver a just-in-time delivery it is very important to remain on schedule. Someone who is stuck in traffic during paid working hours also causes a major cost for the company and indirectly the community. However, recreational traffic has a much lower cost. If you are, for example, visiting a friend or on your way to do some shopping, then lost time in a traffic jam is estimated to be less costly compared to on-the-job lost travel time. According to De Brabander (2006) congestion costs in Belgium in 2002 were 13,282,770. This, however, is an underestimation because it does not take the costs of alternative behavior and unreliability of travel time into account. The variation in the valuation of time and the unreliability of the data related to the lost vehicle hours should also be taken into account. As can be concluded from a previous report in Work Package 5.1, The stakeholders and their criteria in road safety measures: The next step in the development of the MAMCA, one of the most important criteria of road users concerning road safety measures is travel time. In every of the three types of measures that can be taken to improve road safety, i.e. user related measures, vehicle related measures and infrastructure related measures, travel time plays a major role (Van Raemdonck et al., 2010). Congestion costs could be a good indicator of this criterion. This is why it is important to have a good estimation of these congestion costs. This report examines the congestion costs, how they are composed and especially how they can be calculated reliably. The last topic appears to be a problem. First, there are many different figures for the value of time, even if the travel motif is the same. There is also a lot of variation between different individuals and how they experience a loss of time while travelling. Not only the travel motif is important, but also the circumstances in which the journey happens and the type of transport mode can have a major impact. On the other hand, the calculation of the lost vehicle hours seems to be a problem as well. Lost vehicle hours are annually reported for Belgium and Flanders, but these reported figures are not always that reliable. To obtain correct and reliable data for the calculation of the lost vehicle hours it is important that the network is well equipped with accurate measurement systems, i.e. double induction loops to calculate vehicle speeds and intensities. In Flanders, such a quality network of measurement systems is being installed on the main roads since The density of the network however, is only high enough on the Ring of Antwerp and the E313. The extension of a fully redundant quality network of double induction loops, which is dense enough for reliable calculations of the lost vehicle hours, is planned for For secondary roads there is no data available at all, but the development of a measurement system for these roads is also planned. However, this will be a lot more complex than it is for the main road network. This is why, in this report, some methods for the calculation of the lost vehicle hours are being discussed. Based on these theories a new framework to estimate the lost vehicle hours will then be derived. Further research will exist out of the collection of the necessary data, with which the lost vehicle hours can be calculated more accurately. Afterwards, a better figure for the congestion costs can be estimated. The next chapter of this report will deal with the costs resulting from a road accident in general. Next, the value of travel time will be examined in chapter 3. The fourth chapter will consist of a brief literature study about the methodology for the calculation of the lost vehicle hours. Based on this literature study a new framework will be set up in which the different advantages of the various discussed methods will be implemented. Finally, the Steunpunt Mobiliteit & Openbare Werken 6 RA-MOW

7 congestion costs and the calculation of these costs are studied, followed by a thorough discussion on the limitations of these calculations and the used data. Conclusion finalizes the report. Steunpunt Mobiliteit & Openbare Werken 7 RA-MOW

8 2. E C O N O M I C C O S T O F ROAD ACCIDENTS The quantification of the costs of road accidents is a very important step in the efficient spending of resources to improve road safety. It offers an insight in the size of the problem and makes it comparable with other societal problems. Even more relevant, it reflects the magnitude of the benefits to be obtained if road accidents can be avoided (De Brabander, 2006). In other words, it is important for policy makers to take the costs resulting from road accidents into account when deciding on road safety measures. The social costs caused by road accidents include all costs for the compensation and recovery from any injuries and damages and all costs relating to the settlement of damages caused by an accident, but also all other costs resulting from the accident that occurred including production losses, congestion costs (of congestion created by an accident) and intangible suffering (SWOV, 2006). This definition proves that many different types of costs originate from road accidents. The aim of this chapter is to give a brief overview of the different costs resulting from accidents and how they are composed. Costs caused by traffic accidents can be divided into two groups. First there are the injury-related costs such as medical costs, production losses and intangible costs. Second, there are the accident-related costs, which are the tangible costs, handling costs and congestion costs. A schematic representation of the costs incurred by road accidents is shown in figure 1. Figure 1: Classification of the costs resulting from road accidents Source: SWOV, 2006 The injury-related costs are composed of medical costs, costs of production losses and intangible costs. Below, a short explanation of each of these components will be given. Medical costs Medical costs cover the expenses necessary for the transportation to the hospital and care of the victims of an accident. It includes the cost to bring the ambulance to the scene of the accident, the medical care for the wounded, the cost of rehabilitation, etc. Also visiting costs, what visitors must pay to visit a victim in the hospital, and accelerated funeral expenses, which of course only occur in the case of a fatal accident, are considered as medical costs. The medical costs cover about 2% of the total road accident costs in Belgium (De Brabander, 2006). Steunpunt Mobiliteit & Openbare Werken 8 RA-MOW

9 Production losses Victims of accidents will sometimes be incapable to work for while. Some will even never work again, or will work less after their accident. This of course, next to the victim himself and his surroundings, is a cost for the society. These costs are called the "coldblooded component of an accident" (Lindberg et al, 1999). So, production loss is the loss of output because of getting injured or dying prematurely as the result of a road accident. The future loss should also be charged here. This approach for defining production losses is also called the Human Capital Approach (Connelly & Suspangan, 2006). According to the estimates of De Brabander (2006) for 2002 the temporary and permanent production losses account for about 35% of the total costs resulting by road accidents in Belgium. Intangible costs This is, in contrast to the previous, called the "warm-blooded component of an accident" (Lindberg et al, 1999). They are also called human or intangible costs and a previous report about these costs was written by Van Lier et al. in 2009 (RA-MOW ). The premise is that because of all the suffering and pain resulting from an accident, people want to avoid it, even without considering the financial consequences of the accident. These costs are usually calculated based on the value of a statistical life. The value of a statistical life is in fact a return, because by the rescue of a human life, and thus avoiding a fatality, much suffering and pain is avoided. The question to be asked is what road users want to pay to reduce or avoid accident risks. In other words, the willingness to pay for a reduction of the risk for a traffic accident is estimated (De Brabander & Vereeck, 2005; Van Lier et al, 2009). The intangible costs represent the biggest part of the traffic accident costs and covered 41% of the accident costs in Belgium in 2002 (De Brabander, 2006). The accident-related costs are composed of tangible costs, handling costs and congestion costs. Again, a short explanation of these costs is given below. Tangible costs These costs include the total material damage caused by an accident. Both damage to private property, including damage to vehicles, houses, cargo, etc. and damage to public property such as roads and other transport infrastructure are included. These costs accounted for about 19% of the total road accident costs in Belgium in 2002 (De Brabander, 2006). Handling costs These are the costs resulting from the handling of road accidents. They include the costs of the fire departments, police and justice, but also cover the administering costs of insurance companies due to medical and material damage (SWOV, 2006). According to the estimates of De Brabander (2006) for 2002 handling costs covered a little more than 2% of the accident costs in Belgium. Steunpunt Mobiliteit & Openbare Werken 9 RA-MOW

