Economic analysis of bicycle tracks using the HDM-4 model - case study for São Paulo city

In addi1on to technical studies, bicycle paths should be a product of planning and investment policies considering the ability of projects to generate socioeconomic beneﬁts, implemen1ng policie’s objec1ve analysis relevant to the feasibility of projects for the implementa1on of exclusive bike tracks. In this paper the Non-Motorized Traﬃc (NTM) model of the HDM-4 (Highway Development and Management) so@ware is applied for the analysis of diﬀerent alterna1ves for bicycle lanes, evalua1ng aspects such as current and poten1al cyclists demand, opera1on speed, capital costs and economic proﬁle of bicycle users. The combina1on of such variables leads to technical and economic alterna1ves whose analysis results relevant diﬀerences in their proﬁtability indicators. The analyses were carried out considering two scenarios; the ﬁrst comprises the analysis of the proﬁtability of alterna1ves structured with norma1ve guidelines that, in terms of speed, represent ideal opera1ng condi1ons; the second scenario consists of a more realis1c evalua1on for the city of São Paulo, considering speed restric1ons and diversifying the user proﬁle according to the per capita income of the main regions of the city. The results reﬂect interregional diversity about the desirability of bikeway projects based solely on the monetary beneﬁts of reduced travel 1mes.


INTRODUCTION
Road infrastructure, like any other type of infrastructure, is immersed within a planning cycle that begins with the idea of investment and ends with the completion or abandonment of a project for some reason associated with a speci ic technical or socioeconomic criterion. Socioeconomic evaluations must compile all the variables and technical studies of a project; therefore, only when feasibility is determined within the economic and social context of the region, it becomes plausible to state that projects of public agencies go through all the planning stages, regardless of the economic indicators obtained.
In Brazil, highway infrastructure projects meet well-de ined guidelines and methodologies to assess their technical feasibility, as well as to establish levels of service appropriate according to demand. However, in the public planning and investment sector, there is still a gap in the treatment of bicycle infrastructure projects, and it is a common practice to disregard the costs and bene its of non-motorized vehicle users.
Altering such conventional perspective, in this paper it was performed a technical and economic analysis of different situations of bicycle tracks projects in the city of Sao Paulo, which allows the identi ication of investment levels, socially viable, according to the cyclists' demand and other variables inherent to the projects.
The methodological approach of this work is based on the analysis of different alternatives for bicycle projects using the non-motorized traf ic (NMT) model of the Highway Development and Management (HDM-4) software, evaluating factors such as current and potential cyclists demand, operation speed, capital costs and users economic pro ile.
Thus, this study seeks to collaborate in the status quo of socio-economic studies by employing tools that can contribute to manage policies and optimization of public money, facilitating decision making process in bicycle projects within urban roadways in São Paulo city, considering that in recent years the expansion of the city's bicycle network is vertiginous, in contrast to the status of methodologies to assess the technical feasibility of bicycle infrastructure.

