Power over Prosecutors Corrupts Politicians: Cross Country Evidence Using a New Indicator

1
Power over Prosecutors Corrupts
PoliticiansCross Country Evidence Using a New
Indicator

• Stefan VoigtUniversity of Kassel and ICER,
• Lars P. Feld University of Marburg and CESifo,
• Anne van AakenMax Planck Institute Heidelberg

2
1. Introduction

• Possible reforms of prosecution agencies
  discussed in quite a few countries
• One reason members of executive had put
  pressure on prosecutors
• Hypotheses here

• pressure is a function of institutional set-up of
  prosecution agencies
• high degree of government influence on
  prosecutors will, c.p., lead to higher levels of
  government crimes, including corruption

3

• Paper combines economics of prosecution agencies
  with economics of corruption.
• Very little on first topic (Aaken, Salzberger,
  Voigt 2004 first attempt)
• Two branches within the analysis of corruption
• (1) (Economic) consequences of corruption (c.
  exogenous)
• (2) Causes of corruption (c. as endogenous)
• Second branch here relevant Corruption can be
  explained by drawing on
• regulatory policies (Ades/di Tella 1999),
• the level of economic development (Treisman
  2000),
• historical and cultural factors (Treisman 2000)
• electoral institutions (Persson et al. 2003,
  Golden/Chang 2001)

4

• and in addition to these hypotheses by
  drawing on
• the structure of institutions set up for
  prosecuting crimes.
• ? This paper
• Makes organizational structure of prosecution
  agencies comparable by introducing a de jure and
  a de facto indicator
• Estimates effects of organizational set-up on
  (perceived) corruption.
• Main finding
• De facto prosecutorial independence leads to
  lower levels of corruption.

5

• Organization
• Introduction
• Some Theory
• Introducing Two New Indicators
• Estimation Approach
• Estimation Results
• Conclusions and Open Questions

6
2. Some Theory

• Procuracy as generic term for prosecution
  agencies
• Corruption the misuse of entrusted power for
  private benefit
• Institutional Structure of procuracy ? incentives
  of prosecutors
• More specifically prosecutors who are subject to
  pressure from the executive (and/or legislature)
  less likely to prosecute crimes committed by
  government members ? expected utility of
  committing crime ? (? corruption ?)
• Additionally
• ? higher degrees of separation of powers
• ? likelihood prosecution ?
• ? higher degrees of federalism
• ? likelihood prosecution ?

7
3. Two New Indicators for Prosecutorial
Independence

• 3.1 De Jure Prosecutorial Independence
• 16 variables grouped into 5 subindicators
• each variable can take on value between 0 and 1
• sum of variables divided by number of variables
  for which data available
• ? indicator between 0 and 1 with higher values
  indicating more independence.

8

• Subindicator 1 General Institutional Traits of
  Procuracy
• ? Procuracy mentioned in the Constitution?
• ? Formal qualification requirements?
• ? Difficulty of removing prosecutors
• ? General rule for allocating incoming cases?
• Subindicator 2 Personal Independence of
  Prosecutors
• ? Term length
• ? Renewability of term
• ? Appointing organ
• ? Promotion
• ? Removal from office
• ? Transfer against own will

9

• Subindicator 3 Formal Independence of
  Prosecutors
• ? Internal Orders?
• ? External Orders?
• ? Right to Substitute Prosecutors working on
  specific case?
• Subindicator 4 Monopoly to Prosecute?
• ? Monopoly to Prosecute?
• ? Judicial Review of (Non-)Prosecution Decisions?
• Subindicator 5 Degree of Discretion in
  Prosecution
• ? Legality vs. Opportunity Principle

10
3.2 De facto Prosecutorial Independence

• Constructed in the same way as de jure indicator
• But very sticky (1990-2000 1960-2000)
• 6 variables
• Prosecutors forced to retire against their will?
• Prosecutors removed from office against their
  will?
• Number of changes in legal foundations?
• Income of prosecutors at least constant in real
  terms since 1960?
• Budget of procuracy at least constant in real
  terms since 1960?
• Number of cases initiated by others than
  procuracy?