10 Congestion costs These are the costs resulting from the time lost in traffic jams caused by road accidents. To determine these costs, the total congestion costs are divided by the share of the total congestion caused by road accidents. There are three types of congestion costs: costs resulting from direct travel time losses, costs resulting from unreliable travel times and the costs of alternative behavior. Direct travel time losses refer to the economical damage that occurs because within a certain amount of time fewer destinations can be reached and thus people cannot travel as far as they want to. Unreliable travel times arise from the fact that because of traffic jams as a result of road accidents the travel time is difficult to estimate. Arriving too early or too late because of this creates additional costs. Finally, alternative behavior arises because people want to avoid congestion by choosing a different route, leaving sooner or later, choosing another transport mode or even completely renouncing from their journey. According to estimates for 2002, congestion costs are the smallest part of the accident costs in Belgium. They only represent 0.11% of the total costs resulting by traffic accidents (De Brabander, 2006). However, calculating congestion costs often results in an underestimation. Usually only the costs resulting from direct travel time losses are charged, and costs resulting from unreliable travel times and alternative behavior are not taken into account. Besides, data for the lost vehicle hours are often used in the estimation of these costs, but these data are not always reliable enough. Therefore it seems interesting to take a closer look at the calculation of congestion costs. So, the remainder of this report will handle about congestion costs more specifically. It will be studied whether it is true that congestion costs are systematically being underestimated, and whether they account for a larger share of the total costs incurred by road accidents than usually assumed. In order to do so, the congestion costs due to road accidents have to be calculated first. For this calculation the value of travel time has to be multiplied with the lost vehicle hours incurred by road accidents. According to De Brabander and Vereeck (2005) about 12% of the total number of traffic jams is caused by road accidents. This, however, does not mean that also 12% of the lost vehicle hours are caused by traffic jams as a result of accidents. Indeed, traffic jams caused by accidents will, on average, last longer than ordinary structural traffic jams. Hof and Vermeulen (2001) argue that 13.5% of all lost vehicle hours are the result of congestion due to road accidents. Total costs The total cost is the sum of the various sub costs. In 2002, according to De Brabander (2006), the total costs as a result of road accidents in Belgium were This represents 4.6% of Belgium s Gross Domestic Product, which is higher than the average accident cost of developed countries. According to Connelly and Suspangan (2006) the average cost of road accidents in developed countries is about 2-3% of the GDP. The biggest part of the total costs consists of the intangible costs, costs by production losses and material costs. Congestion costs are the smallest part of the total cost (De Brabander, 2006). However, as already mentioned above, congestion costs are being systematically underestimated. In the remainder of this report the calculation of congestion costs will therefore be examined in more detail. In the next chapter the value of time, a first important input variable for the computation of congestion costs, will be discussed. Steunpunt Mobiliteit & Openbare Werken 10 RA-MOW

11 3. V A L U E O F T R AVEL T I M E The value of travel time is a much needed input variable concerning the estimation of congestion costs, because the hours lost due to congestion have to be multiplied with this value of time in order to compute the congestion costs. It would be unrealistic to put a fixed value on the value of time, hence different values of time exist for various journey motifs. This chapter is about this variation in the value of time. The value of travel time (VTT) refers to the cost spent for transportation, including the waiting and the actual transport itself. It consists of the personal, unpaid time lost to transportation, as well as the costs for businesses, being paid working hours that are lost while being on the road. The value of travel time savings (VTTS) consists of the benefits arising from reducing the travel time (Litman, 2009). The costs associated with travel time can be divided into two major groups. First there are the on-the-job time losses; these include the costs to the employer who has to pay the lost hours of employees during work-related travel. Second, there are the off-the-job travel time losses. These are the trips for personal purposes and commuting, and represent the opportunity cost of time spent on journeys that could be spent on doing something else. The total travel cost is the product of the time spent on travelling (expressed in hours) and the unity cost (expressed in euro per hour). TravelCosttotal = Ttravel UnityCost With TravelCost total = the total cost of the trip (expressed in euro) T travel = time spent on the trip (expressed in hours) UnityCost = the cost of one hour travel = value of time (expressed in euro/hour) This unity cost varies according to the purpose of the trip. For example, on-the-job trips will have a higher value of time than off-the-job trips. This is illustrated in table 1. Table 1: Value of time per motif (euro/hour) Freight transport Persons transport commuting business other (1) - 8,37 28,97 5,78 (2) 42,35 8,6 29,77 8,94 (3) 45,78 6,86 24,04 4,58 (4) 45,3 7,5 24,5 6 Sources: (1) De Nocker et al, 2006 (2) Rijkswaterstaat: Dienst Verkeer en Scheepvaart, 2006 (3) Vanhove, 2008 (4) De Brabander, 2006 A lot of other variation occurs in the valuation of travel time. This variation depends on various factors like the type of vehicle, the purpose of the trip, the occupancy of the Steunpunt Mobiliteit & Openbare Werken 11 RA-MOW

12 vehicle 1, the traveler himself, the region, etc. The result of this variation is that, even if sufficient quality data is available, there is uncertainty about the estimates of the value of time and the congestion costs. It should be noted that travel time is not always lost time. Indeed, some travel can be used useful, or people may enjoy a certain amount of time in the car, train or bus. Some types of travel time have very low costs, or even positive values. This can be the case when traveling is a desired activity and people are enjoying the experience of the trip. Mostly these are recreational trips or trips for leisure, which involve e.g. riding with a new car or social activities like joyriding with friends. Travelling can also be combined with another activity. People can for example work on the train, or read a book or the newspaper. The literature however, is still often based on the rather short-sighted idea that travel time is always lost time. There is no model available yet for calculating the value of travel time that takes into account the fact that travel time does not necessarily have to be lost time (Lyons & Urry, 2004). In our case, however, it can be assumed that most travel time has a negative perception, because this report is about the time lost due to congestion. In the next chapter the second important variable to calculate the congestion costs, i.e. the lost vehicle hours, will be discussed. After a brief literature study, a new framework to compute the lost vehicle hours will be presented. 1 The number of occupants in the vehicle, being the driver and the possible passengers. Steunpunt Mobiliteit & Openbare Werken 12 RA-MOW

