land-use impacts in transport appraisal
TRANSCRIPT
lable at ScienceDirect
Research in Transportation Economics 47 (2014) 82e91
Contents lists avai
Research in Transportation Economics
journal homepage: www.elsevier .com/locate/retrec
Land-use impacts in transport appraisal
Maria B€orjesson a, *, R. Daniel Jonsson a, 1, Svante Berglund b, 2, Peter Almstr€om b, 3
a Centre for Transport Studies, Royal Institute of Technology, SE-100 44 Stockholm, Swedenb Centre for Transport Studies, WSP Analysis & Strategy, Sweden
a r t i c l e i n f o
Article history:Available online 4 November 2014
JEL classification:H43C25D61R41R42
Keywords:Cost-benefit analysisTransport planningLand-use planning
* Corresponding author. Tel.: þ46 8 790 68 37.E-mail addresses: [email protected] (M
abe.kth.se (R.D. Jonsson), svante.berglund@[email protected] (P. Almstr€om).
1 Tel.: þ46 8 790 96 37.2 Tel.: þ46 10 722 86 18.3 Tel.: þ46 10 722 86 15.
http://dx.doi.org/10.1016/j.retrec.2014.09.0210739-8859/© 2014 Elsevier Ltd. All rights reserved.
a b s t r a c t
Standard cost-benefit analysis (CBA) does not take into account induced demand due to relocationtriggered by infrastructure investments. Using an integrated transport and land-use model calibrated forthe Stockholm region, we explore whether this has any significant impact on the CBA outcome, and inparticular on the relative ranking of rail and road investments. Our results indicate that induced demandhas a larger impact on the benefit of rail investments than on the benefit of road investments. The effecton the relative ranking is still limited for two reasons. First, the number of houses that are built over 20e30 years is limited in comparison to the size of the existing housing stock. Second, the location of mostof the new houses is not affected by any single infrastructure investment, since the latter has a marginaleffect on total accessibility in a city with a mature transport system. A second aim of this paper is toinvestigate the robustness of the relative CBA ranking of rail and road investments, with respect to theplanning policy in the region 25 years ahead. While the results suggest that this ranking is surprisinglyrobust, there is a tendency that the net benefit of rail investments is more sensitive to the future planningpolicy than road investments. Our results also underscore that the future land-use planning in the regionin general has a considerably stronger impact on accessibility and car use than individual road or railinvestments have.
© 2014 Elsevier Ltd. All rights reserved.
1. Introduction
Most studies analyzing uncertainty in cost-benefit analysis(CBA) outcomes focus on errors in transport model outputs (BeserHugosson, 2005; Brundell-Freij, 2000; De Jong et al., 2007; Zhao &Kockelman, 2002). De Jong et al. (2007) find, however, that theuncertainty induced by modeling errors is small compared to un-certainty induced by future scenario assumptions. Hansson (2007)and Mackie and Preston (1998) discuss uncertainty induced byomitted effects, model errors, input assumptions and valuations.B€orjesson, Eliasson, and Lundberg (2014) analyze robustness of CBAwith respect to various input data and valuations. The main pur-pose of this paper is to explicitly study uncertainty of CBA outcomesdue to uncertainty in future land-use. To the author's knowledgethis is not done in previous research, although the importance of
. B€orjesson), [email protected] (S. Berglund), peter.
land-use impacts in appraisal were identified as a key challenge atthe International Transport Forum (Worsley, 2011).
The transport and land-use systems are mutually dependent oneach other in the transportationeland-use cycle (Kelly, 1994). Thisis, however, not taken into account in standard transport CBA.Swedish and British guidelines for infrastructure appraisal(Department for Transport, 2009; Swedish TransportAdministration, 2012) discuss land-use effects only very briefly.When applying cost-benefit analysis, it is usually the ranking ofmany investments that are most relevant.4 A central hypothesistested in this paper is that the future land-use matters when railand road investments are ranked. We analyze two different land-use effects in separate sub-studies.
In the first sub-study we explore to what extent the evaluatedinvestments influence the future land-use and thereby the traveldemand, as suggested by Goodwin and Noland (2003), Hills (1996),Litman (2007), Noland (2001) and SACTRA (1999). We denote this
4 CBA is usually not used to determine the infrastructure budget and there areglobally uncertain parameters such as discount rate and traffic growth whichsubstantially affect the absolute outcome of all investments. If these change thenthe cut-off rate for what is good value for money will also change, which essentiallymeans that it is the ranking of investments that is relevant.
Table 1Population density (inhabitants/km2) in selected cities in Europe.
City Populationdensity Inh./km2
Source
Stockholm 3597 Statistics Sweden (2013)London 5632 National Statistics, UK (2010a)Manchester 4002 National Statistics, UK (2010b)Munich 4282 Statistisches Bundesamt
Deutschland (2000)Oslo 3192 Statistisk sentralbyrå (2012)
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e91 83
effect induced demand from land-use changes. Induced demandfrom short- and medium-term behavioral responses, such as tripfrequency, route and mode choice and car ownership, is wellestablished.5 In the present study we take them into account in thetransport model and in the CBA, but do not explicitly study theeffect of them on the CBA.6
Smart Growth advocators argue that transit investments mayhelp to form higher density, while new highway investments tendto lead to the opposite, i.e. more urban sprawl (Bernick & Cervero,1997; Newman & Kenworthy, 1989). Litman (2007) and Noland(2001) hypothesize that households tend to locate further awayfrom the city in response to highway capacity expansions,increasing the vehicle kilometers traveled over and above the in-crease due to the short-term responses. These studies suggest thatinduced demand from land-use changes could increase the netbenefit for rail investments more than road investments, becausenegative externalities (congestion, emissions and accidents) arisefrom induced car use and the use of transit infrastructure is usuallymore dependent on a dense and structured land-use.
Using Australian data, Newman and Kenworthy (1988) find thatincreased average car speed increases fuel consumption per capitathrough land-use changes in the urban area. Rodier (2001; 2002;2004) show that land-use changes induced by highway in-vestments account for about 50% of the travel demand increase.Marshall and Grady (2005) find, on the contrary, that land-useimpacts have only a small effect on travel demand in case oflimited road capacity because congestion constrain the urbansprawl in any case. Condor and Lawton (2002) find that the need fornew transport investments is overestimated if not taking land-useeffects into account because strategic planning could be a substi-tute for investments. Cervero and Kockelman (1997) find thatcompact development in terms of high density, pedestrian orientedtransport systems and land-use diversity in the San Francisco BayArea reduces motorized travel significantly, which may indicate areduced benefit from road investments. The varying conclusionsfrom these studies may be caused by different land-use andcongestion conditions. Preferences for high density may also varybetween geographical areas due to self-selection (people withpreferences for high density may to a larger extent live in densecities). Different investments have also different functions in thetransport system.