11
3.3 Some Stock Taking

• De jure and de facto PI deviate strongly from
  each other
• R2 -0.338

12
4. The Estimation Approach
CPI ?0 ?1 de jure PI ?2 de facto PI ?3
JI ?4 LegOr ?5 Regime ?6 X ?

• with
• CPI Corruption Perception Index, average from
  1998 2003,
• (source Transparency International)
• JI vector of de jure and de facto judicial
  independence,
• (source Feld and Voigt 2003)
• LegOr Legal Origin,
• Regime Vector of political regime variables
  (federalism parl./pres.systems)
• X Vector of economic controls (GDP/cap.,
  pop.size, trade openness etc.)

13
5. Estimation Results
14
Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications
Variables (1)
De iure Prosecutorial Independence -2.371(2.52)
De facto Prosecutorial Independence
De iure Judicial Independence
De facto Judicial Independence
Real GDP per capita in 1980 (in 1'000) 0.332(13.16)
Population Size in Million Inhabitants in 1998 0.682(0.84)
Trade Openness in 1998 (in of GDP) 0.008()(1.83)
Constant 2.407
R-2 0.745
SER 1.197
J.-B. 0.512
Observations 65
15
Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications
Variables (1) (2)
De iure Prosecutorial Independence -2.371(2.52)
De facto Prosecutorial Independence 1.146(2.38)
De iure Judicial Independence
De facto Judicial Independence
Real GDP per capita in 1980 (in 1'000) 0.332(13.16) 0.317(11.87)
Population Size in Million Inhabitants in 1998 0.682(0.84) 0.918(0.89)
Trade Openness in 1998 (in of GDP) 0.008()(1.83) 0.009()(1.84)
Constant 2.407 0.788
R-2 0.745 0.737
SER 1.197 1.223
J.-B. 0.512 0.820
Observations 65 62
16
Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications
Variables (1) (2) (3)
De iure Prosecutorial Independence -2.371(2.52) -2.949(2.98)
De facto Prosecutorial Independence 1.146(2.38) 0.954(2.09)
De iure Judicial Independence
De facto Judicial Independence
Real GDP per capita in 1980 (in 1'000) 0.332(13.16) 0.317(11.87) 0.312(12.43)
Population Size in Million Inhabitants in 1998 0.682(0.84) 0.918(0.89) 0.666(0.69)
Trade Openness in 1998 (in of GDP) 0.008()(1.83) 0.009()(1.84) 0.009()(1.96)
Constant 2.407 0.788 2.296
R-2 0.745 0.737 0.769
SER 1.197 1.223 1.146
J.-B. 0.512 0.820 0.555
Observations 65 62 62
17
Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications Table 1 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Baseline Specifications
Variables (1) (2) (3) (4)
De iure Prosecutorial Independence -2.371(2.52) -2.949(2.98) -3.031(2.91)
De facto Prosecutorial Independence 1.146(2.38) 0.954(2.09) 0.861()(1.73)
De iure Judicial Independence -1.431(1.21)
De facto Judicial Independence 0.840()(1.76)
Real GDP per capita in 1980 (in 1'000) 0.332(13.16) 0.317(11.87) 0.312(12.43) 0.303(10.19)
Population Size in Million Inhabitants in 1998 0.682(0.84) 0.918(0.89) 0.666(0.69) 0.611(0.61)
Trade Openness in 1998 (in of GDP) 0.008()(1.83) 0.009()(1.84) 0.009()(1.96) 0.007(1.30)
Constant 2.407 0.788 2.296 2.860
R-2 0.745 0.737 0.769 0.769
SER 1.197 1.223 1.146 1.170
J.-B. 0.512 0.820 0.555 1.520
Observations 65 62 62 53
18
Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis
Variables (1)
De iure Prosecutorial Independence -1.829()(1.96)
De facto PI 0.975(2.43)
De iure Judicial Independence
De facto Judicial Independence
English Legal Origin -2.154(3.75)
Socialist Legal Origin -3.065(5.24)
French Legal Origin -2.640(4.94)
German Legal Origin -2.023(3.47)
Parliamentary System
Checks and Balances
Federalism
R-2 0.840
SER 0.952
J.-B. 2.253
Observations 62
The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998.
19
Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis
Variables (1) (2)
