A New Cross-National
Measure of Corruption
Laarni Escresa and Lucio Picci
European University Institute
University of Bologna
UP School of Economics Seminar
April 23, 2014
Introduction: Measuring corruption
- Corruption is one of the major problems, especially in developing countries
- What causes corruption? What are its consequences? Debate exists in the
literature.
- Studies mostly rely on cross-national perception based measures.
◦ Transparency International Corruption Perception Index (TI-CPI) and World Bank
Control of Corruption Index (WB-CCI)
◦ Perceptions may be weakly correlated with actual corruption
◦
◦
◦
◦
◦
Definition of corruption: petty vs grand, legal, cultural, norms
Depend on level of corruption
Other factors: presence of media density
Individual idiosyncrasies: ethical standards
Foreign and local respondents
Measuring corruption: Alternatives
- Survey: experience-based
◦ Expensive
◦ Reticence: Knowledge or participation in a corrupt transaction, which is illegal
increases reticent behavior. 74% respondents are reticent (Kraay and Murrell, 2013)
◦ Country variation in degree of reticence makes it even more problematic. Gallup
World Poll effective reticence 21% in Indonesia and 48% in India
◦ Can these biases be corrected?
 Use of concrete measures hard or objective data
◦ Golden and Picci (2005) – inventory of public infrastructure in Italy
◦ Glaeser and Saks (2006)– use corruption conviction US Federal level
◦ None at the cross-national level
Cross-national measure of corruption
based on cross-border corruption cases
- Corruption and judicial statistics
◦ “less subjective, cover longer time span, not subject to problems of sampling error and
survey non-response” (Glaeser and Saks)
◦ Enforcement, probability of detection also affect the number of cases
◦ Judicial statistics used within country studies
◦ Glaeser and Saks: corruption cases at the Federal level where assumption of probability of
enforcement is relatively less problematic
- We argue that data on cross-border corruption could also sidestep these issues
- Corruption cases involving the bribery of foreign officials by US firms is not informative of
corruption level in the US, but the distribution of cases wrt nationality of foreign officials is
informative of relative corruption levels abroad.
- Spatial distribution of cases pursued in a given country to evaluate levels of corruption in
other countries
Data
- Cross-border corruption cases: Firms headquartered in country i bribing
officials in country j
◦ 1,138 cases from 1977-2012 (but only use 1998 onwards)
◦ Discard Oil for Food cases
◦ Coding where enforcement action was first pursued allows to control for the
varying country-specific probability of enforcement of anti-bribery legislation
◦ Collection of data made possible by reporting of foreign bribery cases
◦ OECD Foreign Bribery Convention
◦ Treats bribery of a foreign public official as if it were the bribery of a local official
and impose same level of sanctions, prosecuted in home country
◦ Thus, relatively same legal definition and standard of bribery in home country and
same probability of enforcement and detection
Country of firm’s Total Positive Ongoing
headquarter
Cases Cases
Cases
US
344
213
108
Germany
91
55
35
UK
59
26
28
France
52
37
13
Switzerland
40
33
7
Canada
21
6
7
China
19
10
9
Italy
17
2
15
Netherlands
17
8
9
Australia
16
5
11
Japan
16
10
6
Spain
15
1
10
Sweden
11
0
9
Austria
10
0
10
Portugal
9
0
7
:
:
:
TOTAL
816
432
322
Foreign country
China
Nigeria
India
Kazakhstan
Brazil
Korea
Russia
Indonesia
Argentina
Mexico
Poland
Angola
Philippines
Libya
Egypt
TOTAL
Total Cases
96
45
42
26
22
22
21
20
15
17
16
15
15
14
12
:
816
Positive Ongoing
Cases
Cases
49
44
29
6
17
22
7
15
8
11
16
4
8
8
17
3
5
8
5
8
12
1
10
4
9
2
2
9
6
6
:
:
432
322
Total Cases
Positive cases
Ongoing cases
enforced:
enforced:
enforced:
Time period
Tot
hq
US
ot
fo
Tot hq
us
ot
fo
Tot hq
us
ot
fo
1998-2002
308 201 63
27
17 243 162 60
14
7
54
29
3
13
9
2003-2007
256 154 48
13
41 138 82
31
11
14
97
57
12
3
26
2008-2012
252 145 22
22
69
4
9
22 171 102 18
6
45
tot
816 500 133 60 111 432 260 95
34
43 322 188 33
21
80
51
16
An illustration: Cases first pursued in the headquarters country. 1998-2012
Note.
