1 DOCKLESS IN : THE RISE AND DECLINE OF BIKESHARING IN 2 3 4 5 6 Capucine Heymes 7 ENTPE Université de Lyon 8 [email protected] 9 10 11 David Levinson 12 The University of Sydney 13 [email protected] 14 15 16 Word Count: 2523 words + 10 figures× 250 + 5 tables× 250 = 6273 words 17 18 19 20 21 22 23 Submission Date: July 26, 2018 1 Abstract

2 In mid-2017, dockless, (or stationless) bikesharing appeared on the streets of Sydney. The birth 3 of dockless bikesharing, its evolution as well as its consequences, and use habits are studied with 4 review of policies and field investigations. It is found that bicycle use in Sydney is less than hoped 5 for, vandalism is high, regulations unfavourable, and thus, the conditions for successful bikesharing 6 are not met. 7 8 Keywords: Bikesharing, Dockless, Stationless, Public bicycle, Bicycle, Transport, Sharing Econ- 9 omy Heymes and Levinson 1

1 INTRODUCTION 2 Amsterdam’s White Bikes ignobly marked the first generation of bikesharing. Despite, or perhaps 3 because it was free, it failed due to vandalism: bicycles were thrown in canals or stolen for private 4 use. Farsøand Grenå, Denmark in 1991 launched a second generation – this time, fixed docks 5 appeared and a financial dimension was integrated: people had to leave a deposit to use the bike. 6 After, bikesharing appeared in München and Copenhagen, where the first large-scale bike sharing 7 system was launched: . It had many improvements compared to the previous generation: 8 the bikes were stronger and the tracking more accurate. Still operated by a nonprofit organization, 9 the financial incentive and the deposit being were insufficient, and vandalism and the theft of bicy- 10 cles remained. Portsmouth University in England launched Bikeabout in 1996, a third generation 11 of bikesharing (referred to as station-based or docked bikesharing), using credit cards to identify 12 renters. This identification of users, by imposing responsibility, reduced vandalism, as did the sub- 13 sequent GPS tracking of the bikes. The third generation of bikesharing’s first great success was in 14 where the first municipal viable systems started: in Lyon with Vélo’v in 2005, then in 15 with Vélib in 2007. (1–4) 16 Within the third generation, bikesharing has seen real improvements: there are more and 17 more stations, bikes are of better quality, and GPS tracking has become more accurate. But despite 18 the improvements, docked bikesharing faced several difficulties. First, most of these systems are 19 supported by public funds. However, cities, like Sydney, were reticent to put money in this kind 20 of public service because of unprofitability. (5, 6) Second, station-based bikesharing systems are 21 constrained by the capacity of the stations: depending on time of day, stations may be full or empty, 22 preventing users from returning or renting a bike. 23 To overcome these new problems and thanks to the progress of new technologies and the 24 miniaturization of GPS systems, a fourth generation of bikesharing (generally called dockless) 25 appeared. This new generation was brought to the market by Chinese start-ups born in 2014, in 26 particular , , and Gobeebike. (1, 2, 7, 8) 27 Their use is easy, avoiding problems linked to limited capacity of stations and their lim- 28 ited number of bikes. First, taking advantage of the smartphone’s omnipresence, each Chinese 29 company has created application on the Apple iPhone and Android platforms. Thereby, these ap- 30 plications bring all together all the features for using the bike: the position of the nearest bike, 31 its unlocking with the phone’s camera thanks with a scan of the QR code on the bike, and upon 32 completion of the ride, a manual lock of the bike. (9–12) 33 The main asset of this dockless system is the absence of stations, it allows the user to leave 34 a bike literally anywhere (within a large geographical area). 35 This freedom brings abuse and a lack of control of bicycle parking. Some bikes are left in 36 the middle of the footpath, lying and hindering pedestrian traffic. This behavior can be dangerous 37 for people with reduced mobility or blindness. (13) This lack of rigor from some users as well 38 as the excessive abundance of bikes can lead to a visual pollution and cause a strong skepticism 39 for this new system. Furthermore, these colorful bikes have appeared first in Chinese cities where 40 they are now strongly criticized because of the huge open-air bikes dumps. Spectacular pictures of 41 these dumps circulate on the Internet and feed skepticism of inhabitants and governments seeing 42 these bikes arrive in their city, it gives an excuse to people who vandalize them, hoping to get these 43 start-ups out of their city, as happened in France with the Gobeebike company. (6, 14–16) 44 Since 2014, and especially since 2017, these bikes have emerged on most continents. Is- 45 sues like vandalism that confronted the first generation of white bikes, again confront this fourth Heymes and Levinson 2