De iure Prosecutorial Independence -1.829()(1.96) -2.569(2.50)
De facto PI 0.975(2.43) 0.776(1.63)
De iure Judicial Independence
De facto Judicial Independence
English Legal Origin -2.154(3.75)
Socialist Legal Origin -3.065(5.24)
French Legal Origin -2.640(4.94)
German Legal Origin -2.023(3.47)
Parliamentary System 0.282(1.26)
Checks and Balances
Federalism
R-2 0.840 0.771
SER 0.952 1.140
J.-B. 2.253 0.495
Observations 62 62
The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998.
20
Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis
Variables (1) (2) (3)
De iure Prosecutorial Independence -1.829()(1.96) -2.569(2.50) -2.951(3.11)
De facto PI 0.975(2.43) 0.776(1.63) 1.164(2.61)
De iure Judicial Independence
De facto Judicial Independence
English Legal Origin -2.154(3.75)
Socialist Legal Origin -3.065(5.24)
French Legal Origin -2.640(4.94)
German Legal Origin -2.023(3.47)
Parliamentary System 0.282(1.26)
Checks and Balances -0.310(2.41)
Federalism
R-2 0.840 0.771 0.787
SER 0.952 1.140 1.100
J.-B. 2.253 0.495 1.906
Observations 62 62 62
The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998.
21
Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis
Variables (1) (2) (3) (4)
De iure Prosecutorial Independence -1.829()(1.96) -2.569(2.50) -2.951(3.11) -2.638(2.68)
De facto PI 0.975(2.43) 0.776(1.63) 1.164(2.61) 0.966(2.16)
De iure Judicial Independence
De facto Judicial Independence
English Legal Origin -2.154(3.75)
Socialist Legal Origin -3.065(5.24)
French Legal Origin -2.640(4.94)
German Legal Origin -2.023(3.47)
Parliamentary System 0.282(1.26)
Checks and Balances -0.310(2.41)
Federalism -0.766()(1.81)
R-2 0.840 0.771 0.787 0.778
SER 0.952 1.140 1.100 1.124
J.-B. 2.253 0.495 1.906 1.172
Observations 62 62 62 62
The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998.
22
Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis Table 2 OLS-Regressions of the Average Corruption Perception Index between 1998 and 2003 on Prosecutorial Independence and Controls, Robustness Analysis
Variables (1) (2) (3) (4) (5)
De iure Prosecutorial Independence -1.829()(1.96) -2.569(2.50) -2.951(3.11) -2.638(2.68) -2.129()(1.86)
De facto PI 0.975(2.43) 0.776(1.63) 1.164(2.61) 0966(2.16) 0.915()(1.73)
De iure Judicial Independence -0255(0.23)
De facto Judicial Independence 0.581(1.23)
English Legal Origin -2.154(3.75) -2.257(3.10)
Socialist Legal Origin -3.065(5.24) -2.913(4.11)
French Legal Origin -2.640(4.94) -2.521(3.92)
German Legal Origin -2.023(3.47) -2.116(3.04)
Parliamentary System 0.282(1.26) 0.096(0.40)
Checks and Balances -0.310(2.41) -0.034(0.20)
Federalism -0.766()(1.81) 0.029(0.06)
R-2 0.840 0.771 0.787 0.778 0.830
SER 0.952 1.140 1.100 1.124 1.004
J.-B. 2.253 0.495 1.906 1.172 0.310
Observations 62 62 62 62 53
The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998. The numbers in parentheses are the absolute values of the estimated t-statistics. , or () show that the estimated parameter is significantly different from zero on the 1, 5, or 10 percent level, respectively. SER is the standard error of the regression, and J. -B. the value of the Jarque-Bera-test on normality of the residuals. All regressions contain the standard controls real GDP per capita of 1980, openness in 1998 and population in 1998.
23
6 Conclusions and Open Questions

• De facto PI has expected effect on corruption
• But How to explain that de jure PI has opposite
  effect?
• Reverse causality possible reason
• ? (Re-)estimate, possible with useful instrument
• Other possible next steps
• ? Include false positives, i.e. prosecution of
  crimes that have never been committed
• Include the police into analysis.
• Endogenize PI