The total number of cases first pursued in the US is 307, and in DE is 46.
HQ
FO
US
US
Germany
Germany
China
Italy
China
Italy
(1)
Cases,
first
pursued
in HQ
64
5
4
1
(2)
Cases,
as % of total
n. cases first
pursued in HQ
20.8%
1.6%
8.7%
2.2%
(3)
Exports HQ→ FO,
as % of tot
exports of HQ
Ratio
Col. (2) /
Col. (3)
9.2%
1.2%
4.0%
6.7%
2.25
1.27
2.15
0.32
The PACI
N
 cases _ obs _ HQ
i,z
PACI z =
100
i=1
N
 E(cases _ obs _ HQ
i,z
)
i=1
E(cases _ obs _ HQi , z ) =
N
X iz
  cases _ obs _ HQi , j
N
X
j=1
j=1
ij
Comparison between actual no. of corrupt
transactions in z and the expected no. if the
spatial distribution of cases reflect
bilateral trade shares between i and z.
Assumptions
Assumption 1
The probability that a corrupt transaction involving firms from country i and public officials in
country j is observed and first pursued in country i, does not depend on the identity of the
foreign country j: (non discrimination)
pr _ obs _ HQi , j = pr _ obs _ HQi
Assumption 2
The total number of corruption occurrences involving country i’s firms and country j’s officials, ,
is determined by the probability of that a firm proposes a bribe and the probability that official
in j accepts and the number of transactions:
corr _ exchi , j = pr _ corr _ HQi  pr _ corr _ FO j  transactio nsij
Assumption 3:
Bilateral transactions are proportional to the value of exports from country i to country j, ,
according to a constant factor k:
cases _ obs _ HQi , j = pr _ obs _ HQi  pr _ corr _ HQi  pr _ corr _ FO j  k  xij
 probability that a cross border transaction involving z is corrupt/ weighted average of
probabilities for all countries
Measures of dispersion
1. imprecision with which we estimate the index of corruption in countries with
little foreign trade; prob of zero cases
Pr_zero _ cases zPACI = i1 ( Pr_cases _ obs _ HQi , z = 0 | Pr_corr _ FOz = Pr_corr _ FO AVG ) )
N
2. degree of coherence of the information on the level of corruption of country z
coming its different trading partners; coefficient of variation: Std. dev/ ave.
DispersionzPACI =
N
1
2
(cases
_
obs
_
HQ

E(cases
_
obs
_
HQ
|
PACI
)
)

i,z
i,z
z
(N  1 ) i=1
1 N
cases _ obs _ HQi , z

N i 1
PACI
Country
ISO
Finland
FI
Denmark
DK
Sweden
SE
Ireland
IE
Australia
AU
Canada
CA
United Kingdom
UK
Belgium
BE
Japan
JP
Netherlands
NL
Germany
DE
United St. of America US
France
FR
Spain
ES
Switzerland
CH
Singapore
SG
Taiwan
TW
Norway
NO
Israel
IL
Mexico
MX
Italy
IT
Korea
KR
PACI
0.000
0.000
0.000
0.000
0.000
0.000
3.036
4.480
7.057
8.315
9.074
9.797
9.934
9.983
10.581
16.373
19.786
24.454
27.104
35.462
44.965
46.787
Pseudo Pr_zero Rank
RD
SE
WB CCI
4.901
3.073
2.323
1.507
1.651
0.082
0.543
0.873
0.978
