Erin Kelley 1 Gregory Lane 1 David Schönholzer 2 Wagacha Peter Waiganjo 3. CEGA Conference on Infrastructure Monitoring, October 2016

Transcription

1 Mobile-based Matatu Monitoring in Kenya: Using Tracking Devices and Smartphone Technology to Measure Safety and Productivity of Informal Transit in Developing Countries Erin Kelley 1 Gregory Lane 1 David Schönholzer 2 Wagacha Peter Waiganjo 3 1 UC Berkeley, Department of Agricultural and Resource Economics 2 UC Berkeley, Department of Economics 3 University of Nairobi, School of Computing and Informatics CEGA Conference on Infrastructure Monitoring, October 2016

2 Measuring Firm Productivity and Safety Externalities Scarce information on small firms in developing countries: 1 Measuring productivity Low productivity pervasive in developing countries Hsieh and Klenow 2009 Contracting with employees understudied Crude, short-term incentive structure common 2 Measuring externalities Reduced consumer welfare due to agency problems Negative spillovers between firms 2 / 17

3

4 Matatu: A public transport vehicle in Kenya Can be found all over Africa, South Asia, Latin America Common features (Bruun and Behrens 2014): Licensed route networks Small fleet or single-vehicle owners, few owner-drivers Small-scale entrepreneurs Stage fill-and-go system Cash fares Target system for drivers In Kenya, three million people use matatus daily More than 10,000 matatus in metropolitan Nairobi 70% of commuter traffic

5 Asymmetric Information in matatu industry: Matatu owners don t know... (a) how much revenue the driver is making (b) how much risk the driver is taking Information asymmetries between owner and driver Driver hired under a target contract Easing information frictions may change contracts and: (a) increase productivity: higher effort and lower costs (b) decrease negative externalities: owner cares about asset Vehicle Monitoring in developing countries: Rapidly spreading for fuel monitoring and fleet management

6 This Project Objective: Evaluate the use of monitoring technology in firms 1 How do contracts change as information frictions decrease? 2 How does this affect productivity and employee behavior? 3 More generally: What are salient spatio-temporal features of informal transit? What we do: Equip 300 matatus with monitoring devices Focus on single-matatu owners Today: preliminary results from 44 matatus Send info about productivity and safety to treated owners Observe changes in: 1 driving behavior 2 firm and employee outcomes 3 changes in contracts 4 / 17

7 Project Status Development Stage IntervenGon Stage June 2014: Begin project development Pre-pilot sample (8 units) March 2015: First prototype deployed January 2016: Recruitment underway Pilot Stage: 44 units; 22 T, 22 C Full Stage: 253 units (expected*) 120 T, 120 C, 13 pure C (expected*) *Device failure rates have been around 10% in the pilot stage. EsGmaGng a slightly higher rate (15%), we are able to deploy 253 units of the 298 we have available for the full stage. Cross-Treatment Rollout 126 T, 127 C Open Access Stage January 2015: March 2016: 2017 Exploratory surveys Exploratory surveys and focus groups and focus groups 5 / 17

8 Our Monitoring System: Overview Developed a monitoring system suited for matatu industry Hardware: tracking device with: (a) GPS (b) 3D-Accelerometer: 1. Backward acceleration hard braking 2. Forward acceleration strong acceleration 3. Lateral acceleration sharp turning 4. Vertical acceleration offroad driving, speedbumps (c) GPRS data transmission to server Much cheaper than standard fleet management system Validated the system extensively Install tracking device in all participating matatus Android phone with our monitoring app preinstalled to owners Treatment group: Receive safety and productivity info Treatment and control group: collect daily revenue and costs 6 / 17

9 Our Monitoring System: App Overview Features for treatment group: 1 Check location 2 Check real-time productivity 3 Check real-time safety 4 Change notifications Features for all: 1 Submit revenue and costs 2 Set reminder for submission 7 / 17

10 Our Monitoring System: Productivity Information Information items: 1 Location 2 Mileage 3 Off-route 8 / 17

11 Our Monitoring System: Safety Information Information items: 1 Speeding 2 Sharp turns 3 Acceleration 4 Sharp braking 9 / 17

12 Our Monitoring System: Information Preferences Tracking Usage: 1 Logging into the app 2 Show in feed 3 Push notification 10 / 17

13 Our Monitoring System: Data Submission Submit data for: 1 Daily target 2 Revenue received 3 Repair costs 4 Any other info Similar information from driver: 1 Revenue 2 Residual income 3 Any other info 11 / 17

14 Secondary Treatment: Cash Incentives Private versus public enforcement of safe driving: Information treatment may internalize externalities But firm produce externalities with zero information frictions Contrast a market mechanism to regulatory body Mimic a fine system Cash incentives as secondary treatment: Directly provide monetary incentives to drivers for safe driving Bonus for incurring less than x violations Allocate y shillings to the driver Every violation reduces this amount it by z shillings 12 / 17

15 Driving Behavior: Speed density Speed in km/h Treatment Control 13 / 17

16 Driving Behavior: Location for Kiserian Route Matatus -.6 Latitude Longitude Treatment Control 14 / 17

17 Driving Behavior: Hour of Day density Hour of Day Treatment Control 15 / 17

18 Driving Behavior: Acceleration.3.5 density.2.1 density Forward Reverse Treatment Control Treatment Control.25.3 density density Lateral Vertical Treatment Control Treatment Control 16 / 17

19 Preliminary Conclusions and Next Steps Preliminary lessons: High-frequency tracking of informal transit possible Many applications besides ours Some indication of information treatment effect on driving behavior New opportunity to study mapping from driving behavior data to safety Going forward: Begin fitting another 250 minibuses soon Human measurement/validation of driving behavior Improve reporting data quantity and quality: New strategies for high response rates Possible tests for differential truthfulness of drivers 17 / 17