Title: Obtaining a Driving License in India: An Experimental Approach to Studying Corruption
1Obtaining a Driving License in India An
Experimental Approach to Studying Corruption
- Marianne Bertrand (Chicago GSB, CEPR, NBER, IZA)
- Simeon Djankov (World Bank)
- Rema Hanna (NYU Wagner School of Public Service)
- Sendhil Mullainathan (Harvard U and NBER)
2Introduction
- Provision of public services in many developing
countries rife with corruption - Numerous anecdotal evidence
- Newspaper articles/exposes
- Simplest observation
- Aggregate indicators of corruption
- Micro-level indicators through subjective reports
3What are the Efficiency Implications of
Corruption?
- What we Know Empirically
- Cross-country studies tell us there is a negative
correlation between corruption and growth (for
example, Mauro, QJE) - Problem clear omitted variables and reverse
causality questions - (Micro progress on measurement of corruption,
designing and anti-corruption methods)
4What are the Efficiency Implications of
Corruption?
- Theoretically
- Corruption can actually improve efficiency when
used to get around rigid or inefficient rules. - Greasing the wheels, improving allocative
efficiency - But not all rules are inefficient, and corruption
is bad from an economic standpoint when it
subverts an efficient rule - In addition, corrupt bureaucrats may create
additional, inefficient rules to extract higher
surplus - Also, the bargaining process between bureaucrat
and citizen may induce deadweight loss
5Our Objective in This Paper
- Develop an empirical design that allows for a
more direct testing of these different
theoretical efficiency implications of corruption - Consider the following simple set-up
- Bureaucrat involved in screening process
- Good types and bad types apply
- Good and bad are social definition
- Bureaucrat screens for type
- Assume perfect screening technology available
- Should (socially) give permit to only good type
6Questions One Would Like to Ask Within this Setup
- Do bureaucrats respond to individual needs?
- Example I have a need to get a passport faster
than the typical rules. Can I pay more to get it?
- Do only good types get the license?
- Example I want to circumvent background checks
that prevent false passports. Can I pay more to
do this? - Are bureaucrats creating red tape to extract
surplus? - Example Does a bureaucrat invent a rule that you
need to pay to circumvent? All passport
applications must be printed on a particular
(hard to find) size of paper? - Is the bargaining process with the bureaucrat
efficient? - Example I spend 1 hour bargaining on the bribe.
7 What We Do
- Focus on one specific license/screening process
- Getting a driving license in New Delhi
- Survey experiences of actual process
- Number of trips, number of bureaucrats spoke to,
time spent filling in paper work, etc - Payments made through the process
- Experimental design to get at various efficiency
margins - Vary individuals private needs for license
- Vary individuals deservingness of license
- Create a portable research design
- Easy to apply in other places and contexts
8Plan of Talk
- World Bank Project Research Design
- Obtaining a drivers license de jure
- Project Implementation Details
- Sample, experimental design, surveys, etc.
- Obtaining a driving license de facto
- Main Efficiency results
- Relating outcomes to private need for license and
deservingness of license - Focusing on the specifics of the interaction
between bureaucrats and citizens - Endogenous red tape
- Bargaining efficiency
9World Bank ProjectResearch Design
- Choice of service to study.
- Would like
- Process to be short and discrete enough that
people can be tracked - Variation in individual needs
- Large numbers of people regularly going through
process for sample size purposes - An easy measure of social efficiency, whether
someone should get the license or not - Portable. Would like it to exist in many places
10Choice Driving License
- Short, discrete process
- Three to six month process
- Variation in individual needs
- Some people need a license faster than others.
- Time cost differences. Going off to school, etc.
- Variation in social deservingness
- Good and bad drivers
- Can give driving exam to measure deservingness
- Large numbers of people
- In any town or city, large numbers apply
- Portable
- Driving licenses are very common nearly
everywhere in the world
11The Official Licensing Process
Temp Licenses Cost Rs360 18 years of age
Proof of age Proof of residence Medical
certificate (Rs40) Three photos Test on road
signs, traffic rules and traffic
regulation7-day waiting period if fail test
Perm Licenses Cost Rs 90 Practical driving
test as per Rule 15 of the Motor Vehicle Rules
and Forms, 1989 7-day waiting period if fail
test
After 30 days and within 6 months
12World Bank Project Overview
- Follow roughly 800 people through process of
getting driving licenses in New Delhi - Experimental design
- Additional reward for getting license quicker
- Couple of days above official minimum
- Bonus Group
- Offer free driving lessons upfront
- Lesson group
- Surprise test of driving skills at end
- Compensation
- Money for time
- Surprise offer of free lessons to all at end
13Timeline
Random Assignment Initial Survey
Temp. License. Short survey
Learning on Own or w/ School
Recruitment
Survey Surprise Driving Test
Attempt to Get Perm. License
Free Driving School for All
14Recruitment
- June 2004 November 2004
- Continuous recruitment
- Recruiters intercepted individuals who were
entering one of the following four Regional
Transport Offices (RTOs) Southwest, Northwest,
South or New Delhi - Recruitment rules
- Only men (100 attrition among women in pilot)
- Only over 18
- Currently without a permanent or temporary
license but wanting to get one
15Initial Session
- 23 initial sessions, 822 participants in total
- Required to bring formal identification, verify
phone number - Introduction survey
- Demographics, attitudes and beliefs towards
public services, knowledge about driving and
driving license process - Data given to us stripped of all individual
identifiers. - Random assignment into one of 3 groups control,
bonus, lesson
16(Everyone given (unexpected) free driving
lessons upon completion of surveys)
17Follow-up Surveys
- Short phone survey upon obtaining the temporary
license - Experiences through bureaucratic process,
payments - In-person survey upon obtaining the permanent
license - Experiences (length of process, number of
bureaucrats spoke to, etc..), payments - Surprise driving exam (conducted by accredited
school in New Delhi) - Oral exam can the subject safely operate a car?
