Modeling Demand of Bike Share System Using Built Environment Attributes in the City of Buenos Aires, Argentina
By
Ricardo Sanchez Lang
B.S in Economics Universidad Centroamericana, 2009 Managua, Nicaragua
Submitted to the Department of Urban Studies and Planning in partial fulfillment of the requirements for the degree of
Master in City Planning
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2018
0 2018 Ricardo Sanchez Lang. All Rights Reserved
The author hereby grants to MIT the permission to reproduce and to distribute publicly pa er and electronic copies of the thesis document in whole or in part in any medium now own r hereafter created. Signature redacted
Author Department of Urban Studies and Planning May 24, 2018
Certified by Signature redacted, V '7 0 P.?Christopher Zegras Department of Urban Studies and Planning / Thesis Supervisor
Accepted by_ Signature redacted Professor of the Practice, Ceasar McDowell Chair, MCP Committee MASSACHUSETS I NSITUTE Department of Urban Studies and Planning OF TECHNOLOGY
JUN 18 2018 LIBRARIES ARCHIVES 2 Modeling Demand of Bike Share System Using Built Environment Attributes in the City of Buenos Aires, Argentina
by
Ricardo Sainchez Lang
Submitted to the Department of Urban Studies and Planning on May 24th , 2018 in partial fulfillment of the requirements for the degree of Master in City Planning
Abstract
Increasing the share of trips done by bike has become an objective of cities worldwide. At the individual level, biking is associated with better physical and mental health. At the city level, biking contributes to alleviate traffic congestion, reduce commuting times and improve air quality. This research is an initial attempt to measure the relationship between demand of Buenos Aires' bikeshare system and sociodemographic, built environment and transportation attributes using a linear regression model. Departure and arrival counts are used as dependent variables and are aggregated by station over the period of one year. In addition, catchment areas of different sizes were constructed around bike share stations to estimate the relationships between bike-sharing demand and availability of these attributes around stations. Results suggest a positive relationship between population, buses, availability of bike lanes within the buffers, universities and station capacity. This client-based project seeks to shed light on the subject of bike sharing as a mode of transportation in the context of a Latin American country.
Thesis Supervisor: P. Christopher Zegras Title: Associate Professor of Transportation and Urban Planning
3 Acknowledgements
While writing this thesis, a revolution was taking place in my country, Nicaragua, with university students leading a social revolt that embodies the desire of my people for a free, democratic, and just country. As a Nicaraguan and as a MIT student, I share and embrace their cause, which is fundamental for the advancement of our country. This thesis is especially dedicated to the Nicaraguan college students who, in the revolution of April 2018, lost their lives fighting for the Nicaragua that we dream.
I want to start by thanking Chris for his support and guidance throughout the thesis process and throughout these last two years. Thank you for welcoming me in your research team, which has greatly expanded my knowledge and curiosity about urban transportation planning and thank you for advising me during this thesis process.
I also want to express my gratitude to the Government of the City of Buenos Aires for the opportunity to intern at the city's Department of Transportation and for trusting me the exciting task of generating knowledge about the City's bike share system, the Ecobici. In particular, I would like to thank Paula Bisiau, Andres Meyer, Federico Varone and Mariela Vera for providing me with data and guidance throughout this process.
I like to thank my classmates and other MIT friends and professors for sharing their knowledge and making these last two years unforgettable. To He, Rounaq, Eytan and Isabel for patiently sharing with me the technicalities of transport planning. I will not forget our long and fruitful discussions. To Professor Sarah Williams for introducing me to the fascinating world of spatial analysis. To Daniel and Madeline at the GIS Lab for teaching me life-long skills and to my GIS- savvy classmates Laura, Kadeem and my dear friend Toho for answering my questions. To Daniel and Cristian, for your friendship and for sharing your knowledge and passion for urban transportation in Latin America. To Apaar, I am happy to have shared my time with you at MIT.
Finally, I want to thank my mom and my godmother, for the unconditional love and support during all these years and for being a source of moral and intellectual inspiration. To my grandparents, for being an example of integrity, rectitude and humility.