FIGURE 1 Timeline for the introduction and evolution of bikesharing systems in Australia. Source: https://en.wikipedia.org/wiki/List_of_bicycle-sharing_systems

1 generation, compounded with the new issues of visual pollution and congestion of public spaces, 2 and consequently it has seen rejection by some governments and residents. (15, 17). 3 This study investigates the deployment of dockless bikesharing in Sydney in the 2017-2018 4 period. Four dockless bikesharing companies operated in Sydney during the period of study: 5 • Ofo (b. April 2014, ) claimed 600 bikes in Sydney and was the first dockless 6 bikesharing system to appear. Its almost ten million yellow bikes, according to the com- 7 pany, generated over six billion bike trips. It landed in Sydney in October 2017, but it 8 announced July 10, 2018, the end of its Australian operations (in Sydney and ). 9 (11, 18–23) 10 • oBike (b. November 2016, ) claimed 1000 bikes in Sydney. It landed in Mel- 11 bourne was during the summer of 2017. Its arrival in was criticized, bikes 12 were heavily vandalized and the ’s Environment Protection Authority (EPA) 13 threatened large fines if the company didn’t clean damaged bikes. All this led to its 14 departure in June 2018. In Sydney, the yellow bikes operator is still operating (as of July 15 2018) but its longevity is uncertain. (10, 24–27) 16 • Mobike (b. January 2015, Beijing) now has ten million orange bicycles globally. It was 17 the last to arrive in Sydney, in November 2017, and claims 500 bikes, which appears an 18 underestimate. (9, 28, 29) 19 • Reddy Go (b. 2017, Sydney) was the only home-grown company seriously competing 20 for marketshare, but like Ofo, announced its departure on July 10, 2018. It claimed 2000 21 bikes, we are skeptical of the claim. (12, 17, 21, 22) 22 This paper first reviews bicycle and bikeshare regulation in Australia. It then examines the 23 operating territory of the four companies. This is followed by an examination of bicycle utilization 24 and bicycle condition. Heymes and Levinson 3

FIGURE 2 oBike and Ofo bikes thrown in FIGURE 3 A Mobike’s QR code erased, a canal in . Source: http://www. preventing the bike’s use. Source: Photo ecoblog.it by C. Heymes.

1 BICYCLE REGULATION 2 Bicycles and bicyclists are scarce on the auto-oriented streets of most Australian cities, with a 3 regionwide mode share of under 1% for work trips. (30) Sydney, in New South Wales, has several 4 regulations that are considered anti-bicyclist, including mandatory helmet laws and high fines for 5 violations by bicyclists. Sydney also lacks a well-developed separated bike path or bike lane 6 network that would support bicycle demand.(31) 7 When shared bikes arrived in their territory, some councils (local government units): Canada 8 Bay, City of Sydney, Inner West, Randwick, Waverley and Woollahra (figure 4), came together to 9 create the Inner Sydney bike-share guidelines, reviewed in March 2018. It determines the rules 10 and instructions to be followed by the various companies operating in Sydney. For example, after 11 a report of an inappropriate or illegally parked bike, the council can remove the bike and if the 12 company hasn’t picked up it, the council can recycle it. The rules expressed are: 13 • Move bike placed in dangerous spots within three hours and be proactive in the redeploy- 14 ment of the bikes, 15 • Create a phone line on call between 6am and 9pm for bike repair, 16 • Council staff are free to unlock bikes to move them, 17 • Companies must immediately deactivate broken bikes and remove them within a week, 18 • Set up rules to educate users about correct behaviors as correct bike parking and the 19 obligation to wear an helmet, Heymes and Levinson 4

FIGURE 4 The map of the Councils who signed the "Inner Sydney bike-share guidelines", created in December 2017.