BACKGROUND
According to the Inter-American Development Bank (IDB, 2015), there are three types of cycling infrastructure: green or independent, consisting of lanes for the exclusive use of bicycles; segregated, which corresponds to lanes demarcated by paint or other type of separator depending on the speed of motorized traf ic; and shared, which is an open way to other traf ic modes, since the operating conditions allow more than one transport mode. According to Brazilian guidelines, the nomenclature of the cycling infrastructure mentioned correspond to bike track, bike lane and shared lane, respectively.
The classi ication of bicycling infrastructures, according to its level of segregation, is determinant in the cost of implementation and maintenance works, therefore, bike tracks are the most expensive investment within the three options mentioned above, being a scarce infrastructure in São Paulo's network. As published by São Paulo Traf ic Engineering Company (CET, 2021), the city's bicycle network has 684 km, of which only 154.3 km are bicycle tracks, 32.1 km are shared lanes, and the remaining 497.6 km are bicycle lanes.
In terms of planning, São Paulo has the Bicycle Paths Plan 2019-2028 (CET, 2020), released and discussed in 10 workshops that took place between years 2018 and 2019.
Compared to previous plans, its goals are ambitious, aiming to reach 1,800 km of network during the period between 2021 and 2028. The Bicycle Plan, still being in process of implementation, brings together elements of PlanMob (2015) and requests from civil society, prioritizing network connectivity and ef iciency.
With regard to legislation and rules for the cycling sector, a relevant change is attributed to Municipal Law No. 16.738 of November 7, 2017, whose of Article 4 states: "The implementation of bikeways should be preceded by public hearings and the presentation of demand, feasibility, and road impact studies, which should be fully disclosed on the website" (SA@ O PAULO CITY HALL, 2017).
The law has generated concern among organizations promoting bicycle use such as Vá de Bike and Ciclocidade. The apprehension to the law by cycloactivists is sustained in that bicycle infrastructure projects do not have enough demand to make the investment viable, due to the fact that bicycle trips are mainly generated and attracted after the construction of the bikeways (BASILIO, 2017), what at least suggests an existing repressed demand. However, the realities described are not enough to disapprove the practice of feasibility studies in bikeways projects, especially considering that traf ic generated, or potential demand are determining variables in socioeconomic analysis methodologies.