Gracias a la vida que me ha dado tanto - Violeta Parra
4 Table of Contents
A b stra ct...... 3 Acknowledgem ents ...... 4 1. Introduction...... 8 2. Background ...... 9 2.2. General Inform ation about Buenos Aires...... 9 2.3. The Biking Infrastructure ...... 11 3. Literature Review, M ethods and Data ...... 18 3.1 M odeling Bike Sharing Demand ...... 18 3 .2 M eth o d s ...... 2 0 3 .3 D ata ...... 2 2 4. M odeling Framework, Regression Results and Discussion ...... 44 4.1 Linear Regression (OLS) ...... 44 4.2 Transform ations ...... 44 4.3 Spatial Autocorrelation ...... 46 4.4 Regression findings and discussion ...... 48 5. Conclusions...... 52 6. References...... 53 A p p en d ix ...... 5 5
5 List of Figures
Figure 1: G reater Buenos A ires Area ...... 9 Figure 2: City of Buenos Aires' Modal Split...... 10 Figure 3: People U sing Bike Lane N etw ork ...... 11 Figure 4: Buenos Aires' Biking Infrastructure...... 13 Figure 5: No. of Kilometers of Bike Lanes Constructed from 2010-2017...... 14 Figure 6: Buenos Aires Biking Infrastructure and Population...... 16 Figure 7: Buenos Aires Biking Infrastructure and Jobs...... 17 Figure 8: Q ueen and Rook Criterion ...... 21 Figure 9: Departure Counts by Station in 2017...... 24 Figure 10: Arrival Counts by Station in 2018...... 25 Figure 11: Distribution of Departure Counts by Time of Day (2017)...... 27 Figure 12: Distribution of Arrival Counts by Time of Day (2017)...... 27 Figure 13: Distribution of Departure Counts by Month (2017)...... 28 Figure 14: Distribution of Arrival Counts by Month (2017)...... 28 Figure 15: Frequency Distribution of Travel Times...... 29 Figure 16: Distribution of the Original and Transformed Dependent Variable...... 30 Figure 17: Distribution of the Original and Transformed Dependent Variable...... 31 Figure 18: N etw ork-based polygon ...... 32 Figure 19: Buffers around Bike Share Stations...... 33 Figure 20: Jobs within 300m Buffers Around Ecobici Stations ...... 36 Figure 21: Shortest Routes Between Stations and Bike Lane Network ...... 38 Figure 22: Impedance as a Function of Travel Time...... 41 Figure 23 Distribution of Original Departure Counts and Transformations...... 45 Figure 24 Distribution of Original Arrival Counts and Transformations...... 46 Figure 25: Confidence Intervals for Moran's I Spatial Autocorrelation test ...... 47
6 List of Tables
T able 1: V ariable D efinitions...... 22 T able 2: D escriptive Statistics...... 42 Table 3: Shapiro-W ilk N orm ality Test...... 44 Table 4: Location-based analysis for spatial autocorrelation...... 47 Table 5: OLS results: Annual Departure Counts (300 meter buffer zones) ...... 48 Table 6: OLS results: Annual Departure Counts (200 meter buffer zones) ...... 50 Table 7: OLS results: Annual Departure Counts (400 meter buffer zones) ...... 51
Appendix
Table A. 1 Opportunities within 200 meters Buffer...... 55 Table A.2 Opportunities within 300 meters Buffer...... 62 Table A.3 Opportunities within 400 meters Buffer...... 68 Table A.4: Correlation Matrix of the Independent Variables (200 meters buffer zone)...... 75 Table A.5: Correlation Matrix of the Independent Variables (300 meters buffer zone)...... 75 Table A.6: Correlation Matrix of the Independent Variables (400 meters buffer zone)...... 76 Table A.7: OLS results: Annual Departure Counts (200 meter buffer zones, all variables)...... 77 Table A.8: OLS results: Annual Departure Counts (300 meter buffer zones, all variables)...... 78 Table A.9 OLS results: Annual Departure Counts (400 meter buffer zones, all variables)...... 78
7 1. Introduction
This research project aims to quantify the effect of sociodemographic, built environment and transportation attributes on bike sharing demand. The selection of Buenos Aires as an area of analysis is grounded in several reasons. First, the City is committed to continue investing in its biking infrastructure, and authorities are eager to better understand the factors that may be playing a role in the decision to use the bike share system. Also, over the last few years the City has made available relevant information that facilitates research in Buenos Aires, which makes it a better candidate than other Latin American cities with less data availability. This data includes, but is not limited to, bike share counts, shape files of different urban attributes such as transit infrastructure and land use, and sociodemographic data, such as population and concentration of jobs in the city. Finally, the city presents an interesting case because several factors that are thought to be positively associated with bike usage are present such as availability of protected bike lanes, flat topography, high population and job density, among others. Additionally, Buenos Aires features a bike share system free of charge for one hour.