1.130
1.352
1.030
1.944
1.975
2.018
1.716
2.644
4.318
5.006
1.162
4.593
3.611
0.033
0.006
0.001
0.001
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.017
0.025
0.000
0.000
0.000
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
0
0
-2
-10
-2
-5
-3
-9
-10
2
-1
-3
-3
-4
11
13
-7
12
-4
-35
-20
-5
RD
TI CPI
0
0
-1
-12
-2
-6
-2
-9
-9
0
-2
-2
-2
-5
10
13
-7
12
-2
-28
-9
-10
Country
Austria
Colombia
Portugal
South Africa
Luxembourg
El Salvador
Morocco
Malaysia
Turkey
Ukraine
Lithuania
Poland
Kuwait
Trinidad and Tobago
China
Slovakia
Czech Republic
Hungary
Tunisia
Brazil
Venezuela
Qatar
Honduras
PACI
AT
CO
PT
ZA
LU
SV
MA
MY
TR
UA
LT
PL
KW
TT
CN
SK
CZ
HU
TN
BR
VE
QA
HN
56.169
73.778
84.131
91.773
92.093
106.391
123.176
124.491
131.103
131.207
133.812
140.782
158.301
163.236
163.447
166.042
170.155
174.817
215.041
217.255
233.422
235.724
239.167
Pseudo PR zero Rank Rank Rank TI
SE
WB CCI
CPI
3.592
0.000
23
14
15
9.736
0.066
24
-27
-20
10.615 0.028
25
5
5
6.730
0.013
26
-2
-10
18.888 0.338
27
14
16
13.832 0.391
28
-36
-12
14.352 0.197
29
-27
-31
12.179 0.002
30
-8
1
15.347 0.001
31
-14
-18
25.209 0.218
32
-46
-47
26.083 0.474
33
-6
-1
20.725 0.002
34
-6
-19
16.904 0.283
35
10
0
14.447 0.542
36
-11
-10
23.576 0.000
37
-38
-22
17.861 0.164
38
9
1
15.540 0.009
39
9
1
10.952 0.032
40
14
9
25.456 0.395
41
-7
8
15.684 0.000
42
-10
-5
8.908
0.032
43
-51
-55
18.564 0.280
44
22
19
26.010 0.433
45
-37
-39
ALL
PACI
PACI
Pr_
zero
z _ cases z
Country
PACI
Pseduo SE
Pr_zero
Rank
RD
WB CCI
RD
TI CPI
Saudi Arabia
Jordan
Russian
Federation
Romania
Slovenia
Belarus
Thailand
India
Peru
Greece
Ecuador
Pakistan
Brunei
Philippines
Latvia
Afghanistan
Iran
Algeria
Costa Rica
Oman
Bosnia
Croatia
Panama
Yemen
Benin
SA
JO
242.439
243.096
22.404
43.412
0.005
0.663
46
47
-3
12
-8
20
RU
RO
SI
BY
TH
IN
PE
GR
EC
PK
BN
PH
LV
AF
IR
DZ
CR
OM
BA
HR
PA
YE
BJ
247.858
248.097
256.003
258.200
269.053
281.452
318.676
344.656
364.359
367.443
408.515
450.541
461.958
502.132
520.407
526.699
540.679
560.208
565.299
574.016
635.977
649.220
676.352
25.325
34.736
27.805
49.376
30.999
21.186
36.204
38.489
31.576
59.826
96.385
18.712
52.022
87.605
45.083
33.827
51.888
77.058
91.213
49.984
44.612
99.887
109.330
0.001
0.089
0.310
0.679
0.012
0.000
0.285
0.055
0.439
0.337
0.783
0.070
0.649
0.819
0.215
0.219
0.228
0.585
0.838
0.419
0.533
0.735
0.863
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
-37
-5
29
-40
2
-9
-3
22
-25
-43
21
-14
26
-51
-5
-2
32
29
13
25
8
-17
-23
-46
-12
27
-30
7
-12
4
16
-36
-48
30
-32
19
-27
-4
-11
22
43
2
11
16
-8
-1
Macedonia
Congo
Cameroon
MK
CG
CM
679.688
684.171
712.286
133.146
103.282
149.332
0.863
0.864
0.869
71
72
73
3
-26
-28
-5
-23
-30
Country