In the second sub-study we explore to what extent CBA out-comes depend on the future planning policy in the region, i.e. towhat extent the future planning strive for high public transitaccessibility and concentration of new housing, over a period of 25years ahead. The future planning is genuinely uncertain because ofthe highly decentralized planning system. The County of Stockholmcomprises 26 municipalities and each municipality decides aboutits own land-use. There is no long-term land-use plan for the regionthat the municipalities must comply with.
There are some prior studies quantifying the impact of uncertainfuture land-use. Pradhan and Kockelman (2002) find that outputsof the transport model is less variable than outputs of the land-usemodel, because the former seeks equilibrium. Ashley (1980) andZhao and Kockelman (2002) find that uncertainty is likely to inflateover the model steps (i.e. location, trip frequency, destination,mode and route choice) except in the route choice, because larger
5 Næss, Nicolaisen, and Strand (2012) show that they have significant effect onthe CBA outcomes.
6 Our transport model was able to reasonably well predict the responses in termsof frequency, mode, destination and route choice, when the congestion chargeswere introduced (Eliasson, B€orjesson, Brundell Freij, Engelson, & Van Amelsfort,2013), which suggests that the more short term travel behavioral responses aresufficiently taken into account in our analysis.
transport demand leads to more congestion and thereby reduceddemand. They also find that input variables accumulating overtime, such as growth rates, induce larger uncertainty in theoutcome than other input data. Other studies have found that un-certain socioeconomic forecasts are a significant source of uncer-tainty (Harvey & Deakin, 1996; Rodier & Johnston, 2002;Thompson, Baker, & Wade, 1997). In the present study, we do notconsider uncertainty in population growth or demography. Since,however, the growth rate affects all investments in a similar way,their effect on the ranking are limited (B€orjesson et al., 2014).
We apply a large-scale integrated land-use and traffic modelestimated and calibrated for the Stockholm region and evaluate sixrail and road investments in the Stockholm region. Some of themare more peripheral and others more central, affecting accessibilityin large parts of the region.
The impact of induced demand in the first sub-study is esti-mated by simulating two land-use patterns over the period2006e2030, one in which the investment (for which the CBA ismade) is assumed to have been introduced in 2006 and one inwhich it has not. The impact of induced land-use changes on theCBA outcome is then calculated by using different land-use patternsin the traffic forecast model in the build and no-build scenario. Thesecond sub-study investigates the impact of future land-use on thecost-benefit analyses. Three cost-benefit analyses for each invest-ment is made, assuming different planning policies over the years7
2006e2030: one where the planning strives for high density andhigh public transit accessibility, one where the planning is orientedtowards low density and low accessibility to public transit, and onein between, following current planning trend in terms of densityand public transit accessibility of new housing. Three land-usepatterns for 2030, corresponding to each policy, are simulated inthe land-use model.
In Section 2 we describe the transport system and the land-usein the Stockholm region and Section 3 describes the models in use.Section 4 describes the method including experimental setup, thescenario assumptions and description of investments. Section 5contains the results, and Section 6 discusses the results. Section 7concludes.
2. The stockholm region
2.1. Land-use and transport systems
Regions face different degrees of freedom when developing thefuture land-use. In a built up dense region with strong restrictions,few realistic alternative development paths may be open (in suchregions a land-use model would just fill in empty spaces in thelandscape) while several different paths may be open in other
7 25 years is the normal forecasting period, beyond that assumptions input fac-tors become very uncertain. The effects on land-uses changes may be proportion-ally larger with longer forecasting periods.
Table 2Floor space (million m2) in Stockholm in 2010 and potential floors pace increaseaccording to The Regional Planning Office of Stockholm (RUFS, 2012).
Floor space Potential floor space increase
Regional Center 59 115Inner suburb 32 448Outer suburb 39 5603Total 130 6166
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e9184
regions (where a land-use model would be more useful). Thesedifferences raise the issue of to what extent insights from Stock-holm can be generalized to other regions. Table 1 compares thedensity of Stockholm, 3597 inhabitants/km2, with a selection ofother European cities. The table indicates that Stockholm is notextreme in any sense.
The current density is important but for land-use forecasts it isthe potential for additional utilization of the area that is the keyfactor. The Regional Planning Office of Stockholm has made esti-mates of the potential for additional floor space in the differentparts of the region (RUFS, 2012) shown in Table 2. In the land-usemodel we have interpreted these potentials as restrictions.Table 2 shows that the potential for additional floor space iscurrently not a limitation for growth in floor space in any part of theregion.
Compared to many other European cities Stockholm has a highshare of transit trips. During the morning peak, this share reaches75% to and from the inner city. The high share of transit trips is dueto thewell-developed transit system. Other contributing factors arethe congestion charges and that the inner city of Stockholm is builton several islands, connected by bridges, implying relatively highroad congestion for a city of comparable size (2 million in theCounty of Stockholm). The bridges connecting the inner city to theouter city are bottlenecks in the road network. Congestion levelsare highest on arterials, where the relative increase of travel timesin the peak is 150-100 percent compared to free-flow travel times.The congestion levels were substantially higher before congestioncharges were introduced in 2006.
2.2. Investments
In this study we perform cost-benefit analyses for six rail androad investments in the Stockholm region, shown on the map inFig. 1. We selected investments that had previously been evaluatedby the national Road and Rail Administrations, and for which costestimates were available. From the list of projects we thus basedour selection on four criteria 1) both rail and road projects shouldbe represented, 2) projects of different size, 3) projects spreadacross the region and 4) estimates of construction costs should beavailable.
Investment 1, the Stockholm bypass, is amotorway bypass to thewest of Stockholm, mainly in a tunnel. Currently, there is only onehighly congested bypass connecting the northern and southernpart of Stockholm.
The investment cost is estimated to approximately V 2.6billion.8 There is already a decision to build this bypass and it willbe the largest investment in Stockholm since the 1960's.
