A COMPARATIVE ANALYSIS OF THE CLIMATE VALUE OF CYCLING IN DUTCH CITIES

Mark Zuidgeest Mark Brussel Yang Chen Roel Massink Martin van Maarseveen University of Twente, The

ABSTRACT The Climate Value of Cycling (CVoC) model provides a methodology for the assessment of cycling mobility in terms of avoided CO2 emissions. The paper analyses the spatial and transport determinants of the climate value of cycling for Dutch cities.

1. INTRODUCTION Human induced emission of carbon dioxide (CO2) is one of the most important challenges humanity has to deal with in the 21st century. The transport sector is responsible for approximately 23% of global CO2 emissions, a number that is growing, particularly in view of the increasing vehicle ownership and use in developing and emerging economies (IEA, 2008). While clean vehicle technology and cleaner fuels have been adopted as appropriate strategies to reduce greenhouse gas (GHG) emissions in the last few years, the discussion in the 2009 United Nations climate change Conference of the Parties (COP) 15 in Copenhagen concluded that a complete restructuring of the way urban mobility is organized is the only feasible climate mitigation strategy. Non-motorized transport (NMT) development has the potential to reduce CO2 emissions. NMT offers important benefits for society by promoting people‟s health, providing opportunities for economic development and contributing to social inclusion. Sustainable transport projects could induce reductions in CO2 emissions of the road transport sector by: (1) “Avoiding” the need for mobility; (2) “Shifting” mobility to sustainable modes of transport, such as cycling; or (3) “Improving” sustainability of current mobility (Dalkmann and Brannigan, 2007). Cycling projects and programmes are typically developed to accommodate and maintain current cycling levels and/or to expand cycling levels by providing good quality and safe facilities for cycling (i.e. bicycle lanes and protected crossings, traffic calming to enable shared road use, bike parking and rental facilities). In many countries or cities bicycle trips are still decreasing, while motorized transport is on the rise. This is problematic, particularly in view of the current climate debate, where present bicycle mobility avoids more CO2 emissions being © Association For European Transport and Contributors 2011

released, because each bicycle trip could potentially be motorized and emitting. In this context, the “opportunity costs” of a bicycle trip are the additional CO2 emissions that are emitted when the traveller selects an alternative, motorized transport mode for his or her bicycle trip. In terms of avoided CO2 emissions, therefore, cycling provides significant “opportunity benefits”. In The Netherlands, despite increasing trip distances, the bicycle has retained its popularity. Dutch cities, even though all having high bicycle shares, show remarkable differences in modal split figures for cycling ranging from 10 up to 35% of urban trips. In fact for distances up to 7.5 km, the bicycle is the most popular means of transport with 34% of all trips made by bicycle (Ministry of Transport, Public Works and Water Management and Fietsberaad, 2009). Bicycle use very much depends on the distance covered. As 70% of all journeys in the Netherlands are still shorter than 7.5 km, the strong position of the bicycle over short distances (34%) also extends into the total modality split (26% bicycle). At the same time, it is interesting to note that the bicycle is regularly chosen above 7.5 km: 15% of journeys in the category 7.5-15 km (Ministry of Transport, Public Works and Water Management and Fietsberaad, 2009). Even though cycling is very popular, it does not mean that cycling is prevalent all over the country. On average, the Dutch chose the bicycle for 26% of their journeys in 2007. In the cities with the highest bicycle use such as Groningen and Zwolle, the cycling percentage is about 40%. In cities with the lowest bicycle use, it is about 13% (Ministry of Transport, Public Works and Water Management and Fietsberaad, 2009). Being such a popular mode of transport next to various other (motorized) modes, the level of avoided CO2 emissions due to cycling is expected to be significant in Dutch cities. This paper calculates these avoided CO2 emissions for the different municipalities in The Netherlands and shows how these levels of avoided CO2 emissions relate to important physical and mobility characteristics. The Climate Value of Cycling model, introduced in this paper as well, is used to calculate these levels. 2. ATTRIBUTING A CLIMATE VALUE TO CYCLING Assessing the carbon impact of cycling is, ironically, complicated due to the fact that cycling has an intrinsic zero-emission value. Attributing (direct) carbon benefits is therefore difficult. There has been very little scientific research conducted into the estimation of the CO2 reduction potential of cycling. General cost-benefit analysis of bicycle projects has been performed regularly (Cavill et al., 2009; Krizek, 2004; Litman, 2004; Lind, 2005a; Lind et al., 2005b; Saari et al., 2005; Sælensminde, 2004). However, the current state-of-the-art in bicycle evaluation generally ignores the avoidance of potential CO2 emissions. Few studies included CO2 emissions as a variable in their cost-benefit analysis (Browne, et al., 2005; Gotschi, 2008). These studies showed only marginal CO2 reduction effects of bicycle projects, in most cases a result of the relatively small scale of the project such as the improvement of a single bicycle corridor. Browne suggests that CO2 emissions reductions from city-wide cycling projects are more © Association For European Transport and Contributors 2011