The document is organized in five chapters. The second chapter provides background information about Buenos Aires and its transportation network, particularly the biking infrastructure. The third chapter consists of a literature review that briefly describes methods commonly used by scholars and researchers to measure the demand of bike share systems. In this chapter, concepts relevant for this research such as linear regression analysis and spatial autocorrelation are also discussed. The third chapter also describes in detail the steps followed to prepare the variables for analysis. Chapter Four presents the modeling results and discusses possible implications. Finally, the last chapter offers concluding remarks.
I expect this research project will contribute to increasing the City's understanding of the bike share system and its relationship to variables chosen for this study. However, this document represents an initial, as opposed to conclusive, effort to describe the relationship between attributes that may play a role in the demand for the bike share system. As such, its results should be interpreted with caution.
8 2. Background
2.2. General Information about Buenos Aires
The City of Buenos Aires, also known as the Autonomous City of Buenos Aires, is Argentina's most populated city with just over 3 million people (Government of the City of Buenos Aires, 2016). In addition, the City is part of a larger metropolitan area comprised of 24 municipalities known as The Greater Buenos Aires (GBA), which according to the 2010 Census amounts to more than 12 million people, making the GBA the third largest urban agglomerations in Latin America after Mexico City (23 million) and Sao Paulo (20 million). In this thesis, I will refer to the City of Buenos Aires as 'The City' and to the Greater Buenos Aires as GBA. As the country's main political, economic and cultural hub, Buenos Aires experiences an influx of thousands of workers traveling in and out of the city every day. According to the City's official statistics, non-city residents perform almost half of the more than 2 million jobs available in Buenos Aires.
Figure 1: Greater Buenos Aires Area
I - Malvinas Argentinas 2 - ltuzalng6 URUGUAY 3 - Hurlingham 4 - Tres do Febrero 5 - General San Martin -San Fernando RIO DE LA PLATA Vicente L6pez M Greater Buenos Aires (administrative)
Greater Buenos Aires Agglomeration (non-administrative designation) General Avellaneda Rodriguez M Greater La Plate R Other patridos of Buenos Aires ProvinCe
[ City of Buenos Aires
Marcos Paz
Berisso P
Le Plata
San Vicente Cafluelas
Source: Wikimedia.org
9 - -- - -I
The most frequently used modes of transportation are public transit, walking, and personal automobile, respectively. Public transportation can be broken down in the three modes: bus, which accounts for almost 38% of all trips done in Buenos Aires, subway, which consists of 6 underground lines, and rail, connecting the city with the remaining municipalities that comprise the metropolitan area.
Figure 2: City of Buenos Aires' Modal Split
Rail 2% Walking 26%
Taxi. 5%
Subway 9% Bus 38%
Source: National Transportation Observatory (2013)
To curb air pollution stemming from the use of automobiles and reduce traffic congestion, the city adopted a series of steps aimed at increasing the use of public and alternative modes of transportation under the Sustainable Mobility Plan (2011). Among these initiatives was the construction of exclusive bike lanes in highly congested corridors, limiting parking permits, street/sidewalk redesign and improvement, banning automobiles from entering downtown areas,
10 among other measures. One of these measures was the construction of a network of bike lanes and the introduction of a publicly run bike sharing system known as Ecobici.