Investment 2 is a commuter rail line that connects the north-eastern parts of the Stockholm County to the national and regionalrail network and the metro system. The investment cost is esti-mated to V 1.2 billion. Investment 3 is a central road tunnel that
8 Here and throughout the paper we have converted SEK to Euro using a con-version rate of 10 SEK/V.
increases the capacity of the current major north-south highwayarterial with an investment cost of aroundV 0.8 billion. Investment4 is a metro line from central Stockholm to Nacka located east of thecity center. The investment cost is approximately V 0.8 billion. In-vestment 5 is a peripheral main road that provides an improvedeast-west connection in the southern part of Stockholm County.The investment cost is approximatelyV 0.26 billion. Investment 6 isa light rail line that connects Flemingsberg (a major healthcare andeducational center with a regional and commuter train station) tothe large shopping area of Kungens kurva, two of the current metrolines and the commuter train station of €Alvsj€o. The investment costis estimated to aboutV 0.7 billion. All investment costs are given in2006 prices.9
3. Model system
We use an integrated land-use and transport model estimatedand calibrated for the Stockholm region. The transport modelconsists of a nested logit demand model including trip generation,mode and destination choice which is linked to a network assign-ment model. Both the transport model and the land-use modelwork on a zone level. Each zone typically includes less than 2000inhabitants. There are different demand models for commutingtrips and other trips.
The transport model is run iteratively with a land-use model.The accessibility from the transport model in each zone, measuredas the log-sum from the commuting trip model, is fed into the land-use model. The land-use model then assigns the workplaces andpeople (men and women in different age groups) that are relocat-ing within or migrating to the region (see next paragraph) to singleand multifamily houses in the different zones. This new populationis fed into the models for car ownership and license holding (whichboth has a large impact on travel demand and is highly dependenton type of housing and demography), which together form the dataset that goes into the transport model in the next step of iteration.The land-use model is run with a time step of five years, but thetransport model is only run for the years 2005, 2020 and 2030.
The share of single and multifamily houses in the region, theannual numbers of migrants to and from the region, and thenumber of existing households that relocate are given exogenouslyin the land-use model. The annual population growth 2006e2030in the region is fixed to 1% and the annual growth of the number ofjobs is fixed to 1.2%. The resulting increases in population andnumber of jobs are 28% and 35%, respectively, over the 25 years thatwemodel.We disregard the uncertainty in the CBA outcome arisingfrom uncertainties in these growth rates. According to the studiesreferred in the introduction these growth rates may induce a sig-nificant uncertainty in the absolute CBA outcome but much less inthe relative outcomes. Moreover, a higher growth rate implies alarger impact of the planning policy on the resulting land-usewithin the given time span. However, a yearly growth rate of 1%is large for a mature city.
The land-use model is a balance between demand and supply.The supply side is modeled as follows. The lot size is flexible in theland-use model but has a lower bound set by planning restrictions.The model assumes that some land cannot be built on at all, e.g. fornatural, cultural, or military reasons and areas close to roads etc. Inbuilt up areas the model allows for some additional densification ata slow pace. In some old areas in the city center and areas with veryhigh density no additional growth is allowed. In areas that are
9 The costs are taken from decision supports (Swedish Traffic administrationwww.trafikverket.se).
Fig. 1. Map showing the six rail and road investments in the study. Number 1, 3 and 5 are road investments while number 2, 4 and 6 are rail investments.
Fig. 2. In A and B the land-use is the same in both base scenario and investment case,
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e91 85
pointed out by planning authorities as development areas themodel allows faster development.
The demand for housing is governed by an indirect utilityfunction formulated as.
ui;k ¼ bpapi þ bcaci þ gkzi þ dkdi þ Qi; (1)
where ui, k is the utility of land-use of type k (single or multifamilyhousing or work places) in land-use zone i. api and aci are the ac-cessibilities to workplaces (measured as the log-sum from thecommuting trip model) with public transit and car, respectively, inzone i. zi is the local accessibility, represented by the populationdensity of zone i, and di measures crowding, represented by aconvex function of the population density of zone i. The parametergk is positive, capturing the benefit of increased local accessibilityas population density increases, and the parameter dk is negative,representing the disutility of increased crowding. The combinedeffect of gk and dk is that the utility to live in the zone increases withpopulation density up to a certain point. Above this point the utilityinstead declines with population density. The term Qi includes localfeatures of zone i, such as proximity to water and higher educa-tion10 etc.
Crowding parameters gk and di are important, since theydetermine the demand for more dense/sparse residential areas, i.e.the tendency for urban sprawl. The impact of the crowding pa-rameters, however, is moderated by planning regulations, presentin the model, which set an upper limit for densities. All the pa-rameters, gk, di and b, and the global share of one-family houses canbe calibrated to represent stronger or weaker planning policies. Inthe second sub-study these parameters will be varied to modeldifferent land-use policies, see further Section 4.2.
The demand for workplace locations is governed by an indirectutility function identical to Eq. (1), but where api and aci representthe accessibilities to the workforce, also measured as the log-sumfrom the work trip model.
There are no explicit prices in the model. This implies that it isnot modeled how the prices change and that the individuals do nottake prices into account when they make their localization choice.
10 These variables are also highly correlated with school quality.
Instead the lot sizes are flexible; so that higher demand for alocation reduces the lot sizes down to the smallest allowed lot size.
4. Method
This section describes the analysis method in the two sub-studies. In the first sub-study, described in Section 4.1, we exploreto what extent the land-use changes triggered by an investmentinduce new demand for the investment, and thereby affect theresulting CBA outcome. In the second sub-study, described in Sec-tion 4.2, we explore to what extent the uncertainty in the futureland-use induces uncertainty in the CBA outcome. In both sub-studies we simulate the land-use from 2006 to 2030 underdifferent assumptions. Section 4.3 describes the CBA methodology.
4.1. Sub-study 1: induced demand
In the first sub-study only two of the six investments describedin Section 2.2 are analyzed: the largest road investment (theStockholm bypass) and the largest rail investment (the commuterrail line).
with A simulated with a no investment transport system and in B with the investmentin the transport system. In C the base case has a non-investment land-use and theinvestment case an investment-adjusted land-use.
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e9186
The location of households and workplaces in the forecast year2030 is simulated over the period 2006e2030. Two land-use pat-terns are simulated for each analyzed investment, one in which theanalyzed investment has been introduced in 2006 and one inwhichit has not. The impact of the induced land-use changes is thencalculated by assuming different land-use patterns in the build andno-build scenario in the transport model, as illustrated in Fig. 2. Incase A, the land-use pattern simulated without the analyzed in-vestment is applied in the transport model both in the build and theno-build scenario. In case B, the induced land-use effect isaccounted for by assuming the land-use simulated with the in-vestment in the transport model in the build as well as the no-buildscenario. In case C, the induced land-use effect from the investmentis taken into account: different land-use scenarios are assumed inthe build scenario and in the no-build scenario.