Non-Motorized Traffic (NMT) in the Highway Development and Management Tool (HDM-4)
The HDM-4 software, developed to technically and inancially evaluate the initiatives of public and private agencies that manage road networks, quanti ies the pro itability of strategies, programs, or individual projects by weighing the bene its in terms of decreased vehicle operating costs (VOC) and travel time savings.
In this study the tool so-called project within the software was employed, since the segregation of Non-Motorized Traf ic (NMT) is typi ied as a work to improve the existing infrastructure that changes the operating conditions of the different traf ic modes. NMT is considered as a passenger and freight transport mode and not only as a noise factor for motor vehicles (KERALI et al., 2000). Thus, in HDM-4 the costs and bene its of NMT are calculated separately for four types of vehicles: pedestrians, bicycles, carts, and cycle-rickshaws.
Gradually, both pedestrians and Non-Motorized Vehicle (NMV) users have gained space in socioeconomic feasibility analyses of projects of different types. The subject was addressed by the World Bank through the study developed by PADECO (1996) which had as object the inclusion of the NMT module in HDM-4, based on the concepts of Hoban (1987) that had already been applied in ield studies in Indonesia to determine the effect on operating speed caused by the con lict between motorized and non-motorized vehicles. Kerali et. al (2000) revised the NMT model in order to establish the inal speci ications to be adopted; as a result, changes were made regarding the effects of rolling resistance on the speed of non-motorized vehicles.
The NMT models were incorporated into the pre-existing models for Motorized Traf ic (MT) on highways, whose operation is considered uninterrupted and adequate to develop constant speeds. Nevertheless, it is a reality that NMT comes from metropolitan areas, industrial or commercial agglomerations, which suggest the presence of urban settlements adjacent to the highways, and without them, the incorporation of this module in the software would have been unfeasible or unnecessary. Theoretically, NMT travel on the shoulder of the highway or, in the worst case, share lanes with motor vehicles; this interaction generates an impact on the travel times of the entire leet that is quanti ied by the HDM-4 software models.
The purpose and characteristics of the models may prematurely discredit the HDM-4 software as a tool for analyzing NMT on urban roadways; however, the assumptions of the models constitute the technical foundation of some studies that are characterized by having taken advantage of the software's potential beyond its limitations. Kumar (2012), Chopra et al. (2017) and Yogesh et al. (2014) have developed different studies in India on the use or adaptation of the HDM-4 software on urban roads by means of strategic modeling to prioritize investments. In Bogota (Colombia), the Instituto de Desarrollo Urbano (IDU, 2009), calibrated the road deterioration and user cost models of HDM-4 for the management of the city's road network, restricting speed by means of a reduction factor capable of expressing numerically the effects of congestion and traf ic disruptions. In Brazil, Gueller (2012) compared the model developed by Tavakoli et al. (1992), proper management in urban roads of medium-sized cities, with modeling in HDM-4 software.
The studies showed satisfactory results, especially for de ining a schedule for ways rehabilitation or maintenance interventions, which depends on the Road Deterioration (RD) and Road Works Effects (RWE) models.
In HDM-4 the interaction between TM and NMT depends crucially on the existence and size of the shoulder on the road. By making an analogy with urban roads, the level of interaction between vehicular traf ic and bicycle traf ic is governed by the type of bicycle infrastructure.
In this case, shared lane, bike lane, and cycle tracks differentiate the interaction depending on the level of segregation, and it is possible to say that shared lane are infrastructures of high contact between both types of users, on bike lanes interaction decreases according to the separation technique (paint or retrore lective tacks), and on bike tracks there is no interaction between traf ic modes for most of the route. However, even under the best segregation conditions, several impedance factors persist, such as intersections, pedestrian interference, internal con lict between cyclists, bikeway capacity, among others, which reduce the operating speed and, consequently, affect the socioeconomic ef iciency of the infrastructure. Bennett and Greenwood (2000), in the software manuals point out that the mutually generated effects between the TM and NMT models are controlled and de ined through speed reduction factors. The speed of motorized traf ic can be reduced due to the presence of nonmotorized traf ic by means of impedance factor XNMT, and reciprocally the speed of nonmotorized traf ic is reduced by means of the factor XMT. The model limits the value of the XMT factor between 0.4 and 1, where values close to or equal to 1 tend to cancel out the con lict and values close to 0.4 are employed to add model impedance and represent critical operating conditions.
In this study it is assumed null the change in motorized vehicle speed by the implementation of the bikeway, so the effects on its operating costs are not evaluated. This assumption is based on the dif iculty of isolating side friction effects due exclusively to the presence of bicycles on urban roads in the area of in luence of a bicycle path project.
The main equations of the software models correspond to the operating speed, which is given by: (1) TRANSPORTES | ISSN: 2237-1346 5 : Operating speed (km/h).
!: Speed limited by the slope of the track (km/h). !"#$ : Desired speed by vehicle type and surface type (km/h). %& ': Speed limited by the surface irregularity (km/h). ()*: Reduction factor due to motorized traf ic and activities adjacent to the track. The variables in equation (1) are described as follows: (2) !"#: Desired speed (km/h). , -: Model coef icient associated to the irregularity. , . : Coef icient of the model associated to the slope.
$: Irregularity of the road. : Coef icient of the bicycle type vehicle associated with the grade of the road, predetermined in the model (-0.04).
However, the speed as a function of geometry depends on the slope the cyclist experiences, being higher when downhill and with some restriction when uphill, thus: (3) : Coef icient of the bicycle vehicle associated with the grade of the road for uphill (+) or downhill (-) travel: : being the number of descents and ascents per kilometer that is entered into the geometry module as a track characteristic.
Equations (2), (3) and (4) contain calibration coef icients or factors, which were de ined by Odoki and Kerali (2005) during model re inement and their values are shown in Table 1.

STUDY METHODOLOGY
The socioeconomic analysis in this study consists of two scenarios. The irst scenario comprises the analysis of the pro itability of different project alternatives structured with guidelines recommended by manuals or reference documents that, in terms of speed, represent ideal operating conditions. This scenario considers socio-economic characteristics of cyclists that already use bicycle infrastructure; however, it does not represent the diverse socioeconomic conditions of potential users located in other regions in São Paulo city, which do not have expressive development of bicycle infrastructure. The second scenario, on the other hand, consists of a more realistic assessment considering lower operating speeds associated with restrictions such as high slopes, interruptions, space limitations for implementation, con licts with pedestrians, among other impedance factors. In addition, the second scenario evaluates the diversi ication of the user pro ile by analyzing up to six income levels that represent the average per capita of the main regions of the municipality of São Paulo.