2.3. The Biking Infrastructure
As of May 2018, the City had approximately 180km of bike lanes and 200 bike share stations (BSS). Most BSS are located in the city center, which is in the eastern part of the city, while bike lanes extend further out. Data from cyclists counts performed bi-annually by the City suggest that while the number of people using bike lanes significantly increased after their introduction, over time the number of cyclists riding on streets without bike lanes and the number of cyclists using biking facilities have converged. In addition, the overall number of bike lane users has stabilized with respect to 2010, when the first bike lanes were introduced.
Figure 3: People Using Bike Lane Network
16% 242.1% 14.2% 230% S14%
.5 12% IA 180% ED C .3 3 .~10% an .5 VI 130% IA 8.1% In d 2 8% 0 M U 0 0 80% 2 ,M 6% C ED ED 5.1% U' ED ED U .3% C S4% .E 2.8% 30% 17.1% 9.72% S.3% 0% r 2% 10.97% .6% 1.02% 0.4% -8.34% -0.5% 0.0% 0.0% -8.63% 0.0% 11.71% 0% -20% OCT_2011 ABR_2012 OCT_2012 ABR_2013 OCT_2013 ABR_2014 OCT_2014 ABR_2015 %increase of Km of bike lanes -o-% Change in the number of cyclist using bike lanes -e-% Change in the number of cyclist NOT using bike lanes
11 For example, after increasing by more than 200% in October 2011 with respect to October 2010, in April 2012 the number of cyclists using bike lanes increased only by 20% with respect to 2011, by 50% in October 2012 with respect to October 2013, and from there progressively decreased despite the fact that the city continued adding bike lanes. From 2013 onwards, the rate of change in number of cyclist using the bike lane has not increased significantly. This may be normal since bike lanes presumably cannot accommodate all cyclist nor they cover the entire city. However, this rate is close to the percentage change in the number of cyclist not using the bike lanes. By 2015, approximately 35 thousand cyclists were not using the bike lanes versus 21 thousand who were.
12 Figure 4: Buenos Aires' Biking Infrastructure
Legend v 0 Ecobici Station - Me Lane ca"jm..badate Gron Araaa
Source: Govemrnment of the City of Buenos Aires (2017). Prepared by: Ricardo Sanchez, MIT 0 0.5 1 2 3 n~lalomneters
13 Figure 5: No. of Kilometers of Bike Lanes Constructed from 2010-2017
90 -- - -_-_---
80
70
60
) E 50...... 40
30 - 20I I I __l
2010 2011 2012 2013 2014 2015 2016 2017 Unknown
Source: Own elaboration with data from GCBA
Buenos Aires' bike lane network is a mesh of protected bike lanes. Protected bike lanes are positively associated with bike sharing demand as they improve the perception of safety and have the potential to entice non-bikers to consider using the bike (Bert van Wee & Maat, 2010). Paramount for this is a local regulation that limits parking to only one side of the road, thereby freeing up space for dedicated facilities on the other side of the street. Despite the availability of protected bike lanes, cyclists still face the city's fast-paced and oftentimes aggressive driving style. It is relatively common for vehicle drivers not to respect stop signs and even traffic lights, which limits the potential benefits of some of the measures the City has adopted to improve security at intersections such as bike signaling and the adoption of street lights for bikes.
While regulation has facilitated the construction of bike lanes, this has also forced the City to introduce bi-directional bike lanes, which have disadvantages of their own. For example, on narrow streets, which predominate in Buenos Aires' city center, a bi-directional lane limits the maximum potential bike lane width, which jeopardizes both security and comfort since some riders
14 may feel insecure about riding too close to the traffic coming in the opposite direction. In addition, when the slope of the street is pronounced, riding close to the sidewalk could be uncomfortable and unsafe. This seems to be the case in several streets of Buenos Aires. Furthermore, in these places, the surface of the lane is not uniform but rather 'two level,' due to the fact that the road pavement is on top of the drainage infrastructure. Authorities recognize this fact but also acknowledge that providing a solution is complex since the slope is designed for storm water drainage and, in addition, leveling down the surface would increase the costs of the bike lane (and provoke disruption of traffic flow, albeit temporarily).