1 • Provide a bell, helmet, front and rear lights, rear reflector and sturdy kickstand on the 2 bike, 3 • Have public liability insurance, 4 • Send relevant data to councils, whenever requested, for transport and urban planning 5 purpose, 6 • Encourage users to relocate bikes in built-up areas, move inactive bikes after 11 to 14 7 days, if they are not moved: they will be impounded and a fee will be charged, 8 (27, 32) 9 Bikeshare use around Sydney doesn’t respect the limits fixed by the operators, and they are 10 changing continuously, as illustrated in 5. Even a month after, the companies still have problems to 11 contain theirs bikes. It seems that Reddy Go had the most difficulty in containing its bikes within 12 its official geography. This can be explained by the fact that when a bike is outside of the border, it 13 becomes free for the user, maybe in the hope that he will bring it back to the proper area. Mobike 14 is the one which is the closest to its delimited area and Ofo has the most reasonable limit. But, 15 each company has its own policy and tries to keep the bikes in the delimited area and teach the 16 users appropriate behaviors. Also, each company has its own economic model and offers different 17 price for a ride and the amount of the deposit, shown in the table 1. (9–12) 18 Following the vandalism and the disrespect of the bikes, three companies (excluding Reddy Heymes and Levinson 5

FIGURE 5 The evolution of (top left) Mobike, (top right), oBike (bottom left), Ofo and (bottom right) Reddy Go areas between April and July 2018.

1 Go) set up a points system, shown in tables 2 and 3. In order to obtain a viable system, each 2 company has put in place rules to use its bicycles. Good behaviours are rewarded with points, 3 while poor behaviours are punished with a prohibition of using bikes or a higher price. In giving 4 or taking points, users are incentivized to use the bike in a proper way. The consequences differ by 5 company (Table 4). Heymes and Levinson 6

oBike Ofo Mobike Reddy Go Deposit $69 $0 $1 $99 30 minute ride $1.99 $1 $1.20 $1.99 TABLE 1 Deposit and 30 minute ride for each company, May 2018. Source: smartphone apps from Ofo, oBike, Reddy Go and Mobike.

Good Behaviors Pts Poor Behaviors Pts oBike report broken bike 2 report parked bike at non-designated areas 3 share ride to Facebook for the 1st time 1 Ofo report a broken bike 2 report an illegally parked bike 3 share your ride for the first time 1 parked in a preferred parking area 1 complete a trip 1 Mobike ordely use park the bike properly keep it clean maintain a healthy bikesharing community safe riding respect of the traffic rules

outstanding 701-1000 excellent 601 - 700 good 501 -600 TABLE 2 Good behaviours and theirs rewards for each company, May 2018. Source: smart- phone apps from Ofo, oBike and Mobike. Heymes and Levinson 7

Poor Behaviors Pts oBike parking at non-designated bike parking area -20 forget to lock but a bike isn’t lost -20 violation of traffic rules reduce to 0 losing a bike reduce to 0 move bike illegally reduce to 0 Ofo parking outside the geofence -20 parking illegally -50 improperly locking bike -50 violating traffic rules reduce to 0 using an unauthorized bike lock reduce to 0 damaging a bike reduce to 0 Mobike riding bike in an unsafe manner ignore traffic rules vandalize bike (like by putting personal locks) parking bikes in off-limits areas other civil violations obstructing other people

fair 301-500 poor 0 - 300 TABLE 3 Poor behaviors and theirs consequences for each company, May 2018. Source: smartphone apps from Ofo, oBike and Mobike.

