Corruption
• Corruption in general
• Recent examples – overview
– Auditing politicians in Brazil (Ferraz and
Finan)
– Truckers paying bribes in Indonesia (Olken and
Barron)
– Getting a driver’s license in Delhi (Bertrand,
Djankov, Hanna, and Mullainathan)
– Paying parking tickets in the US (Fisman and
Miguel)
Corruption
• Use of public office for private gain
– Bureaucratic corruption
– Political corruption
• Before 1990: a bit of corruption could be good for
growth. Why?
– Kleptocrat cares about the size of the pie.
– Bribery may allow firms to get things done quickly.
• More recent view: corruption is one of the most
important impediments to development => why?
• Corruption wastes taxpayers’ money.
• Policies implemented solely to generate
opportunities for bribery.
• Undermines the rule of law - lax contract
enforcement is hard for businesses. A big
deterrent to firms.
• Wade’s article leads to more uncertainty about
water timing and less well-maintained
irrigation systems.
– Bribes are high where uncertainty is high.
Decentralized VS
Centralized Corruption
• Decentralized - many bribe takers, not a coordinated
effort.
• Centralized - leader organizes all corruption and
determines the shares going to each person.
• With decentralized the total theft rate will be higher
(more road blocks) but may actually have less bribe
revenue because of that (fewer travelers). Thus,
decentralized corruption has the worst incentives for
growth.
• Centralized corruption is not as bad since corrupt leader
cares about growth and the size of the pie and will limit
the amount lower level officials take.
Creating a new firm.
• In Bolivia follow 20 procedures, wait 82
business days, pay $2,696 in fees.
• In Canada follow 2 procedures, wait 2
business days, pay $280 in fees.
• In Georgia (the country) you can go through
12 procedures, wait 70 days, and pay $270,
or you can pay a legal advisory company
that does it for you $450 and have it ready
in 3 days!
Which are the most corrupt
countries?
• Data from 2009 from Transparency
International.
• Based on business people and
country analysts' perceptions of
corruption, defined as abuse of
public office for private gain.
• Questions include bribery, kickbacks
in procurement, embezzlement, and
strength of anticorruption policies.
• Thailand is ranked 84!!
1 New Zealand 9.4
2 Denmark 9.3
3 Sweden, Singapore 9.2
5 Switzerland 9.0
16 Austria 7.9
17 Japan, UK 7.7
19 US 7.5
84 Thailand 3.4
178 Burrma 1.4
179 Afghanistan 1.3
180 Somalia 1.1
Overview of Corruption Studies
• Auditing politicians in Brazil (Ferraz and
Finan)
• Starting in April 2003 as a policy by Lula.
Randomly choose municipalities to audit
every month. Gets info on all funds
transferred to the municipality from the
federal govt from 2001 on and finds out
what happened to them. Post the findings
on the internet and give them to the
media. By July 2005 have reports for 669
municipalities.
• Corruption includes fraud in
procurements (e.g. limiting the number
of bidders to benefit friends), diversion
of funds (giving funds to a phantom
firm that exists only on paper), and
over-invoicing goods and services
(paying more than it is worth).
• Of 669 municipalities 79% had
corruption, and among those the
average number of incidents was 2.45.
• There were elections in Oct 2004. 369
audits before that and 300 audits after
• Compare re-election rates for mayors in
audited vs non-audited municipalities.
– They find no effect of auditing on re-election
rates.
– It doesn't matter whether or not there was an
audit, it mattered what the audit found.
• Compare re-election rates for mayors with
low corruption and those with high
corruption.
– Findings of higher corruption lead to lower re-
election rates. Those mayors for whom no
corruption was found are then much more likely
to be re-elected.
– Looks like voters are expecting around 1
corruption incident per mayor.
• Look at effect of media and having a
local radio station in your municipality.
• (22% of municipalities have a local
radio station.)
• Biggest effects are when there are
pre-election audits and a local radio
station. If there is no local radio station
then the audits are much less useful.
• Did a similar analysis with newspapers
and finds nothing. People don't really
read the newspaper.
• In sum - auditing can be quite helpful
but need some way of getting the
information to the voter (radio?). With
no information transmission there are
no effects of auditing.
• In addition, while the auditing does
have an impact on political
accountability, we still don't know if it
has an effect on reducing corruption.
• Maybe the politician's new best option
is to be super corrupt in a first term
and not get re-elected rather than to
not be corrupt but be in power for two
terms.
Truckers paying bribes in
Indonesia (Olken and
Barron)
• Enumerators ride along with
truckers recording bribes paid on
304 trips (observe 6000 payments).
What do truckers pay for?
• i. Pay military and police to avoid
harassment.
• ii. Pay at weigh stations to avoid
fines for driving overweight.
• iii. Pay protection to criminal
organizations (avoid theft or
• Payments average $40, or 13% of trip costs. (53%
fuel, 14% cargo loading and unloading, 10% salary,
5% food and lodging)
• Road passes through Aceh and N Sumatra.
Indonesian govt withdraws police and military from
roads in Aceh to enforce a peace agreement. No
change in number of road blocks in N Sumatra.
What happens? Average bribe paid in N Sumatra
increases. But, not enough to offset revenue lost
from reduction in checkpoints in Aceh.
• Evidence of price discrimination - surplus that can
be extracted varies depending on checkpoints.
Downstream checkpoints have more bargaining
power and receive higher bribes. Also charge more
to people with newer trucks and more valuable
cargo.