(accelerator versus brake, etc) - If no automatic failure on oral exam, physical
road test - Drop-out surveys upon completion of project
- Those without temporary license
- Those with temporary license but without
permanent license (survey surprise driving
exam) - Data still in the process of being coded
18The Licensing Process in Practice
19A First Look at the Data OLS
- In cross-sectional data, one can ask
- Does paying more help in reducing bureaucratic
rigidity? - Does paying more help bad types get a license?
- Also
- Are good types subject to less bureaucratic
rigidity? - Present results for both full sample and control
group only -
20(No Transcript)
21(No Transcript)
22Observations
- Higher payment associated with
- Less bureaucratic hassle
- Lower likelihood to take the driving test
- Lower driving ability
- Higher driving scores associated with
- More bureaucratic hassle
- Higher likelihood to take the driving test
- Caution with interpretation
- For example those with higher ability to pay may
be systematically worse or better drivers need
not imply that corruption led to subversion of
efficient rule - Experimental design
23Experiment
- Recall 3 randomly assigned groups
- Control group
- Bonus group (e.g. individual need)
- Lesson group (e.g. social deservingness)
- Does bureaucratic process, who gets a license,
and price paid differ across groups?
24(No Transcript)
25Pre-Experiment Characteristics
- Table 3 Socio-economic and past driving
experiences - Could be over-sampling some groups (e.g. young
and student) but unsure - Note random assignment guarantees no bias.
Treatment effect estimated on specific group. - Some small differences across groups
- People willing to talk about bribing
- Assume heavy bureaucracy (6.82 trips/license)
- Little knowledge of procedures
26Impact of Group Assignment on Outcomes
27Observations
- Bureaucratic system responsive to private value
but not public value - Bonus group much more likely to get license
- Lesson group only barely/not
- Many bad licenses given
- Many people who get license automatically fail
our independent exam - Especially in bonus group
28Impact of Group Assignment on Process (I)
29Impact of Group Assignment on Process (II)
30Observations
- Speed group much less likely to
- Be confused by process
- Bother learning how to drive
- Lesson group
- If anything, faces more bureaucratic hassle
- MORE likely to actually take (and retake) exam
31Impact of Group Assignment on Payment
32Observations
- Interestingly, little difference in amount spent,
though lesson group spent a little less. - Driven by lower use of agent
- Little or no bribing except through agent
- No bargaining per se between agents and
bureaucrats - Could it be under-reporting of bribes?
- Are willing to report it elsewhere in survey
- Key variable/actor in this entire process is the
agent
33Agents (I)
- Agents (a.k.a. facilitators, consultants)
help obtain services - Personal connections with the government
bureaucracy knowledge of procedures - India illegal, yet established institution for
obtaining license, passports, and other services
34(No Transcript)
35(No Transcript)
36Agents (II)
- Many questions
- Why agents versus direct bribing?
- How frequent are interactions between agents and
bureaucrats? - What fraction of agents payment do the
bureaucrats get? - What does it take to become an agent?
- No answers provided here
- Still trying to interview a few agents
- But can look at how agents interact with the
licensing process in more details
37Agents Effects on Outcomes
38Agent Experience Across Assignment Groups
39Timing of Agent Hire on Outcomes
40Observations
- Agents allowed for
- Fewer officials to see
- Much less time
- Much less testing
- Amongst those who did not use agent, little
differences between groups. - Amongst those who did, little difference between
groups - Primary difference is in agent utilization rate
- Main reason for switching to agent is confusion
- Higher fraction in lesson group reporting
switching to agent because of failure on the
test!
41What Drives Agents?
- If all corruption takes place through agents,
audit study of agents may lead to more direct
evidence on what agents can/cannot, and at what
price - Performed such an audit (pilot)
- Different treatments
- Missing residential proof
- Missing age proof
- Cannot drive
- Cannot come back
42Audit Results
43Observations
- System is insensitive to ability to drive
- Price is not higher. Claim can still get license
for them. - System is sensitive to objective criteria
- Missing residential proof/age proof
44The New Delhi RTO
- Relocated to a new building surrounded by federal
government buildings - Agent more actively banned at the New Delhi RTO
- New Delhi versus other RTOs what if agents were
banned? - Obvious caveats
- Short-run partial equilibrium
- Omitted variables
45The New Delhi RTO
46Conclusion (I)
- Bureaucrats are responsive to private costs and
benefits - Higher willingness to pay for license associated
with less red tape, higher likelihood to obtain
license - Corruption removes/bends inefficient/rigid rules
- Bureaucrats also pass many who they shouldnt/are
not responsive to social deservingness to get the
license - Corruption also subverts efficient rules
- Limited bargaining inefficiency
- Agents do bargain for them. Could be bargaining
for agents but there is competition here. - Red tape creation by bureaucrats to induce
switching to agents?
47Conclusion (II)
- Is driving unrepresentative?
- Social externalities/good rules exist in the
provision of many other public services
(education, health, regulation of pollution, etc) - Future work
- Currently developing surveys to better understand
nature of the trade-off between direct bribing
and agent usage - Across different kinds of public services
- Across different geographic areas
- Replicating this research design in other
contexts/locations