Buenos Aires' Ecobici system shares similar characteristics to other bike sharing system found in several cities across the world. It consists of stations with several bikes where the user must have either a code or a key to unlock the bike. The number of docks per station varies from station to station, but most have between 15-20 bikes. A particular feature of the Ecobici system is the fact that the service is free of charge for both residents and non-residents. However, in practice, some barriers prevent all potential users from using the Ecobici in equal conditions. To use a bike, a potential user has to create an account on the city's website and provide a proof of residency, which can be any Argentinian address but not an international one. Furthermore, to verify proof of residency, users must, upload a utility bill or a pay stub. This makes it cumbersome for tourists, informal workers, senior citizens, among others to use the bike share system, potentially reducing overall ridership.
The Ecobici system faces several other limitations that could play a role in the demand of the service. The system has not been exempted from technical failures, vandalism, and malfunctioning of stations and bikes, all of which would negatively impact people's perception of the system. According to city officials, in the past technical issues have forced the City to suspend the service for long periods. Furthermore, supply of public bikes is limited. Buenos Aires has a low density of bike share stations (the number of stations per square meters). The National Association of City Transportation Officials (NACTO) recommends a bike share station density of 11 bike share stations per square kilometer. In Buenos Aires, only two neighborhoods (out of forty-eight) meet this recommendation. In fact, the areas with the highest density of bike share stations (and bike lanes) are those located in areas with a high density of jobs but not necessarily in areas with high
15 population density as shown in Figure 6 and Figure 7. For people who do not have bicycles and who live in neighborhoods that have Ecobici stations, this may prevent them from choosing to use Ecobici on their return trip fearing that the stations in their neighborhood may be full, therefore forcing them to travel longer distances to find a station with available docks.
Figure 6: Buenos Aires Biking Infrastructure and Population
V
013 -274E
Legend 0 Eco Abid Station t Lane - Ore PGre en Area Population Density F-- ] 1 m 27,414 - 49,536 49.536 - 81.430 81,430 - 124.440 124,440 - 177.107
Source: Government of the City of Buenos Aires (2017), National Institute of Statistics and Census (2010) Prepared by: Ricardo Sanchez, MIT 0 0.5 1 2 3 Km o- K1omters
16 ------. - -- .1
Figure 7: Buenos Aires Biking Infrastructure and Jobs
N
AA
PERO
Legend 0 Ecobldi Station - SIM Lane Gmen Areas Jobs 20.602 - 50,117 .50118-112,377 112,378 - 207.287 I 207.287 - 331.402 331,402 - 519.317
Source: Government of the City of Buenos Aires (2017), Prepared by: Ricardo Sanchez, MIT 0 0.5 1 2 3 Kilometers
Additionally, a low density of bike share stations at destinations increases the likelihood that people who do not have a bicycle would not even consider using the bicycle as a mode of transportation.
17 3. Literature Review, Methods and Data
The ubiquity of bike sharing systems and the establishment of a culture of biking as a mode of transportation has engendered abundant relevant research. This chapter summarizes the literature and methods consulted to help me to better identify and operationalize my approach to measuring the relationship between bike share station usage and the built environment and sociodemographic features.
3.1 Modeling Bike Sharing Demand
A substantial body of literature examines determinants of bike sharing, particularly in developed countries but also in emerging economies. A less significant amount looks directly at factors influencing the decision to use bike share systems in Latin American countries, probably because such systems are relatively new in the region. In general, authors investigate bike-sharing demand by examining behavioral factors and/or by looking at data produced by bike facilities and attributes around bike share stations. Thus, broadly speaking, two approaches are common: a behavioral- based or a facility-based analysis. This research takes a facility-based approach.