Variables SelecAon
The analysis is based on a combination of explanatory variables that meet the prerequisite of being de ined as input data in the HDM-4 model and that are available in studies and documents with broad technical potential to be used, such as: speed, traf ic, and pavement typology

Traf ic
There are numerous demand scenarios that may represent in the bicycle network of a metropolis, therefore, in order to properly represent the bicycle demand in the city, 30 counting stations from Association of Urban Cyclists of São Paulo - (Ciclocidade, 2018) and 34 stations from CET (2019)   Comparison of the two data sources disclosed similarities in central tendency and position values. Bike tracks have an average traf ic that varies between 1,469 and 1,564 bicycles per day, according to Ciclocidade and CET respectively. The bike lanes have lower average demands, with values around 710 and 786 bicycles per day. In fact, the third quartile shows that 75% of the bike tracks have a traf ic equal or inferior to 1,970 bicycles/day, while in the bike lanes the demand is reduced to 908 bicycles/day. Thus, to ensure coverage of the main demands analyzed, it was adopted as a baseline value for socioeconomic analysis a traf ic of 500 bicycles per day, increasing gradually up to 2,000 bicycles per day. The higher demands were included in the analysis through the generated traf ic, which was de ined through similar statistical analyses, considering the percentage increment of demand in the irst years of bikeways implementation. However, the prominent amplitude of the data set required the evaluation of a wide range of values integrated by the percentages of 10%, 20%, 30%, 40%, 50%, 100% and 150%. The last two values are included to represent success stories such as the Faria Lima Avenue bike track, which in 2019 reached values above 6,000 bikes/day and managed to double demand in the irst years of operation.
In summary, the socioeconomic analyses were performed for six demands corresponding to 500, 750, 1,000, 1,250, 1,500 and 2,000 bicycles per day and each of these values was increased and analyzed with the set of percentages de ined for the generated traf ic.

Traf ic growth rates
According to METRÔ (2019), between 2007 and 2017 there was an increase of 0.1 million bicycle trips, for which, the development of bicycle infrastructure and the implementation of bike share programs were instrumental. In this context, the linear growth rate of demand corresponds to 3.33% (has been approximated to 3%). However, since this is a study that seeks to measure the sensitivity of mutable and exogenous variables such as traf ic, it is considered appropriate to evaluate other rates that represent an optimistic (6%) and pessimistic scenario (0% -no growth). Krizek et al (2006) state the design of bikeways, which depends on several variables that are speci ic to the location, can affect not only the functional life of the bicycle path, but also its economic life.

Pavement types -capital and maintenance costs
Bicycle reference guides such as American manual (AASHTO, 2013) and the Dutch manual (CROW, 2011) do not standardize pavement structures, leaving the technical decision subject to the support conditions of each site; however, both guidelines emphasize the importance of considering the possible overloads that request bikeways, such as maintenance vehicles or infringing vehicles.
The pavement was included by varying the thicknesses of different structures: Asphalt Concrete (AC) with thicknesses of 20 mm, 30 mm and 40 mm, Porous or permeable Friction Courses (PFC) of 40 mm, Double Surface Treatment (DST) and 100 mm non-dowelled plain Concrete (C), and for all cases a 100 mm granular underneath base; subgrade reinforcement varied between 0, 150 and 400 mm. The unit prices were adopted from of icial and governmental databases. In order to establish representative values for bikeways investments, the data were diagrammed as shown in Figure 2. In order to represent different investment possibilities and to simplify the economic scenarios, after a previous analysis of different bicycle infrastructure cost, it was opted to analyze a range of capital costs that oscillate between R$400,000 and R$1,100,000. Since HDM-4 does not have deterioration models that express the effects of maintenance, in this study it was proposed to program global costs for periodic and routine maintenance, both functional and structural. The annual maintenance of the bicycle track amounts to approximately R$20,000 per kilometer with slight differences depending on the type of road surface; the igure is consistent with the reference costs of the National Planning Department (DNP) of Colombia (2017), which gathers experiences of consolidated networks such as the Bogotá bikeways system.