significant in both absolute and relative terms. Similarly, city-wide cycling levels are thought to significantly avoid levels of CO2 emissions.

There are several issues regarding the policy relevance of attributing a climate value to cycling (Massink et al., 2011). First, the importance of attributing a climate value to cycling is not self-evident to policy makers and their professional advisors. Second, the attribution of climate values to transport in general is not very well established. Third, the assessment and validation of the climate value of non-emitting active transport modes like cycling and walking is more complex than for directly or indirectly emitting motorized transport. The policy relevance of attributing climate value to cycling is linked to the absence of the transport sector, non-emitting modes in particular, in Carbon Credit assessment and validation methodologies as applied in Kyoto Protocol mechanisms. The awareness of climate policymakers of the importance of the transport sector has only recently been achieved in debates at the 15th UNFCCC COP in Copenhagen, in December 2009. The bicycle has always been positioned and used as the icon for sustainable climate policies but until very recently not as contributor to emission reduction targets (Huizenga, 2009; Sakamoto et al., 2010). 3. THE CLIMATE VALUE OF CYCLING (CVOC) For the evaluation of the external effects of CO2 emissions of transport, opportunity costs are often referred to as avoidance costs (Bickel, et al., 2001). Avoidance costs here are the monetary costs to avoid a certain level of CO2 emissions. When evaluating the avoidance costs of transport, motorized modes have positive avoidance costs and non-motorized modes have an avoidance cost of zero. The Climate Value of Cycling represents the total amount of avoided CO2 emissions by all bicycle trips, which is the summation of opportunity costs or avoidance costs of each bicycle trip in the study area. The CVoC is calculated based on a prediction of the most likely alternative (substitution) mode for each bicycle trip and the calculation of the additional CO2 emissions for that trip by the alternative mode. The CVoC represents a „risk‟ value. In areas with a relatively high CVoC (total, per capita or per kilometre) failure to maintain the current level of cycling will probably lead to relatively higher level of CO2 emissions, because trips lengths by bicycle are relatively long and likely to be substituted by motorized vehicles, as compared to areas with a low CVoC.

The method for estimating these opportunity costs (or benefits) of cycling is discussed in the next section. 3.1. The Climate Value of Cycling (CVoC) model The CVoC method simulates a virtual substitution of bicycle trips with trips made with an alternative mode. For the research presented in this paper only short term changes are considered, assuming traffic inducement and traffic discouragement on the short and long term to be zero for ease of demonstrating

© Association For European Transport and Contributors 2011

the method. Another application of the CVoC for Bogotá, Colombia, includes such traffic discouragement factors (Massink et al., 2011).

The main focus of the methodology is to estimate the most likely alternative mode for each bicycle trip and to calculate the additional CO2 emissions caused by this induced traffic. The model builds on existing theories of the multinomial logit behavioural model and provides a simple, transparent and data extensive methodology. The behavioural part of the CVoC model defines mode choice situations based on the length and purpose of a trip and the socio-economic background of the trip maker. For this, the model requires an input database describing the present traffic characteristics at trip level, indicating trip length, optionally also socio- economic background and trip purpose. Trips that share the same values for these attributes are clustered together into one class. These clusters are defined as mode choice situations. Irrespective of mode, all trips are clustered into classes of trips sharing the same mode choice situation. For each class, bicycle trips are redistributed to the most likely alternative modes based on observed modal share ratios of the remaining modes in that class. This approach makes the assumption that the probability ratios of choosing one mode over the other remain unchanged when the bicycle mode is excluded from the choice set. This assumption is one of the major properties of the multinomial logit choice model and is described by Luce and Suppes (1965) as the Independence of Irrelevant Alternatives (IIA) axiom, i.e. “Where any two alternatives have a non-zero probability of being chosen, the ratio of one probability over the other is unaffected by the presence or absence of any additional alternative in the choice set”. This means that when cycling is excluded from a choice set, the ratios of probabilities between the other modes remain the same, because utility values for the various mode options can be assumed to be unaffected by the exclusion and substitution of bicycle trips. According to Ortúzar and Willumsen (2001) this axiom is generally perceived to have disadvantages, making the model fail when the bicycle alternative is not independent or when there are taste variations among individuals as a result of different cost perceptions. The assumption of independence of the bicycle mode from the other modes can be justified by the fact that all trips are clustered into small classes (short distance bins for example), share trip characteristics and thereby capturing the taste variations among individuals. In addition, the bicycle mode only has to be independent from the other modes. Unobserved associations between the other modes can still exist in the model and do not hamper the ability of the model. Statistically, IIA violation can and should be tested using the Hausman and/or the Small-Hsiao test (Hausman, 1978; Small and Hsiao, 1985). The modelling procedure consists of three steps: (1) clustering of trips in mode choice situation classes; (2) calculation of induced (substituted) traffic for each mode other than cycling; and (3) calculation of opportunity costs, thus climate value of cycling. In the first step, trips are clustered in specific mode choice