In facility-focused studies' authors use different approaches and data sets to study travel patterns. Authors like Wardman et al. (2007), look at national transportation surveys by different modes, and aggregate them by year. Others like Tran et al (2015) or Rixey (2013) use origin and destination trip data obtained directly from bike share stations and aggregate them at the station level. Nosal and Miranda-Moreno (2014), use automatic counts of trips for commuting and leisure purposes to test the effect of weather on the use of bicycle facilities.
In facility-focused studies, researchers look at elements such as the natural and the built environment, temporal factors, socio-demographic characteristics and flows of cyclist. Natural elements include weather (mainly precipitation) but also the topography of the city, hours of sunlight, temperature, wind and even the presence of water bodies (e.g. Pen et al, 2017). Elements of the built environment that are normally analyzed are functional characteristics such as densities (population, jobs, residential), mixture of land uses, road connectivity, availability of bike lanes, number of stops and street lights, block size, and road width. Authors also look at factors such as
18 distance of homes to work, road network layouts, presence of parks, and even aesthetic characteristics of the built environment (Pikora et al, 2003, and Stefandsdottir, 2014). Temporal factors include specific periods of the year (winter versus summer or fall versus spring), days of the week, and periods of the day (e.g., peak and non-peak hours). Socio-demographic elements often analyzed include population, jobs, income, race, and gender, and household size, availability of private modes of transportation, age, and level of schooling.
The techniques used to count and aggregate these variables vary but in most of the literature analyzed for this research, authors use geographic information systems to construct catchment areas around bike share stations and count the number of opportunities within that area. The size of the catchment area varies. For example, Tran et el (2015) constructed buffer zones of 200, 300 and 400 meters around bike share stations (authors do not clarify whether they use Euclidean or Network-based distance) and examined the relationship between 10 variables such as population, jobs and public transportation availability, among others with bike sharing demand. Rixy (2013) used a buffer of 400 meters (Euclidean distance).
19 3.2 Methods
3.2.1 Linear Regression
Linear regression is a statistical method to establish a relationship between variables. In multiple linear regression, there are two types of variable: a response or dependent variable and a set of predictor or independent variables. I use multivariate linear regression to quantify the relationships between socioeconomic and built environment attributes and bike sharing demand.
3.2.2 Spatial Autocorrelation
Spatial Autocorrelation measures the degree to which a variable we seek to study is correlated due to possible existence of spatial lag and/or error. Spatial error is when the error terms of different spatial units are correlated and treats spatial autocorrelation as an estimation problem. Spatial lag assumes that dependencies exist among the levels of the dependent variable, for example, the usage at one bike share station is impacted by the usage at nearby stations. Anselin and Bera (1998) express the presence of spatial autocorrelation with the following moment condition:
Cov (yi, yj) # 0 for i # j
Where yi, yj are observations of the variable we seek to study at locations i and j. These locations can be states, regions, census tracts, buildings or, as in the case for this research, bike share stations and their buffers. Normally, restrictions are placed on the number of possible locations that could be potentially compared. This is achieved by constructing a spatial weight matrix. Two approaches exist to construct this matrix: the Rook and Queen Criterion. In the Rook Criterion, two observations are close to one another if they share a side. In the Queen Criterion, two observations are close if they share a side or an edge.
20 41
Figure 8: Queen and Rook Criterion Queen Rook
There are different approaches to test for the presence of spatial autocorrelation. A common test, and the one used in this research, is the Moran's I test. Moran's I is defined as:
J1> =1 Wij (xi -X)(x - ) ( = il WOj Dt-(xi- ) Where: N is the number of observations X is the mean of the variable xi is the variable value at a particular location xj is the variable value at another location Wij is a weight indexing location of i relative to j (the weight matrix described above)
A 95 percent confidence interval is commonly used to infer the presence of spatial autocorrelation. At this confidence interval, z-scores required to fail to reject the null hypothesis must be between -1.96 and 1.96. If values obtained from the Moran's I test are within this threshold, then the observed values of the variable under study are likely to be the result of random spatial processes. That is spatial autocorrelation should not be a problem.