13 4. L O S T VEHICLE HOURS The indicator lost vehicle hours (VHL) shows the delays incurred by vehicles because of congestion and delayed traffic flows. In other words, lost vehicle hours draw a picture of the lost travel time of road users and thus are an important indicator for quantifying the economic effects of congestion (AVV, 2004). Congestion costs can be calculated by multiplying these lost vehicle hours with the value of time, which was discussed in the previous chapter. However, the calculation of the VHL is quite complicated. Quality input data is needed, and to collect this data there is a need for a reliable, redundant network of measurement systems. In Flanders such a quality network is being installed on the main roads since Nevertheless, as already mentioned, the density of this network is not always high enough to perform reliable calculations. Therefore the focus in this chapter will lie on the methodology of the calculation of lost vehicle hours and the description of the data needed to perform the calculations. However, the quality and accuracy of the lost vehicle hours does not only depend on the travel time calculation, because the density of the measurement systems, the number of setting parameters, the interpolation method for unavailable data, the level of aggregation (per 10 seconds, minute, 15 minutes, ) etc. also need to be taken into account. Furthermore, the systems work well most of the time in free flow traffic, but situations with traffic delays or unexpected incidents should be handled differently when calculating travel time. It is in this specific types of situations that one algorithm may work better than another. So the methods described in this chapter should be tested and evaluated in different traffic situations to draw more significant conclusions on which method to use in which situation. In this section a brief literature review on the existing methods for computing the lost vehicle hours is presented. First the method used in the Netherlands is explained. Second, a more graphical method based on a study of Tampère et al (2007) is presented. Finally, the better known and similar method of Morales (1986) will be discussed. The last part of this chapter will present a new framework for computing the lost vehicle hours, based on the discussed literature st method: The Netherlands (AVV, 2004) Before 1999 In the Netherlands the former AVV (Advice on Traffic and Transport), now RWS-DVS (Rijkswaterstaat Dienst Verkeer & Scheepvaart), is in charge of the calculation of the lost vehicle hours. Before 1999 the following framework was used. Data were not as accurate as they are now and there was a lot of under registration of congestion. Lost vehicle hours were calculated based on two formulas: = h With: VHL = Lost Vehicle Hours Delay = Delay per kilometer (hours/km) Steunpunt Mobiliteit & Openbare Werken 13 RA-MOW

14 GF = Getaway Flow (number of vehicles per lane that are able to leave the traffic jam during one hour) Lanes = Number of available lanes h = h + + With: Weight = Congestion weight (hours*km) Duration = Duration of the traffic jam (hours) Length = Length of the traffic jam (km) Construction = Construction weight of the traffic jam (hours*km) Reduction = Reduction weight of the traffic jam (hours*km) Duration and length were recorded during the registration of the traffic jam. For the construction and reduction of the traffic jam a fixed value of 2.5 minutes per kilometer was used, with the constraint that the total construction or reduction is no more than 30 minutes. The value for the delay was 4 minutes per kilometer and a getaway flow of 1500 pce/hour/lane was used 2. The number of lost vehicle hours due to unreported and non-observed traffic jams was estimated based on the total number of lost vehicle hours due to the reported traffic jams. The number of lost vehicle hours due to delayed traffic flows was also estimated. This was done by multiplying the lost vehicle hours with a fixed McKinsey-factor of After 1999 The basic ingredients for the present methodology in the Netherlands are data on speeds and intensities on the highway network. For the part of the network that is equipped with signaling infrastructure, this data is used to calculate the lost vehicle hours. For the part that is not equipped with the right infrastructure estimates of a model, called Flowsimulator, are used. This model simulates synthetic data that are similar to the data from the signaling infrastructure. Lost vehicle hours are calculated for each quarter on each measurement site. First the actual speed and the standard speed (normally 100 km per hour) are expressed in hours per kilometer instead of kilometers per hour. By subtracting the standard speed of the actual speed the travel time losses in hours per kilometer are calculated. These travel time losses are then multiplied with the distance over which the measurement site has been declared applicable. This leads to the travel time losses expressed in hours. Finally, this number is multiplied by the number of vehicles, which is determined by multiplying the intensity (in vehicles per hour) with 0.25 hours (one quarter). Or, expressed as an equation: = 2 Pce = passenger car equivalents Steunpunt Mobiliteit & Openbare Werken 14 RA-MOW

15 With: VHL = Vehicle Hours Lost TTL = Travel Time Loss (hours/km) d = Distance over which the data are applicable I = Intensity (vehicles/ hour) t = Time over which the intensity is measured (0.25 hours) And: = With: V actual = The actually driven speed V standard = The standard speed or comparison speed 3 In Flanders a similar method is used to calculate the VHL. The comparison speed is taken as 90% of a normalized speed. The actual speed depends on the speed of the vehicles that enter the road segment and is adapted every minute. This way more accurate speed flows are calculated. The absence of enough measurements systems, however, results in road segments that are so long that calculated speed flows are not reliable anymore. Another problem is that in Flanders not al the roads are well enough equipped with the right measurement systems. The installation of double induction loops started in 2005, but there are still some roads in Flanders that are equipped with single induction loops. These are only capable of counting the passing traffic, not to measure speed, certainly not in a reliable way. To measure speed, a high density of systems of double loops is necessary. At this moment the density is high enough on the Ring of Antwerp and the E313 (the extension from the current network to a fully redundant network in the whole of Flanders is planned for 2012). A third problem with the annually reported VHL-figures in Flanders is that the Ring of Antwerp is not included in these figures, this while a large part of the lost vehicle hours find their origin in and around Antwerp. It is therefore advised that the reported figures are processed in a very critical way. All this strengthens the need for a well founded methodology for the calculation of the lost vehicle hours and a road network that is fully covered with double induction loops nd method: Method of Tampère et al (2007) This model is based on a study of Tampère et al (2007) who has developed a methodology for identifying vulnerable sections in a road network. In a first stage of this model they make a long list of those vulnerable network links, based on the travel time losses and the occurrence of incidents on that particular link. For the simplicity, we consider an accident that only caused congestion on the affected network link. So there is no repercussion of the traffic jam to the upstream links, as can be seen in figure 2. 3 The comparison speed is set at 100 kilometers per hour. If the actual speed exceeds this comparison speed, profit hours are not taken into account. Steunpunt Mobiliteit & Openbare Werken 15 RA-MOW