Points Consequences oBike if the total of points is reduce to 0 user can no longer use a bike Ofo if the total of points is reduce to 0 user can no longer use a bike Mobike 501 - 1000 no impact on the bike’s pricing 301 - 500 (fair behavior) normal ride fee is doubled 0 - 300 (poor behavior) normal ride fee: 100x normal price for both fair and poor behavior no booking, monthly card, ride voucher Reddy Go no rules no consequences TABLE 4 The consequences of user’s behavior for each company, May 2018. Source: smart- phone apps from Ofo, oBike, Reddy Go and Mobike companies. Heymes and Levinson 8

FIGURE 6 Flyers on Mobike bicycles around the University of Sydney. Photo by C. Heymes.

1 METHODS AND DATA 2 Methodology 3 Despite repeated requests, none of the companies was willing to share bike location or utilization 4 data. Therefore the data comes from manually scraping screenshots of bikesharing apps and from 5 field investigations. 6 A focused study on the Pyrmont area, described below, measures the use and mobility 7 of the bikes by company. To have the most accurate data, phone applications were used. Each 8 company has its own application on which we can see the bikes within a geographic perimeter. 9 For one month, every four hours, screenshots from the application of each company located the 10 bikes within this specific perimeter. Afterward, the number of bikes is noted and recorded. We 11 gather several kinds of information, including the distribution of the companies in a specific area 12 and the percentage of stationary bikes (bikes in the same location as at the time of the previous 13 measurement). Heymes and Levinson 9

1 Geographic area 2 The phenomenon of dockless bikesharing is investigated in greater Sydney, and in more detail in 3 the Pyrmont , as shown in Figure 7. While each company has its own area of operation, they 4 tend to overlap and all serve central Sydney. The Pyrmont peninsula, an urban, and in part tourist, 5 area has been chosen for in-depth research. The water provides a natural boundary and delimits 6 the area, and Pyrmont contains a sufficient number of bikes to have interesting results. 7 This area has a significant potential for the use of bikes: although bike mode share is still 8 low in 2016, many employees walk to work. The potential of change of transport mode related to 9 the arrival of a new bikesharing can be interesting to study. (33) 10 For this analysis, we assume that all the bikes are accurately represented in the applications 11 (i.e. all bikes are in the applications, and all bikes reported by the application exist in the field at 12 that location), though we test that assumption later. 13 The operating areas have changed over time. As of April 2018, three companies (exclud- 14 ing Obike) reported their own official operating area within the app, by July 2018, all four apps 15 reported boundaries for the use of their bikes, shown in Figure 8. Ofo and Mobike reduced their 16 area to control the geographic spread of bikes, Reddy Go did not. Heymes and Levinson 10

FIGURE 7 (top) Sydney and the peninsula of Pyrmont. (bottom) Detail of Pyrmont. Source: Openstreetmap. Heymes and Levinson 11

FIGURE 8 Map of the Dockless Bikesharing companies in Sydney in (top) April and (bottom) July 2018. Heymes and Levinson 12

FIGURE 9 The Distribution of dockless bikes in the peninsula of Pyrmont. Source: field investigations.

1 RESULTS 2 Bicycle Use 3 As mentioned previously, Pyrmont was a convenient place to study bikes for a month. The results 4 of this observation follow the intuitions at the scale of Sydney. Mobike and oBike are the most 5 represented dockless bikesharing companies, as shown in Figure 9. Ofo is the least represented, in 6 contrast with Beijing, for example, where yellow bikes seems to be the most numerous. 7 However, despite the small number of Ofo bikes, they are the most mobile (least static), 8 see Figure 10. Mobike has a good score too. Recall that the more the bikes move, the more viable 9 is the system, so the aim is to maximize trips per day per bike. 10 Reddy Go, despite being first to enter the Sydney market, deployed the fewest bikes and 11 sees little use of them. 12 The two graphs highlight which company has its bikes that are the most mobile. Neverthe- 13 less, it is a percentage, so depends on the number of the bikes that companies leave at disposal. 14 For example, Ofo has, on average, the smallest percentage of static bikes, but it is also the com- 15 pany which has the fewest bikes in Pyrmont. However, Mobike represents almost half of the bikes Heymes and Levinson 13