21 3.3 Data
This section explains in detail the selection of variables for this research as well as the processes used to obtain them. Table 1 summarize the variables used. Table 1: Variable Definitions
Variable Definition Source Dependent Square root of the number of departure and arrival Government of the Counts counts during 2017, by station. City of Buenos Aires
Independent Demographic Factors Population* Total population (in 1 Os of persons) National Institute of Statistics and Census Jobs* Number of jobs (in IOs of jobs) Argentinian Ministry of Transportation Built Environment Attributes Recreational Facilities Number of restaurants, cinemas, theaters, bars and Government of the clubs City of Buenos Aires Universities 1 if a university is located within the BSS buffer, 0 Government of the otherwise City of Buenos Aires Transportation Network Factors Ecobici Stations 1 if an Ecobici station falls within the buffer of Government of the another station, 0 otherwise City of Buenos Aires Bike Lanes Length of bike lanes within the BSS buffer Government of the City of Buenos Aires Prop of Bike Lanes Proportion of bike lane availability on the shortest Government of the route between one station and all other stations City of Buenos Aires Bus Stops Number of bus lines serving one bus stop Government of the City of Buenos Aires Rail Stops 1 if a rail stop falls within the buffer of a bike share Government of the station, 0 otherwise City of Buenos Aires
Subway Stops 1 if a rail stop falls within the buffer of a bike share Government of the station, 0 otherwise City of Buenos Aires Station Capacity Number of docks per station Government of the City of Buenos Aires Accessibility-based Indicators Accessibility to Accessibility to persons from one BSS to all other Own elaboration Population BSSs Accessibility to Jobs Accessibility to jobs from one BSS to all other BSSs Own elaboration Accessibility to All Accessibility to Persons and Jobs from one BSS to Own elaboration all other BSSs
*Aggregated proportionally by area intersecting 2010 Census Tracts
22 3.3.1 Selection of Dependent Variable
The dependent variable is the bike share counts or trips made using the Ecobici service during 2017. Data comes from Buenos Aires's open data website and consists of a comma separated value (CSV) file containing more than 1.8 million observations with seven variables. These variables include, among others, departure and arrival stations, departure time, and total travel time. It does not include personal information about the users such as gender, race or income. To obtain the arrival time, I calculated it by adding the total travel time to the departure time. Not all the data in the original raw dataset is used in this analysis. Stations with less than a year of operations, trips with the same origin and destination and trips with travel times beyond 2 hours and less than three minutes were excluded from the analysis. The resulting dataset consists of approximately 650 thousand observations, aggregated by station and shown in Figure 9 and Figure 10.
23 No. of Counts