Speed (S)
Bikeway projects must satisfy the primary requirement of providing as direct a route as possible, and all factors affecting travel time, such as speed, traf ic lows, delays, and detours, must be evaluated. In this study, the impedance is expressed numerically by the XMT factor which, under speci ic model conditions, reduces the design speed until the selected operating speeds are obtained as discrete values to integrate the factorial of project alternatives.
In the normative scenario, a design speed of 30 km/h is adopted as recommended by the global organization Institute for Transportation and Development Policy (ITDP, 2017), as well as other international guidelines such as IDU (1999), CROWN (2011), AASHTO (2012), DPTI (2015), while the operating speed is set with systematic reductions through the XMT iteration. In the regional scenario, the selected speeds are more critical and correspond to the results of the research analysis of the Association of Urban Cyclists of São Paulo -Transporte Ativo and Ciclocidade (2016), from the selected operating speeds, and the XMT factor and design speed are iterated. It should be noted that the factor that governs the model results is not directly associated with the operating speed, but rather with the user's speed gain. Thus, the bene its in the normative context in terms of speed are higher when operations above 20km/h are evaluated. Table 1 presents the XMT factors, design and operating speeds and speed gains for each of the scenarios evaluated.
Another factor that regulates the speed in the HDM-4 model is the surface roughness in terms of IRI (International Roughness Index), which increases gradually depending on an environmental coef icient (selected by default). In this study an initial IRI of 3 m/km is The roughness model increases gradually as a function of weather conditions. It is worth noting that, to roads, in HDM-4 the rate of deterioration of roughness is conditioned to the sum of other degradations, such as loss of structural capacity, cracks, wear, potholes, among others, which are strictly associated with traf ic loads. In the non-motorized traf ic models, traf ic loads are not considered, so degradations do not impact the irregularity and the IRI value is only affected by the weather factor predetermined in the software according to Table 1. In conclusion, for non-motorized traf ic paths, the software simpli ies the roughness model and provides a factor that increases the user's impedance conditions.

Vector Space Size
Since the sampling is being performed in 3 sampling variables, there are several possibilities of sampling vectors, i.e., each bikeway section can receive a discrete value of each sampling variable, as shown in the following equation: In other words, the i-th bicycle track can belong to the j-th traf ic category, have a k-th operating speed, and have an associated with a l-th pavement cost. With these variables it was performed a joint-distribution analysis (more commonly known as crosstab analysis). In addition to the base alternative, each section is evaluated in 168 different situations, which correspond to the combination of 7 possibilities of generated traf ic, 8 investment values for build works and 3 operation speeds. Each combination corresponds to a project alternative, therefore, there are 1008 different combinations or vector possibilities for the normative scenario.
In the regional scenario it was considered pertinent to limit the demand according to the statistical analyses carried out, being possible to reach values of up to 2500 bicycles/day in the irst years of operation service when incorporating the maximum percentage of generated traf ic. The evaluation by regions implies performing six times the modeling of each of the projects with the described combinations, since the user's average income is inserted in the software con iguration. For each of the scenarios evaluated (normative and regional), the variables for the factorial of the project alternatives are presented in Table 2 and inserted into the HDM-4 as shown in Figure 3.