16 Figure 2: Accident with no repercussion to the upstream network links Source: Tampère et al (2007) The model is based on the following figure: Figure 3: Calculation of the lost vehicle hours due to an incident Source: Tampère et al (2007) The line under slope I from the origin in figure 3 shows the inflow in a certain road network link. The line under the same slope, but shifted along the time axis, represents the outflow on that particular network link. An incident reduces the outflow to zero. If, after a certain amount of time t 1 the capacity has fully recovered, the traffic jam will resolve at a getaway flow C. This continues until after a period T all the accumulated traffic at the road network link is dissolved. From that moment, the outflow is back at its original level (slope I). The gray area A represents the total vehicle hours that are lost due to the accident. For A we find the following expression (Tampère et al, 2007): = Δ 2 1 The formula shows (and this is illustrated in figure 3 as well) that the effect of an incident measured in lost vehicle hours is proportional to the square of the incident duration. Namely, if t 1 increases to t 1, the amount of vehicle hours lost will increase from A to A+A. Steunpunt Mobiliteit & Openbare Werken 16 RA-MOW

17 4.3 3 th method: Method of Morales (1986) The premise of this theory is that an incident affects the road capacity. An incident reduces the capacity of the part of the road where it occurs. Congestion will arise when the incident reduces the capacity in that extent that the intensity exceeds the reduced road capacity. If the intensity already exceeds the road capacity before the accident, then congestion will increase even more due to the incident (De Ceuster, 2003). For any road, delays due to incidents depend on the following data: - Demand (λ: the number of arrivals per hour) at the time of the incident - The reduced capacity (µ r ) of the road after the incident - The total incident duration - The capacity after the incident (µ: drive off capacity or getaway flow) Figure 4: Method of Morales Source: De Ceuster, 2003 Figure 4 can be used to calculate the incident delay (the area AB 1 B 2 C). The horizontal axis represents time, the vertical axis the cumulative traffic flow. The demand, in vehicles per minute, is shown by the line segment AC. If an incident takes place, the traffic flow gets stopped because capacity reduces (line segment AB 1 ). At that moment demand (λ) is greater than supply (reduced capacity µ r ). The reduction of the capacity gets smaller when the vehicles that were involved in the accident are cleared (line segment B 1 B 2 ). After the complete recovery of the incident, traffic can continue its path (line segment B 2 C). At that moment full capacity is restored, but it will still take some time for the congestion to be resolved. It is thus possible to determine the total incident duration on the basis of this theory. Steunpunt Mobiliteit & Openbare Werken 17 RA-MOW

18 In figure 5 the cumulative volume (expressed in number of vehicles) is again plotted against time. λ indicates the arrival intensity (vehicles per unit of time) and µ indicates the departure capacity. The reduced capacity (µ R1 and µ R2 ) can still vary throughout the incident. Once µ falls below λ congestion is created. Total delay is then equal to the polygon AB 1 C 1 D 1 (expressed in lost vehicle hours). The length of the traffic jam (t Q ) is a function of the incident duration (t R ), λ, µ and µ R : = λ The total delay (TD) caused by an incident is equal to: = λ 2 These equations show, like in the previous method, that the total delay increases in proportional fashion with the square of the incident duration. If the duration of the incident can be decreased, then the total delay will be reduced (as can be seen in figure 5). Figure 5: Incident duration and delay Source: De Ceuster, 2003 Steunpunt Mobiliteit & Openbare Werken 18 RA-MOW

19 4.4 New Framework Each of the previous methods has some strengths and weaknesses. Therefore a new framework, which combines the strengths of the previous methods, is presented in figure 6. This model is applicable on any given road segment, for which accurate and reliable measurement systems (i.e. double loops) are available to measure the in- and outflow on that given segment. If an incident happens, the outflow (or capacity) reduces, but it does not reduce to zero. Namely, an incident does not always block the entire road, so that a reduced outflow is still possible. This reduced outflow is shown by the slope A in the figure. The inflow is given by slope I. After a while the capacity will regain its normal value (this can also happen in different steps if different driving lanes have to be cleared), and the congested traffic will get away at a getaway flow with slope C. Like in the two methods discussed above, the problem arises when, because of the accident, the capacity can no longer fulfill the demand. Because of this, the inflow and outflow curve are not parallel anymore, which results in lost vehicle hours. These lost vehicle hours are presented as the shaded area in figure 6. Figure 6 Own Framework Source: Own setup Because in this model inflow, outflow, reduced capacity and getaway flow are all linear functions, it is known that: Outflow (Capacity): = + Reduced capacity: = + Getaway flow: = + With K 1, K 2 and K 3 being constants. Steunpunt Mobiliteit & Openbare Werken 19 RA-MOW

20 The lost vehicle hours can then be calculated using integrals: = + This formula can be easily expanded if the recovery of the road capacity takes place in various stages. If the accident would reduce the outflow to zero, the formula proposed in the method of Tampère can be used (only if capacity recovers in just one step). Steunpunt Mobiliteit & Openbare Werken 20 RA-MOW

21 5. R E L AT I O N B E T W E E N ACCIDENT C O ST S AND T H E V A L U E O F T I M E: CONGESTION C O S T S This part will cover costs incurred by the time lost due to congestion and traffic delays because of road accidents. By valuating these time losses, the cost of being in a traffic jam will be calculated. As mentioned before, multiplying the value of time and the lost vehicle hours, which have been discussed in the previous chapters, results in the costs incurred by congestion. It should be noted that not all congestion costs are being discussed here, but only those resulting from an accident. Structural, daily traffic jams caused by the overload of the road network or traffic jams caused by roadwork are not included. First, incidental congestion in general is discussed. Afterwards this incidental congestion will be limited to congestion caused by road accidents. In a third paragraph, the various costs associated with these traffic jams resulting from road accidents are commented. Finally, the costs incurred by direct travel time losses are calculated, followed by a critical discussion of the limitations of this calculation and the used data. 5.1 Incidental traffic jams According to Van Reisen (2006) incidental traffic jams, in the broadest form, can be defined as all traffic jams that are not situated at the everyday bottlenecks. In other words, incidental traffic jams are traffic jams that differ in cause, location, time, length and duration from the congestion at the fixed structural bottlenecks. These include both incidental congestion caused by problems in the capacity supply, such as congestion as a result of weather conditions, accidents and road works, as incidental congestion caused by problems in transport demand such as congestion due to seasonality and special events. Hereby, it has to be noted that most of the incidental traffic jams are caused by the first group of reasons, i.e. problems in capacity supply. 5.2 Congestion due to road accidents This type of congestion falls under the category of incidental traffic jams. If it is known that congestion due to road accidents represents 60% of the incidental traffic jams, and the incidental traffic jams account for 20% of the total number of congestion, it can be stated that the number of traffic jams caused by an accident is about 12% of all congestion (De Brabander & Vereeck, 2005). This does not mean that also 12% of lost vehicle hours are caused by traffic jams as a result of accidents. Indeed, traffic jams caused by accidents will, on average, last longer than ordinary structural traffic jams. According to Wismans and Knibbe (2007) in The Netherlands approximately 20% of the number of lost vehicle hours is caused by incident congestion. Hof and Vermeulen (2001) argue that 13.5% of all lost vehicle hours are the result of congestion due to road accidents. Congestion caused by accidents, along with congestion caused by weather conditions, are the least predictable types of traffic jams. These traffic jams are therefore often very unexpected and thence the cost of unreliability will play an important role in the calculation of the cost of this type of congestion. Steunpunt Mobiliteit & Openbare Werken 21 RA-MOW