Getting a drivers license in Delhi
(Bertrand, Djankov, Hanna, and
Mullainathan)
• Followed 822 people while they got a
driver's license. Average license getter pays
1080 rupees, or 2.5 times the official 450
fee. 60% don't take the exam.54% are
unqualified to drive according to an
independent test.
• People often go through agents to get their
license. Only 23% of those with agents take
driving test while 89% of those without
agent do. 53% of those with agent fail
independent test vs 25% of those who did
not have agent.
• 822 participants divided into one of
three groups.
• i. Control - just asked to return when
they got their license to fill out survey.
• ii. Bonus - offered a bonus of 2,000
rupees if they could get their license in
32 days (minimum is 30 days).
• iii. Lesson - provided with up to 15
free driving lessons.
• Results - 61% got license within 180 days.
On average took 42 days.
• i. Bonus group more likely to get license and
get it quickly, don't take test and drive
badly. Lesson group also more likely to get
license and pay less. Most extra money goes
to agents.
• ii. How do agents affect outcomes?
• iii. New Delhi prohibits agents so people just
(illegally) go somewhere else. You should go
based on your address.
• iv. Comparison of New Delhi livers with and
without agents.
Paying parking tickets in
the US (Fisman and Miguel)
• Is corruption cultural? Diplomats in
NYC had immunity until 2002 so they
didn't get into trouble if they didn't pay
their parking tickets.
• High correlation between country's
corruption measure and number of
unpaid parking tickets. In 2002 when
immunity ends, the relationship
disappears.
Corruption (II)
Spotlight on Indonesian
Road Building
How best to reduce corruption?
• Corruption creates efficiency cost.
• It has been suggested that corruption may
be a major contributor to the low growth
rates of many developing countries.
• Despite the importance of the problem, it is
difficult to measure corruption.
• Becker and Stigler (1974) suggests that the right
combination of monitoring and punishments can
control corruption.
• Problem? Transfer corruption between officials.
• An alternative approach which has gained
prominence in recent years is to increase
grassroots participation by community members in
local-level monitoring.
• Monitoring public projects is a public good =>
free riding problem
• Olken (2007) examine these 2 alternative
approaches to fighting corruption. He designed
and conducted a randomized, controlled field
experiment.
Details
• Setting
• Implementation
• Accountability
• Audits
Setting
• National Indonesian government
sponsored a development program
funded through a loan from the World
Bank.
• Finances 15,000 villages each year.
• Usually for small scale infrastructure or
for startup capital for micro-credit.
• Average grant is $8,800. Average annual
village budget is only $7,800 so this is
big.
Implementation
• Each year a village puts together a
proposal.
• Most common project is to cover a dirt
road with sand, rock, and gravel.
• Data comes from 608 road projects from
2003-2004.
• If project is funded, meeting is held to plan
construction.
• No contractors are allowed and no more
than 3% of costs can go to implementation
team for their time.
Accountability
• Lump sum grant at three points in
time.
• 40% of money is disbursed at first,
then accountability meeting, then
another 40%, then another meeting,
then another 20%.
• Must submit final accountability
report to be eligible for grant in next
year.
• Meetings attended by 40-50 adults
Audits
• Types of corruption include inflating the cost of inputs,
inflating the quantity of inputs, putting false workers
on the payroll, etc.
• After the projects were completed, engineers dug core
samples in each road to estimate the quantity of
materials used, surveyed local suppliers to estimate
prices, and interviewed villagers to determine the wage
paid.
• Missing expenditure = difference between what the
village claimed the road cost to build and what the
engineers estimated it actually cost to build = 24%
across the villages in the study
• Missing expenditure => corruption
Experiments
• External monitoring
• i. Audit Treatment - villages were told
at first meeting that they'd be audited
for sure, either during or after
construction. Told that results would go
to elected officials and to villagers.
Could affect the probability of getting a
grant in the future and villagers could be
angry. 1-4 months after start of
construction, audit 1 village in each
subdistrict to show them this was for
real. All villages were told they would
be audited again later.
• Local-level Monitoring
• ii. Invitation Treatment - Anybody
could go to the meeting, but in
practice they don't go unless they're
formally invited. Randomly gave
300-500 invitations. Sent home with
kids from school or gave to elders to
distribute.
• iii. Invitation with Comment
Treatment - in some invitation
villages they also added an
anonymous comment form to the
Experiment
• Audit randomized at the sub-district
level.
• Invitations randomized at the village
level.
• Did the randomization work?
• Not so well for audit treatment.
Results
• Main independent variable is the
reported amount spent minus the
independent account spent, as a
percent.
• Audits decrease corruption.
• But, still 20% missing when audits
are 100% certain to happen.
• Where does corruption come from?
Comes from quantities. Why?
• Main results for invitations. They do
have an effect on attendance and
speaking up.
• They do have an effect on the number
of corruption related issues brought up
and whether a `serious' response will
be taken.
But, find no effect whatsoever of invitations on
corruption.
• Does this imply that government
auditing can reduce corruption while
grass-roots monitoring is not
helpful?
• Are audits cost effective? They cost
$355 each and reduce the amount
lost by $468, plus you probably get a
better road too, etc.
Conclusion
• Increasing the prob of external audits substantially
reduced missing funds from 27.7 to 19.2%.
• Why not larger?
• Increasing in grassroots participation in
monitoring reduced only missing labor
expenditures, with no impact on materials and
overall.
• Why? What does this imply?
• Issuing anonymous comment forms reduced
missing fund only if the forms were distributed via
schools, completely bypassing village officials.