Con iguration and parameterization
The HDM-4 can be con igured for local conditions in order to re lect the reality of the case to be studied. In this research, the con iguration consists in the insertion of variables of interest that represent factors such as climate, geometry and currency type.
The parameterization of the system involves the values concerning the physical and economic characteristics of the leet and, with greater relevance, includes the passenger time value which, in this research, is represented by the user's income, a factor that according to Brilhante (2012), determines the ownership of vehicles and impacts on the individual choice of transport mode.
Two analyses were developed to represent with igures the cyclist's economic pro ile. The irst evaluation consisted in considering income values from the research on cyclist pro ile whose sample corresponds to 1800 cyclists that, naturally, already use the city's bikeways and was conducted by Ciclocidade and Transporte Ativo (2016), members of the Association of Urban Cyclists of São Paulo. The second analysis focuses on the diversity of the inter-regional per capita income in the city of São Paulo, being applicable, for instance, for mobility and regional development projects when the demand is still repressed, and the target user is dif icult to identify. According to the São Paulo State Data Analysis System Foundation (SEADE, 2020), in terms of per capita income, São Paulo has a wide diversity, considering ive main regions as shown in Figure 4 (North, South, East 1, East 2 and Expanded Center). A sixth income category was included aiming at representing regions with critical social conditions (favela), which have a high Social Vulnerability Index (from Portuguese, "IPVS"). Finally, the value of passenger time in R$/hour was calculated considering the divisor of 220 monthly hours, which includes the remunerated weekly rest period, besides being widely accepted and used in the Brazilian labor regime.   Table 5 summarizes the values de ined to parameterize the aspects related to the user and the type of vehicle and presents a comparison with the parameters predetermined by Odoki and Kerali (2005) in relation to the vehicle characteristics.

Conditions of the Socioeconomic Analysis
The horizon adopted for this analysis corresponds to 20 years, a period that is in consonance with the long useful life that bicycle paths can offer under adequate maintenance conditions, and the analysis period is consistent with Australian guidelines given by Department of Planning Transport and Infrastructure (DPTI, 2015), which adopts 20 years for bicycle paths with lexible surfaces.
Regarding the Social Discount Rate (SDR), this study uses the value of 8.5% recommended by the Ministry of Economy of Brazil (2020) for cost-bene it analyses of infrastructure investment projects; in addition, analyses are carried out with SDRs of 10% and 12%.
To estimate the salvage value, the concept of average economic life was applied, which corresponds to the average value of the works as a function of the useful life and weighted cost of the elements of the infrastructure evaluated. Thus, the calculated residual value of the works corresponds to 39% and 60% for lexible and rigid pavement, respectively.

Bene its and Economic Indicators
For each alternative, pro itability indicators were obtained relative to Net Present Value (NPV), Internal Rate of Return (IRR), Bene it/Cost ratio (B/C), NPV/Investment ratio (NPV/I), the latter being selected for the comparative analyses.
In this research indirect bene its are not considered, since the analyses strictly adhere to the HDM-4 economic model, which quanti ies economic bene its based on the Travel Cost Method (TCM), widely applied in the area of social feasibility of road projects. Other approaches may include additional bene its, although there is not necessarily a monetary transaction, there is a gain associated with the sense of well-being and health resulting from bicycling that is related to the enjoyment of the activity. According to Wang et al. (2004), the overall bene it increases as the environment and conditions for cyclists improve, increasing the frequency of use and the spectrum of associated bene its. Table 5 presents a quantitative analysis of the variation of pro itable alternatives as a function of traf ic growth rates and the project's social discount rate (SDR), showing that both rates impact the results, but the SDR incidence is higher.