1 available on the peninsula and, at the end of the experiment, has the smallest percentage of static 2 bikes. So, either the bikes are more used by the inhabitants, or the company moves its bikes and 3 redeploys them. 4 Consistent with the graphical intuition, a Fisher test to verify the difference of the variances, 5 and a Student t-test, corroborate that all companies differ in their percent static at a p-value at 5%. 6 Reddy Go doesn’t have a large dispersion range, apart from a few exceptions, the number 7 of static bicycles during four hours periods remains between 81% and 87% when its median is at 8 86%. Ofo has the largest variation: its first quartile is at 40%, its third one at 67% and its median 9 at 50%. There are no strong trends by day of week. Heymes and Levinson 14

FIGURE 10 (top) One and (bottom) two-day percentage of unmoved bikes in Pyrmont by Dockless bikesharing company. Source: field investigations. Heymes and Levinson 15

1 Bicycle Condition 2 A review of nearly 400 deployed shared bikes on the streets of Sydney has shown that oBike and 3 Reddy Go are the most vandalized: respectively 34% and 50% and that Mobike is the first when 4 it comes to helmet on bikes, with 14% of its bicycles with helmets. Finally, field investigations 5 highlight that all the companies have GPS tracking defects: from 67% for Reddy Go to 10% for 6 Ofo (table 5).

oBike Ofo Mobike Reddy Go Total Bikes in Sydney (claimed) 1000 600 500 2000 Average Bikes in Pyrmont (observed) 34 5 44 15 Bikes Used Daily in Pyrmont (observed (%)) 25% 68% 53% 18% Vandalized bikes (field observation) 3% 34% 10% 50% With helmet (field observation) 14% 3% 10% 0% GPS defects (field observation) 22% 31% 10% 67%

TABLE 5 Bike characteristics Heymes and Levinson 16

1 DISCUSSION AND CONCLUSIONS 2 The mobility of the bikes are one of the most important data for the dockless bikesharing system, 3 it represents the use and retention of riders for the system. This study shows that shared bikes are 4 not moving as much as expected or hoped for by the companies deploying bikes. Indeed, Australia 5 is one of the countries with the lowest rate of trips per bike and per day: only 0.3 compared to 2-6 6 for countries overseas. (22) 7 This can explain why two of the four companies operating in Sydney decided to leave the 8 city in July 2018: the low rate of trips per day per bike, a high level of vandalism, and a threat 9 to heavy fines from Councils made the system a system without financial profit. While dockless 10 bikesharing appears to be successful in many cities globally, the factors leading to its success have 11 not been replicated in Sydney to date. Heymes and Levinson 17