22 5.3 Costs of congestion As can be seen in Figure 7, there are different types of congestion costs. First of all they are divided into direct and external costs. These external costs include environmental nuisance (emissions, noise,, ) and road safety. The direct costs can in turn also be separated into two parts. On the one hand there are the welfare losses caused by alternative driver behavior. Alternative behavior covers detours, leaving earlier or later to avoid traffic jams, using a different transport mode or completely renouncing from the journey. On the other hand there are the congestion costs on the road itself, or observed congestion costs. These are then further divided into the costs for the average extra travel time, costs for the unreliability of travel time and extra fuel costs. The costs for the average extra travel time, which are calculated later in this report, correspond to the average delay or cost for travel time losses and are expressed in Euros per hour travel time. The costs of unreliability of travel time are equal to the dispersion around the mean delay and are expressed in Euros per minute standard deviation of travel time. These latter costs play a very important role in congestion due to road accidents, because these traffic jams are very unpredictable, and therefore the travel time losses are very difficult to estimate. Finally, there are the additional fuel costs, which will also play a role while standing in a traffic jam (Van Reisen, 2006). Figure 7: Overview of the social costs due to congestion Social costs due to congestion Direct costs External costs On the road Alternative behavior Environmental nuisance Road safety Costs for average extra travel time Costs for unreliablility of travel time Extra fuel costs Source: Van Reisen, Alternative behavior Koopmans and Kroes (2003) argue that the measurement of congestion costs calculated solely based on the observed congestion, is probably not accurate since there exist alternative travel routes and modes. Such alternatives may include using other transportation ation modes, but also travelling during off-peak hours, travelling to another destination or using another travel route, and even completely renouncing from the journey. It should be noted that alternative behavior will occur less frequently in the case of traffic jams resulting from accidents, because they are more difficult to anticipate due Steunpunt Mobiliteit & Openbare Werken 22 RA-MOW

23 to their incidental nature. However, if good information coverage is provided, there will still be alternative behavior, but in lesser extent than there would be in the case of structural traffic jams. Dynamic speed signs can partly resolve this problem of information coverage. The unobserved congestion costs, i.e. the costs for alternative travel behavior, are clarified by Koopmans and Kroes based on the following figure: Figure 8: Supply and demand on the mobility market Source: Koopmans and Kroes, 2003 In this figure, the amount of travel on a given road at a given time (horizontal axis) is plotted against the generalized costs (vertical axis). The marginal private cost curve (MPC) is horizontal at low traffic intensity, because traffic can drive without delay and is not faced with additional costs. The curve starts to rise at a critical amount of traffic, and then only gets steeper until point F, the maximal road capacity, is reached. At that point the marginal cost curve goes to infinity. Because there are several alternative options to avoid the congestion, the demand curve D1 is rather price elastic. Indeed, if costs rise, demand will fall. The observed congestion costs (lost time and fuel) are represented by the rectangle ABCD. The unobserved congestion costs are equal to the triangle CED. It is worth noting that, compared with the rectangle, the triangle is relatively small, what would mean that the observed congestion costs are greater than the unobserved. This would result in the fact that total congestion costs (in the case of moderate congestion) are not much higher than the observed congestion costs, and thus the costs for alternative behavior, in this case, only play a minor role. However, if demand increases and the road capacity remains the same (the demand curve shifts from D1 to D2), the unobserved congestion costs become more important. The new demand curve intersects the marginal cost curve at a higher level, allowing both the rectangle and the triangle to grow. It can be seen that the costs by alternative behavior (the triangle) increase compared to the costs by the remaining travelers (the Steunpunt Mobiliteit & Openbare Werken 23 RA-MOW

24 rectangle). Koopmans and Kroes note that the rectangle s surface increases in proportionate fashion with the capacity shortage, but the increase of the area of the triangle is proportional to the square of the capacity shortage. It can be concluded that the observed congestion costs can seriously underestimate the real congestion costs in a situation with a large capacity deficit, because the costs of alternative travel behavior are not observable and therefore they are not often included in the calculation. In the situation of congestion by road accidents this is, as previously mentioned, less of a problem, because in this case alternative behavior is more difficult as a result from the incidental nature of accidents (van Reisen, 2006) Costs by the unreliability of travel times Concerning the (un)reliability of travel time, both the ability to estimate travel time correctly and the variation in travel time are of great importance. The reliability will be high when the predictability of travel time is high or when the variation in travel time is small. Only if the predictability of travel time is small and the variation in travel time is high, there will be unreliability. Traffic jams due to accidents in particular, are a major cause for fluctuations in the expected travel time, because they are more difficult to anticipate. Thanks to the good estimation of the expected delay, people may be able to anticipate better on these traffic jams, which would reduce the congestion costs due to accidents (van Reisen, 2006). If there are many, or high, fluctuations in the travel time, travel time will be unreliable. A person than can arrive too late, or he can anticipate on the unreliability and leave earlier (with the result that he will sometimes arrive too early at his destination). In both cases a situation arises where time is not used ideally. Next to the utility loss because of the unreliability of travel time, the individual can face additional utility losses by arriving too early or too late at his destination (Rietveld et al, 2005). In other words, the difference between the actual arriving time and the preferred arrival time can play an important role in the decisions of the traveler. This difference between the actual and the preferred arrival time is defined as schedule delay, so that S = PAT + D ( t T ( t )) h h With S D = schedule delay PAT = preferred arrival time T h = time of departure T(t h ) = amount travel time (dependent of t h ) In general, people will appreciate too early and too late arrival times differently. Namely, if you arrive somewhere much too early, it can cost you some time you could have spent on doing something else. In contrary, if you are only 5 minutes early, this is not of great importance. However, if you are 5 minutes late, in many situations this can pose major problems. You can, for example, miss your train connection, arrive too late in a meeting, etc. And the later you arrive, the worse it gets. Hence the curve in the following picture is flatter while arriving too early, and it becomes very steep when arriving too late (Rietveld et al, 2005). Steunpunt Mobiliteit & Openbare Werken 24 RA-MOW