© Association For European Transport and Contributors 2011

situation classes. In this paper all trips are clustered in discrete trip length bins with bin sizes depending on the sample. Second, to estimate the alternative modes, discrete probability distributions are estimated for each class based on the observation within that class. The probability that a mode is the alternative mode for a bicycle trip in the mode choice class is then based on the probability mass function for each subclass (Johnson et al., 1993). The sample space consists of all modes in that subclass excluding the bicycle mode. The induced traffic effect (the most likely substitution mode) for the mode in each subclass is accordingly calculated. Finally, the total opportunity costs or CVoC (kg CO2) is calculated by multiplying the induced traffic per mode with a modal emission factor (kg CO2 per km). These steps have been described in detail accompanied by their mathematical descriptions in Massink et al. (2011). 4. THE CLIMATE VALUE OF CYCLING FOR DUTCH CITIES The Dutch Mobility Survey (Mobiliteitsonderzoek Nederland or MON) collects annual information regarding the mobility of the Dutch population. It is carried out since 2004 through the Directorate-General for Public Works and Water Management. Data are collected on the transport movement of individuals, such as the reason for the movement, the place of departure and the destination, the transport means and the time. Data about the households are also recorded, such as composition and size, as well as age, sex and education of the travellers, and some basic data on the municipality of the trip makers. The survey‟s sample size is 50,000 persons. Weighting and scaling up through expansion factors renders the numbers representative for the entire Dutch population (DANS, 2009). With the ready-to-use MON database and the CVoC model, which was implemented in the mathematical modelling software MATLAB, the CVoC was successfully calculated for 437 out of 443 municipalities in the year 2008. Six municipalities have no results partly due to their very small sample size, and were left out. A 5 km trip length bin size was initially chosen, while all trips longer than 30 km were clustered in one final class. Ten travel modes were distinguished: walking, cycling, motorcycle, car, taxi, bus, BRT, LRT, MRT, and others. Each mode was assigned an average CO2 emission factor per kilometre travelled from ADB (2010). These factors are also used in an international city comparison of CVoC values, hence non-Dutch average emission factors were used for ease of conducting the study. 4.1 Analysing CVoC per city

For each Dutch municipality, the total CVoC, per capita CVoC, and unit CVoC (the CVoC per cycling kilometre) have been calculated. Based on these calculations all 437 municipalities together avoid 1.41 Mtons of CO2 emissions for the whole year. Amsterdam, Utrecht, Groningen, Eindhoven,„s-Gravenhage, , Tilburg, Enschede, Breda, and Amersfoort are in the top 10 municipalities in terms of total CVoC, as can be seen in Table 1. Underlying maps and social-economic data have been obtained from CBS (2009).

© Association For European Transport and Contributors 2011

The results can also be presented in maps, see Figures 1, 2 and 3. The total CVoC for each city, depicted in Figure 1, shows the regional differences for Dutch cities. Obviously the large cities in the West and South show higher absolute CVoC on average, while the less urbanized North and East parts show lower absolute levels of CVoC.