1 REFERENCES 2 [1] Wikipedia contributors, Bicycle-sharing system — Wikipedia, The Free Encyclo- 3 pedia. https://en.wikipedia.org/w/index.php?title=Bicycle-sharing_system& 4 oldid=847623027, 2018. 5 [2] DeMaio, P., Bike-sharing: History, impacts, models of provision, and future. Journal of Pub- 6 lic Transportation, Vol. 12, No. 4, 2009, p. 3. 7 [3] Beroud, B., Les expériences de vélos en libre service en Europe. Transports urbains, Vol. 8 111, 2007, pp. 1–5. 9 [4] Baca, A., What Cities Need to Understand About Bikeshare Now. Citylab, 2018. 10 [5] Razemon, O., Le vélo en libre-service a 10 ans mais toujours pas de modèle économique. Le 11 Monde, 2015. 12 [6] O’Sullivan, M., Sydneysiders want shared bikes but not the 25 million cost. Sydney Morning 13 Herald, 2018. 14 [7] Wikibooks contributors, Transportation Planning Casebook / Stationless Bikesharing in 15 Sydney — Wikipedia, The Free Encyclopedia. https://en.m.wikibooks.org/wiki/ 16 Transportation_Planning_Casebook/Stationless_Bikesharing_in_Sydney, 2018. 17 [8] Lévêque, F., Vélos en libre-service : la lutte pour survivre. La Tribune, 2018. 18 [9] Mobike, website. https://www.mobike.com/au/, 2018. 19 [10] Obike, website. https://www.o.bike/au/, 2018. 20 [11] Ofo, website. http://ofobicycle.com/, 2018. 21 [12] ReddyGo, website. https://www.reddygo.com.au/, 2018. 22 [13] Noyes, J., It’s definitely escalated the unseen danger of share bikes. Sydney Morning Herald, 23 2018. 24 [14] Schmitt, A., Five Ground Rules to Help Cities Get the Most Out of Dockless Bike-Share. 25 Streetsblog, 2018. 26 [15] Ehrenkranz, M., Can People Ever Be Trusted With Dockless Bike Sharing? Gizmodo, 2018. 27 [16] France Info, L’entreprise Gobee.bike quitte la France après la "destruction en masse" 28 de ses vélos en libre-service. https://www.francetvinfo.fr/economie/transports/ 29 gobee-bike-quitte-la-france-apres-la-destruction-en-masse-de-ses-velos-en-libre-service_ 30 2627378.html, 2018. 31 [17] Zhou, N., Dockless bike share: privacy and safety concerns voiced ahead of Sydney launch. 32 , 2017. 33 [18] Crunchbase. https://www.crunchbase.com/organization/ofo#section-overview, 34 2018. 35 [19] Wikipedia contributors, Ofo (company) — Wikipedia, The Free Encyclopedia. https://en. 36 wikipedia.org/w/index.php?title=Ofo_(company)&oldid=848909257, 2018. 37 [20] Daly, N., Bike-sharing company ofo cycles out of Australia, others remain in doubt. ABC 38 News, 2018. 39 [21] Joyce, E., Bike share companies Reddy Go and Ofo are leaving Sydney. Timeout Sydney, 40 2018. 41 [22] 9news, Ofo bike-sharing company to close in Australia. 9 News, 2018. 42 [23] Gorrey, M., Bump in the road as bike share operators Reddy Go, ofo quit Sydney. Sydney 43 Morning Herald, 2018. 44 [24] Wikipedia contributors, OBike — Wikipedia, The Free Encyclopedia. https://en. 45 wikipedia.org/w/index.php?title=OBike&oldid=848650615, 2018. Heymes and Levinson 18

1 [25] Crunchbase. https://www.crunchbase.com/organization/obike# 2 section-overview, 2018. 3 [26] abc.net, oBike to leave Melbourne after crackdown on bicy- 4 cle share company. http://www.abc.net.au/news/2018-06-12/ 5 -dockless-bicycle-scheme-to-leave-melbourne/9860314, 2018. 6 [27] Sas, N., Bike-sharing phenomenon set to stay in Sydney, but for how long? ABC News, 2018. 7 [28] Wikipedia contributors, Mobike — Wikipedia, The Free Encyclopedia. https://en. 8 wikipedia.org/w/index.php?title=Mobike&oldid=847201888, 2018. 9 [29] Crunchbase. https://www.crunchbase.com/organization/mobike# 10 section-overview, 2018. 11 [30] Loader, C., Trends in journey to work mode shares in Australian cities to 2011. Charting 12 Transport, 2012. 13 [31] Schoner, J. E. and D. M. Levinson, The missing link: Bicycle infrastructure networks and 14 ridership in 74 US cities. Transportation, Vol. 41, No. 6, 2014, pp. 1187–1204. 15 [32] Cormican, L., Bike sharing companies have three months to comply with new council rules. 16 Sydney Morning Herald, 2017. 17 [33] .idcommunity demographic resources, City of Sydney, Population density, 2016. City of Syd- 18 ney, 2016.