Presentation by Rema Hanna "Transparency is Power: Identification Cards and Food Subsidy Programs in Indonesia" at the SITE Corruption Conference, 31 August 2015.
Find more at: https://www.hhs.se/site
The Core Functions of the Bangko Sentral ng Pilipinas
Transparency is Power: Identification Cards and Food Subsidy Programs in Indonesia
1. Transparency
is
Power:
Iden2fica2on
Cards
and
Food
Subsidy
Programs
in
Indonesia
Rema
Hanna,
Harvard,
NBER,
BREAD
JPAL
Joint
with
Abhijit
Banerjee,
MIT;
Jordan
Kyle,
Columbia
University;
Benjamin
Olken,
MIT;
and
Sudarno
Sumarto,
TNP2K
and
SMERU
2. • Central
governments
have
a
goal
and
design
a
program
accordingly
• Local
officials
who
implement
the
programs
may
have
incen2ves
that
run
counter
to
the
central
government’s
goal
• Difficult
for
central
government
to
monitor
local
officials
Principal-‐agent-‐ci2zen
problem
3. • Subsidized
food
program
where
the
central
government
sets
the
rules,
but
the
leader:
– Has
leeway
in
implementa2on
– Has
more
informa2on
than
ci2zens
about
program
rules
• Would
providing
more
informa*on
to
eligible
households
about
the
rules
improve
program
efficiency?
– Does
form
of
informa2on
maXer?
In
our
context
4. • Background,
Experimental
Design
and
Data
• Overall
effect
of
iden2fica2on
cards
• Effect
of
Varying
Level
and
Type
of
Informa2on
• Conclusions
(and
scale-‐up)
Outline
6. • Distributes
15kg
of
subsidized
rice/month
at
a
subsidized
price
of
Rp.
1600/kg
to
over
17
million
eligible
households
– Largest
targeted
social
program:
• $1.5
billion
per
year
in
subsidies
• 3.4
million
tons
of
rice
• Distribu2on
run
by
the
village
officials,
who
pick
up
rice
from
sub-‐district,
distribute
it,
and
collect
the
co-‐payment
The
Raskin
Program
7. • Targe2ng:
84
percent
of
eligible
bought
some
rice;
67
percent
of
ineligible
did
so
as
well
• Leakages:
comparing
surveys
to
administra2ve
data
suggests
about
23%
of
rice
disappears
• Rice-‐markups:
mean
co-‐payment
in
our
data
is
Rp.
2,276
per
kg,
instead
of
Rp.
1,600
• Eligible
households
buy
5.3
kg/month
at
Rp.
2,276:
32
percent
of
value
of
intended
subsidy
Problems
8. • 572
villages
in
Indonesia
across
six
districts
(three
provinces)
Experimental
Design
9. • 194
villages
status
quo
• 378
villages
randomly
assigned
to
receive
cards
– Cards
included
name,
quan2ty,
and
instruc2ons
– Send
through
mail
service
(although
15%
who
received
it
got
it
directly
from
postman)
• Within
the
378
card
villages,
we
also
randomly
varied
cards
on
4
dimensions
(discuss
2
due
to
2me
constraints)
Experimental
Design
11. • Common
knowledge
facilitate
collec2ve
ac2on?
– Standard:
village
head
gets
list
and
one
copy
posted
– Public:
posted
many
copies
of
list
+
posters
Public
Versus
Private
Informa2on
12. • Control
households
think
that
the
copay
is
25%
higher
than
it
actually
is
and
pay
40%
more
in
reality
• Test
precisely
whether
varying
informa2on
on
the
cards
maXered
Price
Versus
No-‐Price
13. • Two
follow-‐up
surveys:
– About
two
and
eight
distribu2ons
aner
cards
were
mailed
(two
different
hamlets)
– 18
month
results
(complicated
due
to
other
programs
occurring),
but
in
paper
as
well
• Administra2ve
data
on
eligibility
status
Data
14. Outcomeivs
=
β0+
β1Cardsivs
+
αs+δv+εivs
• Pre-‐analysis
plan
for
government
• All
analysis
is
intent
to
treat
• Conduct
analysis
separately
by
eligibility
status
• Stack
2
month
and
8
month
observa2ons
in
same
regressions
(effects
similar
across
both)
• Cluster
by
village
• Randomiza2on
check
confirms
balance
Empirical
Analysis
16. • Increase
in
cards
for
eligible,
albeit
imperfect
• Eligible
households
9
percentage
points—or
30
percent—more
likely
to
know
their
status
Do
People
Receive
the
Cards?