25 DE MAYO ADUANA AGUERO AIME PAINE ALSINA ARAOZ ARENALES AYACUCHO AZUCENA VILLAFLOR BALCARCE BELGRANO BILLINGHURST BOUCHARD CARLOS GARDEL CATEDRAL CEMENTERIO DE LA RECOLETA CERRITO CHILE COLEGIO NACIONAL BUENOS AIRES CONGRESO CORDOBA CORONEL DIAZ DELLA PAOLE RA DIAGONAL NORTE DISTRITO AUDIOVISUAL DOBLAS ECUADOR ESME RALDA FACULTAD DE DERECHO FACULTAD DE MEDICINA GALERIAS PACIFICO 9T GUAYAQUIL GUZMAN HOSPITAL ITALIANO HOSPITAL RAMOS MEJIA HOSPITAL RIVADAVLA INDE PENDENCIA INGENIERO BUTTY INSTITUTO LELOIR JUANA MANSO JULIAN ALVAREZ LAVALLE LEGISLATURA MADERO UCA MAIPU MALABIA MEXICO MINISTERIO DE ECONOMIA MINISTRO CARRANZA MISIONES MONTEVIDEO 9o MORENO OB E LISCO ONCE PACIFICO PADILLA PARQUE CENTENARIO PARQUE LAS HERAS PARQUE LEZAMA PARQUE PATRICIOS PASCO PERIA PERON PIE DRAS PLAZA ALMAGRO PLAZA BOEDO PLAZA GUEMES PLAZA HOUSSAY PLAZA ITALIA PLAZA LIBERTAD PLAZA PALERMO VIEJO PLAZA PRIMERO DE MAYO PLAZA ROMA PLAZA SAN MARTIN PLAZA VICENTE LOPEZ QUINTANA RECONQUISTA RETIRO RICARDO ROJAS RINCON RIOBAMBA RIVAROLA SAAVE D RA SALCE DO SANCHEZ DE BUSTAMANTE SARANDI SARMIENTO SUIPACHA SUIPACHA Y ARROYO TREINTA Y TRES ORIENTALES TRIBUNALES TUCUMAN URQUIZA VENEZUELA VERA PENALOZA YATAY ZOOLOGICO YATAY ZOOLOGICO
k- No. of Counts
25 DE MAYO ADUANA AGUERO AIME PAINE ALSINA ARAOZ ARENALES AYACUCHO AZUCENA VILLAFLOR BALCARCE BELGRANO BILLINGHURST BOUCHARD CARLOS GARDEL CATEDRAL CEMENTERIO DE LA RECOLETA CERRITO CHILE COLEGIO NACIONAL BUENOS AIRES CONGRESO CORDOBA CORONEL DIAZ DELLA PAOLERA DIAGONAL NORTE DISTRITO AUDIOVISUAL DOBLAS ECUADOR ESMERALDA FACULTAD DE DERECHO FACULTAD DE MEDICINA GALERIAS PACIFICO GUAYAQUIL GUZMAN HOSPITAL ITALIANO HOSPITAL RAMOS MEJIA HOSPITAL RIVADAVIA INDEPENDENCIA INGENIERO BUTTY INSTITUTO LELOIR sI,. JUANA MANSO JULIAN ALVAREZ LAVALLE LEGISLATURA MADE RO UCA MAIPU MALARIA MEXICO MINISTERIC) DE ECONOMIA MINISTRO CARRANZA MISIONES MONTEVIDEO MORENO OBELISCO ONCE PACIFICO 00 PADILLA PARQUE CENTENARIO PARQUE LAS HERAS PARQUE LEZAMA PARQUE PATRICIOS PASCO 00 PENA PERON PIE DRAS PLAZA ALMAGRO PLAZA BOE DO PLAZA GUEMES PLAZA HOUSSAY PLAZA ITALIA PLAZA LIBERTAD PLAZA PALERMO VIEJO PLAZA PRIMERO DE MAYO PLAZA ROMA PLAZA SAN MARTIN PLAZA VICENTE LOPEZ QUINTANA RECONQUISTA RETIRO RICARDO ROJAS RINCON RIOBAMBA RIVAROLA SAAVEDRA SALCE DO SANCHEZ DE BUSTAMANTE SARANDI SAR MIE NTO SUIPACHA SUIPACHA Y ARROYO TREINTA YTRES ORIENTALES TRIBUNALES TUCUMAN URQUIZA VENEZUELA VERA PENALOZA YATAY ZOOLOGICO
I. 3.3.2 Processing of the Dependent Variable
Aggregating the counts by time of day reveals the time-based patterns of trips. Figure 11 and Figure 12 show the frequency distribution of trips during 2017 (departures and arrivals). They show that the majority of trips take place between the midday and pm-peak period (64%). This may be a first indication of the type of use people make of the Ecobici system (recreational, commuting, and so forth). The date-time format in which count data has been collected by the City also permits studying the frequency distribution of trips by month. Figure 13 and Figure 14, for example, which show trip counts by month, reveal a small fluctuation in counts from January to July, but then a sharp decrease from July onwards. This is counterintuitive, as one would expect demand to pick up after the winter months (June-August). However, according to city officials interviewed for this project, this was due to an interruption in the supply of bikes.
26 Figure 11: Distribution of Departure Counts by Time of Day (2017)
C C CL 0 CS 0 Cq,
0 0 0 0
0 m
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
Hours
Figure 12: Distribution of Arrival Counts by Time of Day (2017)
0 0
S
44 Am
0 0 -E 0 C