First, we disregard benefits or losses arising from changes in theattractiveness of origins (terms 3e5 in Eq. (1)); this is further dis-cussed in the next paragraph. Second, the land-use and transportmodel is not perfectly consistent, because only the accessibility forcommuting trips governs the zonal location choice and commutingtrips constitute less than half of all trips.11 Third, even if an inducedland-use change would increase the total benefit for the mover, itmay also induce negative externalities in terms of congestion andemissions, reducing benefits for others. For these reasons it cannotbe a priori determined how the net benefit of cases A, B and C willbe ranked.
The consumer surplus on the transport market is well definedby the log-sum from the transport model, since it is a logit modelwithout income effects. It is commonly approximated with therule-of-a-half (RoH):
CS ¼Xi;j;m
12
�T0ijm þ T1ijm
��c1ijm � c0ijm
�; (2)
where Tijm is number of trips and cijm is the generalized transportcost in the OD relation ij bymodem. In our case the utility functionsalso includes the attractiveness of destinations, which is straight-forward to add to (2), by deriving an identical expression but withattractiveness12 wjwj expressed in monetary units, instead of cijm:
CS ¼Xi;j;m
12
�T0ijm þ T1ijm
��c1ijm � c0ijm
�
þXj
12
�T0ijm þ T1ijm
��w1
j �w0j
�:
(3)
When land-use is changed in response to some new infra-structure Equation (3) still omits changes in origin attractiveness(see Minken et al. (2003) and Martínez and Araya (2000) for amathematical derivation of RoH in an integrated land-use andtransport model). In this paper, however, we omit the utility changearising from changes in origin attractiveness and calculate CS by (3).
4.2. Sub-study 2: uncertainty in planning future
In the second sub-study, we neglect the effect of land-usechanges induced by the investments. Instead we concentrate onthe robustness of the CBA outcomes with regard to the planningpolicy in the region.
11 It is difficult to estimate the effect on location choices of accessibility forcommuting trips and other trips explicitly because they are highly correlated.12 Attractiveness wj consists of size variables (e.g. number of work places) in thetransport demand model, which changes between land-use scenarios because thelocation of workplaces changes.
Three different planning policies are constructed and imple-mented in the land-usemodel: Central, Trend and Peripheral. Thesepolicies were used by authorities when evaluating and forecastingthe regional development plan13 for Stockholm for 2010. Theresulting plan is to a large extent an outcome of a negotiation be-tween the regional level and the municipalities.
We perform one cost-benefit analysis for each investment andplanning policy. Since the land-use remains unchanged betweenthe build and the no-built scenario in the transport network in thisanalysis, the origin and destination attractiveness remains un-changed and do not affect the CS calculation.
The planning policies are implemented in the land-usemodel bymodifying parameters in the utility function, Eq. (1), the exoge-nously given global share of single and multifamily housing in theregion and the utility function corresponding to Eq. (1) for worktrips. The utility function is modified with respect to the tolerancefor housing density (dk) and sensitivity for public transit and caraccessibility (bp and bc) for both households and work places.
The Trend policy is calibrated to reproduce current long-termtrends in land-use pattern, in terms of how the population den-sity declines with distance to the central business district, tolerancefor housing density, and demand for public transit accessibility. 58%of the additional population (the population increase) is assumedto move into multifamily dwellings, which also reflects the trend inthe region over the past 30 years. In the Central policy 78% of theadditional population is assumed to move into multifamily housesand the tolerance for high density is considerable. A higher weightis assigned to the accessibility with public transit (bp). In the Pe-ripheral policy we assume that a minority, 27%, of the new housingunits are multifamily houses. A higher weight is assigned to theaccessibility with car (bc) and there is a low tolerance for denseland-use.
Table 3 describes the land-use pattern simulated for 2030,resulting from the three planning policies. In the Trend scenario,the resulting population growth in the regional center is 30%. In theCentral and Peripheral policies, the land-use model calculates that60% and 14%, respectively, of the growth in population will takeplace in the regional center. These numbers indicate a substantialdifference between the resulting land-use patterns.
The realism of the different scenarios can be assessed from theperspective of historical development trends and current plans. Thecurrent regional development plan for Stockholm14 is a mix of theCentral and Trend planning, with regard to regional distribution ofpopulation and density. The peripheral planning differs most fromthe current regional plan but since themunicipalities are not boundto the regional plan this scenario is still not unrealistic. The freedomof a decentralized planning system can be and is used by somemunicipalities to depart from regional plans, which in some casesresults in a more dispersed settlement pattern.
4.3. CBA details
The CBA in this paper follows in general Swedish guidelines forinfrastructure appraisal (Swedish Transport Administration, 2012).
4.3.1. InvestmentFor the purpose of comparison between the investments we
have assumed that the whole investment cost is paid in the year2019, that the investment is in use from 2020 and the benefits are
13 Regional plans in Sweden are not legally binding and provide only guidance.The national infrastructure plans use the land-use in the regional plan.14 Regional Utvecklingsplan f€or Stockholm 2010. http://www.tmr.sll.se/Global/Dokument/publ/2010/RUFS10_hela.pdf.
Table 3Output from the land-use model in the base scenario: population distribution in 2010 and in the scenarios studied. The total population in the region is assumed to be fixedbetween scenarios.
Percent of population 2010 Percent of additional population 2010e2030 Percent of population 2030
Central Trend Peripheral Central Trend Peripheral
Regional center 46 59 30 14 48 44 41Inner suburb 25 21 36 38 24 26 27Outer suburb 29 20 34 48 28 30 32
Table 4CBA under different assumptions of induced land-use changes. Scenario definitions(A, B and C) from Fig. 2.