Received
Card
Used
Card
Knows
Own
Status
(1)
(2)
(3)
Eligible
Card
0.30***
0.15***
0.09***
(0.02)
(0.02)
(0.02)
Control
Mean
0.07
0.06
0.30
Ineligible
Card
0.03**
0.04***
0.05**
(0.01)
(0.01)
(0.02)
Control
Mean
0.05
0.04
0.36
17. B oug ht
las t
two
months
Quantity P ric e S ubs idy
E ligible
C ard 0.02 1.25*** -‐57*** 7,455***
(0.01) (0.24) (18) (1,328)
C ontrol
Mean 0.79 5.29 2,276 28,605
Ineligible
C ard -‐0.06*** 0.07 -‐35 526
(0.02) (0.19) (24) (1,035)
C ontrol
Mean 0.63 3.46 2,251 18,754
• Subsidy
increases
by
about
~26%
for
eligible
• No
overall
decrease
in
quan2ty
for
ineligible
– Rice
increases
by
17%;
36%
decrease
in
rice
leakage
Effect
of
Cards
on
Raskin?
18. • General
protests
increased
• Those
who
bought
rice
complained
less;
complaints
about
distribu2on
fell
• Those
who
did
not
get
to
buy
rice
complained
more,
with
more
complaints
about
targe2ng
Impacts
on
Complaints?
Protest
Occurred
Bought Rice
Did not buy
Rice
Targeting Distribution
Card 0.07*** -0.09*** 0.07*** 0.07*** -0.07**
(0.02) (0.03) (0.03) (0.03) (0.03)
Mean Control 0.11 0.43 0.22 0.18 0.41
Complaints Occurred Type
19. E ligible Ineligble
C ard 3,175* 205
(1,622) (909)
L og
C onsumption 950 -‐3,107***
(1,078) (651)
C ard
x
L og
C onsumption -‐1,938 -‐176
(1,573) (798)
S ubs idy
• If
distor2ng,
we
should
see
a
shin
from
the
poor,
ineligible
to
the
rich,
eligible
• But
treatment
effect
of
card,
by
consump2on,
on
subsidy
cannot
be
rejected
from
zero
Undoing
a
“local”
fix
of
Targe2ng?
20. • Eligible
households
receive
26%
larger
subsidy
– This
is
with
only
30
pp
increase
in
cards
–
effect
may
differ
with
larger
penetra2on
of
cards
– Cost
effec2ve:
subsidy
increase
is
more
than
5
2mes
the
cost
of
the
cards
over
the
study
period
• Poorer
ineligibles
not
hurt
by
cards
–
suggests
that
cards
are
not
undoing
a
local
“fix”
of
a
bad
targe2ng
alloca2on
Summary
21. EFFECT
OF
VARYING
LEVEL
AND
TYPE
OF
INFORMATION:
(1)
STANDARD
VS.
PUBLIC
(2)
PRICE
VS.
NO
PRICE
22. • Common
Knowledge:
Cards
provide
individual
informa2on
on
one’s
eligibility
status
– If
I
want
to
protest,
I
may
need
to
know
if
others
would
join
– The
village
head
may
need
to
know
that
everyone
“knows”
to
keep
him
in
check
• Varied
informa2on
about
program:
– Standard
Informa*on:
List
sent
to
village
head
and
one
poster
with
beneficiary
lists
posted
– Public
Informa*on:
Posters
for
both
the
full
lis2ng
and
for
cards
(3
posters
per
hamlet)
and
mosque
radio
announcements
1)
Public
Versus
Private
Informa2on
23.
Eligible
Ineligible
Village
officials
Informal
Leaders
(1)
(2)
(3)
(4)
Panel
A:
Respondent
has
seen
the
list
Public
0.14***
0.10***
0.24***
0.12***
(0.02)
(0.02)
(0.06)
(0.05)
Standard
0.02*
0.01
0.05
-‐0.01
(0.01)
(0.01)
(0.05)
(0.04)
Difference:
Public
-‐
Standard
0.12***
0.10***
0.18***
0.13***
(0.02)
(0.02)
(0.06)
(0.05)
Control
Mean
0.07
0.06
0.36
0.12
Panel
B:
Respondent
correctly
identifies
own
status
Public
0.12***
0.08***
0.25***
0.00
(0.02)
(0.03)
(0.05)
(0.07)
Standard
0.06***
0.01
0.14***
-‐0.02
(0.02)
(0.03)
(0.05)
(0.07)
Difference:
Public
-‐
Standard
0.06**
0.07***
0.11**
0.02
(0.03)
(0.03)
(0.05)
(0.07)
Control
Mean
0.30
0.36
0.44
0.48
Greater
Knowledge
with
Public
24.
