Is corruption good for your health?

Is corrup)on good for your
health?

Guilherme
Lichand

Marcos
Lopes

Marcelo
Medeiros

Harvard
University

CEPESP/FGV

PUC-‐Rio

FighBng
CorrupBon
in
Developing
and
TransiBon
Countries
–
SITE
Conference,
September,
1st

Every
dollar
that
a
corrupt

oﬃcial
or
a
corrupt
business

person
puts
in
their
pocket
is

a
dollar
stolen
from
a

pregnant
woman
who
needs

health
care.

In
the
developing
world,

corrup&on
is
public
enemy

number
1”

“

Even
in
the
case
of
peQy

bribery
or
extorBon,
it
is

relevant
to
ask,
what
is
the

alterna&ve?”

“

What are the causal eﬀects of corrup)on?
•  Experiments:

•  Bertrand,
Djankov,
Hanna
and
Mullainathan
(2004);
Zamboni
and
Litschig

(2013)

•  Natural
experiments:

•  Reinikka
and
Svensson
(2004);
DiTella
and
Schargrodsky
(2003)

This paper
•  Assesses
the
impacts
of
the
introducBon
of
a
random-‐audits
program

in
Brazil
on
the
incidence
of
corrupBon
within
Health
transfers;

•  Assesses
how
such
impacts
were
transmiQed
to
Health
outputs
and

outcomes.

Preview of findings
•  Random-‐audits
program
significantly
reduced
procurement

problems;

•  Outputs
and
outcomes
become
worse.

•  Mismanagement
increased
one-‐to-‐one
with
the
decrease
in

corrupBon.

•  Sharp
decrease
in
spending
points
to
procurement
staff
being

paralyzed
by
the
risk
of
misprocuring.

Audit reports
[Ghost
ﬁrm]
“When
evalua+ng
the
procurement
process
to
purchase

two
mobile
health
units
in
2002
by
the
municipal
government
of

Dourados,
auditors
found
evidence
of
fraud.
The
only
public
outbidder,

Santa
Maria
Comércio
e
Representações
Ltda.,
did
not
legally
exist

according
to
the
local
branch
of
the
Federal
Revenue
Secretariat
in

Cuiabá,
State
of
Mato
Grosso
do
Sul.
Despite
this
fact,
the
municipal

government
has
concluded
the
public
bid
and
paid
the
company
the

agreed-‐upon
amount.”

Audit reports
[Over-‐invoicing]
“When
analyzing
a
procurement
process
to
purchase

medical
supplies
in
2004,
auditors
found
that
the
municipal

government
of
Poloni
had
paid
higher
prices
for
medica+on
than
the

one
agreed
upon
the
public-‐bid
contract.
For
example,
according
to

receipt
number
115655
(Procurement
number
2004/01696),
the
correct

price
150
mg
of
the
medica+on
Rani+dine
was
R$
0.18
per
tablet,
but

the
municipality
paid
R$
0.28
per
tablet.
No
further
documenta+on
was

presented
by
the
municipal
government,
and
the
outbidder
Empresa

Soquímica
Laboratórios
Ltda.
embezzled
the
resources.”

What is corrup)on?
Table A1 – List of irregularities
Panel A: Corruption
Category Irregularity
Procurement Irregular receipts
Procurement Evidence for ghost firms
Procurement Contracts not signed or falsified signatures
Procurement Favored vendor
Procurement Lack of publicity
Procurement Documents set with different dates
Procurement Other procurement problems
Procurement Irregular class
Procurement No realization
Resource diversion Over-invoicing
Resource diversion Off-the-record invoice

Challenges
•  An
anB-‐corrupBon
program,
right
at
the
mid-‐point
of
mayors’
term,

but…

•  What
about
pre-‐treatment
data?

•  What
about
a
control
group?

•  Novel
dataset
with:

•  Pre-‐treatment
data:

•  Auditors
follow
transfers’
paper
trail,
usually
3
years
(but
someBmes
more)
prior
to
the

Bme
of
the
audit.

•  Very
detailed
informaBon
about
each
transfer:

•  In
parBcular,
extent
to
which
there
are
heterogeneous
opportuniBes
for
embezzlement.

Procurement-‐intensity
•  We
code
the
share
of
acBons
listed
under
Health
Ministry’s

descripBon
of
the
transfer
involving
procurement-‐related
terms,
such

as:

•  AcquisiBon;

•  Purchase;

•  ModernizaBon;

•  Enlargement;

•  ConstrucBon.