Argentina
Indonesia
Viet Nam
Iraq
Bulgaria
Serbia
Syrian Arab
Republic
Egypt
Sudan
Senegal
Gabon
Côte d'Ivoire
Haiti
Bahrain
Madagascar
Bolivia
Ghana
Georgia
Kenya
Bangladesh
Azerbaijan
Turkmenistan
Niger
Chad
Mozambique
Libya
Equatorial
Guinea
PACI
Pseudo SE
Pr zero
Rank RD WB RD TI CPI
CCI
74
8
2
75
-15
-25
76
-8
-7
77
-36
-25
78
34
35
79
18
17
AR
ID
VN
IQ
BG
RS
758.380
771.134
847.251
879.072
880.925
986.975
34.228
71.228
98.604
109.339
90.900
108.122
0.138
0.097
0.307
0.505
0.452
0.603
SY
EG
SD
SN
GA
CI
HT
BH
MG
BO
GH
GE
KE
BD
AZ
TM
NE
TD
MZ
LY
993.677
1042.357
1043.116
1172.681
1282.609
1316.462
1322.406
1340.288
1541.492
1698.260
1744.195
1834.977
1857.055
1860.802
1982.937
2108.595
2156.377
2438.516
2453.064
2685.393
98.935
71.320
141.505
110.206
140.093
115.333
80.269
179.296
315.927
136.430
155.411
185.225
167.168
177.201
168.353
281.971
435.470
499.616
374.126
212.317
0.669
0.316
0.826
0.843
0.791
0.796
0.797
0.689
0.937
0.889
0.751
0.849
0.724
0.724
0.817
0.910
0.955
0.960
0.922
0.594
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
1
12
-29
37
13
-20
-24
56
45
6
32
32
-3
-14
-2
-14
16
-11
28
7
23
26
-25
25
14
-24
-27
61
15
3
39
-6
-14
-22
-7
-17
3
-17
23
10
GQ
2707.104
404.723
0.895
100
-15
-8
The PACI
Country
Nigeria
Djibouti
Tanzania
Kazakstan
Mongolia
Mali
Zimbabwe
Burma
Nepal
Cambodia
Uzbekistan
Sao Tome and P.
Zambia
Liberia
Uganda
PACI
2753.113
3017.412
3032.193
3175.626
3264.273
3728.458
4076.816
4164.303
4774.961
4822.404
5144.979
5609.297
6285.890
6399.922
9018.875
PseudoS.E.
172.999
512.072
425.489
217.250
608.289
352.223
837.928
886.369
855.104
773.506
532.195
352.414
557.955
899.955
801.283
Pr zero
0.202
0.967
0.876
0.500
0.970
0.948
0.952
0.976
0.979
0.940
0.907
0.982
0.938
0.896
0.925
Rank
Rank
difference difference
Country Rank WB CCI
TI CPI
101
-1
-9
102
25
34
103
27
34
104
7
24
105
33
42
106
43
39
107
1
25
108
-6
-3
109
35
19
110
6
14
111
8
12
112
23
34
113
26
28
114
15
10
115
27
28
Spearman Rank Correlations between different indexes of corruption
PACI1
PACI1
1
PACI2
0.861
PACI2
PACI3
PACI4
TI-CPI
WB-CC
1
(115)
PACI3
PACI4
TI-CPI
WB-CC
TI-GCB
0.934
0.806
1
(115)
(115)
0.797
0.935
0.859
(115)
(115)
(115)
-0.787
-0.706
-0.743
-0.675
(115)
(115)
(115)
(115)
-0.781
-0.692
-0.718
-0.642
-0.956
(114)
(114)
(114)
(114)
(114)
0.772
0.717
0.753
0.715
-0.795
1
1
1
-0.763
(54) (the same
(54)values shown
(54)in Table 5;(54)
(54)
1: All cases, (54)
all administrations
our preferred
index)
2: All cases, with the exclusion of health and telecom administration.
3: Only “positive” and “ongoing” cases, all administrations.
4: Only “positive” and “ongoing” cases, with the exclusion of health and telecom administrations.