Eligible
Ineligible
Village
officials
Informal
Leaders
(1)
(2)
(3)
(4)
Respondent
believes
that
the
stated
category
of
individuals
has
seen
the
list
Public
0.36***
0.27***
0.24***
0.24***
(0.05)
(0.04)
(0.06)
(0.05)
Standard
0.08**
0.02
0.04
0.05
(0.04)
(0.02)
(0.05)
(0.04)
Difference:
Public
-‐
Standard
0.28***
0.25***
0.20***
0.19***
(0.05)
(0.04)
(0.06)
(0.05)
Control
Mean
0.31
0.15
1.04
0.47
Respondent
correctly
identifies
status
of
other
households
Public
-‐0.01
0.01
-‐0.00
-‐0.03
(0.01)
(0.01)
(0.03)
(0.04)
Standard
-‐0.00
0.03**
0.03
0.00
(0.01)
(0.01)
(0.04)
(0.04)
Difference:
Public
-‐
Standard
-‐0.00
-‐0.02
-‐0.03
-‐0.04
(0.01)
(0.02)
(0.04)
(0.04)
Control
Mean
0.66
0.32
0.60
0.63
Knowledge
About
Others?
25. Eligible
Households
Ineligible
Households
Bought
in
the
Last
2m
Quantity
Price)
Subsidy
Bought
in
the
Last
2m
Quantity
Price)
Subsidy
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Public
0.01
1.64***
-‐81***
9,666***
-‐0.08***
0.12
-‐46
764
(0.02)
(0.30)
(26)
(1,703)
(0.03)
(0.24)
(30)
(1,293)
Standard
0.02
0.83***
-‐24
4,839***
-‐0.03
0.10
-‐19
623
(0.02)
(0.31)
(29)
(1,764)
(0.03)
(0.25)
(30)
(1,347)
Public
-‐
Standard
-‐0.01
0.81**
-‐58**
4,827**
-‐0.06*
0.02
-‐27
140
(0.02)
(0.36)
(28)
(2,031)
(0.03)
(0.26)
(30)
(1,419)
Control
Mean
0.79
5.29
2,276
28,605
0.63
3.46
2,251
18,754
• Public
informa2on
nearly
doubles
the
impact
of
the
cards
for
eligible
• Effect
on
ineligible
households
driven
by
public
informa2on
Effect
of
Public
on
Subsidy
26. • Part
of
the
impact
is
that
cards
are
more
likely
to
be
distributed
by
6
pp
• But,
not
enough
to
explain
the
full
impact
of
the
public
info
treatment
on
the
subsidy
– If
you
compute
the
IV
effect
of
receiving
a
card
on
the
subsidy,
it
is
Rp
17,000
in
standard
and
Rp
32,600
in
public
(p-‐value
of
0.03)
– Suggests
that
public
informa2on
treatment
had
broader
effect
than
just
the
card
Discussion
of
Public
Results
27. Eligible
Households
Ineligible
Households
Bought
in
the
Last
2m
Quantity
Price
Subsidy
Bought
in
the
Last
2m
Quantity
Price
Subsidy
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Price
0.03
1.17***
-‐49
6,802***
-‐0.03
0.12
-‐43
861
(0.03)
(0.36)
(35)
(2,017)
(0.03)
(0.28)
(34)
(1,555)
No
Price
0.02
0.55
-‐6
3,200*
-‐0.03
0.14
18
664
(0.03)
(0.35)
(29)
(1,935)
(0.03)
(0.28)
(32)
(1,514)
Difference:
Price
-‐
No
Price
0.01
0.62*
-‐43
3,602*
-‐0.00
-‐0.02
-‐62**
197
(0.02)
(0.35)
(28)
(1,974)
(0.03)
(0.26)
(28)
(1,436)
Control
Mean
0.79
5.29
2,276
28,605
0.63
3.46
2,251
18,754
• Adding
price
to
the
card
increases
the
subsidy
for
eligible
• Quan2ty
decreases
with
printed
price:
– If
just
accountability,
we
would
expect
difference
in
price
– Quan2ty
“bump”
easier
to
give
to
some,
than
having
to
lower
price
for
all
2)
Price
Versus
No
Price
28. • Providing
informa2on
to
ci2zens
reduced
leakages
in
Raskin
– Public
informa2on
nearly
double
subsidy
gain
– Return
to
increased
informa2on
wriXen
on
cards
• More
broadly,
transparency
could
play
a
role
in
improving
program
func2oning,
but
it
is
important
to
think
about
the
form
of
it
• Policy
Impact:
in
response
to
vp
request,
influenced
Indonesia’s
card
program
Conclusion