Figure 1. Total CVoC of Dutch municipalities in 2008 (base map: Centraal Bureau voor de Statistiek / Kadaster, Zwolle, 2011)

© Association For European Transport and Contributors 2011

Figure 2. Per capita CVoC of Dutch municipalities in 2008 (base map: Centraal Bureau voor de Statistiek / Kadaster, Zwolle, 2011)

© Association For European Transport and Contributors 2011

Name Pop- Total CVoC Bicycle Total Average Cycling ulation CVoC per Passenger Bicycle cycling share in per capita Kilometer PKT per distance modal year per Travelled day [km] per split [tons year [kg (PKT) per person [%]

CO2] CO2] capita per per day year [km] [km]

Amsterdam 747,090 41,091 55 1,003 2,053,496 2.8 21%

Utrecht 294,740 27,140 92 1,290 1,041,470 3.5 22%

Groningen 182,480 26,055 14 3 1,644 821,832 4.5 36%

Eindhoven 210,330 25,986 12 4 1,284 739,869 3.5 26%

The Hague 475,680 22,064 46 735 957,249 2.0 18%

Rotterdam 582,950 20,014 34 538 859,363 1.5 14%

Tilburg 202,090 19,921 99 846 468,615 2.3 25%

Enschede 154,750 17,588 11 4 1,023 433,900 2.8 32%

Breda 170,960 15,137 89 913 427,714 2.5 24%

Amersfoort 141,210 14,721 104 999 386,524 2.7 27%

Table 1. Top ten municipalities in terms of total CVoC (data source: MON 2008, emission factors ADB (2010))

The top 10 CVoC cities in Table 1 are cities that have relatively similar characteristics in terms of their cycling modal share, which is generally high because of the rather compact urban spatial structure of most Dutch cities in combination with cycling friendly policies. However, interesting differences can be seen from the table. The three largest cities in terms of their population (Amsterdam, Rotterdam and ) do not rank 1, 2 and 3 in terms of their overall CVoC. The reasons for this are that, despite their high cycling modal share, cycling trip lengths are relatively short, e.g. only 1.5 km in the case of Rotterdam and they have a well-established public transport system.

A city like Groningen, that combines a very high modal share with long cycling trip lengths and a less well developed public transport system therefore arrives at a per capita CVoC triple that of Amsterdam and The Hague.

© Association For European Transport and Contributors 2011

What lessons can we draw from the interpretation of the CVoC? The interpretation of per capita CVoC is rather complicated, since the CVoC itself is calculated based on the total cycling distance per trip and the emissions of its most likely alternative mode for this trip. When comparing the per capita CVoC with cycling modal share, we can come to a more meaningful interpretation. In doing so, we distinguish 4 combinations, as indicated in Table 2 below.

CVoC Cycling Interpretation and potential policy response per modal Capita share

High High These cities, although doing well in terms of cycling modal share are facing a high risk of increased CO2 emissions in case cycling levels cannot be maintained. If, however, they succeed to increase cycling modal share even further, high values of CO2 emissions can be avoided. High Low Fewer people are making longer trips in combination with a transport system that offers less sustainable alternatives. This combination occurs not very often, only in a few smaller and more rural municipalities. Policies should address the increase of modal share for cycling while at the same time creating more sustainable transport options for longer trips. Low High A low risk situation where we have mostly a combination of many short distance cycling trips and sufficiently sustainable alternatives. Policies should be directed at maintaining the current status, or even improving the modal share of cycling further. Low Low A situation where there are opportunities for the use of cycling to grow, preferably while keeping the performance of the other transport modes as green as possible. Usually characterized by short trip lengths, not necessarily by sustainable substitution modes Table 2. Comparing levels of CVoC per capita and Cycling Modal Share

© Association For European Transport and Contributors 2011

Figure 3. Unit CVoC (base map: Centraal Bureau voor de Statistiek / Kadaster, Zwolle, 2011)

© Association For European Transport and Contributors 2011

Figure 4. Unit CVoC vs. Population density

Examples of the high-high category of the bigger cities are Groningen and Eindhoven, but you may also find smaller municipalities that have high cycling modal shares such as Reeuwijk or Twenterand with modal shares close to 40% and per capita CVoC values of around 200 kg CO2. These are smaller towns with higher income residents that do cycle a lot, but where almost all cycling trips would be replaced by car trips. In the high – low category we do not really find the larger urban agglomerations, none of the 10 largest cities in terms of CVoC falls into this category. In fact there are not so many municipalities in The Netherlands that have low cycling shares and even fewer that combine this with high climate values. Examples in this category are Lopik, Westvoorne, Ubbergen and some others. The low- high category is probably the best place for a municipality to be in. These cities combine a low climate value with a high modal share which indicates both the important role of cycling as a sustainable mode, but also the sustainable nature of the substituted trips. Examples of these