•  High
procurement-‐intensity
transfers
have
higher
opportuniBes
for

embezzlement.

Diﬀerences-‐in-‐diﬀerences
Corruption
Low High
Difference
[Low - High]procurement-
intensity
procurement-
intensity
Before
2003
0.17 0.37 -0.19***
[0.02] [0.03] [0.04]
After
2003
0.12 0.12 0.00
[0.01] [0.01] [0.01]
Difference
[Before - After]
0.05** 0.25*** -0.20***
[0.03] [0.03] (0.03)

Iden)ﬁca)on assump)on
•  Iden&cal
poten&al
(reported)
outcomes

•  Suﬃcient
condiBons:

Ø 
Parallel
trends
(between
treatment
and
control)
for
actual
outcomes;

Ø 
Parallel
trends
for
misreporBng
by
auditors
for
reported
outcomes.

Corrup)on: Before vs. aFer
18.2%

22.3%

21.6%

37.1%

28.0%

30.6%

14.1%

12.3%

14.3%

15.2%

12.1%

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

40.0%

1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

Corrup)on
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
45.0%
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Low
procurement-‐intensity High

Regression framework
𝐶 𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛↓𝑗, 𝑚, 𝑡  =

𝛽𝑃𝑜𝑠𝑡𝑃𝑟𝑜𝑔𝑟𝑎 𝑚↓𝑡  𝑥 𝑃𝑟𝑜𝑐𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑣 𝑒↓𝑗 + 𝑃𝑟𝑜𝑐𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑣 𝑒↓𝑗 + 𝛼↓𝑚 + 𝛼↓𝑡 + 𝛾 𝑋↓𝑗, 𝑚, 𝑡 +
𝜖↓𝑗, 𝑚, 𝑡

Results: Corrup)on
(1) (2) (3) (4) (5)
Municipal-
level
[2001-2004] [2001-2004] [1997-2007] [1997-2007] [1997-2007]
Post-2003 x -0.178*** -0.169*** -0.167*** -0.196*** -0.195***
High procurement intensity [0.0387] [0.0339] [0.0330] [0.0382] [0.0382]
High procurement intensity 0.214*** 0.174*** 0.135*** 0.135*** 0.135***
[0.0394] [0.0405] [0.0384] [0.0384] [0.0389]
Procurement intensity 0.0610** 0.0765*** 0.0752*** 0.0728**
[0.0304] [0.0290] [0.0289] [0.0294]
ln(transfer amount) 0.00379 0.00386 0.00338 0.00295
[0.00478] [0.00367] [0.00372] [0.00368]
Second-half of the term x 0.0421** 0.0434**
High procurement intensity [0.0208] [0.0207]
Second-half of the term -0.239*** -0.279***
[0.0487] [0.0462]
Municipality fixed-effects Yes Yes Yes Yes Yes
Year fixed-effects Yes Yes Yes Yes Yes
Term fixed-effects No No Yes Yes Yes
Draw fixed-effects No No No No Yes
Observations 1,932 7,072 10,538 10,538 10,538
R-squared 0.555 0.243 0.196 0.196 0.201
Robust standard errors clustered at the municipal level in brackets
*** p<0.01, ** p<0.05, * p<0.1
Transfer-level Transfer-level Transfer-level Transfer-level

Diff-‐in-‐diff coefficient by year

Diff-‐in-‐diff coefficient by draw

Health outputs and outcomes

𝑍↓𝑗, 𝑚, 𝑡 =1/ 𝑛↓𝑗  ∑𝑖↑▒ 𝑌↓𝑖, 𝑗, 𝑚, 𝑡 −¯ˉ 𝑌 ↓𝑖, 𝑗, 𝑡=0 / 𝜎↓𝑖, 𝑗, 𝑡=0   

⇒ 𝐸[ 𝑍↓𝑗, 𝑚, 𝑡=0 ]=0, 𝑉𝑎𝑟[ 𝑍↓𝑗, 𝑚, 𝑡=0 ]=1

High
Low

-‐  Hospital
beds;

-‐  ImmunizaBon;

-‐  %
adequate
water;

-‐  %
adequate
sanitaBon.

-‐  Coverage
of
Family
Health
program;

-‐  Medical
consultaBons.