25 Figure 9: Valuation of 'schedule delay' Source: Rietveld et al, 2005 Following figure 9, the previous formula can then be divided into S DE and S DL : S = Max, + S DE DL = Max, ( 0 PAT ( th T ( th ))) ( 0 ( t + T ( t ) ) PAT ) h h With S DE = Schedule delay early S DL = Schedule delay late The cost for average extra travel time The costs for average extra travel time are all time losses incurred by the road users on the main and secondary roads summed over a certain time horizon, expressed in money. These costs only represent the waste of time while underway. So they do not cover the total economic impact of congestion and delays, nor the maximal benefits to be obtained by solving a traffic jam. In these costs the relationship between the value of time and the economic costs of road accidents is the most apparent, because they are about the additional travel time caused by congestion due to accidents. To calculate these costs different data are needed. The number of lost vehicle hours due to traffic jams by accidents for both the main roads and the secondary roads are necessary. For further specification of these data it should be known which part of lost vehicle hours is undergone by trucks and which part by passenger cars, what the travel motif of the traveler is and what the occupancy (the number of occupants, being the driver and possible passengers) and value of time for the different motifs are. With all this data it will then be possible to calculate the overall costs of extra travel time by congestion due to road accidents. Below, the costs for extra travel time are calculated in six steps, based on Van Reisen (2006) and Maerivoet and Yperman (2008). Steunpunt Mobiliteit & Openbare Werken 25 RA-MOW

26 1. First of all, the lost time has to be calculated. These time losses (V) can be defined as the difference between the average travel time (T) and the travel time in an unloaded network with free flowing traffic (T ff ). This free flow traffic is usually taken at 90% of a normalized speed on that part of the road network. The time losses show how much time an average vehicle loses per kilometer on an average hour in a given period and can be calculated by using the following formula (Maerivoet & Yperman, 2008): V = T T ff If the average travel time is less than T ff, then the loss of time will be equal to zero. This may occur in time periods characterized by very low travel times, e.g. during the night. 2. In the second step the total number of lost vehicle hours is specified. Lost vehicle hours are defined as the product of the traffic and the lost time. They show how much lost time is incurred by all vehicles on an average hour in a given period. The formula to calculate the lost vehicle hours is the following 4 (Maerivoet & Yperman, 2008): VHL = ( q V ) 3600 With VHL = Vehicle hours lost V = Time losses q = Traffic volume Note that the first two steps are similar to the first method used to calculate the lost vehicle hours (i.e. the method used in the Netherlands). If another method is applied, the obtained VHL from this method can be used directly in the following step, and the time losses do not need to be calculated separately. 3. In a third step the proportion of the traffic jams due to accidents in the total number of lost vehicle hours will be determined. The part of traffic jams by an incident is, as earlier stated, about 13.5% of the total amount of lost vehicle hours (Hof & Vermeulen, 2001). Travel motifs are not yet taken into account in this step. The annually reported figures for the lost vehicle hours in Flanders are shown in the following table. These figures, however, have to contend with some serious limitations. These restrictions are thoroughly discussed further in this report. 4 The division by 3600 is done to convert the lost vehicle hours from seconds to hours. Steunpunt Mobiliteit & Openbare Werken 26 RA-MOW

27 Table 2: Lost vehicle hours in Flanders Year VHL VHL incurred by accidents (13.5%) Source: Flemish Government, MOW, Traffic Centre & Hof en Vermeulen, In the fourth step the division per motif is applied to the lost vehicle hours. For each travel motif, the lost vehicle hours are multiplied with the share of the relevant motif. Table 3: Travel Motif Distributions Persons transport Freight Commuting Business other transport De Brabander (Vl) 23,31% 4,32% 57,85% 14,52% OVG (Vl) 23,5% 3,8% 56,7% 16 % Van Reisen (NL) 29% 22% 41,50% 7,50% SWOV (NL) 39% 28% 25% 8% Source: Own setup As can be seen in the table above, variation in motif distribution exists. The Netherlands traditionally have a higher proportion of business transport, which is partly compensated with less freight transport. In the calculation presented here, each of the previous scenarios will be examined. The lost vehicle hours due to road accidents per travel motif are given in the next table: Table 4: VHL per travel motif Persons transport commuting business other De Brabander OVG Freight transport , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,53 5 The factor to calculate VHL incurred by accidents is based on a Dutch study (Hof & Vermeulen, 2001) Steunpunt Mobiliteit & Openbare Werken 27 RA-MOW

28 van Reisen SWOV , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,76 Source: Own setup An interesting remark is that travel times can be directly determined per vehicle category, which would mean that a global percentage classification is no longer necessary. This could improve the quality, because the distribution of vehicle categories and travel motifs are not constant on different highways and regions. This can be an interesting research perspective for the future. 5. Thereafter, the next formula (van Reisen, 2006) has to be applied for the different travel motifs to determine the costs for travel time losses per motif. CostExtraTrave ltime = VHLaccidents occupancy VoT With Cost ExtraTravelTime = Cost for the average extra travel time VHL accidents = Lost vehicle hours due to congestion by accidents VoT = Value of Time As can be seen in the table below, not only the value of time differs according to the motif (as mentioned in chapter 3), but also the average occupancy of the vehicles will be different for each motif. This is an important difference between the indicator congestion costs and the lost vehicle hours. With the lost vehicle hours this distinction in travel motif is not made, so the economic impact of congestion is underestimated (Vanhove, 2008). Steunpunt Mobiliteit & Openbare Werken 28 RA-MOW

29 Table 5: Occupancy and VoT per motif Persons transport commuting business other Occupancy Value of Time Freight transport ,16 1,12 1, ,16 1,12 1, ,14 1,11 1, ,14 1,11 1, ,14 1,11 1, ,37 28,97 5,78 40, ,34 28,9 5,76 40, ,41 29,14 5,81 41, ,48 29,39 5,86 42, ,55 29,63 5,91 42,94 Source: Own setup based on Van Reisen (2006) and RWS-DVS (2006) In the case of freight transport it is assumed that the occupancy is always equal to 1, so it is not necessary to multiply with the occupancy for this motif. This is not because of the fact that there is only one driver, but because the freight is important, and not the occupants. Time losses are thus not inflicted to the occupants, but to the freight (De Ceuster & De Schrijver, 2002). 6. If the costs of the travel time losses are calculated for all motifs, these costs per motif need to be summed to obtain the total costs for travel time losses for a given period. If the calculations for the main roads and the secondary roads are conducted separately, they should obviously also be added together. The table below shows the results of the calculation for the main roads in Flanders for each scenario in motif distribution. Table 6: Costs of direct travel time losses due to road accidents in Flanders (in euros) Persons transport Freight TOTAL commuting business other transport , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,18 De Brabander OVG van Reisen Steunpunt Mobiliteit & Openbare Werken 29 RA-MOW