© Association For European Transport and Contributors 2011

municipalities are Amsterdam and to a lesser extent The Hague, but (not surprisingly) also some of the cities that are contending for the title of best cycling city (see below and table 2), such as , Houten, Harderwijk and . The low-low category generally pertains to cities where little cycling is combined with a low per capita substitution value. e.g. Kerkrade with a 7% modal share and a per capita CVoC value of 44 kg CO2 per year. Here generally trip lengths are low, which accounts for a low substitution value. Also in this category we find smaller rural villages that combine low mode shares with short trip lengths and thereby achieving a limited CVoC, particularly in the South and East of the country. Overall speaking, cities with large population and good performance of public transport have a lower per capita CVoC, see Figure 2.

The overall sustainability of a cities‟ transport system can be even better illustrated with the Unit CVoC, see Figure 3. This measure reflects for each municipality the induced CO2 emissions of losing one kilometer cycling trip to an alternative transportation mode. Basically, it shows the sustainability of the modal split when cycling is excluded. In the map, the greener the municipality, the more sustainable its transport system is, which indicates more walking and public transport and less use of private vehicles, and vice versa. For example, of all the 25 municipalities with population more than 100,000 inhabitants Emmen (with a unit CVoC of 0.129 kg/km, almost comparable to the emission factor of a passenger car) is two times less sustainable than Amsterdam and The Hague (with their unit CVoC 0.055 kg/km and 0.063 kg/km, respectively). Meanwhile, the Pearson Correlation between unit CVoC and population density of those municipalities is -.717 at the 0.01 significance level (two-tailed). A scatter plot of unit CVoC (y-axis) and population density (x-axis) is depicted in Figure 4. Since the unit CVoC can be regarded as a sustainability indicator, we can conclude that the sustainability of the transport system (without cycling) of Dutch cities is negatively correlated with population density. A higher population density usually means short distance between activities and good public transport service, which would make the unit CVoC comparatively lower than cities with lower population density.

4.2 Analysis of CVoC based on best cycling city contest Organized by the “Fietsersbond”, a Netherlands based cycling NGO that advocates cycling inclusive planning in the Netherlands, Dutch cities compete for the title Fietsstad 2011 (Cycling City of the year) (Fietsersbond, 2011). Based on criteria such as the role of cycling in spatial planning policy, the provision of cycling infrastructure and routes the title best cycling city is given to several of them. This year 16 cities participated in the contest (see Table 3) out of which 5 got finally nominated. These are Groningen, Harderwijk, Houten, Pijnacker-

© Association For European Transport and Contributors 2011

Nootdorp, and 's-Hertogenbosch. Our indicators of CVoC, per capita CVoC and unit CVoC provide a better understanding of these cities in terms of their performance of cycling besides the criteria used in the competition. First of all, it must be said that these cities have substantial to very high cycling shares, and some of them like Delft, Houten, Harderwijk, and Alphen aan den Rijn combine this with low per capita climate values. If we look at the unit value (see Figure 5), which is a measure of the sustainability of the non-cycling part of the system, we can see that Delft has the lowest Unit CVoC, but links this also to a very high cycling share, substantial cycling trip length and a low per capita CVoC. On this basis, we can argue that Delft is the most sustainable cycling city.

© Association For European Transport and Contributors 2011

Figure 5. Unit CVoC for municipalities competing for cycling city of the year (base map: Centraal Bureau voor de Statistiek / Kadaster, Zwolle, 2011)

Name Pop- Total C CVoC Bicycle Total Average Cycling ulation VoC per Passenger Bicycle cycling share in per capita Kilometer PKT per distance modal year per Travelled day [km] per split [tons year (PKT) per person [%]

CO2] [kg capita per per day

CO2] year [km] [km]