Modeling Scenario I -NormaAve Context
The table above shows the number of pro itable alternatives and their respective percentage in relation to the 168 possible alternatives for each traf ic category. Naturally, traf ic represents the most determinant variable in terms of pro itability, a fact that is corroborated by the wide differences obtained among the six demand categories herein analyzed. When traf ic does not grow over time, with an initial demand of 500 bicycles/day, the number of pro itable alternatives is only 44% of the total analyzed. It is important to note that most of the unfeasible alternatives correspond to the minimum operating speed (15 km/h).
With normal traf ic of 750 bicycles per day, there is still an important level of risk, and it is evident that with more challenging SDR and TGR, alternatives are pro itable in a proportion TRANSPORTES | ISSN: 2237-1346 13 lower than 61% of the total. For demands equal to or greater than 1,000 bicycles per day, economic pro itability is a predominant condition, and under the most favorable rates, socioeconomic feasibility is achieved in 100% of the possible alternatives. Regarding the most incident variables and their combinations, it was evident that there are borderline situations in which one of the variables evaluated is a determining factor for the feasibility of an alternative. For example, the feasibility of alternatives with normal traf ic of 500 bicycles/day is subject to investments ranging from R$ 400,000 to R$ 700,000 per kilometer, always when the operating speed exceeds the value of 20km/h, in other words, it is essential that the project represents a minimum gain of 10km/hour for the user. This situation is maintained even considering the generated traf ic in the irst years of the project.
Alternatives with higher traf ic gradually and favorably increase the results of the economic indicators, but important risks remain for alternatives with normal traf ic of 750 bicycles/day, whose viability is tied to investments less than or equal to R$ 900,000 per kilometer. In addition, the operating speed should be higher than 20 km/h to guarantee positive results. However, high percentages of generated traf ic (50%, 100% and 150%) economically compensate for the disadvantages of an operating speed of 15 km/hour, and projects with low to medium investments (R$ 400,000 -R$ 900,000 per kilometer) are pro itable under these conditions. However, just the maximum initial demand considered (2,000 bicycles per day) reaches full ef iciency in terms of pro itability.

Scenario II Modeling -Regional Analysis
Considering the range of investments addressed in this study, the maximum investment allowed for each project alternative is presented in terms of regional per capita income, and the analysis is carried out with the volume of normal and generated traf ic in the irst two years of operation for the three situations of traf ic growth and a SDR of 8.5%. The viability of projects with restricted speed gains are dependent on demand and, relevantly, the economic conditions of the bicycles users as shown in chromatic Figure 5.
According to Figure 5 (a), with an average operating speed of 8 km/h and no growth in demand, bikeway projects are not feasible in most of the city of São Paulo when traf ic levels range between 550 and 750 bicycles/day. Only in the Extended Center, whose population was TRANSPORTES | ISSN: 2237-1346 14 characterized with an average per capita income of 10.10 R$/hour, the projects admit investments above 600,000 R$/km for the minimum demand studied (550 bicycles/day) and projects quickly become feasible for demands above 1,000 bicycles/day with the highest value of the investment range analyzed. Naturally, the most critical situation is presented for the region with high IPVS, whose economic bene its are insuf icient to make viable projects even with a volume of 2,000 bicycles/day.  According to Figure 5 (b) and (c), the economic scenarios with growth rates of 3% and 6% per year, derive more favorable situations for the viability of the projects. In this case, the region with the lowest per capita income reaches social pro itability only when demand exceeds 2,000 bicycles/day. With respect to the operating speed of 10 km/h, the most critical situation is preserved by the region whose users have high IPVS. The admissible investment corresponds to R$ 600,000/km when the highest demand is projected with a static behavior in the analysis period ( Figure 5 (d)).
However, the high non-motorized traf ic (2,000 bicycles/day), in the most optimistic condition (annual growth rate of 6%), makes projects with investments equivalent to R$ 1,000,000/km feasible. In the North, South, East 1, East 2 regions, traf ic exceeding 750 bicycles/day admits the maximum investment established in this analysis ( Figure 5 (f)). In the Extended Center region, the best pro itability conditions are maintained without any risk for any level of investment.
When the operating speed is signi icantly increased to 15 km/h and a gain of up to 10 km/h is achieved, the user reduces travel time in such proportions that the pro itability of the projects is not strictly associated with cycling demand, except in the high IPVS region that represents the greatest risks for cycling projects from the perspective of user direct bene its.
According to Figures 5 (g), (h), and (i), only the traf ic of 550 bicycles/day represents budget restrictions for the North, South, East1 and East 2 regions, but it is a range of vast investments that allow the implementation of infrastructure appropriate to the current standards.
Other variables were sensitized to determine the risks of the projects, such as the value of the new vehicle. According to the results, this variable is not very representative for projects developed with users whose pro ile its good socioeconomic conditions; however, in projects with the lowest income per capita (High IPVS), even a variable with little incident in the model can imply favorable or unfavorable results, given that these are projects in constant inancial risk that become sensitive to slight variations.
Thus, the variation of the unit price of the vehicle (new bicycle) was performed, with decreases of 25% (R$ 594) and 50% (R$ 396) in relation to the value adopted in the total analyses, which corresponds to R$ 792. The sensitivity analyses were performed for nonmotorized traf ic of 500, 750 and 1000 bicycles/day. The other conditions of the analysis were maintained.
The results showed that this variable has little bearing on the feasibility of projects located in regions with high IPVS and low traf ic (500 bicycles/day). However, with traf ic higher than 750 bicycles/day and halving the cost of a new vehicle, there are important and positive variations that suggest the need for a proper user pro ile characterization for speci ic bicycle projects, especially when bikeways allow a minimum operating speed of 15 km/h and have a high potential to increase user demand.