30 SWOV Source: Own setup , , , , , , , , , , , , , , , , , , , , , , , , , Discussion As can be seen in table 6 and in the figure below, the congestion costs in 2007 were significantly higher in comparison with the other years. According to the Mobility Report for Flanders the reason for this peak is unknown (MORA, 2009). Another notable finding from the figure is that the Dutch motif distributions (Van Reisen and SWOV) result in higher congestion costs compared with the Flemish distributions. The reason for this difference is the higher proportion of business traffic in the Dutch motif distributions. For the Dutch scenarios this business traffic also accounts for the highest costs, while for the Flemish scenarios freight transport and other traffic account for the biggest part of the congestion costs. A third remark in relation with the results is that if another source for the occupancies would be used (e.g. Zwerts and Nuyts, 2004) the congestion costs would rise. However, these data were not used because they do not contain a specific value for the occupancy of business traffic and the occupancies are assumed to be constant over time. Figure 10: Costs of direct travel time losses due to road accidents for Flanders , , , , , , , ,00 0, De Brabander OVG van Reisen SWOV 2008 Source: Own setup Overall, these figures seem quite low compared with the figures from De Brabander in the introduction. This is because De Brabander calculated the congestion costs for the whole of Belgium, while these results are based on the lost vehicle hours reported for Flanders only. Moreover, the lost vehicle hours incurred on the Ring around Antwerp are not included in these figures, which can cause a significant underestimation. If it is Steunpunt Mobiliteit & Openbare Werken 30 RA-MOW

31 known that the costs of alternative behavior and unreliability of travel times still need to be added, it seems clear that the congestion costs of accidents are a lot higher than generally assumed. Furthermore, in these calculations, only the traffic jams on the main road network are charged. The lost hours on the local network are not taken into account in this calculation due to the absence of the necessary data. As already mentioned, the reported lost vehicle hours for Flanders and the current methodology for their calculation have some important restrictions to deal with. Not all traffic measurement systems on the main road network in Flanders have yet developed sufficiently to visualize the lost vehicle hours in an accurate and reliable way, though this is planned for Moreover, the traffic jams on the R1 around Antwerp, which surely are a significant source of congestion in Flanders, are not included in these figures. The necessary measurement systems on the R1 were removed and replaced by double loops during road works, while the calculations are still done by the old measurement network with single loops, which are not ideal for calculating speeds. Nonetheless, this old monitoring system has to estimate the speeds, because the algorithm for the calculation with the new systems is not yet adapted, causing uncertainty in the figures. Because of these limitations, the reliability of the data should be put in perspective and they should therefore only be considered as a general indication of the congestion problem. Nevertheless it seemed interesting to estimate the congestion costs for Flanders, to obtain a general view on the economic burden of congestion. Steunpunt Mobiliteit & Openbare Werken 31 RA-MOW

32 6. C O N C L U S I O N This report examined congestion costs due to road accidents in Flanders and studies whether it is true that congestion costs are being systematically underestimated. To valuate these costs, different input data are needed: most importantly the value of travel time and the lost vehicle hours. Concerning the value of travel time, the problem is that even if sufficient quality data exists, the estimations will still be subject to a lot of variation. This variation depends on various factors such as the type of vehicle, the purpose of the journey, the occupancy of the vehicle, the traveler himself, the region, etc. As for the lost vehicle hours, there is an even bigger problem: the reliability of these figures should be put in perspective because the lost vehicle hours are calculated with data from single induction loops, which are not accurate enough to calculate vehicle speeds. Therefore the focus in this report was on the methodology of the calculation of lost vehicle hours and the description of the data needed to perform the calculations. After a brief literature study, a new framework was proposed in which the different advantages of the various methods that have been discussed were implemented. This framework however, could not yet be used for the calculations in this report because Flanders is not yet fully equipped with the appropriate infrastructure: in Flanders, single loops are used to measure speeds and derive lost vehicle hours, while for a reliable estimation of the lost vehicle hours double loops (in a redundant network) are needed. Currently, such a redundant quality network of double induction loops is only present on the Ring around Antwerp and the E313, while the extension of the network to provide the whole of Flanders with the necessary infrastructure is planned for The reported lost vehicle hours for Flanders, calculated with single loops and Antwerp not included, should therefore be viewed at with the necessary criticism. Nonetheless it seemed interesting to calculate the congestion costs based on these figures, to give a general indication about the congestion problem in Flanders. What can be concluded from these calculations is that congestion costs were significantly higher in 2007, although no specific reason can be given for this. Also notable is that the scenarios with the Dutch motif distributions result in higher congestion costs in comparison with the Belgian scenarios. The figures from table 6 also seem much lower than the figures of De Brabander, but this is because the calculations by De Brabander were executed for the whole of Belgium, while in this report only Flanders was considered. In addition, the R1 around Antwerp, a major source of congestion, was not included in these figures due to the obsolete measurement network. Also, only the main road network is taken into account, while ideally the congestion on the local roads should also be charged. Finally, the costs of the unreliability of travel times and these of alternative behavior should also be added. So, it can be concluded that congestion due to road accidents is indeed often underestimated and that it is an even bigger problem economically than generally assumed. How big this underestimation really is, is difficult to assess because of the difficulty of drawing conclusions based on the calculated lost vehicle hours for Flanders. In addition, for a perfect estimation of the congestion costs, data of the delays on the local road network are still missing. 6.1 Policy Recommendations The extension of the monitoring network should certainly be further elaborated, so that the whole network will exist out of double induction loops and the less reliable data from single loops are no longer necessary for the calculation of lost vehicle hours. In order to do so, the current algorithm for calculating VHL must be adapted to the use of these double loops, since the current algorithm can only calculate VHL based on single induction loops, which can make the estimates less accurate. By adjusting the algorithm, Steunpunt Mobiliteit & Openbare Werken 32 RA-MOW