Alphen aan den Rijn 71,66 4,758 66 623 122,266 1.7 29%

Breda 170,96 15,137 89 913 427,714 2.5 24%

Culemborg 27,3 3,71 136 1,485 111,043 4.1 43%

Delft 96,17 6,445 67 967 254,766 2.6 31%

Doetinchem 56,25 4,901 87 812 125,129 2.2 25%

Goes 36,71 5,075 138 1,379 138,721 3.8 33%

Groningen 182,48 26,055 143 1,644 821,832 4.5 36%

Harderwijk 42,33 1,976 47 398 46,149 1.1 20%

Houten 46,48 1,776 38 412 52,515 1.1 19%

Leeuwarden 92,86 8,77 94 1,011 257,153 2.8 30%

Littenseradiel 10,88 1,15 106 888 26,471 2.4 24%

Pijnacker- Nootdorp 43,76 5,431 124 1274 152,728 3.5 29%

Roermond 54,45 4,508 83 854 127,364 2.3 27%

's-Hertogen- bosch 136,48 10,986 80 856 320,083 2.3 23%

Utrecht 294,74 27,14 92 1,29 1,041,470 3.5 22%

Zoetermeer 119,5 8,06 67 785 256,882 2.1 23%

Table 3. CVoC for municipalities competing for cycling city of the year

© Association For European Transport and Contributors 2011

5. LIMITATIONS OF THE STUDY The results of the CVoC analysis presented in this paper are limited by the following factors that are worth discussing. First of all, the sample size of the MON Mobility Survey in 2008 is large for the whole country, but differs (obviously) per municipality. Even though expansion factors have been applied, errors due to sample sizes may occur. Stacking different years of MON data can solve this problem. Related to this is the choice of bin size in the calculation of the CVoC. Obviously it would be better to have smaller bins of for example 2 km, given typical cycling trip distance; however the limited number of observations per distance bin might distort the calculation of modal split per bin. Stacking MON data may again solve this problem. Social-economic data and trip purpose data can be added to the analysis, creating multi-dimensional bins based on distance, social-economic and purpose class membership. The MATLAB model is ready for this (see Massink et al. (2011), but this has not been applied for the analysis in this paper. IIA violation can and should be tested using the Hausman test for each municipality as discussed in section 3.1. In Massink et al. (2011) this is successfully tested for the city of Bogotá, Colombia. Doing that for all 437 municipalities is very time consuming though. Finally, in the current modelling international emission factors, based on ADB (2010) have been applied. Obviously, Dutch average emission factors would improve the CVoC calculations significantly. 6. CONCLUSIONS In this article, the concept of the Climate Value of Cycling (CVoC) has been introduced and applied to 437 cities in The Netherlands using MON Mobility Survey data. The CVoC concept allows the calculation of the opportunity benefits of cycling in terms of avoided CO2 emissions. The methodology shows that bicycle mobility contributes to climate change mitigation. The CVoC is dependent on the amount of cycling mobility in the case study area, the competitive relation of the bicycle with the other modes and mode specific CO2 emission factors. The model provides an intuitive, straightforward approach, which allows urban planners, politicians and scientists to assess the value of current cycling mobility based on minimal input data and with minimal transport modelling knowledge. It assists in the redefinition of urban mobility planning by demonstrating the value of avoided CO2 emissions that cycling and non-motorized transport intrinsically provide. Application of the CVoC concept to Dutch cities has shown that some cities with a high cycling modal share have a low CVoC, because cycling trip lengths are relatively short and a well-established public transport system is available, implying substitution by walking or public transport.

© Association For European Transport and Contributors 2011

Combining the CVoC level with the cycling modal share allows the allocation of risk levels of transport unsustainability and describes appropriate policy responses. The unit CVoC, indicating the induced CO2 emissions of losing one kilometer cycling trip, can be used as a measure of the sustainability of the non- cycling part of the system. Cities with a low unit CVoC and a high bicycle share seem to be at lowest risk of becoming less sustainable in terms of transport related CO2 emissions. Further statistical analysis is expected to further confirm the results presented in this paper. Stacking several years of MON data will allow for adding social- economic and trip purpose dimensions to the calculation and interpretation. 7. REFERENCES ADB (2010), Reducing Carbon Emissions from Transport Project, Available at: http://www.adb.org/Documents/Evaluation/Knowledge-Briefs/REG/EKB-REG- 2010-16/default.asp, Accessed: 4 June 2011.

Bickel, P. and Friedrich, R. (2001) Environmental External Costs of Transport. Springer-Verlag Telos.

Browne, J., Sanhueza, E., Silsbe, E., Winkelman, S. and Zegras, C. (2005) Getting on Track: Finding a Path for Transport in the CDM, International Institute for Sustainable Development, Winnipeg, Manitoba.

Cavill, N., Kahlmeier, S., Rutter, H., Racioppi, F. and Oja, P. (2009) Economic analyses of transport infrastructure and policies including health effects related to cycling and walking: A systematic review. Transport Policy, 15(5) 291-304.

