Crowdsourcing Bikeshare Transit Planning: An Empirical
Analysis of Washington DC and New York City
R² = 0.0042
0
5000
10000
15000
0 0.2 0.4 0.6 0.8 1 1.2 1.4V
ote
Co
un
t
Average # of Bike Transactions (10 min. interval)
Weekday Transactions vs. Votes per Voronoi Region
R² = 0.0135
-5000
0
5000
10000
15000
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Vo
te C
ou
nt
Average # Bike Transactions (10 min. interval)
Weekend Transactions vs. Votes per Voronoi Region
Voronoi Regions
Zipcode
R² = 0.2709
0
1000
2000
3000
4000
0 2 4 6 8 10 12
Vo
te C
ou
nt
Average # of Bike Transactions (10 min. interval)
Weekday Transactions vs. Votes per Zipcode
R² = 0.334
0
1000
2000
3000
4000
0 2 4 6 8 10 12
Vo
te C
ou
nt
Average # of Bike Transactions (10 min. interval)
Weekend Transactions vs. Votes per Zipcode
Voronoi Regions
R² = 0.0235
-100
0
100
200
300
400
0 0.1 0.2 0.3 0.4 0.5 0.6
Vo
te C
ou
nt
Average # of Bike Transactions (2 min. interval)
Weekday Transactions vs. Votes per Voronoi Region
R² = 0.0002
-100
0
100
200
300
400
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
Vo
te C
ou
nt
Average # of Bike Transactions (2 min. interval)
Weekend Transactions vs. Votes per Voronoi Region
Zipcode
R² = 0.0005
0
20
40
60
80
100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Vo
te C
ou
nt
Average # of Bike Transactions (2 min. interval)
Weekday Transactions vs. Vote Count (400m)
R² = 0.1477
0
20
40
60
80
100
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Vo
te C
ou
nt
Average # of Bike Transactions (2 min. interval)
Weekend Transactions vs. Vote Count (400m)
324 existing stations; 10,000 suggested stations; 65,000 votes238 existing stations; 3,000 suggested stations; 11,000 votes
Stations and suggestions were partitioned in different ways for analysis
We studied the historical usage data for both Capital Bikeshare and CitiBike to find the answer to
two questions about the effects of crowdsourcing urban planning on bikeshare systems
Does bikeshare system usage reflect crowdsourced suggestions….?
Does the placement of new stations reflect crowdsourced suggestions…?
FindingsActual system usage by region does not strongly correlate with crowdsourced requests and suggestions
Larger groupings of suggestions had stronger correlations in general with actual station usage
Station placement over time does indicate an effort to respond to the crowdsourced data
Human-Computer
Interaction Lab
Cy Neita | [email protected]
Joseph Owen | [email protected]
Jon Froehlich | [email protected]