33 the R1 around Antwerp can also be included again, allowing a more complete estimation of the lost vehicle hours for Flanders. Moreover, this network of double induction loops must be dense enough, making it a closed and redundant, and in this way also a more reliable, measurement system. Along with the elaboration of the measurement system on the main road network, it is also very important to calculate the lost vehicle hours incurred on secondary roads. This will be far more complex than on the main road network, but since the estimated time losses incurred on the secondary roads are four times as big as the time losses on the main roads (Maerivoet & Yperman, 2008), this is a very important issue for further examination. Once all this data can be collected, it must be examined which method is the best for actually calculating the lost vehicle hours in a reliable way. 6.2 Future Research To further improve the calculation of the lost vehicle hours it may be interesting to examine some of the following issues: The different methods to calculate the VHL have to be tested and evaluated in different traffic situations (free flow traffic, stop and go traffic jams, accidental traffic jams, ). This way the algorithm that works best in a specific situation can be selected. It has to be noted that the existing algorithm, which is being used at this moment, can still work properly after the adjustment to the new measurement systems of double loops. Moreover, other travel time algorithms, such as the trajectory method, the input-output method, can be studied in more detail. New technologies that can be used to determine the lost vehicle hours, such as the use of ALPR, Bluetooth, signature detection with double induction loops, FCD (GPS), are also worth exploring and the future may lie in the combinations of some of these systems. Finally, the interpolation techniques for unavailable data are very important for the calculation of the lost vehicle hours because of their influence on the accuracy of the figures and can also be studied in more detail. Besides, in this report data from different, sometimes foreign, studies are frequently used. Interesting further research could therefore include a deeper examination of the different calculations and figures of each study, certainly if multiple sources provide different results for the same parameter. In this way, the accuracy, reliability and representativeness of the used figures can be examined. Steunpunt Mobiliteit & Openbare Werken 33 RA-MOW

34 7. R E F E R E N C E S ADVIESDIENST VERKEER & VERVOER (AVV) Voertuigverliesuren op het autosnelwegennet: Voor de periode Rotterdam: AVV - Ministerie van Verkeer en Waterstaat. CONNELLY L.B. & SUSPANGAN R The economic costs of road traffic crashes: Australia, states and territories. Accident Analysis and Prevention, Vol. 38, p DE BRABANDER B. & VEREECK L Verkeersongevallen in België kosten jaarlijks 12,5 miljard. Verkeersspecialist 122: DE BRABANDER B Valuing the reduced risk of road accidents: Empirical estimates for Flanders based on stated preference methods. Diepenbeek: Universiteit Hasselt. DE CEUSTER M.J.C Incident Management op autosnelwegen in België. Leuven: Transport & Mobility Leuven. DE CEUSTER M.J.C. & DE SCHRIJVER M Verkeersindices: Congestie- en Milieukosten. Brussel: Transport & Mobility Leuven (in opdracht van Ministerie van Verkeer en Infrastructuur). DE NOCKER L., INT PANIS L. & MAYERES I De externe kosten van personenvervoer. Brussel: Vlaams Wetenschappelijk Economisch Congres VWEC. HOF A. & VERMEULEN J Maatschappelijke kosten van ongevallen met vrachtauto s buiten de bebouwde kom. Delft: CE. JIANG Y Estimation of Traffic Delays and Vehicle Queues at Freeway Work Zones. Washington D.C.: Transportation Research Board. KOOPMANS C. & KROES E Estimation of congestion costs in the Netherlands. Amsterdam: Stichting voor Economisch Onderzoek des Universiteit van Amsterdam. LINDBERG et al Final report of the export advisors to the high level group on infrastructure charging. Brussel. (cited in STAES H. & DE BRABANDER B Inleiding tot economische afwegingsmethoden op verkeersveiligheidsmaatregelen. Diepenbeek: Universiteit Hasselt.) LITMAN T.A Transportation cost en benefit analysis: Techniques, estimates and implications. Victoria (Canada): Victoria Transport Policy Institute. LYONS G. & URRY J The use and value of travel time. Unpublished paper. Steunpunt Mobiliteit & Openbare Werken 34 RA-MOW

35 MAERIVOET S. & YPERMAN I Analyse van de verkeerscongestie in België. Leuven: Transport & Mobility Leuven. MORA Mobiliteitsrapport van Vlaanderen Brussel: Mobiliteitsraad van Vlaanderen. MORALES J Analytical procedures for estimating freeway traffic congestion. Public Roads 50, RIETVELD P., VERHOEF E. & TSENG Y A meta-analysis of valuation of travel time reliability. Amsterdam: Department of Spatial Economics, Vrije Universiteit Amsterdam. RIJKSWATERSTAAT: DIENST VERKEER & SCHEEPVAART (RWS-DVS) Personenvervoer: Groei reistijdwaardering in de tijd. Rotterdam: Rijkswaterstaat. STICHTING WETENSCHAPPELIJK ONDERZOEK VERKEERSVEILIGHEID (swov) Kosten van verkeersongevallen in Nederland: Ontwikkelingen Rotterdam: SWOV Ministerie van Verkeer en Waterstaat. TAMPERE C.M.J., STADA J. & IMMERS L.H Een methodiek voor het vaststellen van kwetsbare wegvakken in een wegennetwerk. Tijdschrift Vervoerswetenschap, Vol. 43, March 2007, pp VANHOVE F Analyse van de mobiliteit op de Belgische autosnelwegen: Verkeersindices Leuven: Transport & Mobility Leuven. VAN LIER T., F. VAN MALDEREN & C. MACHARIS Indicatoren bij de beoordeling van verkeersveiligheidsmaatregelen: Knelpunten en mogelijke oplossingen. Steunpuntrapport RA-MOW Diepenbeek, België: Steunpunt Mobiliteit en Openbare Werken. VAN RAEMDONCK K., NOVIKOVA E., VAN MALDEREN F. & C. MACHARIS The Stakeholders and their criteria in road safety measures: The next step in the development of the MAMCA. Steunpuntrapport RA-MOW Diepenbeek, België: Steunpunt Mobiliteit en Openbare Werken. VAN REISEN M Incidentele Files: De kenmerken, de kosten en het beleid. Amsterdam: SEO Economisch Onderzoek. WISMANS L. & KNIBBE W.J Voertuigverliesuren door incidenten. Rotterdam: AVV Ministerie van Verkeer en Waterstaat. ZWERTS E. & NUYTS E Onderzoek verplaatsingsgedrag Vlaanderen 2. Diepenbeek/Brussel: Onderzoek in opdracht van het ministerie van de Vlaamse Gemeenschap departement Leefmilieu en Infrastructuur, Mobiliteitscel. Steunpunt Mobiliteit & Openbare Werken 35 RA-MOW