CBS (2009) Buurtkaart met cijfers 2008. Centraal Bureau voor de Statistiek / Kadaster, Zwolle.

Dalkmann, H. and Brannigan, C. (2007) Module 5e. Sustainable Transport: A Sourcebook for Policy-makers in Developing Cities, GTZ, Frankfurt.

DANS. (2009) The Dutch Mobility Survey 2009. Data Archiving and Networked Services (DANS). Available at: http://www.dans.knaw.nl, Accessed: 4 June 2011.

Fietsersbond (2011) Cycling City of the Year. Available at: http://www.fietsersbond.nl/fietsstad2011, Accessed: 1 June 2011.

Gotschi, T. and Mills, K. (2008) Active Transport for America: The Case for Increased Federal Investment in Walking and Cycling, Rails-to-trails Conservancy, Washington, DC.

Hausman, J.A. (1978) Specification tests in econometrics. Econometrica, 46(6) 1251-1271.

© Association For European Transport and Contributors 2011

Huizenga, C. (2009) Common Policy Framework on Transport and Climate Change. Available at: http://www.slocat.net, Accessed 25 October 2010.

International Energy Agency (IEA). (2008) World Energy Outlook, Paris, France. Johnson, N.L., Kotz, S., Kemp A. (1993) Univariate Discrete Distributions (2nd Edition), Wiley.

Krizek, K. (2004) Estimating the economic benefits of bicycling and bicycle facilities: An interpretive review and proposed methods, In TRB (Transport Research Board), TRB 2004 Annual Meeting, Washington D.C., January 11–15, 2004. TRB: Washington, DC.

Litman, T. (2004) Quantifying the Benefits of Nonmotorized Transport for Achieving Mobility Management Objectives, Victoria Transport Policy Institute, Victoria, Canada.

Lind, G. (2005) CBA of cycling: summary of conference proceedings. In Nordic Council of Ministers, CBA of Cycling: Nordic Council of Ministers’ seminar. Stockholm, Sweden, February 1–2, 2005. Nordic Council of Ministers: Copenhagen.

Lind, G., Hydén, C., and Persson, U. (2005) Benefits and costs of bicycle infrastructure in Sweden, In Nordic Council of Ministers, CBA of Cycling: Nordic Council of Ministers’ seminar. Stockholm, February 1–2, 2005. Nordic Council of Ministers: Copenhagen.

Luce, R.D. and Suppes, P. (1965) Preference, utility and subjective probability. In R.D. Luce, R.R. Bush and E. Galanter (ed.), Handbook of Mathematical Psychology, John Wiley & Sons ,New York.

Massink, R., Zuidgeest, M.H.P., Rijnsburger, J., Sarmiento, O.L. and van Maarseveen, M.F.A.M., 2011. The Climate Value of Cycling. Natural Resources Forum, 35(2011)2: 100-111.

Ministry of Transport, Public Works and Water Management and Fietsberaad. (2009) Cycling in The Netherlands, Ministry of Transport, Public Works and Water Management and Fietsberaad.

MON (2008). Mobility Survey Netherlands. Data source year 2008. Rijkswaterstaat Dienst Verkeer en Scheepvaart (RWS DVS), The Netherlands.

Ortúzar, J. de D. and Willumsen, L.G. (2001) Modelling Transport. Third ed., John Wiley & Sons, Chichester.

Saari, R., Metsäranta, H., and Tervonen, J. (2005) Finnish guidelines for the assessment of walking and cycling projects. In Nordic Council of Ministers, CBA

© Association For European Transport and Contributors 2011

of Cycling: Nordic Council of Ministers’ seminar. Stockholm, February 1–2, 2005. Nordic Council of Ministers, Copenhagen.

Sakamoto, K., Dalkmann, H. and Palmer, D. (2010) A Paradigm Shift Towards Sustainable Low-Carbon Transport – Financing the Vision ASAP, Institute for Transport & Development Policy, New York, United States.

Sælensminde, K. (2004) Cost-benefit analyses of walking and cycling track networks taking into account insecurity, health effects and external costs of motorized traffic, Transport Research Part A: Policy and Practice, 38(3) 593–606.

Small. K.A. and Hsiao C. (1985) Multinomial logit specification tests, International Economic Review, 26: 619.

© Association For European Transport and Contributors 2011