CONCLUSIONS
Overall, the non-motorized traf ic model of the HDM-4 software proved to be practical and effective in determining the feasibility of bicycle infrastructure projects, quantifying travel time, user costs and bene its as a function of the exogenous and endogenous projects variables. Through the socioeconomic indicators obtained, it is proven that the direct economic bene its of bicycle tracks users can generate feasible projects when the demand is compatible with the investment and the desired speed of operation.
Operating speed is the heart of the Non-motorized traf ic model of HDM-4 software, since saving users' travel time becomes a decisive factor for project viability, and its impact is greater than the incidence of variations in traf ic demand or investment in the works. This is demonstrated by analyzing the different travel times and costs through the comparison of different per capita income levels, the results of which show the budgetary constraints when some project variable results in some risk to the project.
Regarding the pro itability and bene its of project alternatives, the following results and considerations stand out below: • The best results, in terms of pro itability, correspond to traf ic above 1,000 bicycles/day. From this category of traf ic on, situations that do not suggest high risks in terms of investment recovery are predominant. In any case, it should be mentioned that only the demand equivalent to 2,000 bicycles/day guaranteed absolute economic ef iciency in all the scenarios evaluated. • As for the low traf ic (500 -750 bicycles/day), the results raise a re lection on the operating conditions that justify the construction of bike lanes in places with low demand whose implementation work corresponds to high investments. Even with high percentages of traf ic generated in the irst years of implementation, the investments are in a risky situation whose return depends on very good operating conditions and challenging traf ic growth. • Such assessment alludes to some realities presented in this document about the average volume of bicycles that circulate on bicycle paths in São Paulo. It is pertinent to mention that according to available CET records, 50% of bicycle paths have a traf ic lower than 1,000 bicycles/day and some elements have such low traf ic that currently they do not suggest very optimistic situations about demand behavior. • The importance of the correct de inition of the social discount rate is emphasized, since its variation generate the pro itability or rejection of a prominent number of project alternatives. In this sense, it is convenient that the public or private agency evaluates planning policies and inancing conditions of bicycle paths projects. Finally, it should be noted that since this is an analysis at the network level, the modeling results should be interpreted in a referential way, and not as an absolute technical-economic behavior, since the results of speci ic projects will always depend on the pavement typology, the network and regional characteristics, the user pro ile and the objectives of the particular project or network. In addition, the socioeconomic analysis must constitute a phase of planning, being preceded, necessarily, by interdisciplinary technical studies as any road project.