Stockholm bypass Commuting rail line
A B C A B C
Reduced traveltimes Car
28216 28596 28455 142 �30 1122
Reduced travelcost, Car
22102 22241 22194 17 18 57
Reduced traveltimes Transit
11256 11310 11287 8514 8612 8582
Destinationattractiveness
0 0 488 0 0 649
Freight 29,532 29,580 29579 46 17 374Total CS 91106 91727 92002 8719 8617 10784Running cost �2095 �2077 �2101 �3861 �3859 �4787Ticket revenues 256 261 269 1012 1006 1921Total PS ¡1839 ¡1817 ¡1832 ¡2848 ¡2853 ¡2866CO2 �4741 �4770 �4797 100 92 339Traffic safety �2661 �2694 �2768 180 179 687Other �453 �456 �458 10 9 33Total externalities ¡7855 ¡7920 ¡8023 289 280 1059Congestion Charges �20332 �20376 �20371 0 0 0VAT tickets 64 65 67 253 252 480Fuel tax 9530 9589 9643 �201 �186 �681Sum Government ¡10738 ¡10722 ¡10661 52 66 ¡201Net benefit 70673 71268 71487 6212 6110 8776Investment cost 25000 25000 25000 9000 9000 9000MCPF 8052 8045 8050 3555 3556 3560BCR 2.3 2.4 2.4 0.2 0.2 0.4
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e91 87
assumed to increase by 2% per year due to traffic increase over thisperiod. According to the CBA guidelines the discount rate is 3.5%and the economic life of the investments is 60 years and the mar-ginal cost of public funds (MCPF) is equal to 1.3.
4.3.2. Consumer surplus (CS)The consumer surplus is calculated using RoH defined by (3),
with values of time from Swedish guidelines. They amount to 8.7(commuting) and 5.9 (other purposes) V/h for car and 6.9(commuting) and 5.3 (other purposes) V/h for public transit in-vehicle time. Public transit auxiliary and waiting time is weighted1.5 of in-vehicle travel time. Commercial traffic has a value of timeof 39 V/h.
Distance cost for cars is 0.18 V/km. Commercial transport con-sists of both heavy and light trucks, and is assumed to on averagehave a distance cost of 2.5 times a private car.
4.3.3. Producer surplus (PS)The cost of operating a public transit line is based on the number
of departures per day, the occupancy rate and vehicle type. Costs forrolling stock and maintenance are included, but not infrastructurere-investment. Different costs are assigned to buses, light rail,commuter rail, and the metro.
4.3.4. ExternalitiesExternalities are computed link by link, with emission rates and
fuel consumption differentiated by link type (but not by congestionlevel). The Swedish guidelines differentiate values depending onlocation (urban, rural, suburban etc.). For simplification, we haveassumed that the entire Stockholm County is an urban area.
4.3.5. GovernmentGovernment revenues are affected by fuel taxes, congestion
charges paid, and VAT paid for public transit fares.
5. Results
5.1. Result sub-study 1: induced demand
The magnitude of the relocation response induced by the twoinvestments, the Stockholm bypass and the commuter rail line, issimilar and relatively small. For both investments 25-30 000 peopleand 15e20 000 jobs are located in different zones in the 2030 buildand in the no-build scenarios. These numbers account for about 5%of the population and employment growth in Stockholm Countyand for about 1e1.5% of the total population and number of jobs.
5.1.1. Vehicle kilometers traveledThe increase in vehicle kilometers in response to the introduc-
tion of the Stockholm bypass in the county is similar in case A(land-use adapted to the traffic system in the no-build scenario,both in the build and no-build scenario), B (land-use adapted to thetraffic system in the build and no-build scenario) and in C (land-useare assumed to adapt to the traffic system between the build and
the no-build scenario); it is 3.68% in case A and 3.84% in both B andC. The effect in case A and B arise due to the short-term responses infrequency, mode, destination and route choices (since the land-useis fixed). The effect is similar in case C, indicating that the increasein car travel due to induced land-use changes is small compared tothe increase due to the more short term responses.
Not surprisingly, the commuter rail line has a much smallerimpact on the vehicle kilometers in the county (the effect on cartravel is more indirect and the also more local than the bypass), butthe effect does also differ more between scenarios: it is �0.07% inA, �0.10% in B and �0.42% in C. The larger reduction in vehiclekilometers in case C indicates that the more transit oriented land-use induced by the commuter rail line reduces vehicle kilometersmore than the short-term responses to the investment itself (interms of trip frequency, mode, destination and route choices).
The results we describe here are obtained by assuming theTrend planning, but the effects are similar in the Central and Pe-ripheral planning.
5.1.2. Cost-benefit analysisThe full cost-benefit analyses are shown in Table 4. For the
Stockholm Bypass, the CS is very similar in all cases: they differ lessthan 1% between A, B and C. Attractiveness of destinations (due torelocation of workplaces and households) changes only in C.
The other components of the CBA, producer surplus (PS), ex-ternalities, and Government budget effects, do not show any
Table 5Relative changes in total travel distance with car and public transit in Stockholm County. Comparisons are made with the Trend base case.
Base (no investments) Central road Peripheral main road Stockholm bypass Metro Light rail Commuter train
Car travel distance Trend 0% 1% 0% 4% 0% 0% 0%Central �5% �4% �5% �1% �5% �5% �5%Peripheral 4% 5% 4% 9% 4% 4% 4%
Transit# trips Trend 0% 0% 0% 0% 1% 0% 1%Central 5% 5% 5% 5% 5% 5% 6%Peripheral �4% �4% �4% �4% �3% �4% �2%
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e9188
systematic differences between A, B and C. The increased cost fromexternalities in cases B and C are balanced by higher revenues of thegovernment from fuel taxes, since the externalities (exceptcongestion) arising from car traffic to is internalized. The benefit-cost ratios (BCR)15 are almost the same in case A, B and C.
For the Commuter train, both BCR and CS are approximately 20%higher for case C than for A and B. Moreover, the benefit fromreduced externalities, are more than 70% higher in case C. Againthese results indicate that the long-term responses in locationpatterns per se increases accessibility and reduce car use. Still theBCR changes only from 0.2 to 0.4, indicating that the effect on theranking of this investment within a transport plan is limited (seefurther discussion in the end of Section 5.2).
5.2. Result sub-study 2: uncertainty in planning future
In the next analysis we neglect the effect of land-use changestriggered by the investments and concentrate on the robustness ofa CBA ranking with regard to the general planning future in theregion over 25 years. The presented effect on in travel distance andtrips are only due to short and medium term responses in tripfrequency, mode, destination and route choices since land-use iskept fixed in the build and no-build scenario.
5.2.1. Vehicle kilometers traveledThe upper rows of Table 5 show the percentages increase in
vehicle kilometers in the Stockholm County for all combinations ofinvestments and land-uses, relative to a base case where no in-vestments and the Trend scenario are assumed. In the base case,where no investments are added, the vehicle kilometers traveled is5% lower in the Central scenario and 4% higher in the Peripheralscenario in. These differences are in general larger than those be-tween investment scenarios. It is only the Stockholm bypass thatgives an increase in vehicle kilometers of the same magnitude asthe difference between the planning scenarios. The public trans-port investments and the peripheral road do not have any notableeffect on the car travel in the region.