Variable
Ln GDP per capita recent year
Democratic since 1995
WB-CCI
(Treisman 2007
Table 3 Column 4)
1.095***
(0.107)
Ln(Paci)
(All cases, all administrations)
-2.065***
(0.218)
Dependent variable
WB-CCI
(Treisman 2007
Table 4 Column 8)
0.518***
(0.072)
0.091
(0.192)
Ln(Paci)
(All cases, all administrations)
-1.099***
(0.201)
-1.323***
(0.480)
FH Press Freedom
Newspaper circulation 1996
0.015***
(0.003)
0.001
(0.001)
-0.002
(0.007)
-0.002
(0.002)
-0.074
(0.048)
0.006
(0.144)
0.304*
(0.174)
-0.078
(0.379)
0.266*
(0.147)
-0.586***
(0.246)
-0.052
(0.149)
-0.383
(0.333)
0.038
(0.258)
-0.912*
(0.459)
0.04
(0.175)
-0.549
(0.367)
0.298
(0.325)
0.135
(0.388)
-0.841
(0.592)
-1.882**
(0.920)
Presidential democracy
Former British colony
Former French colony
Former colony of other power, except
ES, PT
British legal tradition
French legal tradition
German legal tradition
Scandinavian legal tradition
Variable
Ln GDP per capita
recent year
WB-CCI
(Treisman 2007
Table 3 Column 4)
Ln(Paci)
(All cases, all
administrations)
WB-CCI
(Treisman 2007
Table 4 Column 8)
Ln(Paci)
(All cases, all
administrations)
1.095***
(0.107
-2.065***
(0.218)
0.518***
(0.072)
0.091
(0.192)
-1.099***
(0.201)
-1.323***
(0.480)
0.015***
(0.003)
-0.002
(0.007)
0.001
(0.001)
-0.002
(0.002)
-0.074
(0.048)
0.006
(0.144)
0.304*
(0.174)
-0.078
(0.379)
0.266*
(0.147)
-0.586***
(0.246)
-0.052
(0.149)
-0.383
(0.333)
0.038
(0.258)
-0.912*
(0.459)
0.04
(0.175)
-0.549
(0.367)
0.298
(0.325)
-0.841
(0.592)
0.135
(0.388)
-1.882**
(0.920)
0.007)
(0.004)
0.010
(0.011)
-0.000
(0.002)
-0.006
(0.005)
0.003
(0.002)
-0.000
(0.005)
-4.171***
(0.725)
16.349***
(1.813)
Democratic since 1995
FH Press Freedom
Newspaper circulation
1996
Presidential democracy
Former British colony
Former French colony
Former colony of other
power, except ES, PT
British legal tradition
French legal tradition
German legal tradition
Scandinavian legal
tradition
Percent protestant 1980
Percent catholic 1980
Percent muslim 1980
Constant
-9.603***
(1.035)
23.433***
(2.038)
Bribe Payer´s Index
Country
ISO
BPI
Pr_zero
Rank diff, WB CCI
Portugal
Bulgaria
Slovakia
Norway
Denmark
Hungary
Poland
Finland
Belgium
Korea
Japan
Italy
Sweden
Turkey
Australia
Argentina
Germany
Canada
Spain
PT
BG
SK
NO
DK
HU
PL
FI
BE
KR
JP
IT
SE
TR
AU
AR
DE
CA
ES
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
24.012
27.957
30.803
38.860
46.096
47.689
48.345
89.071
92.218
94.297
0.753
0.673
0.436
0.342
0.337
0.255
0.233
0.218
0.024
0.000
0.000
0.002
0.076
0.114
0.015
0.126
0.000
0.013
0.014
6.47
United St. of America
Netherlands
France
Switzerland
Austria
United Kingdom
Czech Republic
US
NL
FR
CH
AT
UK
CZ
102.513
105.907
121.359
171.184
222.063
253.290
338.839
0.000
0.009
0.001
0.030
0.105
0.006
0.229
7.22
7.28
6.5
7.81
7.5
7.39
7.22
5.83
7.1
5.94
7.62
5.23
7.59
7.3
7.34
7.46
6.63
Conclusion
o New cross national measure of corruption based on new set of judicial
statistics on occurrence of corruption
o Particular informational content allows to control for cross country variability
in the probability of detection of corruption
oNarrow definition of corruption: propensity of public officials to accept
bribe from firms
o Also informative of general level of corruption in the country
o Allows for easy interpretation vs perception and subjective based measure
o Based on clear assumptions
oOpen data would lead to development of corruption and governance
measures based on hard data
Discussion of Assumptions
1. In the real world, countries may apply a discriminatory policy when pursuing crossborder corruption cases
-
Target heavy trade partners, strategic trade policy
Results: Existence of bilateral agreements modest (McLean). See also Choi and Davis
However, include all countries as reference hence heterogeneity
2. Sector-specific characteristics of certain corrupt transactions
-
Ex. Different exposure to trade
How to address: consider sectoral decomposition but data problems
3. Exports vs FDI
4. Differences in the size and scope of the public sector
Conclusion
oPACI has high rank correlation with WB CCI and TI CPI
oAllow to understand different measures
oComplementarity
oContribute to understanding of causes and consequences of
corruption
End.
Thank you for your attention.