Regression framework
𝑍↓𝑗, 𝑚, 𝑡  =

𝛿𝑃𝑜𝑠𝑡𝑃𝑟𝑜𝑔𝑟𝑎 𝑚↓𝑡  𝑥 𝑃𝑟𝑜𝑐𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑣 𝑒↓𝑗 + 𝜃↓𝑗 + 𝜃↓𝑚 + 𝜃↓𝑡 + 𝜂 𝑋↓𝑚, 𝑡 +
𝜐↓𝑗, 𝑚, 𝑡

Z-‐scores
-‐1.5
-‐1
-‐0.5
0
0.5
1
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Low

Results: Z-‐scores
(1) (2) (3)
[2001-2004] [1997-2007] [1997-2007]
Post-2003 x -0.309*** -0.494*** -0.495***
High procurement-intensity [0.107] [0.0948] [0.103]
High proc. intensity x 0.00140
Audit within 100 km in previous year [0.0977]
Audit within 100 km in previous year -0.109
[0.0897]
Municipality fixed-effects Yes Yes Yes
Year fixed-effects Yes Yes Yes
Term fixed-effects No Yes Yes
Observations 1,870 2,699 2,699
R-squared 0.556 0.513 0.514

Mechanism
(1) (2) (3)
Corruption Mismanagement Compliance
Post-2003 x -0.150*** 0.153*** -0.00262
Audit within 100 km in t-1 x -0.0424** 0.0247 0.0177
High proc. intensity [0.0187] [0.0253] [0.0199]
Audit within 100 km in t-1 0.136** -0.116 -0.0198
[0.0652] [0.0734] [0.0314]
R-squared 0.245 0.251 0.301
Robust standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1

Mismanagement categories
(1) (2) (3) (4) (5) (6)
Resource
diversion
Health
council
problems
Performance
problems
Infrastructure
and stock
problems
Human
resources
problems
Documentation
problems
Post-2003 x -0.0614** 0.0289* -0.0404 0.156*** 0.0169 0.0906**
High procurement-intensity [0.0260] [0.0162] [0.0324] [0.0280] [0.0147] [0.0360]
High proc. intensity x -0.0328* 0.0042 0.0188 0.0106 0.0079 0.0414*
Audit within 100 km in t-1 [0.0175] [0.00875] [0.0250] [0.0251] [0.0147] [0.0235]
Audit within 100 km in t-1 -0.0346 -0.0142 -0.0215 0.0398 -0.00578 -0.0915**
[0.0385] [0.0221] [0.0456] [0.0482] [0.0222] [0.0424]
Municipality fixed-effects Yes Yes Yes Yes Yes Yes
Year fixed-effects Yes Yes Yes Yes Yes Yes
Observations 7,072 7,072 7,072 7,072 7,072 7,072
R-squared 0.237 0.125 0.136 0.21 0.124 0.165
Robust standard errors clustered at the municipal level in brackets
*** p<0.01, ** p<0.05, * p<0.1

Spending
8
9
10
11
12
13
14
15
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Low

Spending
ln(transfer amount)
(1) (2) (3)
Post-2003 x -0.490*** -0.455*** -0.562**
Number of investigations 0.0234*** 0.0253***
[0.00749] [0.00866]
ln(supply of funds by transfer group) -0.0279
[0.0277]
R-squared 0.416 0.419 0.446
Robust standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1

A simple decomposi)on
𝑑ln⁠( 𝐶↓𝑡 ) /𝑑 𝑡 = 𝑑 𝑙 𝑛( 𝑌↓𝑡 / 𝑋↓𝑡  )/𝑑 𝑡 + 𝑑 𝑙 𝑛( 𝑋↓𝑡 / 𝑇↓𝑡  )/𝑑 𝑡 

−0.2 = 𝑑 𝑙 𝑛( 𝑌↓𝑡 / 𝑋↓𝑡  )/𝑑 𝑡  − 0.5

⇒ 𝑑 𝑙 𝑛( 𝑌↓𝑡 / 𝑋↓𝑡  )/𝑑 𝑡 =0.3

“Big ideas”
•  Procurement
mistakes
might
be
interesBng
to
incorporate
in
a
model

Ø 
Capacity-‐building
intervenBons;

Ø 
SelecBve
monitoring:
larger
transfers,
where
incenBves
to
embezzle
are

much
higher;

Ø 
InteracBon
with
asymmetric
punishments.

•  “Cap
and
trade”
for
corrupBon

Ø 
Diﬀerent
ciBes
have
diﬀerent
costs
of
reducing
corrupBon.

Is corruption good for your health?

Recommended

Recommended

More Related Content

Similar to Is corruption good for your health?

Similar to Is corruption good for your health? (20)

More from Stockholm Institute of Transition Economics

More from Stockholm Institute of Transition Economics (20)

Recently uploaded

Recently uploaded (20)

Is corruption good for your health?