The total travel distance with public transit (still relative to theTrend base case) is shown in the bottom rows of Table 5. Again theland-use has a substantially larger effect than any of the in-vestments. Relative to the Trend base case scenario, the publictransit travel is 4% lower in the Peripheral scenario and 5% higher inthe Central scenario. The road investments have no notable effecton the transit use in an of the land-use scenarios. The transit in-vestments (the metro and the commuter train) have the largesteffects on transit use in the Peripheral land-use. In this land-use thelarger population in the suburbs uses these investments tocommute to the city Centre.
15 This ratio is here defined as the net benefit (including the marginal cost ofpublic funds) divided by the investment and running cost. Swedish appraisal usesthe net present value ratios (NPVR), which equals the BCRþ1.
5.2.2. Cost-benefit analysisNext we turn to how the Cost-Benefit analyses for the in-
vestments differ depending on land-use future. Table 6 shows theannual CS for the six investments, for each of the three land-usescenarios. The table shows also the change in CS relative to theTrend planning future and the change in total net benefit of theinvestments (including all effects such as produced surplus,external effects and Government budget effects) relative to theTrend planning future.
Comparing the Trend and the Central planning futures, there is atendency that the benefit for road investments reduces in theCentral planning and that the benefits for transit investments in-creases. The light rail is an exception; but this investment is ratherperipheral. The differences in CS between Central and Trend arelargest for the peripheral main road and themetro. Not surprisinglythe benefit of the peripheral main road is lowest in the Centralplanning, because this planning reduces car use in general and caruse in the periphery in particular. As expected, the metro (which isa relatively central investment) has the highest benefit in theCentral land-use.
For all investments the differences in CS between the Trend andthe Peripheral planning future are small (less than 5%). For both theperipheral main road and the central road the benefits are slightlyhigher in the Trend scenario than the Peripheral scenario. This mayseem counterintuitive, but higher levels of road congestion incentral part of the road network in the former could be oneexplanation. The public transit investments are more beneficial inthe Peripheral land-use than in the Trend, again because a largerpopulation in the suburbs uses them for commuting.
The Stockholm bypass increases accessibility in most parts ofthe county, as opposed to the other investments that increaseaccessibility more locally. For this reason the benefit is less sen-sitive to different planning policies; the difference in CS betweenthe planning policies is within 5%. As expected the benefit isslightly higher for the Peripheral scenario with highest vehiclekilometers traveled in the county and slightly lower for theCentral scenario with lowest vehicle kilometers traveled in thecounty.
The main conclusion from these results is that the benefit isfairly stable across future planning policies. Moreover, the benefit ofrather peripheral transit investments does not in general increasewith a more transit- and high density oriented planning. Similarly,benefit of road investments does not always increase in a moreperipheral, less transit oriented, land-use; in particular not in thepresence of road congestion. In a denser land-use pattern conges-tion will increase, increasing the benefit of higher road capacityparticular in central locations.
The BCR for the six investments, given each of the three plan-ning policies, are shown in Table 7. Only the road investmentsindicate a net benefit (BCR>1). There are only small differencesacross planning policies and the pattern is consistent with Table 6.The ranking of investments are stable, except that in the Centralplanning the central road has a higher BCR than the Peripheralroad, and this order is reversed in the other scenarios.
Table 6Annual consumer surplus (CS) [MV/year] for the six investments, for the three land-use scenarios. Relative differences of CS and total net benefit of the investments in percentbetween Central and Trend and Peripheral, respectively.
Trend Central Peripheral
CS CS CS relative to trend Net benefit relative to trend CS CS relative to trend Net benefit relative to trend
Central road 37 37 �2% 2% 36 �5% �2%Peripheral main road 14 11 �20% �17% 14 �3% �4%Stockholm Bypass 322 312 �3% �4% 329 2% 2%Metro 27 29 9% 11% 28 4% 4%Light rail 16 15 �6% �9% 16 1% 1%Commuter rail line 30 31 1% 3% 30 0% 0%
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e91 89
Finally we interpret the results in the light of the ranking ofobjects in the Swedish national investments plan 2010e2020. Inthe investment plan process, complete CBAs were carried out for480 investments, 417 road investments and 63 rail investments.These investments can be divided into approximately equally sizedsegments based on their BCR: i) 11% of the investments had a BCRabove 3 ii) 19% had a BCR in the range of 2e3, iii) 18% had a BCR inthe range of 1.5e2, iv) 25% had a BCR in the range of 1e1.5, v) 15% inhad a BCR in the range of 0.7e1 and vi) 11% had a BCR in the rangeof �4 to 0.7.
If the benefit of the peripheral main road with a BCR of 1.4 (inTrend) changed ±10% due to errors in the predicted future plan-ning, the BCR would increase to 1.5 or decrease 1.2. The BCR of themetro would change from 0.5 to 0.6 or 0.4 and the BCR of theStockholm bypass would change from 2.5 to 2.8 or 2.2 if the ben-efits changed ±10%. Hence, assuming an uncertainty interval ofabout 20% (±10%) of the benefits due to uncertainty in futureplanning (which is a larger change of net benefit between planningpolicies than for any of the investments in Table 6) the investmentswe have analyzed would still stay within the same segment of theBCR ranking.
6. Discussion
Despite good intentions, integrated land-use and investmentplanning this is hard to accomplish in Sweden, and in many othercountries, because responsibilities and decision power are dividedbetween actors. Normally the state or the region funds infrastruc-ture investments while municipalities have power over the land-use planning. A common problem is that politicians in municipal-ities in the periphery are less concerned with congestion and otherexternalities within the city center. Land-use forecasts are alsoinherently uncertain, usually more uncertain than traffic forecasts.The transport system is to a larger extent than the housing marketan outcome of forces working in different directions seekingequilibrium. Moreover, housing supply is uncertain because con-struction costs are uncertain and it is not only developed bymarketforces, but also by planners and politicians, whose decisions aredifficult to forecast and model.
Table 7Benefit-cost ratio (BCR) for the six investments assuming the three different land-use scenarios.
Investment Trend Central Peripheral
Central road 1.2 1.2 1.1Peripheral main road 1.4 1.1 1.3Stockholm Bypass 2.5 2.4 2.6Metro 0.5 0.6 0.5Light rail 0.1 0.1 0.1Commuter rail line 0.3 0.3 0.3
For these reasons it is a reassuring result that the CBA rankingsseem to be relatively stable with respect to future planning andinduced demand due to land-use adaptations. There are severalfactors adding to this result. First, the existing demand constitutesthe largest part of future demand because land-use patterns changeslowly. Over a period of 25 years the population is forecast to in-crease 28% in the region, but 78% of the future population wouldstill be living in the existing housing stock. This holds for mostmature cities, but may be different for investments built in previ-ously very sparsely or unpopulated locations, where existing de-mand would be limited compared to new demand or demandinduced by the investment.
Second, induced car use does often increase the benefit of roadinvestments, but this is balanced by negative effects in terms ofcongestion and emissions. Third, most investments are used bymany different categories of traffic which are affected in differentways. Fourth, the difference in BCR for the investments in a nationalplan is usually large. Hence, even if the net benefit changes 20%, theeffect on the ranking of investments is limited. 20% is the largestimpact of different assumptions of future planning policies or ofinduced demand that is found for any of the investments in thispaper.
In general, lowdensity increases cardependencyand reduces thepossibility of efficient public transport provision. In the introductionwe therefore hypothesized that high density planning tends to favorrail investments relative to road investment. Even if the benefit ofthe Stockholm bypass increases slightly in the Peripheral planningscenario and decreases slightly in the Central, our results do, how-ever, in general not support this hypothesis. One reason is thatcongestion increases in a denser land-use, which tends to increasethe benefit of road capacity expansion. Another reason is that thebenefit of peripheral public transit investments does not increasewith higher density. And most of the suggested public transit in-vestments are not located in the central parts of the region, simplybecause the transit system is less developed in the periphery.
The induced land-use changes from the largest planned roadinvestment in Stockholm since the 1960's, the Stockholm bypass,have almost no effect on the CBA outcome. One reason is that thebypass increases car accessibility in most parts of the county, sothat the relative change in accessibility between different locationsremains rather stable. Note though that this result holds under theassumption that the land-use demand and supply are ruled bymarket forces. If, however, an investment decision is well-inte-grated with an extensive land-use plan for the concerned area, theoutcome could be different from the land-use model forecast. Inthat case, however, the planned land-use effects could easily betaken into account in the CBA evaluation, since it is known. Thepoint is that large land-use adaptation, structuring or dispersing,cannot be expected just to appear but has to be deliberately plan-ned for.
However, the results are somewhat different in case of theCommuter rail line. The induced land-use adaptation reduces the
Fig. 3. Summary of the results on the impact of the land-use on the transport systemand vice versa.
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e9190
vehicle kilometers traveled four times as much as the shorter-termadaptations and the additional accessibility benefit due to the land-use adaptation is approximately 20% of the benefit of thecommuting rail line. This suggests that transit investments indeeddo have additional benefits, by increasing demand for denser andmore public transit oriented land-use development, as argued bySmart Growth advocators (Bernick & Cervero, 1997; Newman &Kenworthy, 1989). However, for the reasons discussed above, theimpact does not seem to be large enough to have any substantialimpact on the ranking of investments in a national plan.
The findings are summed up in Fig. 3. The top arrow in the figureindicates that this study gives rather weak support to the hypoth-esis that road infrastructure, without well-integrated land-useplanning, induces either land-use structure or sprawl. The addi-tional accessibility arising from the new investment is inmost casesmarginal compared to the original accessibility. We have onlystudied the effect of the bypass to underpin this result, but this is avery large investment. The relatively small effects from inducedland-use must, however, be seen in the perspective of a mature citythat has found its form and does not grow faster than 1% per year.On the other hand, no other cities in Sweden, and few cities inEurope, grow faster than Stockholm, which indicates that theseresults are representative for many other cities. For public transportinvestments, our result indicates that effect on land-use is larger.
As indicated by the lower arrow in Fig. 3, the land-use impactsachieved by planning have an important effect on the function ofthe transport system in terms of vehicle kilometers, public trans-port use, congestion and accessibility (the vehicle kilometers andpublic transport use are strongly dependent on the future land-usepolicy).
7. Conclusion
This paper studies whether uncertainty of future planning andlong-term induced land-use impacts from infrastructure in-vestments give rise to significant uncertainty in the CBA ranking ofinvestments, for instance in a national plan. The impact of short-and medium-term responses in form of trip frequency, mode,destination and route choices are not studied explicitly, but aretaken into account.
Our results suggest that the uncertainty in future planning doesnot influence the CBA ranking to any large extent. Moreover, wefind no support for the hypothesis that the benefit of rail in-vestments, relative to road investments, is systematically higher ifthe future planning stresses higher density land-use and publictransit accessibility.
Our results further suggest that the induced land-use changesfrom transit investments in fact add a significant reduction of caruse and increase the accessibility benefit. This may be valuableinformation to decision makers and supports the use of land-usemodeling in appraisal. However, the magnitude of this effect isstill not large enough to change the ranking of investments in anational plan to any large extent.
Another insight from this paper is that the vehicle kilometerstraveled in a region is to a larger extent determined by land-usepolicies than by infrastructure investments. This is consistentwith the conclusions of Zhao and Kockelman (2002) and Pradhanand Kockelman (2002). This suggests that it is the future land-useplanning strategy that, to a larger extent than transport in-vestments, leads to a more car or public transit oriented society.That is, even though the public transport system in Stockholm iswell-developed its competitiveness is reduced if the planning al-lows more sprawl.
To summarize, there is little evidence in this study suggestingthat it is important to run land-use models for transport CBA. Incase, however, that new settlements are planned as part of inte-grated land-use and infrastructure projects this should obviouslybe taken into account, but doing so does not require land-usemodeling.
References
Ashley, D. J. (1980). Uncertainty in the context of highway appraisal. Transportation,9(3), 249e267. http://dx.doi.org/10.1007/BF00153867.
Bernick, M., & Cervero, R. (1997). Transit villages in the 21st century. New York:McGraw-Hill.
Beser Hugosson, M. (2005). Quantifying uncertainties in a national forecastingmodel. Transportation Research A, 39(6), 531e547.
B€orjesson, M., Eliasson, J., & Lundberg, M. (2014). Is CBA ranking of transport in-vestments robust? Journal of Transport Economics and Policy, 48(2), 189e204.
Brundell-Freij, K. (2000). Sampling, specification and estimation as sources of in-accuracy in complex transport modelsdsome examples analysed by MonteCarlo simulation and bootstrap. In Proceedings of the European TransportConference.
Cervero, R., & Kockelman, K. (1997). Travel demand and the 3Ds: density, diversity,and design. Transportation Research Part D: Transport and Environment, 2(3),199e219. http://dx.doi.org/10.1016/S1361-9209(97)00009-6.
Condor, S., & Lawton, K. (2002). Alternative futures for tansportation and landuseeintegrated models contrasted with “Trend Delphi” Methods: the Portlandmetro results. Transportation Research Record, 1805, 99e107.
De Jong, G., Daly, A., Pieters, M., Miller, S., Plasmeijer, S., & Hofman, F. (2007). Un-certainty in traffic forecasts: literature review and new results for TheNetherlands. Transportation, 34, 375e395.
Department for Transport. (2009). Department for transport e Transport analysisguidance e WebTAG.
Eliasson, J., B€orjesson, M., Brundell Freij, K., Engelson, L., & Van Amelsfort, D. (2013).Accuracy of congestion pricing forecasts. Transportation Research A, 52, 34e46.http://dx.doi.org/10.1016/j.tra.2013.04.004.
Goodwin, P., & Noland, R. B. (2003). Building new roads really does create extratraffic: a response to Prakash et al. Applied Economics, 35(13), 1451. http://dx.doi.org/10.1080/0003684032000089872.
Hansson, S.-O. (2007). Philosophical problems in costebenefit analysis. Economicsand Philosophy, 23(2), 163e183.
Harvey, G. W., & Deakin, E. (1996). Description of the STEP analysis package. DHS,Incorporated. Retrieved from http://www.mtc.ca.gov.
Hills, P. (1996). What is induced traffic? Transportation, 23(1) http://dx.doi.org/10.1007/BF00166216.
Kelly, E. D. (1994). The transportation land-use link. Journal of Planning Literature,9(2), 128e145. http://dx.doi.org/10.1177/088541229400900202.
Litman, T. (2007). Generated traffic and induced travel. ITE Journal, 71(4), 38e47.Mackie, P., & Preston, J. (1998). Twenty-one sources of error and bias in transport
project appraisal. Transport Policy, 5(1), 1e7. http://dx.doi.org/10.1016/S0967-070X(98)00004-3.
Marshall, N., & Grady, B. (2005). Travel demand modeling for regional visioning andscenario analysis. Transportation Research Record: Journal of the TransportationResearch Board, 1921(1), 44e52. http://dx.doi.org/10.3141/1921-06.
Martínez, F., & Araya, C. (2000). Transport and land-use benefits under locationexternalities. Environment and Planning A, 32(9), 1611e1624.
Minken, H., Jonsson, D., Shepherd, S., J€arvi, T., May, A., Page, M., et al. (2003).A. Methodological Guidebook; prospects. Retrieved from http://www.infra.kth.se/courses/1H1402/Litteratur/pr_del14mg.pdf.
M. B€orjesson et al. / Research in Transportation Economics 47 (2014) 82e91 91
Næss, P., Nicolaisen, M. S., & Strand, A. (2012). Traffic forecasts ignoring induceddemand: a shaky fundament for cost-benefit analyses. European Journal ofTransport and Infrastructure Research, 12(3), 291e309.
National Statistics, UK. (2010a). Usual resident population: Census 2001, Key Statisticsfor urban areas (excel-file). Downloaded January 6, 2010.
National Statistics, UK. (2010b). Population estimates for UK, England and Wales,Scotland and Northern Ireland, Mid-2008 (excel-files). Downloaded January 6,2010.
Newman, P. W. G., & Kenworthy, J. R. (1988). The transport energy trade-off: fuel-efficient traffic versus fuel-efficient cities. Transportation Research Part A: Gen-eral, 22(3), 163e174. http://dx.doi.org/10.1016/0191-2607(88)90034-9.
Newman, P. W. G., & Kenworthy, J. R. (1989). Gasoline consumption and cities.Journal of the American Planning Association, 55(1), 24e37. http://dx.doi.org/10.1080/01944368908975398.
Noland, R. B. (2001). Relationships between highway capacity and induced vehicletravel. Transportation Research A, 35(1), 47e72. http://dx.doi.org/10.1016/S0965-8564(99)00047-6.
Pradhan, A., & Kockelman, K. M. (2002). Uncertainty propagation in an integratedland use-transportation modeling framework: output variation via UrbanSim.Transportation Research Record: Journal of the Transportation Research Board,1805(1), 128e135. http://dx.doi.org/10.3141/1805-15.
Rodier. (2001). Anatomy of induced travel using an integrated land use and trans-portation model in the Sacramento Region.
Rodier, C. J. (2002). Heuristic policy analysis of regional land use, transit, and travelpricing scenarios using two urban models. Transportation Research. Part E, Lo-gistics and Transportation Review, 7(4), 243.
Rodier, C. (2004). Verifying accuracy of regional models used in transportation andair quality planning: case study in Sacramento, California, region. Transportation
Research Record: Journal of the Transportation Research Board, 1898(1), 45e51.http://dx.doi.org/10.3141/1898-06.
Rodier, C. J., & Johnston, R. A. (2002). Uncertain socioeconomic projections used intravel demand and emissions models: could plausible errors result in air qualitynonconformity? Transportation Research A, 36(7), 613e631. http://dx.doi.org/10.1016/S0965-8564(01)00026-X.
RUFS. (2012). Regional growth, environment and planning. In Befolkning och Sys-selsatta på små områden år 2030 enligt RUFS 2010. Teknisk dokumentation.
SACTRA. (1999). Transport and the economy. London: Standing Advisory Committeeon Trunk Road Assessment.
Statistics Sweden. (2013). Land- och vattenareal per den 1 januari efter region ocharealtyp. År 2012e2013 (excel-file). Downloaded December 15, 2013.
Statistisches Bundesamt Deutschland. (2000). Gemeindeverzeichnis GV 2000,Gebietsstand: 31.12.2009 (Jahr) (excel-file). Downloaded January 16, 2011.
Statistisk sentralbyrå, Norway. (2012). Tettsteder. Folkemengde og areal, etter kom-mune. January 2012. Report. Downloaded september 23, 2012.
Swedish Transport Administration. (2012). ASEK 5-rapporter.Thompson, D., Baker, M., & Wade, D. (1997). Conformity: long-term prognoses for
selected ozone nonattainment areas in California. Transportation Research Re-cord: Journal of the Transportation Research Board, 1587(1), 44e51. http://dx.doi.org/10.3141/1587-06.
Worsley, T. (2011). The evolution of London's crossrail scheme and the development ofthe department for transport's economic appraisal methods. Discussion Paper2011-27.
Zhao, Y., & Kockelman, K. M. (2002). The propagation of uncertainty through traveldemand models: an exploratory analysis. Annals of Regional Science, 36(1),145e163.